* [RFC PATCH v2 01/48] netfs: Fix netfs_extract_iter_to_sg() for ITER_UBUF/IOVEC
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 02/48] iov_iter: Remove last_offset member David Howells
` (46 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Steve French,
Shyam Prasad N, Rohith Surabattula, linux-cachefs, linux-cifs
Fix netfs_extract_iter_to_sg() for ITER_UBUF and ITER_IOVEC to set the size
of the page to the part of the page extracted, not the remaining amount of
data in the extracted page array at that point.
This doesn't yet affect anything as cifs, the only current user, only
passes in non-user-backed iterators.
Fixes: 018584697533 ("netfs: Add a function to extract an iterator into a scatterlist")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: Steve French <sfrench@samba.org>
cc: Shyam Prasad N <nspmangalore@gmail.com>
cc: Rohith Surabattula <rohiths.msft@gmail.com>
cc: linux-cachefs@redhat.com
cc: linux-cifs@vger.kernel.org
cc: linux-fsdevel@vger.kernel.org
---
fs/netfs/iterator.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/fs/netfs/iterator.c b/fs/netfs/iterator.c
index e9a45dea748a..8a4c86687429 100644
--- a/fs/netfs/iterator.c
+++ b/fs/netfs/iterator.c
@@ -139,7 +139,7 @@ static ssize_t netfs_extract_user_to_sg(struct iov_iter *iter,
size_t seg = min_t(size_t, PAGE_SIZE - off, len);
*pages++ = NULL;
- sg_set_page(sg, page, len, off);
+ sg_set_page(sg, page, seg, off);
sgtable->nents++;
sg++;
len -= seg;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 02/48] iov_iter: Remove last_offset member
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
2023-03-29 14:13 ` [RFC PATCH v2 01/48] netfs: Fix netfs_extract_iter_to_sg() for ITER_UBUF/IOVEC David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 03/48] iov_iter: Add an iterator-of-iterators David Howells
` (45 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, linux-nfs
With the removal of ITER_PIPE, the last_offset member of struct iov_iter is
no longer used, so remove it and un-unionise the remaining member.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: Alexander Viro <viro@zeniv.linux.org.uk>
cc: Jeff Layton <jlayton@kernel.org>
cc: linux-nfs@vger.kernel.org
cc: linux-fsdevel@vger.kernel.org
cc: linux-mm@kvack.org
cc: netdev@vger.kernel.org
---
include/linux/uio.h | 5 +----
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/include/linux/uio.h b/include/linux/uio.h
index 74598426edb4..2d8a70cb9b26 100644
--- a/include/linux/uio.h
+++ b/include/linux/uio.h
@@ -43,10 +43,7 @@ struct iov_iter {
bool nofault;
bool data_source;
bool user_backed;
- union {
- size_t iov_offset;
- int last_offset;
- };
+ size_t iov_offset;
size_t count;
union {
const struct iovec *iov;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 03/48] iov_iter: Add an iterator-of-iterators
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
2023-03-29 14:13 ` [RFC PATCH v2 01/48] netfs: Fix netfs_extract_iter_to_sg() for ITER_UBUF/IOVEC David Howells
2023-03-29 14:13 ` [RFC PATCH v2 02/48] iov_iter: Remove last_offset member David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag David Howells
` (44 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Trond Myklebust,
Anna Schumaker, linux-nfs
Add a new I/O iterator type, ITER_ITERLIST, that allows iteration over a
series of I/O iterators, provided the iterators are all the same direction
(all ITER_SOURCE or all ITER_DEST) and none of them are themselves
ITER_ITERLIST (this function is recursive).
To make reversion possible, I've added an 'orig_count' member into the
iov_iter struct so that reversion of an ITER_ITERLIST can know when to go
move backwards through the iter list. It might make more sense to make the
iterator list element, say:
struct itervec {
struct iov_iter iter;
size_t orig_count;
};
rather than expanding struct iov_iter itself and have iov_iter_iterlist()
set vec[i].orig_count from vec[i].iter->count.
Also, for the moment, I've only permitted its use with source iterators
(eg. sendmsg).
To use this, you allocate an array of iterators and point the list iterator
at it, e.g.:
struct iov_iter iters[3];
struct msghdr msg;
iov_iter_bvec(&iters[0], ITER_SOURCE, &head_bv, 1,
sizeof(marker) + head->iov_len);
iov_iter_xarray(&iters[1], ITER_SOURCE, xdr->pages,
xdr->page_fpos, xdr->page_len);
iov_iter_kvec(&iters[2], ITER_SOURCE, &tail_kv, 1,
tail->iov_len);
iov_iter_iterlist(&msg.msg_iter, ITER_SOURCE, iters, 3, size);
This can be used by network filesystem protocols, such as sunrpc, to glue a
header and a trailer on to some data to form a message and then dump the
entire message onto the socket in a single go.
[!] Note: I'm not entirely sure that this is a good idea: the problem is
that it's reasonably common practice to copy an iterator by direct
assignment - and that works for the existing iterators... but not this
one. With the iterator-of-iterators, the list of iterators has to be
modified if we recurse. It's probably fine just for calling sendmsg()
from network filesystems, but I'm not 100% sure of that.
Suggested-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Trond Myklebust <trond.myklebust@hammerspace.com>
cc: Anna Schumaker <anna@kernel.org>
cc: Chuck Lever <chuck.lever@oracle.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: Alexander Viro <viro@zeniv.linux.org.uk>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: linux-nfs@vger.kernel.org
cc: linux-fsdevel@vger.kernel.org
cc: linux-mm@kvack.org
cc: netdev@vger.kernel.org
---
include/linux/uio.h | 13 ++-
lib/iov_iter.c | 254 ++++++++++++++++++++++++++++++++++++++++++--
2 files changed, 260 insertions(+), 7 deletions(-)
diff --git a/include/linux/uio.h b/include/linux/uio.h
index 2d8a70cb9b26..6c75c94566b8 100644
--- a/include/linux/uio.h
+++ b/include/linux/uio.h
@@ -27,6 +27,7 @@ enum iter_type {
ITER_XARRAY,
ITER_DISCARD,
ITER_UBUF,
+ ITER_ITERLIST,
};
#define ITER_SOURCE 1 // == WRITE
@@ -45,12 +46,14 @@ struct iov_iter {
bool user_backed;
size_t iov_offset;
size_t count;
+ size_t orig_count;
union {
const struct iovec *iov;
const struct kvec *kvec;
const struct bio_vec *bvec;
struct xarray *xarray;
void __user *ubuf;
+ struct iov_iter *iterlist;
};
union {
unsigned long nr_segs;
@@ -101,6 +104,11 @@ static inline bool iov_iter_is_xarray(const struct iov_iter *i)
return iov_iter_type(i) == ITER_XARRAY;
}
+static inline bool iov_iter_is_iterlist(const struct iov_iter *i)
+{
+ return iov_iter_type(i) == ITER_ITERLIST;
+}
+
static inline unsigned char iov_iter_rw(const struct iov_iter *i)
{
return i->data_source ? WRITE : READ;
@@ -235,6 +243,8 @@ void iov_iter_bvec(struct iov_iter *i, unsigned int direction, const struct bio_
void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count);
void iov_iter_xarray(struct iov_iter *i, unsigned int direction, struct xarray *xarray,
loff_t start, size_t count);
+void iov_iter_iterlist(struct iov_iter *i, unsigned int direction, struct iov_iter *iterlist,
+ unsigned long nr_segs, size_t count);
ssize_t iov_iter_get_pages(struct iov_iter *i, struct page **pages,
size_t maxsize, unsigned maxpages, size_t *start,
iov_iter_extraction_t extraction_flags);
@@ -342,7 +352,8 @@ static inline void iov_iter_ubuf(struct iov_iter *i, unsigned int direction,
.user_backed = true,
.data_source = direction,
.ubuf = buf,
- .count = count
+ .count = count,
+ .orig_count = count,
};
}
/* Flags for iov_iter_get/extract_pages*() */
diff --git a/lib/iov_iter.c b/lib/iov_iter.c
index fad95e4cf372..8a9ae4af45fc 100644
--- a/lib/iov_iter.c
+++ b/lib/iov_iter.c
@@ -282,7 +282,8 @@ void iov_iter_init(struct iov_iter *i, unsigned int direction,
.iov = iov,
.nr_segs = nr_segs,
.iov_offset = 0,
- .count = count
+ .count = count,
+ .orig_count = count,
};
}
EXPORT_SYMBOL(iov_iter_init);
@@ -364,6 +365,26 @@ size_t _copy_from_iter(void *addr, size_t bytes, struct iov_iter *i)
if (WARN_ON_ONCE(!i->data_source))
return 0;
+ if (unlikely(iov_iter_is_iterlist(i))) {
+ size_t copied = 0;
+
+ while (bytes && i->count) {
+ size_t part = min(bytes, i->iterlist->count), n;
+
+ if (part > 0)
+ n = _copy_from_iter(addr, part, i->iterlist);
+ addr += n;
+ copied += n;
+ bytes -= n;
+ i->count -= n;
+ if (n < part || !bytes)
+ break;
+ i->iterlist++;
+ i->nr_segs--;
+ }
+ return copied;
+ }
+
if (user_backed_iter(i))
might_fault();
iterate_and_advance(i, bytes, base, len, off,
@@ -380,6 +401,27 @@ size_t _copy_from_iter_nocache(void *addr, size_t bytes, struct iov_iter *i)
if (WARN_ON_ONCE(!i->data_source))
return 0;
+ if (unlikely(iov_iter_is_iterlist(i))) {
+ size_t copied = 0;
+
+ while (bytes && i->count) {
+ size_t part = min(bytes, i->iterlist->count), n;
+
+ if (part > 0)
+ n = _copy_from_iter_nocache(addr, part,
+ i->iterlist);
+ addr += n;
+ copied += n;
+ bytes -= n;
+ i->count -= n;
+ if (n < part || !bytes)
+ break;
+ i->iterlist++;
+ i->nr_segs--;
+ }
+ return copied;
+ }
+
iterate_and_advance(i, bytes, base, len, off,
__copy_from_user_inatomic_nocache(addr + off, base, len),
memcpy(addr + off, base, len)
@@ -411,6 +453,27 @@ size_t _copy_from_iter_flushcache(void *addr, size_t bytes, struct iov_iter *i)
if (WARN_ON_ONCE(!i->data_source))
return 0;
+ if (unlikely(iov_iter_is_iterlist(i))) {
+ size_t copied = 0;
+
+ while (bytes && i->count) {
+ size_t part = min(bytes, i->iterlist->count), n;
+
+ if (part > 0)
+ n = _copy_from_iter_flushcache(addr, part,
+ i->iterlist);
+ addr += n;
+ copied += n;
+ bytes -= n;
+ i->count -= n;
+ if (n < part || !bytes)
+ break;
+ i->iterlist++;
+ i->nr_segs--;
+ }
+ return copied;
+ }
+
iterate_and_advance(i, bytes, base, len, off,
__copy_from_user_flushcache(addr + off, base, len),
memcpy_flushcache(addr + off, base, len)
@@ -514,7 +577,31 @@ EXPORT_SYMBOL(iov_iter_zero);
size_t copy_page_from_iter_atomic(struct page *page, unsigned offset, size_t bytes,
struct iov_iter *i)
{
- char *kaddr = kmap_atomic(page), *p = kaddr + offset;
+ char *kaddr, *p;
+
+ if (unlikely(iov_iter_is_iterlist(i))) {
+ size_t copied = 0;
+
+ while (bytes && i->count) {
+ size_t part = min(bytes, i->iterlist->count), n;
+
+ if (part > 0)
+ n = copy_page_from_iter_atomic(page, offset, part,
+ i->iterlist);
+ offset += n;
+ copied += n;
+ bytes -= n;
+ i->count -= n;
+ if (n < part || !bytes)
+ break;
+ i->iterlist++;
+ i->nr_segs--;
+ }
+ return copied;
+ }
+
+ kaddr = kmap_atomic(page);
+ p = kaddr + offset;
if (!page_copy_sane(page, offset, bytes)) {
kunmap_atomic(kaddr);
return 0;
@@ -585,19 +672,49 @@ void iov_iter_advance(struct iov_iter *i, size_t size)
iov_iter_bvec_advance(i, size);
} else if (iov_iter_is_discard(i)) {
i->count -= size;
+ }else if (iov_iter_is_iterlist(i)) {
+ i->count -= size;
+ for (;;) {
+ size_t part = min(size, i->iterlist->count);
+
+ if (part > 0)
+ iov_iter_advance(i->iterlist, part);
+ size -= part;
+ if (!size)
+ break;
+ i->iterlist++;
+ i->nr_segs--;
+ }
}
}
EXPORT_SYMBOL(iov_iter_advance);
+static void iov_iter_revert_iterlist(struct iov_iter *i, size_t unroll)
+{
+ for (;;) {
+ size_t part = min(unroll, i->iterlist->orig_count - i->iterlist->count);
+
+ if (part > 0)
+ iov_iter_revert(i->iterlist, part);
+ unroll -= part;
+ if (!unroll)
+ break;
+ i->iterlist--;
+ i->nr_segs++;
+ }
+}
+
void iov_iter_revert(struct iov_iter *i, size_t unroll)
{
if (!unroll)
return;
- if (WARN_ON(unroll > MAX_RW_COUNT))
+ if (WARN_ON(unroll > i->orig_count - i->count))
return;
i->count += unroll;
if (unlikely(iov_iter_is_discard(i)))
return;
+ if (unlikely(iov_iter_is_iterlist(i)))
+ return iov_iter_revert_iterlist(i, unroll);
if (unroll <= i->iov_offset) {
i->iov_offset -= unroll;
return;
@@ -641,6 +758,8 @@ EXPORT_SYMBOL(iov_iter_revert);
*/
size_t iov_iter_single_seg_count(const struct iov_iter *i)
{
+ if (iov_iter_is_iterlist(i))
+ i = i->iterlist;
if (i->nr_segs > 1) {
if (likely(iter_is_iovec(i) || iov_iter_is_kvec(i)))
return min(i->count, i->iov->iov_len - i->iov_offset);
@@ -662,7 +781,8 @@ void iov_iter_kvec(struct iov_iter *i, unsigned int direction,
.kvec = kvec,
.nr_segs = nr_segs,
.iov_offset = 0,
- .count = count
+ .count = count,
+ .orig_count = count,
};
}
EXPORT_SYMBOL(iov_iter_kvec);
@@ -678,7 +798,8 @@ void iov_iter_bvec(struct iov_iter *i, unsigned int direction,
.bvec = bvec,
.nr_segs = nr_segs,
.iov_offset = 0,
- .count = count
+ .count = count,
+ .orig_count = count,
};
}
EXPORT_SYMBOL(iov_iter_bvec);
@@ -706,6 +827,7 @@ void iov_iter_xarray(struct iov_iter *i, unsigned int direction,
.xarray = xarray,
.xarray_start = start,
.count = count,
+ .orig_count = count,
.iov_offset = 0
};
}
@@ -727,11 +849,47 @@ void iov_iter_discard(struct iov_iter *i, unsigned int direction, size_t count)
.iter_type = ITER_DISCARD,
.data_source = false,
.count = count,
+ .orig_count = count,
.iov_offset = 0
};
}
EXPORT_SYMBOL(iov_iter_discard);
+/**
+ * iov_iter_iterlist - Initialise an I/O iterator that is a list of iterators
+ * @iter: The iterator to initialise.
+ * @direction: The direction of the transfer.
+ * @iterlist: The list of iterators
+ * @nr_segs: The number of elements in the list
+ * @count: The size of the I/O buffer in bytes.
+ *
+ * Set up an I/O iterator that just discards everything that's written to it.
+ * It's only available as a source iterator (for WRITE), all the iterators in
+ * the list must be the same and none of them can be ITER_ITERLIST type.
+ */
+void iov_iter_iterlist(struct iov_iter *iter, unsigned int direction,
+ struct iov_iter *iterlist, unsigned long nr_segs,
+ size_t count)
+{
+ unsigned long i;
+
+ BUG_ON(direction != WRITE);
+ for (i = 0; i < nr_segs; i++) {
+ BUG_ON(iterlist[i].iter_type == ITER_ITERLIST);
+ BUG_ON(!iterlist[i].data_source);
+ }
+
+ *iter = (struct iov_iter){
+ .iter_type = ITER_ITERLIST,
+ .data_source = true,
+ .count = count,
+ .orig_count = count,
+ .iterlist = iterlist,
+ .nr_segs = nr_segs,
+ };
+}
+EXPORT_SYMBOL(iov_iter_iterlist);
+
static bool iov_iter_aligned_iovec(const struct iov_iter *i, unsigned addr_mask,
unsigned len_mask)
{
@@ -879,6 +1037,15 @@ unsigned long iov_iter_alignment(const struct iov_iter *i)
if (iov_iter_is_xarray(i))
return (i->xarray_start + i->iov_offset) | i->count;
+ if (iov_iter_is_iterlist(i)) {
+ unsigned long align = 0;
+ unsigned int j;
+
+ for (j = 0; j < i->nr_segs; j++)
+ align |= iov_iter_alignment(&i->iterlist[j]);
+ return align;
+ }
+
return 0;
}
EXPORT_SYMBOL(iov_iter_alignment);
@@ -1078,6 +1245,18 @@ static ssize_t __iov_iter_get_pages_alloc(struct iov_iter *i,
}
if (iov_iter_is_xarray(i))
return iter_xarray_get_pages(i, pages, maxsize, maxpages, start);
+ if (iov_iter_is_iterlist(i)) {
+ ssize_t size;
+
+ while (!i->iterlist->count) {
+ i->iterlist++;
+ i->nr_segs--;
+ }
+ size = __iov_iter_get_pages_alloc(i->iterlist, pages, maxsize, maxpages,
+ start, extraction_flags);
+ i->count -= size;
+ return size;
+ }
return -EFAULT;
}
@@ -1126,6 +1305,31 @@ ssize_t iov_iter_get_pages_alloc2(struct iov_iter *i,
}
EXPORT_SYMBOL(iov_iter_get_pages_alloc2);
+static size_t csum_and_copy_from_iterlist(void *addr, size_t bytes, __wsum *csum,
+ struct iov_iter *i)
+{
+ size_t copied = 0, n;
+
+ while (i->count && i->nr_segs) {
+ struct iov_iter *j = i->iterlist;
+
+ if (j->count == 0) {
+ i->iterlist++;
+ i->nr_segs--;
+ continue;
+ }
+
+ n = csum_and_copy_from_iter(addr, bytes - copied, csum, j);
+ addr += n;
+ copied += n;
+ i->count -= n;
+ if (n == 0)
+ break;
+ }
+
+ return copied;
+}
+
size_t csum_and_copy_from_iter(void *addr, size_t bytes, __wsum *csum,
struct iov_iter *i)
{
@@ -1133,6 +1337,8 @@ size_t csum_and_copy_from_iter(void *addr, size_t bytes, __wsum *csum,
sum = *csum;
if (WARN_ON_ONCE(!i->data_source))
return 0;
+ if (iov_iter_is_iterlist(i))
+ return csum_and_copy_from_iterlist(addr, bytes, csum, i);
iterate_and_advance(i, bytes, base, len, off, ({
next = csum_and_copy_from_user(base, addr + off, len);
@@ -1236,6 +1442,21 @@ static int bvec_npages(const struct iov_iter *i, int maxpages)
return npages;
}
+static int iterlist_npages(const struct iov_iter *i, int maxpages)
+{
+ ssize_t size = i->count;
+ const struct iov_iter *p;
+ int npages = 0;
+
+ for (p = i->iterlist; size; p++) {
+ size -= p->count;
+ npages += iov_iter_npages(p, maxpages - npages);
+ if (unlikely(npages >= maxpages))
+ return maxpages;
+ }
+ return npages;
+}
+
int iov_iter_npages(const struct iov_iter *i, int maxpages)
{
if (unlikely(!i->count))
@@ -1255,6 +1476,8 @@ int iov_iter_npages(const struct iov_iter *i, int maxpages)
int npages = DIV_ROUND_UP(offset + i->count, PAGE_SIZE);
return min(npages, maxpages);
}
+ if (iov_iter_is_iterlist(i))
+ return iterlist_npages(i, maxpages);
return 0;
}
EXPORT_SYMBOL(iov_iter_npages);
@@ -1266,11 +1489,14 @@ const void *dup_iter(struct iov_iter *new, struct iov_iter *old, gfp_t flags)
return new->bvec = kmemdup(new->bvec,
new->nr_segs * sizeof(struct bio_vec),
flags);
- else if (iov_iter_is_kvec(new) || iter_is_iovec(new))
+ if (iov_iter_is_kvec(new) || iter_is_iovec(new))
/* iovec and kvec have identical layout */
return new->iov = kmemdup(new->iov,
new->nr_segs * sizeof(struct iovec),
flags);
+ if (WARN_ON_ONCE(iov_iter_is_iterlist(old)))
+ /* Don't allow dup'ing of iterlist as the cleanup is complicated */
+ return NULL;
return NULL;
}
EXPORT_SYMBOL(dup_iter);
@@ -1759,6 +1985,22 @@ ssize_t iov_iter_extract_pages(struct iov_iter *i,
return iov_iter_extract_xarray_pages(i, pages, maxsize,
maxpages, extraction_flags,
offset0);
+ if (iov_iter_is_iterlist(i)) {
+ ssize_t size;
+
+ while (i->nr_segs && !i->iterlist->count) {
+ i->iterlist++;
+ i->nr_segs--;
+ }
+ if (!i->nr_segs) {
+ WARN_ON_ONCE(i->count);
+ return 0;
+ }
+ size = iov_iter_extract_pages(i->iterlist, pages, maxsize, maxpages,
+ extraction_flags, offset0);
+ i->count -= size;
+ return size;
+ }
return -EFAULT;
}
EXPORT_SYMBOL_GPL(iov_iter_extract_pages);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (2 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 03/48] iov_iter: Add an iterator-of-iterators David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-30 14:28 ` Willem de Bruijn
2023-03-30 15:07 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 05/48] mm: Move the page fragment allocator from page_alloc.c into its own file David Howells
` (43 subsequent siblings)
47 siblings, 2 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Willem de Bruijn
Declare MSG_SPLICE_PAGES, an internal sendmsg() flag, that hints to a
network protocol that it should splice pages from the source iterator
rather than copying the data if it can. This flag is added to a list that
is cleared by sendmsg and recvmsg syscalls on entry.
This is intended as a replacement for the ->sendpage() op, allowing a way
to splice in several multipage folios in one go.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
include/linux/socket.h | 3 +++
net/socket.c | 7 +++++++
2 files changed, 10 insertions(+)
diff --git a/include/linux/socket.h b/include/linux/socket.h
index 13c3a237b9c9..c2fa0f800999 100644
--- a/include/linux/socket.h
+++ b/include/linux/socket.h
@@ -327,6 +327,7 @@ struct ucred {
*/
#define MSG_ZEROCOPY 0x4000000 /* Use user data in kernel path */
+#define MSG_SPLICE_PAGES 0x8000000 /* Splice the pages from the iterator in sendmsg() */
#define MSG_FASTOPEN 0x20000000 /* Send data in TCP SYN */
#define MSG_CMSG_CLOEXEC 0x40000000 /* Set close_on_exec for file
descriptor received through
@@ -337,6 +338,8 @@ struct ucred {
#define MSG_CMSG_COMPAT 0 /* We never have 32 bit fixups */
#endif
+/* Flags to be cleared on entry by sendmsg, recvmsg, sendmmsg and recvmmsg syscalls */
+#define MSG_INTERNAL_FLAGS (MSG_SPLICE_PAGES)
/* Setsockoptions(2) level. Thanks to BSD these must match IPPROTO_xxx */
#define SOL_IP 0
diff --git a/net/socket.c b/net/socket.c
index 6bae8ce7059e..dfb912bbed62 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -2139,6 +2139,7 @@ int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
msg.msg_name = (struct sockaddr *)&address;
msg.msg_namelen = addr_len;
}
+ flags &= ~MSG_INTERNAL_FLAGS;
if (sock->file->f_flags & O_NONBLOCK)
flags |= MSG_DONTWAIT;
msg.msg_flags = flags;
@@ -2192,6 +2193,7 @@ int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
if (!sock)
goto out;
+ flags &= ~MSG_INTERNAL_FLAGS;
if (sock->file->f_flags & O_NONBLOCK)
flags |= MSG_DONTWAIT;
err = sock_recvmsg(sock, &msg, flags);
@@ -2579,6 +2581,7 @@ long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
return -EINVAL;
+ flags &= ~MSG_INTERNAL_FLAGS;
sock = sockfd_lookup_light(fd, &err, &fput_needed);
if (!sock)
@@ -2627,6 +2630,7 @@ int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
entry = mmsg;
compat_entry = (struct compat_mmsghdr __user *)mmsg;
err = 0;
+ flags &= ~MSG_INTERNAL_FLAGS;
flags |= MSG_BATCH;
while (datagrams < vlen) {
@@ -2775,6 +2779,7 @@ long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
struct user_msghdr __user *umsg,
struct sockaddr __user *uaddr, unsigned int flags)
{
+ flags &= ~MSG_INTERNAL_FLAGS;
return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
}
@@ -2787,6 +2792,7 @@ long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
return -EINVAL;
+ flags &= ~MSG_INTERNAL_FLAGS;
sock = sockfd_lookup_light(fd, &err, &fput_needed);
if (!sock)
@@ -2839,6 +2845,7 @@ static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
goto out_put;
}
}
+ flags &= ~MSG_INTERNAL_FLAGS;
entry = mmsg;
compat_entry = (struct compat_mmsghdr __user *)mmsg;
^ permalink raw reply related [flat|nested] 77+ messages in thread* RE: [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag
2023-03-29 14:13 ` [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag David Howells
@ 2023-03-30 14:28 ` Willem de Bruijn
2023-03-30 15:07 ` David Howells
1 sibling, 0 replies; 77+ messages in thread
From: Willem de Bruijn @ 2023-03-30 14:28 UTC (permalink / raw)
To: David Howells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Willem de Bruijn
David Howells wrote:
> Declare MSG_SPLICE_PAGES, an internal sendmsg() flag, that hints to a
> network protocol that it should splice pages from the source iterator
> rather than copying the data if it can. This flag is added to a list that
> is cleared by sendmsg and recvmsg syscalls on entry.
>
> This is intended as a replacement for the ->sendpage() op, allowing a way
> to splice in several multipage folios in one go.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: netdev@vger.kernel.org
> ---
> include/linux/socket.h | 3 +++
> net/socket.c | 7 +++++++
> 2 files changed, 10 insertions(+)
>
> diff --git a/include/linux/socket.h b/include/linux/socket.h
> index 13c3a237b9c9..c2fa0f800999 100644
> --- a/include/linux/socket.h
> +++ b/include/linux/socket.h
> @@ -327,6 +327,7 @@ struct ucred {
> */
>
> #define MSG_ZEROCOPY 0x4000000 /* Use user data in kernel path */
> +#define MSG_SPLICE_PAGES 0x8000000 /* Splice the pages from the iterator in sendmsg() */
> #define MSG_FASTOPEN 0x20000000 /* Send data in TCP SYN */
> #define MSG_CMSG_CLOEXEC 0x40000000 /* Set close_on_exec for file
> descriptor received through
> @@ -337,6 +338,8 @@ struct ucred {
> #define MSG_CMSG_COMPAT 0 /* We never have 32 bit fixups */
> #endif
>
> +/* Flags to be cleared on entry by sendmsg, recvmsg, sendmmsg and recvmmsg syscalls */
> +#define MSG_INTERNAL_FLAGS (MSG_SPLICE_PAGES)
This is fine, but there is no real need to cover both send and receive.
The sendpage internal flags only ensure that those flags cannot enter sendpage
code from any unintentional path. Indeed those "internal" flags can end up in
sendmsg, at least for UDP.
Similarly, this flag set only has to protect sendto and sendmsg. That
can simplify the patch a bit.
> /* Setsockoptions(2) level. Thanks to BSD these must match IPPROTO_xxx */
> #define SOL_IP 0
> diff --git a/net/socket.c b/net/socket.c
> index 6bae8ce7059e..dfb912bbed62 100644
> --- a/net/socket.c
> +++ b/net/socket.c
> @@ -2139,6 +2139,7 @@ int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
> msg.msg_name = (struct sockaddr *)&address;
> msg.msg_namelen = addr_len;
> }
> + flags &= ~MSG_INTERNAL_FLAGS;
> if (sock->file->f_flags & O_NONBLOCK)
> flags |= MSG_DONTWAIT;
> msg.msg_flags = flags;
> @@ -2192,6 +2193,7 @@ int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
> if (!sock)
> goto out;
>
> + flags &= ~MSG_INTERNAL_FLAGS;
> if (sock->file->f_flags & O_NONBLOCK)
> flags |= MSG_DONTWAIT;
> err = sock_recvmsg(sock, &msg, flags);
> @@ -2579,6 +2581,7 @@ long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
>
> if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
> return -EINVAL;
> + flags &= ~MSG_INTERNAL_FLAGS;
>
> sock = sockfd_lookup_light(fd, &err, &fput_needed);
> if (!sock)
> @@ -2627,6 +2630,7 @@ int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
> entry = mmsg;
> compat_entry = (struct compat_mmsghdr __user *)mmsg;
> err = 0;
> + flags &= ~MSG_INTERNAL_FLAGS;
> flags |= MSG_BATCH;
>
No need to modify __sys_sendmmsg explicitly, as it ends up calling
__sys_sendmsg?
Also, sendpage does this flags masking in the internal sock_FUNC
helpers rather than __sys_FUNC. Might be preferable.
> while (datagrams < vlen) {
> @@ -2775,6 +2779,7 @@ long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
> struct user_msghdr __user *umsg,
> struct sockaddr __user *uaddr, unsigned int flags)
> {
> + flags &= ~MSG_INTERNAL_FLAGS;
> return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
> }
>
> @@ -2787,6 +2792,7 @@ long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
>
> if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
> return -EINVAL;
> + flags &= ~MSG_INTERNAL_FLAGS;
>
> sock = sockfd_lookup_light(fd, &err, &fput_needed);
> if (!sock)
> @@ -2839,6 +2845,7 @@ static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
> goto out_put;
> }
> }
> + flags &= ~MSG_INTERNAL_FLAGS;
>
> entry = mmsg;
> compat_entry = (struct compat_mmsghdr __user *)mmsg;
>
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag
2023-03-29 14:13 ` [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag David Howells
2023-03-30 14:28 ` Willem de Bruijn
@ 2023-03-30 15:07 ` David Howells
2023-03-30 17:51 ` Willem de Bruijn
1 sibling, 1 reply; 77+ messages in thread
From: David Howells @ 2023-03-30 15:07 UTC (permalink / raw)
To: Willem de Bruijn
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm
Willem de Bruijn <willemdebruijn.kernel@gmail.com> wrote:
> No need to modify __sys_sendmmsg explicitly, as it ends up calling
> __sys_sendmsg?
>
> Also, sendpage does this flags masking in the internal sock_FUNC
> helpers rather than __sys_FUNC. Might be preferable.
I was wondering whether other flags, such as MSG_BATCH should be added to the
list. Is it bad if userspace sets that in sendmsg()? AF_KCM, at least, looks
at it.
David
^ permalink raw reply [flat|nested] 77+ messages in thread
* Re: [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag
2023-03-30 15:07 ` David Howells
@ 2023-03-30 17:51 ` Willem de Bruijn
0 siblings, 0 replies; 77+ messages in thread
From: Willem de Bruijn @ 2023-03-30 17:51 UTC (permalink / raw)
To: David Howells, Willem de Bruijn
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm
David Howells wrote:
> Willem de Bruijn <willemdebruijn.kernel@gmail.com> wrote:
>
> > No need to modify __sys_sendmmsg explicitly, as it ends up calling
> > __sys_sendmsg?
> >
> > Also, sendpage does this flags masking in the internal sock_FUNC
> > helpers rather than __sys_FUNC. Might be preferable.
>
> I was wondering whether other flags, such as MSG_BATCH should be added to the
> list. Is it bad if userspace sets that in sendmsg()? AF_KCM, at least, looks
> at it.
That flag was added exactly for AF_KCM. A process that explicitly
sets it might experience bad behavior (increased latency), but
there are no legacy AF_KCM applications that precede the flag.
^ permalink raw reply [flat|nested] 77+ messages in thread
* [RFC PATCH v2 05/48] mm: Move the page fragment allocator from page_alloc.c into its own file
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (3 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 04/48] net: Declare MSG_SPLICE_PAGES internal sendmsg() flag David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 06/48] mm: Make the page_frag_cache allocator use multipage folios David Howells
` (42 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Bernard Metzler,
Tom Talpey, linux-rdma
Move the page fragment allocator from page_alloc.c into its own file
preparatory to changing it.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Bernard Metzler <bmt@zurich.ibm.com>
cc: Tom Talpey <tom@talpey.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-rdma@vger.kernel.org
cc: netdev@vger.kernel.org
---
mm/Makefile | 2 +-
mm/page_alloc.c | 126 -----------------------------------------
mm/page_frag_alloc.c | 131 +++++++++++++++++++++++++++++++++++++++++++
3 files changed, 132 insertions(+), 127 deletions(-)
create mode 100644 mm/page_frag_alloc.c
diff --git a/mm/Makefile b/mm/Makefile
index 8e105e5b3e29..4e6dc12b4cbd 100644
--- a/mm/Makefile
+++ b/mm/Makefile
@@ -52,7 +52,7 @@ obj-y := filemap.o mempool.o oom_kill.o fadvise.o \
readahead.o swap.o truncate.o vmscan.o shmem.o \
util.o mmzone.o vmstat.o backing-dev.o \
mm_init.o percpu.o slab_common.o \
- compaction.o \
+ compaction.o page_frag_alloc.o \
interval_tree.o list_lru.o workingset.o \
debug.o gup.o mmap_lock.o $(mmu-y)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index ac1fc986af44..c08847308907 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -5694,132 +5694,6 @@ void free_pages(unsigned long addr, unsigned int order)
EXPORT_SYMBOL(free_pages);
-/*
- * Page Fragment:
- * An arbitrary-length arbitrary-offset area of memory which resides
- * within a 0 or higher order page. Multiple fragments within that page
- * are individually refcounted, in the page's reference counter.
- *
- * The page_frag functions below provide a simple allocation framework for
- * page fragments. This is used by the network stack and network device
- * drivers to provide a backing region of memory for use as either an
- * sk_buff->head, or to be used in the "frags" portion of skb_shared_info.
- */
-static struct page *__page_frag_cache_refill(struct page_frag_cache *nc,
- gfp_t gfp_mask)
-{
- struct page *page = NULL;
- gfp_t gfp = gfp_mask;
-
-#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- gfp_mask |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY |
- __GFP_NOMEMALLOC;
- page = alloc_pages_node(NUMA_NO_NODE, gfp_mask,
- PAGE_FRAG_CACHE_MAX_ORDER);
- nc->size = page ? PAGE_FRAG_CACHE_MAX_SIZE : PAGE_SIZE;
-#endif
- if (unlikely(!page))
- page = alloc_pages_node(NUMA_NO_NODE, gfp, 0);
-
- nc->va = page ? page_address(page) : NULL;
-
- return page;
-}
-
-void __page_frag_cache_drain(struct page *page, unsigned int count)
-{
- VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
-
- if (page_ref_sub_and_test(page, count))
- free_the_page(page, compound_order(page));
-}
-EXPORT_SYMBOL(__page_frag_cache_drain);
-
-void *page_frag_alloc_align(struct page_frag_cache *nc,
- unsigned int fragsz, gfp_t gfp_mask,
- unsigned int align_mask)
-{
- unsigned int size = PAGE_SIZE;
- struct page *page;
- int offset;
-
- if (unlikely(!nc->va)) {
-refill:
- page = __page_frag_cache_refill(nc, gfp_mask);
- if (!page)
- return NULL;
-
-#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- /* if size can vary use size else just use PAGE_SIZE */
- size = nc->size;
-#endif
- /* Even if we own the page, we do not use atomic_set().
- * This would break get_page_unless_zero() users.
- */
- page_ref_add(page, PAGE_FRAG_CACHE_MAX_SIZE);
-
- /* reset page count bias and offset to start of new frag */
- nc->pfmemalloc = page_is_pfmemalloc(page);
- nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
- nc->offset = size;
- }
-
- offset = nc->offset - fragsz;
- if (unlikely(offset < 0)) {
- page = virt_to_page(nc->va);
-
- if (!page_ref_sub_and_test(page, nc->pagecnt_bias))
- goto refill;
-
- if (unlikely(nc->pfmemalloc)) {
- free_the_page(page, compound_order(page));
- goto refill;
- }
-
-#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- /* if size can vary use size else just use PAGE_SIZE */
- size = nc->size;
-#endif
- /* OK, page count is 0, we can safely set it */
- set_page_count(page, PAGE_FRAG_CACHE_MAX_SIZE + 1);
-
- /* reset page count bias and offset to start of new frag */
- nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
- offset = size - fragsz;
- if (unlikely(offset < 0)) {
- /*
- * The caller is trying to allocate a fragment
- * with fragsz > PAGE_SIZE but the cache isn't big
- * enough to satisfy the request, this may
- * happen in low memory conditions.
- * We don't release the cache page because
- * it could make memory pressure worse
- * so we simply return NULL here.
- */
- return NULL;
- }
- }
-
- nc->pagecnt_bias--;
- offset &= align_mask;
- nc->offset = offset;
-
- return nc->va + offset;
-}
-EXPORT_SYMBOL(page_frag_alloc_align);
-
-/*
- * Frees a page fragment allocated out of either a compound or order 0 page.
- */
-void page_frag_free(void *addr)
-{
- struct page *page = virt_to_head_page(addr);
-
- if (unlikely(put_page_testzero(page)))
- free_the_page(page, compound_order(page));
-}
-EXPORT_SYMBOL(page_frag_free);
-
static void *make_alloc_exact(unsigned long addr, unsigned int order,
size_t size)
{
diff --git a/mm/page_frag_alloc.c b/mm/page_frag_alloc.c
new file mode 100644
index 000000000000..bee95824ef8f
--- /dev/null
+++ b/mm/page_frag_alloc.c
@@ -0,0 +1,131 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Page fragment allocator
+ *
+ * Page Fragment:
+ * An arbitrary-length arbitrary-offset area of memory which resides within a
+ * 0 or higher order page. Multiple fragments within that page are
+ * individually refcounted, in the page's reference counter.
+ *
+ * The page_frag functions provide a simple allocation framework for page
+ * fragments. This is used by the network stack and network device drivers to
+ * provide a backing region of memory for use as either an sk_buff->head, or to
+ * be used in the "frags" portion of skb_shared_info.
+ */
+
+#include <linux/export.h>
+#include <linux/init.h>
+#include <linux/mm.h>
+
+static struct page *__page_frag_cache_refill(struct page_frag_cache *nc,
+ gfp_t gfp_mask)
+{
+ struct page *page = NULL;
+ gfp_t gfp = gfp_mask;
+
+#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
+ gfp_mask |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY |
+ __GFP_NOMEMALLOC;
+ page = alloc_pages_node(NUMA_NO_NODE, gfp_mask,
+ PAGE_FRAG_CACHE_MAX_ORDER);
+ nc->size = page ? PAGE_FRAG_CACHE_MAX_SIZE : PAGE_SIZE;
+#endif
+ if (unlikely(!page))
+ page = alloc_pages_node(NUMA_NO_NODE, gfp, 0);
+
+ nc->va = page ? page_address(page) : NULL;
+
+ return page;
+}
+
+void __page_frag_cache_drain(struct page *page, unsigned int count)
+{
+ VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
+
+ if (page_ref_sub_and_test(page, count - 1))
+ __free_pages(page, compound_order(page));
+}
+EXPORT_SYMBOL(__page_frag_cache_drain);
+
+void *page_frag_alloc_align(struct page_frag_cache *nc,
+ unsigned int fragsz, gfp_t gfp_mask,
+ unsigned int align_mask)
+{
+ unsigned int size = PAGE_SIZE;
+ struct page *page;
+ int offset;
+
+ if (unlikely(!nc->va)) {
+refill:
+ page = __page_frag_cache_refill(nc, gfp_mask);
+ if (!page)
+ return NULL;
+
+#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
+ /* if size can vary use size else just use PAGE_SIZE */
+ size = nc->size;
+#endif
+ /* Even if we own the page, we do not use atomic_set().
+ * This would break get_page_unless_zero() users.
+ */
+ page_ref_add(page, PAGE_FRAG_CACHE_MAX_SIZE);
+
+ /* reset page count bias and offset to start of new frag */
+ nc->pfmemalloc = page_is_pfmemalloc(page);
+ nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
+ nc->offset = size;
+ }
+
+ offset = nc->offset - fragsz;
+ if (unlikely(offset < 0)) {
+ page = virt_to_page(nc->va);
+
+ if (page_ref_count(page) != nc->pagecnt_bias)
+ goto refill;
+ if (unlikely(nc->pfmemalloc)) {
+ page_ref_sub(page, nc->pagecnt_bias - 1);
+ __free_pages(page, compound_order(page));
+ goto refill;
+ }
+
+#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
+ /* if size can vary use size else just use PAGE_SIZE */
+ size = nc->size;
+#endif
+ /* OK, page count is 0, we can safely set it */
+ set_page_count(page, PAGE_FRAG_CACHE_MAX_SIZE + 1);
+
+ /* reset page count bias and offset to start of new frag */
+ nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
+ offset = size - fragsz;
+ if (unlikely(offset < 0)) {
+ /*
+ * The caller is trying to allocate a fragment
+ * with fragsz > PAGE_SIZE but the cache isn't big
+ * enough to satisfy the request, this may
+ * happen in low memory conditions.
+ * We don't release the cache page because
+ * it could make memory pressure worse
+ * so we simply return NULL here.
+ */
+ return NULL;
+ }
+ }
+
+ nc->pagecnt_bias--;
+ offset &= align_mask;
+ nc->offset = offset;
+
+ return nc->va + offset;
+}
+EXPORT_SYMBOL(page_frag_alloc_align);
+
+/*
+ * Frees a page fragment allocated out of either a compound or order 0 page.
+ */
+void page_frag_free(void *addr)
+{
+ struct page *page = virt_to_head_page(addr);
+
+ __free_pages(page, compound_order(page));
+}
+EXPORT_SYMBOL(page_frag_free);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 06/48] mm: Make the page_frag_cache allocator use multipage folios
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (4 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 05/48] mm: Move the page fragment allocator from page_alloc.c into its own file David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 07/48] mm: Make the page_frag_cache allocator use per-cpu David Howells
` (41 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Change the page_frag_cache allocator to use multipage folios rather than
groups of pages. This reduces page_frag_free to just a folio_put() or
put_page().
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
include/linux/mm_types.h | 13 ++----
mm/page_frag_alloc.c | 88 +++++++++++++++++++---------------------
2 files changed, 45 insertions(+), 56 deletions(-)
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 0722859c3647..49a70b3f44a9 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -420,18 +420,13 @@ static inline void *folio_get_private(struct folio *folio)
}
struct page_frag_cache {
- void * va;
-#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- __u16 offset;
- __u16 size;
-#else
- __u32 offset;
-#endif
+ struct folio *folio;
+ unsigned int offset;
/* we maintain a pagecount bias, so that we dont dirty cache line
* containing page->_refcount every time we allocate a fragment.
*/
- unsigned int pagecnt_bias;
- bool pfmemalloc;
+ unsigned int pagecnt_bias;
+ bool pfmemalloc;
};
typedef unsigned long vm_flags_t;
diff --git a/mm/page_frag_alloc.c b/mm/page_frag_alloc.c
index bee95824ef8f..c3792b68ce32 100644
--- a/mm/page_frag_alloc.c
+++ b/mm/page_frag_alloc.c
@@ -16,33 +16,34 @@
#include <linux/init.h>
#include <linux/mm.h>
-static struct page *__page_frag_cache_refill(struct page_frag_cache *nc,
- gfp_t gfp_mask)
+/*
+ * Allocate a new folio for the frag cache.
+ */
+static struct folio *page_frag_cache_refill(struct page_frag_cache *nc,
+ gfp_t gfp_mask)
{
- struct page *page = NULL;
+ struct folio *folio = NULL;
gfp_t gfp = gfp_mask;
#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- gfp_mask |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY |
- __GFP_NOMEMALLOC;
- page = alloc_pages_node(NUMA_NO_NODE, gfp_mask,
- PAGE_FRAG_CACHE_MAX_ORDER);
- nc->size = page ? PAGE_FRAG_CACHE_MAX_SIZE : PAGE_SIZE;
+ gfp_mask |= __GFP_NOWARN | __GFP_NORETRY | __GFP_NOMEMALLOC;
+ folio = folio_alloc(gfp_mask, PAGE_FRAG_CACHE_MAX_ORDER);
#endif
- if (unlikely(!page))
- page = alloc_pages_node(NUMA_NO_NODE, gfp, 0);
-
- nc->va = page ? page_address(page) : NULL;
+ if (unlikely(!folio))
+ folio = folio_alloc(gfp, 0);
- return page;
+ if (folio)
+ nc->folio = folio;
+ return folio;
}
void __page_frag_cache_drain(struct page *page, unsigned int count)
{
- VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
+ struct folio *folio = page_folio(page);
- if (page_ref_sub_and_test(page, count - 1))
- __free_pages(page, compound_order(page));
+ VM_BUG_ON_FOLIO(folio_ref_count(folio) == 0, folio);
+
+ folio_put_refs(folio, count);
}
EXPORT_SYMBOL(__page_frag_cache_drain);
@@ -50,54 +51,47 @@ void *page_frag_alloc_align(struct page_frag_cache *nc,
unsigned int fragsz, gfp_t gfp_mask,
unsigned int align_mask)
{
- unsigned int size = PAGE_SIZE;
- struct page *page;
- int offset;
+ struct folio *folio = nc->folio;
+ size_t offset;
- if (unlikely(!nc->va)) {
+ if (unlikely(!folio)) {
refill:
- page = __page_frag_cache_refill(nc, gfp_mask);
- if (!page)
+ folio = page_frag_cache_refill(nc, gfp_mask);
+ if (!folio)
return NULL;
-#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- /* if size can vary use size else just use PAGE_SIZE */
- size = nc->size;
-#endif
/* Even if we own the page, we do not use atomic_set().
* This would break get_page_unless_zero() users.
*/
- page_ref_add(page, PAGE_FRAG_CACHE_MAX_SIZE);
+ folio_ref_add(folio, PAGE_FRAG_CACHE_MAX_SIZE);
/* reset page count bias and offset to start of new frag */
- nc->pfmemalloc = page_is_pfmemalloc(page);
+ nc->pfmemalloc = folio_is_pfmemalloc(folio);
nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
- nc->offset = size;
+ nc->offset = folio_size(folio);
}
- offset = nc->offset - fragsz;
- if (unlikely(offset < 0)) {
- page = virt_to_page(nc->va);
-
- if (page_ref_count(page) != nc->pagecnt_bias)
+ offset = nc->offset;
+ if (unlikely(fragsz > offset)) {
+ /* Reuse the folio if everyone we gave it to has finished with it. */
+ if (!folio_ref_sub_and_test(folio, nc->pagecnt_bias)) {
+ nc->folio = NULL;
goto refill;
+ }
+
if (unlikely(nc->pfmemalloc)) {
- page_ref_sub(page, nc->pagecnt_bias - 1);
- __free_pages(page, compound_order(page));
+ __folio_put(folio);
+ nc->folio = NULL;
goto refill;
}
-#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- /* if size can vary use size else just use PAGE_SIZE */
- size = nc->size;
-#endif
/* OK, page count is 0, we can safely set it */
- set_page_count(page, PAGE_FRAG_CACHE_MAX_SIZE + 1);
+ folio_set_count(folio, PAGE_FRAG_CACHE_MAX_SIZE + 1);
/* reset page count bias and offset to start of new frag */
nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
- offset = size - fragsz;
- if (unlikely(offset < 0)) {
+ offset = folio_size(folio);
+ if (unlikely(fragsz > offset)) {
/*
* The caller is trying to allocate a fragment
* with fragsz > PAGE_SIZE but the cache isn't big
@@ -107,15 +101,17 @@ void *page_frag_alloc_align(struct page_frag_cache *nc,
* it could make memory pressure worse
* so we simply return NULL here.
*/
+ nc->offset = offset;
return NULL;
}
}
nc->pagecnt_bias--;
+ offset -= fragsz;
offset &= align_mask;
nc->offset = offset;
- return nc->va + offset;
+ return folio_address(folio) + offset;
}
EXPORT_SYMBOL(page_frag_alloc_align);
@@ -124,8 +120,6 @@ EXPORT_SYMBOL(page_frag_alloc_align);
*/
void page_frag_free(void *addr)
{
- struct page *page = virt_to_head_page(addr);
-
- __free_pages(page, compound_order(page));
+ folio_put(virt_to_folio(addr));
}
EXPORT_SYMBOL(page_frag_free);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 07/48] mm: Make the page_frag_cache allocator use per-cpu
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (5 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 06/48] mm: Make the page_frag_cache allocator use multipage folios David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 08/48] tcp: Support MSG_SPLICE_PAGES David Howells
` (40 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Lorenzo Bianconi,
Felix Fietkau, John Crispin, Sean Wang, Mark Lee, Keith Busch,
Christoph Hellwig, Sagi Grimberg, Chaitanya Kulkarni, linux-nvme,
linux-mediatek
Make the page_frag_cache allocator have a separate allocation bucket for
each cpu to avoid racing. This means that no lock is required, other than
preempt disablement, to allocate from it, though if a softirq wants to
access it, then softirq disablement will need to be added.
Make the NVMe and mediatek drivers pass in NULL to page_frag_cache() and
use the default allocation buckets rather than defining their own.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Lorenzo Bianconi <lorenzo@kernel.org>
cc: Felix Fietkau <nbd@nbd.name>
cc: John Crispin <john@phrozen.org>
cc: Sean Wang <sean.wang@mediatek.com>
cc: Mark Lee <Mark-MC.Lee@mediatek.com>
cc: Keith Busch <kbusch@kernel.org>
cc: Christoph Hellwig <hch@lst.de>
cc: Sagi Grimberg <sagi@grimberg.me>
cc: Chaitanya Kulkarni <kch@nvidia.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
cc: linux-nvme@lists.infradead.org
cc: linux-mediatek@lists.infradead.org
---
drivers/net/ethernet/mediatek/mtk_wed_wo.c | 19 +--
drivers/net/ethernet/mediatek/mtk_wed_wo.h | 2 -
drivers/nvme/host/tcp.c | 19 +--
drivers/nvme/target/tcp.c | 22 +--
include/linux/gfp.h | 17 +-
mm/page_frag_alloc.c | 182 +++++++++++++++------
net/core/skbuff.c | 32 ++--
7 files changed, 164 insertions(+), 129 deletions(-)
diff --git a/drivers/net/ethernet/mediatek/mtk_wed_wo.c b/drivers/net/ethernet/mediatek/mtk_wed_wo.c
index 69fba29055e9..859f34447f2f 100644
--- a/drivers/net/ethernet/mediatek/mtk_wed_wo.c
+++ b/drivers/net/ethernet/mediatek/mtk_wed_wo.c
@@ -143,7 +143,7 @@ mtk_wed_wo_queue_refill(struct mtk_wed_wo *wo, struct mtk_wed_wo_queue *q,
dma_addr_t addr;
void *buf;
- buf = page_frag_alloc(&q->cache, q->buf_size, GFP_ATOMIC);
+ buf = page_frag_alloc(NULL, q->buf_size, GFP_ATOMIC);
if (!buf)
break;
@@ -286,7 +286,6 @@ mtk_wed_wo_queue_free(struct mtk_wed_wo *wo, struct mtk_wed_wo_queue *q)
static void
mtk_wed_wo_queue_tx_clean(struct mtk_wed_wo *wo, struct mtk_wed_wo_queue *q)
{
- struct page *page;
int i;
for (i = 0; i < q->n_desc; i++) {
@@ -297,20 +296,11 @@ mtk_wed_wo_queue_tx_clean(struct mtk_wed_wo *wo, struct mtk_wed_wo_queue *q)
skb_free_frag(entry->buf);
entry->buf = NULL;
}
-
- if (!q->cache.va)
- return;
-
- page = virt_to_page(q->cache.va);
- __page_frag_cache_drain(page, q->cache.pagecnt_bias);
- memset(&q->cache, 0, sizeof(q->cache));
}
static void
mtk_wed_wo_queue_rx_clean(struct mtk_wed_wo *wo, struct mtk_wed_wo_queue *q)
{
- struct page *page;
-
for (;;) {
void *buf = mtk_wed_wo_dequeue(wo, q, NULL, true);
@@ -319,13 +309,6 @@ mtk_wed_wo_queue_rx_clean(struct mtk_wed_wo *wo, struct mtk_wed_wo_queue *q)
skb_free_frag(buf);
}
-
- if (!q->cache.va)
- return;
-
- page = virt_to_page(q->cache.va);
- __page_frag_cache_drain(page, q->cache.pagecnt_bias);
- memset(&q->cache, 0, sizeof(q->cache));
}
static void
diff --git a/drivers/net/ethernet/mediatek/mtk_wed_wo.h b/drivers/net/ethernet/mediatek/mtk_wed_wo.h
index dbcf42ce9173..6f940db67fb8 100644
--- a/drivers/net/ethernet/mediatek/mtk_wed_wo.h
+++ b/drivers/net/ethernet/mediatek/mtk_wed_wo.h
@@ -210,8 +210,6 @@ struct mtk_wed_wo_queue_entry {
struct mtk_wed_wo_queue {
struct mtk_wed_wo_queue_regs regs;
- struct page_frag_cache cache;
-
struct mtk_wed_wo_queue_desc *desc;
dma_addr_t desc_dma;
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 7723a4989524..fa32969b532f 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -147,8 +147,6 @@ struct nvme_tcp_queue {
__le32 exp_ddgst;
__le32 recv_ddgst;
- struct page_frag_cache pf_cache;
-
void (*state_change)(struct sock *);
void (*data_ready)(struct sock *);
void (*write_space)(struct sock *);
@@ -470,9 +468,8 @@ static int nvme_tcp_init_request(struct blk_mq_tag_set *set,
struct nvme_tcp_queue *queue = &ctrl->queues[queue_idx];
u8 hdgst = nvme_tcp_hdgst_len(queue);
- req->pdu = page_frag_alloc(&queue->pf_cache,
- sizeof(struct nvme_tcp_cmd_pdu) + hdgst,
- GFP_KERNEL | __GFP_ZERO);
+ req->pdu = page_frag_alloc(NULL, sizeof(struct nvme_tcp_cmd_pdu) + hdgst,
+ GFP_KERNEL | __GFP_ZERO);
if (!req->pdu)
return -ENOMEM;
@@ -1288,9 +1285,8 @@ static int nvme_tcp_alloc_async_req(struct nvme_tcp_ctrl *ctrl)
struct nvme_tcp_request *async = &ctrl->async_req;
u8 hdgst = nvme_tcp_hdgst_len(queue);
- async->pdu = page_frag_alloc(&queue->pf_cache,
- sizeof(struct nvme_tcp_cmd_pdu) + hdgst,
- GFP_KERNEL | __GFP_ZERO);
+ async->pdu = page_frag_alloc(NULL, sizeof(struct nvme_tcp_cmd_pdu) + hdgst,
+ GFP_KERNEL | __GFP_ZERO);
if (!async->pdu)
return -ENOMEM;
@@ -1300,7 +1296,6 @@ static int nvme_tcp_alloc_async_req(struct nvme_tcp_ctrl *ctrl)
static void nvme_tcp_free_queue(struct nvme_ctrl *nctrl, int qid)
{
- struct page *page;
struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
struct nvme_tcp_queue *queue = &ctrl->queues[qid];
unsigned int noreclaim_flag;
@@ -1311,12 +1306,6 @@ static void nvme_tcp_free_queue(struct nvme_ctrl *nctrl, int qid)
if (queue->hdr_digest || queue->data_digest)
nvme_tcp_free_crypto(queue);
- if (queue->pf_cache.va) {
- page = virt_to_head_page(queue->pf_cache.va);
- __page_frag_cache_drain(page, queue->pf_cache.pagecnt_bias);
- queue->pf_cache.va = NULL;
- }
-
noreclaim_flag = memalloc_noreclaim_save();
sock_release(queue->sock);
memalloc_noreclaim_restore(noreclaim_flag);
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 66e8f9fd0ca7..d6cc557cc539 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -143,8 +143,6 @@ struct nvmet_tcp_queue {
struct nvmet_tcp_cmd connect;
- struct page_frag_cache pf_cache;
-
void (*data_ready)(struct sock *);
void (*state_change)(struct sock *);
void (*write_space)(struct sock *);
@@ -1312,25 +1310,25 @@ static int nvmet_tcp_alloc_cmd(struct nvmet_tcp_queue *queue,
c->queue = queue;
c->req.port = queue->port->nport;
- c->cmd_pdu = page_frag_alloc(&queue->pf_cache,
- sizeof(*c->cmd_pdu) + hdgst, GFP_KERNEL | __GFP_ZERO);
+ c->cmd_pdu = page_frag_alloc(NULL, sizeof(*c->cmd_pdu) + hdgst,
+ GFP_KERNEL | __GFP_ZERO);
if (!c->cmd_pdu)
return -ENOMEM;
c->req.cmd = &c->cmd_pdu->cmd;
- c->rsp_pdu = page_frag_alloc(&queue->pf_cache,
- sizeof(*c->rsp_pdu) + hdgst, GFP_KERNEL | __GFP_ZERO);
+ c->rsp_pdu = page_frag_alloc(NULL, sizeof(*c->rsp_pdu) + hdgst,
+ GFP_KERNEL | __GFP_ZERO);
if (!c->rsp_pdu)
goto out_free_cmd;
c->req.cqe = &c->rsp_pdu->cqe;
- c->data_pdu = page_frag_alloc(&queue->pf_cache,
- sizeof(*c->data_pdu) + hdgst, GFP_KERNEL | __GFP_ZERO);
+ c->data_pdu = page_frag_alloc(NULL, sizeof(*c->data_pdu) + hdgst,
+ GFP_KERNEL | __GFP_ZERO);
if (!c->data_pdu)
goto out_free_rsp;
- c->r2t_pdu = page_frag_alloc(&queue->pf_cache,
- sizeof(*c->r2t_pdu) + hdgst, GFP_KERNEL | __GFP_ZERO);
+ c->r2t_pdu = page_frag_alloc(NULL, sizeof(*c->r2t_pdu) + hdgst,
+ GFP_KERNEL | __GFP_ZERO);
if (!c->r2t_pdu)
goto out_free_data;
@@ -1438,7 +1436,6 @@ static void nvmet_tcp_free_cmd_data_in_buffers(struct nvmet_tcp_queue *queue)
static void nvmet_tcp_release_queue_work(struct work_struct *w)
{
- struct page *page;
struct nvmet_tcp_queue *queue =
container_of(w, struct nvmet_tcp_queue, release_work);
@@ -1460,9 +1457,6 @@ static void nvmet_tcp_release_queue_work(struct work_struct *w)
if (queue->hdr_digest || queue->data_digest)
nvmet_tcp_free_crypto(queue);
ida_free(&nvmet_tcp_queue_ida, queue->idx);
-
- page = virt_to_head_page(queue->pf_cache.va);
- __page_frag_cache_drain(page, queue->pf_cache.pagecnt_bias);
kfree(queue);
}
diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index 65a78773dcca..b208ca315882 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -304,14 +304,17 @@ extern void free_pages(unsigned long addr, unsigned int order);
struct page_frag_cache;
extern void __page_frag_cache_drain(struct page *page, unsigned int count);
-extern void *page_frag_alloc_align(struct page_frag_cache *nc,
- unsigned int fragsz, gfp_t gfp_mask,
- unsigned int align_mask);
-
-static inline void *page_frag_alloc(struct page_frag_cache *nc,
- unsigned int fragsz, gfp_t gfp_mask)
+extern void *page_frag_alloc_align(struct page_frag_cache __percpu *frag_cache,
+ size_t fragsz, gfp_t gfp,
+ unsigned long align_mask);
+extern void *page_frag_memdup(struct page_frag_cache __percpu *frag_cache,
+ const void *p, size_t fragsz, gfp_t gfp,
+ unsigned long align_mask);
+
+static inline void *page_frag_alloc(struct page_frag_cache __percpu *frag_cache,
+ size_t fragsz, gfp_t gfp)
{
- return page_frag_alloc_align(nc, fragsz, gfp_mask, ~0u);
+ return page_frag_alloc_align(frag_cache, fragsz, gfp, ULONG_MAX);
}
extern void page_frag_free(void *addr);
diff --git a/mm/page_frag_alloc.c b/mm/page_frag_alloc.c
index c3792b68ce32..7844398afe26 100644
--- a/mm/page_frag_alloc.c
+++ b/mm/page_frag_alloc.c
@@ -16,25 +16,23 @@
#include <linux/init.h>
#include <linux/mm.h>
+static DEFINE_PER_CPU(struct page_frag_cache, page_frag_default_allocator);
+
/*
* Allocate a new folio for the frag cache.
*/
-static struct folio *page_frag_cache_refill(struct page_frag_cache *nc,
- gfp_t gfp_mask)
+static struct folio *page_frag_cache_refill(gfp_t gfp)
{
- struct folio *folio = NULL;
- gfp_t gfp = gfp_mask;
+ struct folio *folio;
#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
- gfp_mask |= __GFP_NOWARN | __GFP_NORETRY | __GFP_NOMEMALLOC;
- folio = folio_alloc(gfp_mask, PAGE_FRAG_CACHE_MAX_ORDER);
+ folio = folio_alloc(gfp | __GFP_NOWARN | __GFP_NORETRY | __GFP_NOMEMALLOC,
+ PAGE_FRAG_CACHE_MAX_ORDER);
+ if (folio)
+ return folio;
#endif
- if (unlikely(!folio))
- folio = folio_alloc(gfp, 0);
- if (folio)
- nc->folio = folio;
- return folio;
+ return folio_alloc(gfp, 0);
}
void __page_frag_cache_drain(struct page *page, unsigned int count)
@@ -47,41 +45,68 @@ void __page_frag_cache_drain(struct page *page, unsigned int count)
}
EXPORT_SYMBOL(__page_frag_cache_drain);
-void *page_frag_alloc_align(struct page_frag_cache *nc,
- unsigned int fragsz, gfp_t gfp_mask,
- unsigned int align_mask)
+/**
+ * page_frag_alloc_align - Allocate some memory for use in zerocopy
+ * @frag_cache: The frag cache to use (or NULL for the default)
+ * @fragsz: The size of the fragment desired
+ * @gfp: Allocation flags under which to make an allocation
+ * @align_mask: The required alignment
+ *
+ * Allocate some memory for use with zerocopy where protocol bits have to be
+ * mixed in with spliced/zerocopied data. Unlike memory allocated from the
+ * slab, this memory's lifetime is purely dependent on the folio's refcount.
+ *
+ * The way it works is that a folio is allocated and fragments are broken off
+ * sequentially and returned to the caller with a ref until the folio no longer
+ * has enough spare space - at which point the allocator's ref is dropped and a
+ * new folio is allocated. The folio remains in existence until the last ref
+ * held by, say, an sk_buff is discarded and then the page is returned to the
+ * page allocator.
+ *
+ * Returns a pointer to the memory on success and -ENOMEM on allocation
+ * failure.
+ *
+ * The allocated memory should be disposed of with folio_put().
+ */
+void *page_frag_alloc_align(struct page_frag_cache __percpu *frag_cache,
+ size_t fragsz, gfp_t gfp, unsigned long align_mask)
{
- struct folio *folio = nc->folio;
+ struct page_frag_cache *nc;
+ struct folio *folio, *spare = NULL;
size_t offset;
+ void *p;
- if (unlikely(!folio)) {
-refill:
- folio = page_frag_cache_refill(nc, gfp_mask);
- if (!folio)
- return NULL;
-
- /* Even if we own the page, we do not use atomic_set().
- * This would break get_page_unless_zero() users.
- */
- folio_ref_add(folio, PAGE_FRAG_CACHE_MAX_SIZE);
+ if (!frag_cache)
+ frag_cache = &page_frag_default_allocator;
+ if (WARN_ON_ONCE(fragsz == 0))
+ fragsz = 1;
+ align_mask &= ~3UL;
- /* reset page count bias and offset to start of new frag */
- nc->pfmemalloc = folio_is_pfmemalloc(folio);
- nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
- nc->offset = folio_size(folio);
+ nc = get_cpu_ptr(frag_cache);
+reload:
+ folio = nc->folio;
+ offset = nc->offset;
+try_again:
+
+ /* Make the allocation if there's sufficient space. */
+ if (fragsz <= offset) {
+ nc->pagecnt_bias--;
+ offset = (offset - fragsz) & align_mask;
+ nc->offset = offset;
+ p = folio_address(folio) + offset;
+ put_cpu_ptr(frag_cache);
+ if (spare)
+ folio_put(spare);
+ return p;
}
- offset = nc->offset;
- if (unlikely(fragsz > offset)) {
- /* Reuse the folio if everyone we gave it to has finished with it. */
- if (!folio_ref_sub_and_test(folio, nc->pagecnt_bias)) {
- nc->folio = NULL;
+ /* Insufficient space - see if we can refurbish the current folio. */
+ if (folio) {
+ if (!folio_ref_sub_and_test(folio, nc->pagecnt_bias))
goto refill;
- }
if (unlikely(nc->pfmemalloc)) {
__folio_put(folio);
- nc->folio = NULL;
goto refill;
}
@@ -91,27 +116,56 @@ void *page_frag_alloc_align(struct page_frag_cache *nc,
/* reset page count bias and offset to start of new frag */
nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
offset = folio_size(folio);
- if (unlikely(fragsz > offset)) {
- /*
- * The caller is trying to allocate a fragment
- * with fragsz > PAGE_SIZE but the cache isn't big
- * enough to satisfy the request, this may
- * happen in low memory conditions.
- * We don't release the cache page because
- * it could make memory pressure worse
- * so we simply return NULL here.
- */
- nc->offset = offset;
+ if (unlikely(fragsz > offset))
+ goto frag_too_big;
+ goto try_again;
+ }
+
+refill:
+ if (!spare) {
+ nc->folio = NULL;
+ put_cpu_ptr(frag_cache);
+
+ spare = page_frag_cache_refill(gfp);
+ if (!spare)
return NULL;
- }
+
+ nc = get_cpu_ptr(frag_cache);
+ /* We may now be on a different cpu and/or someone else may
+ * have refilled it
+ */
+ nc->pfmemalloc = folio_is_pfmemalloc(spare);
+ if (nc->folio)
+ goto reload;
}
- nc->pagecnt_bias--;
- offset -= fragsz;
- offset &= align_mask;
+ nc->folio = spare;
+ folio = spare;
+ spare = NULL;
+
+ /* Even if we own the page, we do not use atomic_set(). This would
+ * break get_page_unless_zero() users.
+ */
+ folio_ref_add(folio, PAGE_FRAG_CACHE_MAX_SIZE);
+
+ /* Reset page count bias and offset to start of new frag */
+ nc->pagecnt_bias = PAGE_FRAG_CACHE_MAX_SIZE + 1;
+ offset = folio_size(folio);
+ goto try_again;
+
+frag_too_big:
+ /*
+ * The caller is trying to allocate a fragment with fragsz > PAGE_SIZE
+ * but the cache isn't big enough to satisfy the request, this may
+ * happen in low memory conditions. We don't release the cache page
+ * because it could make memory pressure worse so we simply return NULL
+ * here.
+ */
nc->offset = offset;
-
- return folio_address(folio) + offset;
+ put_cpu_ptr(frag_cache);
+ if (spare)
+ folio_put(spare);
+ return NULL;
}
EXPORT_SYMBOL(page_frag_alloc_align);
@@ -123,3 +177,25 @@ void page_frag_free(void *addr)
folio_put(virt_to_folio(addr));
}
EXPORT_SYMBOL(page_frag_free);
+
+/**
+ * page_frag_memdup - Allocate a page fragment and duplicate some data into it
+ * @frag_cache: The frag cache to use (or NULL for the default)
+ * @fragsz: The amount of memory to copy (maximum 1/2 page).
+ * @p: The source data to copy
+ * @gfp: Allocation flags under which to make an allocation
+ * @align_mask: The required alignment
+ */
+void *page_frag_memdup(struct page_frag_cache __percpu *frag_cache,
+ const void *p, size_t fragsz, gfp_t gfp,
+ unsigned long align_mask)
+{
+ void *q;
+
+ q = page_frag_alloc_align(frag_cache, fragsz, gfp, align_mask);
+ if (!q)
+ return q;
+
+ return memcpy(q, p, fragsz);
+}
+EXPORT_SYMBOL(page_frag_memdup);
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index eb7d33b41e71..0506e4cf1ed9 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -222,13 +222,13 @@ static void *page_frag_alloc_1k(struct page_frag_1k *nc, gfp_t gfp_mask)
#endif
struct napi_alloc_cache {
- struct page_frag_cache page;
struct page_frag_1k page_small;
unsigned int skb_count;
void *skb_cache[NAPI_SKB_CACHE_SIZE];
};
static DEFINE_PER_CPU(struct page_frag_cache, netdev_alloc_cache);
+static DEFINE_PER_CPU(struct page_frag_cache, napi_frag_cache);
static DEFINE_PER_CPU(struct napi_alloc_cache, napi_alloc_cache);
/* Double check that napi_get_frags() allocates skbs with
@@ -250,11 +250,9 @@ void napi_get_frags_check(struct napi_struct *napi)
void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask)
{
- struct napi_alloc_cache *nc = this_cpu_ptr(&napi_alloc_cache);
-
fragsz = SKB_DATA_ALIGN(fragsz);
- return page_frag_alloc_align(&nc->page, fragsz, GFP_ATOMIC, align_mask);
+ return page_frag_alloc_align(&napi_frag_cache, fragsz, GFP_ATOMIC, align_mask);
}
EXPORT_SYMBOL(__napi_alloc_frag_align);
@@ -264,15 +262,12 @@ void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask)
fragsz = SKB_DATA_ALIGN(fragsz);
if (in_hardirq() || irqs_disabled()) {
- struct page_frag_cache *nc = this_cpu_ptr(&netdev_alloc_cache);
-
- data = page_frag_alloc_align(nc, fragsz, GFP_ATOMIC, align_mask);
+ data = page_frag_alloc_align(&netdev_alloc_cache,
+ fragsz, GFP_ATOMIC, align_mask);
} else {
- struct napi_alloc_cache *nc;
-
local_bh_disable();
- nc = this_cpu_ptr(&napi_alloc_cache);
- data = page_frag_alloc_align(&nc->page, fragsz, GFP_ATOMIC, align_mask);
+ data = page_frag_alloc_align(&napi_frag_cache,
+ fragsz, GFP_ATOMIC, align_mask);
local_bh_enable();
}
return data;
@@ -656,7 +651,6 @@ EXPORT_SYMBOL(__alloc_skb);
struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int len,
gfp_t gfp_mask)
{
- struct page_frag_cache *nc;
struct sk_buff *skb;
bool pfmemalloc;
void *data;
@@ -681,14 +675,12 @@ struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int len,
gfp_mask |= __GFP_MEMALLOC;
if (in_hardirq() || irqs_disabled()) {
- nc = this_cpu_ptr(&netdev_alloc_cache);
- data = page_frag_alloc(nc, len, gfp_mask);
- pfmemalloc = nc->pfmemalloc;
+ data = page_frag_alloc(&netdev_alloc_cache, len, gfp_mask);
+ pfmemalloc = folio_is_pfmemalloc(virt_to_folio(data));
} else {
local_bh_disable();
- nc = this_cpu_ptr(&napi_alloc_cache.page);
- data = page_frag_alloc(nc, len, gfp_mask);
- pfmemalloc = nc->pfmemalloc;
+ data = page_frag_alloc(&napi_frag_cache, len, gfp_mask);
+ pfmemalloc = folio_is_pfmemalloc(virt_to_folio(data));
local_bh_enable();
}
@@ -776,8 +768,8 @@ struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, unsigned int len,
} else {
len = SKB_HEAD_ALIGN(len);
- data = page_frag_alloc(&nc->page, len, gfp_mask);
- pfmemalloc = nc->page.pfmemalloc;
+ data = page_frag_alloc(&napi_frag_cache, len, gfp_mask);
+ pfmemalloc = folio_is_pfmemalloc(virt_to_folio(data));
}
if (unlikely(!data))
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 08/48] tcp: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (6 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 07/48] mm: Make the page_frag_cache allocator use per-cpu David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 09/48] tcp: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data David Howells
` (39 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Make TCP's sendmsg() support MSG_SPLICE_PAGES. This causes pages to be
spliced from the source iterator if possible (the iterator must be
ITER_BVEC and the pages must be spliceable).
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Eric Dumazet <edumazet@google.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/ipv4/tcp.c | 67 ++++++++++++++++++++++++++++++++++++++++++++------
1 file changed, 60 insertions(+), 7 deletions(-)
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 288693981b00..910b327c236e 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1220,7 +1220,7 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
int flags, err, copied = 0;
int mss_now = 0, size_goal, copied_syn = 0;
int process_backlog = 0;
- bool zc = false;
+ int zc = 0;
long timeo;
flags = msg->msg_flags;
@@ -1231,17 +1231,22 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
if (msg->msg_ubuf) {
uarg = msg->msg_ubuf;
net_zcopy_get(uarg);
- zc = sk->sk_route_caps & NETIF_F_SG;
+ if (sk->sk_route_caps & NETIF_F_SG)
+ zc = 1;
} else if (sock_flag(sk, SOCK_ZEROCOPY)) {
uarg = msg_zerocopy_realloc(sk, size, skb_zcopy(skb));
if (!uarg) {
err = -ENOBUFS;
goto out_err;
}
- zc = sk->sk_route_caps & NETIF_F_SG;
- if (!zc)
+ if (sk->sk_route_caps & NETIF_F_SG)
+ zc = 1;
+ else
uarg_to_msgzc(uarg)->zerocopy = 0;
}
+ } else if (unlikely(msg->msg_flags & MSG_SPLICE_PAGES) && size) {
+ if (sk->sk_route_caps & NETIF_F_SG)
+ zc = 2;
}
if (unlikely(flags & MSG_FASTOPEN || inet_sk(sk)->defer_connect) &&
@@ -1304,7 +1309,7 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
goto do_error;
while (msg_data_left(msg)) {
- int copy = 0;
+ ssize_t copy = 0;
skb = tcp_write_queue_tail(sk);
if (skb)
@@ -1345,7 +1350,7 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
if (copy > msg_data_left(msg))
copy = msg_data_left(msg);
- if (!zc) {
+ if (zc == 0) {
bool merge = true;
int i = skb_shinfo(skb)->nr_frags;
struct page_frag *pfrag = sk_page_frag(sk);
@@ -1390,7 +1395,7 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
page_ref_inc(pfrag->page);
}
pfrag->offset += copy;
- } else {
+ } else if (zc == 1) {
/* First append to a fragless skb builds initial
* pure zerocopy skb
*/
@@ -1411,6 +1416,54 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
if (err < 0)
goto do_error;
copy = err;
+ } else if (zc == 2) {
+ /* Splice in data. */
+ struct page *page = NULL, **pages = &page;
+ size_t off = 0, part;
+ bool can_coalesce;
+ int i = skb_shinfo(skb)->nr_frags;
+
+ copy = iov_iter_extract_pages(&msg->msg_iter, &pages,
+ copy, 1, 0, &off);
+ if (copy <= 0) {
+ err = copy ?: -EIO;
+ goto do_error;
+ }
+
+ can_coalesce = skb_can_coalesce(skb, i, page, off);
+ if (!can_coalesce && i >= READ_ONCE(sysctl_max_skb_frags)) {
+ tcp_mark_push(tp, skb);
+ iov_iter_revert(&msg->msg_iter, copy);
+ goto new_segment;
+ }
+ if (tcp_downgrade_zcopy_pure(sk, skb)) {
+ iov_iter_revert(&msg->msg_iter, copy);
+ goto wait_for_space;
+ }
+
+ part = tcp_wmem_schedule(sk, copy);
+ iov_iter_revert(&msg->msg_iter, copy - part);
+ if (!part)
+ goto wait_for_space;
+ copy = part;
+
+ if (can_coalesce) {
+ skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
+ } else {
+ get_page(page);
+ skb_fill_page_desc_noacc(skb, i, page, off, copy);
+ }
+ page = NULL;
+
+ if (!(flags & MSG_NO_SHARED_FRAGS))
+ skb_shinfo(skb)->flags |= SKBFL_SHARED_FRAG;
+
+ skb->len += copy;
+ skb->data_len += copy;
+ skb->truesize += copy;
+ sk_wmem_queued_add(sk, copy);
+ sk_mem_charge(sk, copy);
+
}
if (!copied)
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 09/48] tcp: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (7 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 08/48] tcp: Support MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 10/48] tcp: Convert do_tcp_sendpages() to use MSG_SPLICE_PAGES David Howells
` (38 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
If sendmsg() with MSG_SPLICE_PAGES encounters a page that shouldn't be
spliced - a slab page, for instance, or one with a zero count - make
tcp_sendmsg() copy it.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Eric Dumazet <edumazet@google.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/ipv4/tcp.c | 28 +++++++++++++++++++++++++---
1 file changed, 25 insertions(+), 3 deletions(-)
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 910b327c236e..6ef0518eb706 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1417,10 +1417,10 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
goto do_error;
copy = err;
} else if (zc == 2) {
- /* Splice in data. */
+ /* Splice in data if we can; copy if we can't. */
struct page *page = NULL, **pages = &page;
size_t off = 0, part;
- bool can_coalesce;
+ bool can_coalesce, put = false;
int i = skb_shinfo(skb)->nr_frags;
copy = iov_iter_extract_pages(&msg->msg_iter, &pages,
@@ -1447,12 +1447,34 @@ int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size)
goto wait_for_space;
copy = part;
+ if (!sendpage_ok(page)) {
+ const void *p = kmap_local_page(page);
+ void *q;
+
+ q = page_frag_memdup(NULL, p + off, copy,
+ sk->sk_allocation, ULONG_MAX);
+ kunmap_local(p);
+ if (!q) {
+ iov_iter_revert(&msg->msg_iter, copy);
+ err = copy ?: -ENOMEM;
+ goto do_error;
+ }
+ page = virt_to_page(q);
+ off = offset_in_page(q);
+ put = true;
+ can_coalesce = false;
+ }
+
if (can_coalesce) {
skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
} else {
- get_page(page);
+ if (!put)
+ get_page(page);
+ put = false;
skb_fill_page_desc_noacc(skb, i, page, off, copy);
}
+ if (put)
+ put_page(page);
page = NULL;
if (!(flags & MSG_NO_SHARED_FRAGS))
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 10/48] tcp: Convert do_tcp_sendpages() to use MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (8 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 09/48] tcp: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 11/48] tcp_bpf: Inline do_tcp_sendpages as it's now a wrapper around tcp_sendmsg David Howells
` (37 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Convert do_tcp_sendpages() to use sendmsg() with MSG_SPLICE_PAGES rather
than directly splicing in the pages itself. do_tcp_sendpages() can then be
inlined in subsequent patches into its callers.
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Eric Dumazet <edumazet@google.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/ipv4/tcp.c | 161 +++----------------------------------------------
1 file changed, 10 insertions(+), 151 deletions(-)
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 6ef0518eb706..a1c5a6d9419c 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -971,163 +971,22 @@ static int tcp_wmem_schedule(struct sock *sk, int copy)
return min(copy, sk->sk_forward_alloc);
}
-static struct sk_buff *tcp_build_frag(struct sock *sk, int size_goal, int flags,
- struct page *page, int offset, size_t *size)
-{
- struct sk_buff *skb = tcp_write_queue_tail(sk);
- struct tcp_sock *tp = tcp_sk(sk);
- bool can_coalesce;
- int copy, i;
-
- if (!skb || (copy = size_goal - skb->len) <= 0 ||
- !tcp_skb_can_collapse_to(skb)) {
-new_segment:
- if (!sk_stream_memory_free(sk))
- return NULL;
-
- skb = tcp_stream_alloc_skb(sk, 0, sk->sk_allocation,
- tcp_rtx_and_write_queues_empty(sk));
- if (!skb)
- return NULL;
-
-#ifdef CONFIG_TLS_DEVICE
- skb->decrypted = !!(flags & MSG_SENDPAGE_DECRYPTED);
-#endif
- tcp_skb_entail(sk, skb);
- copy = size_goal;
- }
-
- if (copy > *size)
- copy = *size;
-
- i = skb_shinfo(skb)->nr_frags;
- can_coalesce = skb_can_coalesce(skb, i, page, offset);
- if (!can_coalesce && i >= READ_ONCE(sysctl_max_skb_frags)) {
- tcp_mark_push(tp, skb);
- goto new_segment;
- }
- if (tcp_downgrade_zcopy_pure(sk, skb))
- return NULL;
-
- copy = tcp_wmem_schedule(sk, copy);
- if (!copy)
- return NULL;
-
- if (can_coalesce) {
- skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy);
- } else {
- get_page(page);
- skb_fill_page_desc_noacc(skb, i, page, offset, copy);
- }
-
- if (!(flags & MSG_NO_SHARED_FRAGS))
- skb_shinfo(skb)->flags |= SKBFL_SHARED_FRAG;
-
- skb->len += copy;
- skb->data_len += copy;
- skb->truesize += copy;
- sk_wmem_queued_add(sk, copy);
- sk_mem_charge(sk, copy);
- WRITE_ONCE(tp->write_seq, tp->write_seq + copy);
- TCP_SKB_CB(skb)->end_seq += copy;
- tcp_skb_pcount_set(skb, 0);
-
- *size = copy;
- return skb;
-}
-
ssize_t do_tcp_sendpages(struct sock *sk, struct page *page, int offset,
size_t size, int flags)
{
- struct tcp_sock *tp = tcp_sk(sk);
- int mss_now, size_goal;
- int err;
- ssize_t copied;
- long timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);
-
- if (IS_ENABLED(CONFIG_DEBUG_VM) &&
- WARN_ONCE(!sendpage_ok(page),
- "page must not be a Slab one and have page_count > 0"))
- return -EINVAL;
-
- /* Wait for a connection to finish. One exception is TCP Fast Open
- * (passive side) where data is allowed to be sent before a connection
- * is fully established.
- */
- if (((1 << sk->sk_state) & ~(TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)) &&
- !tcp_passive_fastopen(sk)) {
- err = sk_stream_wait_connect(sk, &timeo);
- if (err != 0)
- goto out_err;
- }
-
- sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
-
- mss_now = tcp_send_mss(sk, &size_goal, flags);
- copied = 0;
-
- err = -EPIPE;
- if (sk->sk_err || (sk->sk_shutdown & SEND_SHUTDOWN))
- goto out_err;
-
- while (size > 0) {
- struct sk_buff *skb;
- size_t copy = size;
-
- skb = tcp_build_frag(sk, size_goal, flags, page, offset, ©);
- if (!skb)
- goto wait_for_space;
-
- if (!copied)
- TCP_SKB_CB(skb)->tcp_flags &= ~TCPHDR_PSH;
-
- copied += copy;
- offset += copy;
- size -= copy;
- if (!size)
- goto out;
-
- if (skb->len < size_goal || (flags & MSG_OOB))
- continue;
-
- if (forced_push(tp)) {
- tcp_mark_push(tp, skb);
- __tcp_push_pending_frames(sk, mss_now, TCP_NAGLE_PUSH);
- } else if (skb == tcp_send_head(sk))
- tcp_push_one(sk, mss_now);
- continue;
-
-wait_for_space:
- set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
- tcp_push(sk, flags & ~MSG_MORE, mss_now,
- TCP_NAGLE_PUSH, size_goal);
-
- err = sk_stream_wait_memory(sk, &timeo);
- if (err != 0)
- goto do_error;
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = flags,
+ .msg_kflags = MSG_SPLICE_PAGES,
+ };
- mss_now = tcp_send_mss(sk, &size_goal, flags);
- }
+ bvec_set_page(&bvec, page, size, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
-out:
- if (copied) {
- tcp_tx_timestamp(sk, sk->sk_tsflags);
- if (!(flags & MSG_SENDPAGE_NOTLAST))
- tcp_push(sk, flags, mss_now, tp->nonagle, size_goal);
- }
- return copied;
+ if (flags & MSG_SENDPAGE_NOTLAST)
+ msg.msg_flags |= MSG_MORE;
-do_error:
- tcp_remove_empty_skb(sk);
- if (copied)
- goto out;
-out_err:
- /* make sure we wake any epoll edge trigger waiter */
- if (unlikely(tcp_rtx_and_write_queues_empty(sk) && err == -EAGAIN)) {
- sk->sk_write_space(sk);
- tcp_chrono_stop(sk, TCP_CHRONO_SNDBUF_LIMITED);
- }
- return sk_stream_error(sk, flags, err);
+ return tcp_sendmsg_locked(sk, &msg, size);
}
EXPORT_SYMBOL_GPL(do_tcp_sendpages);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 11/48] tcp_bpf: Inline do_tcp_sendpages as it's now a wrapper around tcp_sendmsg
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (9 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 10/48] tcp: Convert do_tcp_sendpages() to use MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 12/48] espintcp: Inline do_tcp_sendpages() David Howells
` (36 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, John Fastabend,
Jakub Sitnicki, bpf
do_tcp_sendpages() is now just a small wrapper around tcp_sendmsg_locked(),
so inline it. This is part of replacing ->sendpage() with a call to
sendmsg() with MSG_SPLICE_PAGES set.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: John Fastabend <john.fastabend@gmail.com>
cc: Jakub Sitnicki <jakub@cloudflare.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
cc: bpf@vger.kernel.org
---
net/ipv4/tcp_bpf.c | 20 ++++++++++++--------
1 file changed, 12 insertions(+), 8 deletions(-)
diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
index cf26d65ca389..7f17134637eb 100644
--- a/net/ipv4/tcp_bpf.c
+++ b/net/ipv4/tcp_bpf.c
@@ -72,11 +72,13 @@ static int tcp_bpf_push(struct sock *sk, struct sk_msg *msg, u32 apply_bytes,
{
bool apply = apply_bytes;
struct scatterlist *sge;
+ struct msghdr msghdr = { .msg_flags = flags | MSG_SPLICE_PAGES, };
struct page *page;
int size, ret = 0;
u32 off;
while (1) {
+ struct bio_vec bvec;
bool has_tx_ulp;
sge = sk_msg_elem(msg, msg->sg.start);
@@ -88,16 +90,18 @@ static int tcp_bpf_push(struct sock *sk, struct sk_msg *msg, u32 apply_bytes,
tcp_rate_check_app_limited(sk);
retry:
has_tx_ulp = tls_sw_has_ctx_tx(sk);
- if (has_tx_ulp) {
- flags |= MSG_SENDPAGE_NOPOLICY;
- ret = kernel_sendpage_locked(sk,
- page, off, size, flags);
- } else {
- ret = do_tcp_sendpages(sk, page, off, size, flags);
- }
+ if (has_tx_ulp)
+ msghdr.msg_flags |= MSG_SENDPAGE_NOPOLICY;
+ if (flags & MSG_SENDPAGE_NOTLAST)
+ msghdr.msg_flags |= MSG_MORE;
+
+ bvec_set_page(&bvec, page, size, off);
+ iov_iter_bvec(&msghdr.msg_iter, ITER_SOURCE, &bvec, 1, size);
+ ret = tcp_sendmsg_locked(sk, &msghdr, size);
if (ret <= 0)
return ret;
+
if (apply)
apply_bytes -= ret;
msg->sg.size -= ret;
@@ -398,7 +402,7 @@ static int tcp_bpf_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
long timeo;
int flags;
- /* Don't let internal do_tcp_sendpages() flags through */
+ /* Don't let internal sendpage flags through */
flags = (msg->msg_flags & ~MSG_SENDPAGE_DECRYPTED);
flags |= MSG_NO_SHARED_FRAGS;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 12/48] espintcp: Inline do_tcp_sendpages()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (10 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 11/48] tcp_bpf: Inline do_tcp_sendpages as it's now a wrapper around tcp_sendmsg David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 13/48] tls: " David Howells
` (35 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Steffen Klassert,
Herbert Xu
do_tcp_sendpages() is now just a small wrapper around tcp_sendmsg_locked(),
so inline it, allowing do_tcp_sendpages() to be removed. This is part of
replacing ->sendpage() with a call to sendmsg() with MSG_SPLICE_PAGES set.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Steffen Klassert <steffen.klassert@secunet.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: Eric Dumazet <edumazet@google.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/xfrm/espintcp.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)
diff --git a/net/xfrm/espintcp.c b/net/xfrm/espintcp.c
index 872b80188e83..3504925babdb 100644
--- a/net/xfrm/espintcp.c
+++ b/net/xfrm/espintcp.c
@@ -205,14 +205,16 @@ static int espintcp_sendskb_locked(struct sock *sk, struct espintcp_msg *emsg,
static int espintcp_sendskmsg_locked(struct sock *sk,
struct espintcp_msg *emsg, int flags)
{
+ struct msghdr msghdr = { .msg_flags = flags | MSG_SPLICE_PAGES, };
struct sk_msg *skmsg = &emsg->skmsg;
struct scatterlist *sg;
int done = 0;
int ret;
- flags |= MSG_SENDPAGE_NOTLAST;
+ msghdr.msg_flags |= MSG_SENDPAGE_NOTLAST;
sg = &skmsg->sg.data[skmsg->sg.start];
do {
+ struct bio_vec bvec;
size_t size = sg->length - emsg->offset;
int offset = sg->offset + emsg->offset;
struct page *p;
@@ -220,11 +222,13 @@ static int espintcp_sendskmsg_locked(struct sock *sk,
emsg->offset = 0;
if (sg_is_last(sg))
- flags &= ~MSG_SENDPAGE_NOTLAST;
+ msghdr.msg_flags &= ~MSG_SENDPAGE_NOTLAST;
p = sg_page(sg);
retry:
- ret = do_tcp_sendpages(sk, p, offset, size, flags);
+ bvec_set_page(&bvec, p, size, offset);
+ iov_iter_bvec(&msghdr.msg_iter, ITER_SOURCE, &bvec, 1, size);
+ ret = tcp_sendmsg_locked(sk, &msghdr, size);
if (ret < 0) {
emsg->offset = offset - sg->offset;
skmsg->sg.start += done;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 13/48] tls: Inline do_tcp_sendpages()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (11 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 12/48] espintcp: Inline do_tcp_sendpages() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 14/48] siw: " David Howells
` (34 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Boris Pismenny,
John Fastabend
do_tcp_sendpages() is now just a small wrapper around tcp_sendmsg_locked(),
so inline it, allowing do_tcp_sendpages() to be removed. This is part of
replacing ->sendpage() with a call to sendmsg() with MSG_SPLICE_PAGES set.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Boris Pismenny <borisp@nvidia.com>
cc: John Fastabend <john.fastabend@gmail.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
include/net/tls.h | 2 +-
net/tls/tls_main.c | 25 ++++++++++++++++---------
2 files changed, 17 insertions(+), 10 deletions(-)
diff --git a/include/net/tls.h b/include/net/tls.h
index 154949c7b0c8..d31521c36a84 100644
--- a/include/net/tls.h
+++ b/include/net/tls.h
@@ -256,7 +256,7 @@ struct tls_context {
struct scatterlist *partially_sent_record;
u16 partially_sent_offset;
- bool in_tcp_sendpages;
+ bool splicing_pages;
bool pending_open_record_frags;
struct mutex tx_lock; /* protects partially_sent_* fields and
diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c
index 3735cb00905d..9a1e77de6a1f 100644
--- a/net/tls/tls_main.c
+++ b/net/tls/tls_main.c
@@ -124,7 +124,11 @@ int tls_push_sg(struct sock *sk,
u16 first_offset,
int flags)
{
- int sendpage_flags = flags | MSG_SENDPAGE_NOTLAST;
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = flags | MSG_SENDPAGE_NOTLAST,
+ .msg_kflags = MSG_SPLICE_PAGES,
+ };
int ret = 0;
struct page *p;
size_t size;
@@ -133,16 +137,19 @@ int tls_push_sg(struct sock *sk,
size = sg->length - offset;
offset += sg->offset;
- ctx->in_tcp_sendpages = true;
+ ctx->splicing_pages = true;
while (1) {
if (sg_is_last(sg))
- sendpage_flags = flags;
+ msg.msg_flags = flags;
/* is sending application-limited? */
tcp_rate_check_app_limited(sk);
p = sg_page(sg);
retry:
- ret = do_tcp_sendpages(sk, p, offset, size, sendpage_flags);
+ bvec_set_page(&bvec, p, size, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
+
+ ret = tcp_sendmsg_locked(sk, &msg, size);
if (ret != size) {
if (ret > 0) {
@@ -154,7 +161,7 @@ int tls_push_sg(struct sock *sk,
offset -= sg->offset;
ctx->partially_sent_offset = offset;
ctx->partially_sent_record = (void *)sg;
- ctx->in_tcp_sendpages = false;
+ ctx->splicing_pages = false;
return ret;
}
@@ -168,7 +175,7 @@ int tls_push_sg(struct sock *sk,
size = sg->length;
}
- ctx->in_tcp_sendpages = false;
+ ctx->splicing_pages = false;
return 0;
}
@@ -246,11 +253,11 @@ static void tls_write_space(struct sock *sk)
{
struct tls_context *ctx = tls_get_ctx(sk);
- /* If in_tcp_sendpages call lower protocol write space handler
+ /* If splicing_pages call lower protocol write space handler
* to ensure we wake up any waiting operations there. For example
- * if do_tcp_sendpages where to call sk_wait_event.
+ * if splicing pages where to call sk_wait_event.
*/
- if (ctx->in_tcp_sendpages) {
+ if (ctx->splicing_pages) {
ctx->sk_write_space(sk);
return;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 14/48] siw: Inline do_tcp_sendpages()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (12 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 13/48] tls: " David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 15/48] tcp: Fold do_tcp_sendpages() into tcp_sendpage_locked() David Howells
` (33 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Bernard Metzler,
Tom Talpey, linux-rdma
do_tcp_sendpages() is now just a small wrapper around tcp_sendmsg_locked(),
so inline it, allowing do_tcp_sendpages() to be removed. This is part of
replacing ->sendpage() with a call to sendmsg() with MSG_SPLICE_PAGES set.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Bernard Metzler <bmt@zurich.ibm.com>
cc: Tom Talpey <tom@talpey.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-rdma@vger.kernel.org
cc: netdev@vger.kernel.org
---
drivers/infiniband/sw/siw/siw_qp_tx.c | 17 ++++++++++++-----
1 file changed, 12 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/sw/siw/siw_qp_tx.c b/drivers/infiniband/sw/siw/siw_qp_tx.c
index 05052b49107f..fa5de40d85d5 100644
--- a/drivers/infiniband/sw/siw/siw_qp_tx.c
+++ b/drivers/infiniband/sw/siw/siw_qp_tx.c
@@ -313,7 +313,7 @@ static int siw_tx_ctrl(struct siw_iwarp_tx *c_tx, struct socket *s,
}
/*
- * 0copy TCP transmit interface: Use do_tcp_sendpages.
+ * 0copy TCP transmit interface: Use MSG_SPLICE_PAGES.
*
* Using sendpage to push page by page appears to be less efficient
* than using sendmsg, even if data are copied.
@@ -324,20 +324,27 @@ static int siw_tx_ctrl(struct siw_iwarp_tx *c_tx, struct socket *s,
static int siw_tcp_sendpages(struct socket *s, struct page **page, int offset,
size_t size)
{
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = (MSG_MORE | MSG_DONTWAIT | MSG_SENDPAGE_NOTLAST |
+ MSG_SPLICE_PAGES),
+ };
struct sock *sk = s->sk;
- int i = 0, rv = 0, sent = 0,
- flags = MSG_MORE | MSG_DONTWAIT | MSG_SENDPAGE_NOTLAST;
+ int i = 0, rv = 0, sent = 0;
while (size) {
size_t bytes = min_t(size_t, PAGE_SIZE - offset, size);
if (size + offset <= PAGE_SIZE)
- flags = MSG_MORE | MSG_DONTWAIT;
+ msg.msg_flags = MSG_MORE | MSG_DONTWAIT;
tcp_rate_check_app_limited(sk);
+ bvec_set_page(&bvec, page[i], bytes, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
+
try_page_again:
lock_sock(sk);
- rv = do_tcp_sendpages(sk, page[i], offset, bytes, flags);
+ rv = tcp_sendmsg_locked(sk, &msg, size);
release_sock(sk);
if (rv > 0) {
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 15/48] tcp: Fold do_tcp_sendpages() into tcp_sendpage_locked()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (13 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 14/48] siw: " David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES David Howells
` (32 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Fold do_tcp_sendpages() into its last remaining caller,
tcp_sendpage_locked().
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Eric Dumazet <edumazet@google.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
include/net/tcp.h | 2 --
net/ipv4/tcp.c | 26 ++++++++------------------
2 files changed, 8 insertions(+), 20 deletions(-)
diff --git a/include/net/tcp.h b/include/net/tcp.h
index db9f828e9d1e..844bc8e6a714 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -333,8 +333,6 @@ int tcp_sendpage(struct sock *sk, struct page *page, int offset, size_t size,
int flags);
int tcp_sendpage_locked(struct sock *sk, struct page *page, int offset,
size_t size, int flags);
-ssize_t do_tcp_sendpages(struct sock *sk, struct page *page, int offset,
- size_t size, int flags);
int tcp_send_mss(struct sock *sk, int *size_goal, int flags);
void tcp_push(struct sock *sk, int flags, int mss_now, int nonagle,
int size_goal);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index a1c5a6d9419c..a8f8ccaed10e 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -971,14 +971,16 @@ static int tcp_wmem_schedule(struct sock *sk, int copy)
return min(copy, sk->sk_forward_alloc);
}
-ssize_t do_tcp_sendpages(struct sock *sk, struct page *page, int offset,
- size_t size, int flags)
+int tcp_sendpage_locked(struct sock *sk, struct page *page, int offset,
+ size_t size, int flags)
{
struct bio_vec bvec;
- struct msghdr msg = {
- .msg_flags = flags,
- .msg_kflags = MSG_SPLICE_PAGES,
- };
+ struct msghdr msg = { .msg_flags = flags | MSG_SPLICE_PAGES, };
+
+ if (!(sk->sk_route_caps & NETIF_F_SG))
+ return sock_no_sendpage_locked(sk, page, offset, size, flags);
+
+ tcp_rate_check_app_limited(sk); /* is sending application-limited? */
bvec_set_page(&bvec, page, size, offset);
iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
@@ -988,18 +990,6 @@ ssize_t do_tcp_sendpages(struct sock *sk, struct page *page, int offset,
return tcp_sendmsg_locked(sk, &msg, size);
}
-EXPORT_SYMBOL_GPL(do_tcp_sendpages);
-
-int tcp_sendpage_locked(struct sock *sk, struct page *page, int offset,
- size_t size, int flags)
-{
- if (!(sk->sk_route_caps & NETIF_F_SG))
- return sock_no_sendpage_locked(sk, page, offset, size, flags);
-
- tcp_rate_check_app_limited(sk); /* is sending application-limited? */
-
- return do_tcp_sendpages(sk, page, offset, size, flags);
-}
EXPORT_SYMBOL_GPL(tcp_sendpage_locked);
int tcp_sendpage(struct sock *sk, struct page *page, int offset,
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (14 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 15/48] tcp: Fold do_tcp_sendpages() into tcp_sendpage_locked() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-30 14:20 ` Willem de Bruijn
` (2 more replies)
2023-03-29 14:13 ` [RFC PATCH v2 17/48] ip, udp: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data David Howells
` (31 subsequent siblings)
47 siblings, 3 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Willem de Bruijn
Make IP/UDP sendmsg() support MSG_SPLICE_PAGES. This causes pages to be
spliced from the source iterator if possible (the iterator must be
ITER_BVEC and the pages must be spliceable).
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/ipv4/ip_output.c | 85 +++++++++++++++++++++++++++++++++++++++++---
1 file changed, 81 insertions(+), 4 deletions(-)
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index 4e4e308c3230..07736da70eab 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -973,11 +973,11 @@ static int __ip_append_data(struct sock *sk,
int hh_len;
int exthdrlen;
int mtu;
- int copy;
+ ssize_t copy;
int err;
int offset = 0;
bool zc = false;
- unsigned int maxfraglen, fragheaderlen, maxnonfragsize;
+ unsigned int maxfraglen, fragheaderlen, maxnonfragsize, xlength;
int csummode = CHECKSUM_NONE;
struct rtable *rt = (struct rtable *)cork->dst;
unsigned int wmem_alloc_delta = 0;
@@ -1017,6 +1017,7 @@ static int __ip_append_data(struct sock *sk,
(!exthdrlen || (rt->dst.dev->features & NETIF_F_HW_ESP_TX_CSUM)))
csummode = CHECKSUM_PARTIAL;
+ xlength = length;
if ((flags & MSG_ZEROCOPY) && length) {
struct msghdr *msg = from;
@@ -1047,6 +1048,14 @@ static int __ip_append_data(struct sock *sk,
skb_zcopy_set(skb, uarg, &extra_uref);
}
}
+ } else if ((flags & MSG_SPLICE_PAGES) && length) {
+ struct msghdr *msg = from;
+
+ if (inet->hdrincl)
+ return -EPERM;
+ if (!(rt->dst.dev->features & NETIF_F_SG))
+ return -EOPNOTSUPP;
+ xlength = transhdrlen; /* We need an empty buffer to attach stuff to */
}
cork->length += length;
@@ -1074,6 +1083,50 @@ static int __ip_append_data(struct sock *sk,
unsigned int alloclen, alloc_extra;
unsigned int pagedlen;
struct sk_buff *skb_prev;
+
+ if (unlikely(flags & MSG_SPLICE_PAGES)) {
+ skb_prev = skb;
+ fraggap = skb_prev->len - maxfraglen;
+
+ alloclen = fragheaderlen + hh_len + fraggap + 15;
+ skb = sock_wmalloc(sk, alloclen, 1, sk->sk_allocation);
+ if (unlikely(!skb)) {
+ err = -ENOBUFS;
+ goto error;
+ }
+
+ /*
+ * Fill in the control structures
+ */
+ skb->ip_summed = CHECKSUM_NONE;
+ skb->csum = 0;
+ skb_reserve(skb, hh_len);
+
+ /*
+ * Find where to start putting bytes.
+ */
+ skb_put(skb, fragheaderlen + fraggap);
+ skb_reset_network_header(skb);
+ skb->transport_header = (skb->network_header +
+ fragheaderlen);
+ if (fraggap) {
+ skb->csum = skb_copy_and_csum_bits(
+ skb_prev, maxfraglen,
+ skb_transport_header(skb),
+ fraggap);
+ skb_prev->csum = csum_sub(skb_prev->csum,
+ skb->csum);
+ pskb_trim_unique(skb_prev, maxfraglen);
+ }
+
+ /*
+ * Put the packet on the pending queue.
+ */
+ __skb_queue_tail(&sk->sk_write_queue, skb);
+ continue;
+ }
+ xlength = length;
+
alloc_new_skb:
skb_prev = skb;
if (skb_prev)
@@ -1085,7 +1138,7 @@ static int __ip_append_data(struct sock *sk,
* If remaining data exceeds the mtu,
* we know we need more fragment(s).
*/
- datalen = length + fraggap;
+ datalen = xlength + fraggap;
if (datalen > mtu - fragheaderlen)
datalen = maxfraglen - fragheaderlen;
fraglen = datalen + fragheaderlen;
@@ -1099,7 +1152,7 @@ static int __ip_append_data(struct sock *sk,
* because we have no idea what fragment will be
* the last.
*/
- if (datalen == length + fraggap)
+ if (datalen == xlength + fraggap)
alloc_extra += rt->dst.trailer_len;
if ((flags & MSG_MORE) &&
@@ -1206,6 +1259,30 @@ static int __ip_append_data(struct sock *sk,
err = -EFAULT;
goto error;
}
+ } else if (flags & MSG_SPLICE_PAGES) {
+ struct msghdr *msg = from;
+ struct page *page = NULL, **pages = &page;
+ size_t off;
+
+ copy = iov_iter_extract_pages(&msg->msg_iter, &pages,
+ copy, 1, 0, &off);
+ if (copy <= 0) {
+ err = copy ?: -EIO;
+ goto error;
+ }
+
+ err = skb_append_pagefrags(skb, page, off, copy);
+ if (err < 0)
+ goto error;
+
+ if (skb->ip_summed == CHECKSUM_NONE) {
+ __wsum csum;
+ csum = csum_page(page, off, copy);
+ skb->csum = csum_block_add(skb->csum, csum, skb->len);
+ }
+
+ skb_len_add(skb, copy);
+ refcount_add(copy, &sk->sk_wmem_alloc);
} else if (!zc) {
int i = skb_shinfo(skb)->nr_frags;
^ permalink raw reply related [flat|nested] 77+ messages in thread* RE: [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES David Howells
@ 2023-03-30 14:20 ` Willem de Bruijn
2023-03-30 14:39 ` David Howells
2023-03-30 15:11 ` David Howells
2 siblings, 0 replies; 77+ messages in thread
From: Willem de Bruijn @ 2023-03-30 14:20 UTC (permalink / raw)
To: David Howells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Willem de Bruijn
David Howells wrote:
> Make IP/UDP sendmsg() support MSG_SPLICE_PAGES. This causes pages to be
> spliced from the source iterator if possible (the iterator must be
> ITER_BVEC and the pages must be spliceable).
>
> This allows ->sendpage() to be replaced by something that can handle
> multiple multipage folios in a single transaction.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: netdev@vger.kernel.org
> ---
> net/ipv4/ip_output.c | 85 +++++++++++++++++++++++++++++++++++++++++---
> 1 file changed, 81 insertions(+), 4 deletions(-)
>
> diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
A non-RFC version would require the same for ipv6, of course.
> index 4e4e308c3230..07736da70eab 100644
> --- a/net/ipv4/ip_output.c
> +++ b/net/ipv4/ip_output.c
> @@ -973,11 +973,11 @@ static int __ip_append_data(struct sock *sk,
> int hh_len;
> int exthdrlen;
> int mtu;
> - int copy;
> + ssize_t copy;
> int err;
> int offset = 0;
> bool zc = false;
> - unsigned int maxfraglen, fragheaderlen, maxnonfragsize;
> + unsigned int maxfraglen, fragheaderlen, maxnonfragsize, xlength;
Does x here stand for anything?
> int csummode = CHECKSUM_NONE;
> struct rtable *rt = (struct rtable *)cork->dst;
> unsigned int wmem_alloc_delta = 0;
> @@ -1017,6 +1017,7 @@ static int __ip_append_data(struct sock *sk,
> (!exthdrlen || (rt->dst.dev->features & NETIF_F_HW_ESP_TX_CSUM)))
> csummode = CHECKSUM_PARTIAL;
>
> + xlength = length;
> if ((flags & MSG_ZEROCOPY) && length) {
> struct msghdr *msg = from;
>
> @@ -1047,6 +1048,14 @@ static int __ip_append_data(struct sock *sk,
> skb_zcopy_set(skb, uarg, &extra_uref);
> }
> }
> + } else if ((flags & MSG_SPLICE_PAGES) && length) {
> + struct msghdr *msg = from;
> +
> + if (inet->hdrincl)
> + return -EPERM;
> + if (!(rt->dst.dev->features & NETIF_F_SG))
> + return -EOPNOTSUPP;
> + xlength = transhdrlen; /* We need an empty buffer to attach stuff to */
> }
>
> cork->length += length;
> @@ -1074,6 +1083,50 @@ static int __ip_append_data(struct sock *sk,
> unsigned int alloclen, alloc_extra;
> unsigned int pagedlen;
> struct sk_buff *skb_prev;
> +
> + if (unlikely(flags & MSG_SPLICE_PAGES)) {
> + skb_prev = skb;
> + fraggap = skb_prev->len - maxfraglen;
> +
> + alloclen = fragheaderlen + hh_len + fraggap + 15;
> + skb = sock_wmalloc(sk, alloclen, 1, sk->sk_allocation);
> + if (unlikely(!skb)) {
> + err = -ENOBUFS;
> + goto error;
> + }
> +
> + /*
> + * Fill in the control structures
> + */
> + skb->ip_summed = CHECKSUM_NONE;
> + skb->csum = 0;
> + skb_reserve(skb, hh_len);
> +
> + /*
> + * Find where to start putting bytes.
> + */
> + skb_put(skb, fragheaderlen + fraggap);
> + skb_reset_network_header(skb);
> + skb->transport_header = (skb->network_header +
> + fragheaderlen);
> + if (fraggap) {
> + skb->csum = skb_copy_and_csum_bits(
> + skb_prev, maxfraglen,
> + skb_transport_header(skb),
> + fraggap);
> + skb_prev->csum = csum_sub(skb_prev->csum,
> + skb->csum);
> + pskb_trim_unique(skb_prev, maxfraglen);
> + }
> +
> + /*
> + * Put the packet on the pending queue.
> + */
> + __skb_queue_tail(&sk->sk_write_queue, skb);
> + continue;
> + }
> + xlength = length;
> +
> alloc_new_skb:
> skb_prev = skb;
> if (skb_prev)
> @@ -1085,7 +1138,7 @@ static int __ip_append_data(struct sock *sk,
> * If remaining data exceeds the mtu,
> * we know we need more fragment(s).
> */
> - datalen = length + fraggap;
> + datalen = xlength + fraggap;
> if (datalen > mtu - fragheaderlen)
> datalen = maxfraglen - fragheaderlen;
> fraglen = datalen + fragheaderlen;
> @@ -1099,7 +1152,7 @@ static int __ip_append_data(struct sock *sk,
> * because we have no idea what fragment will be
> * the last.
> */
> - if (datalen == length + fraggap)
> + if (datalen == xlength + fraggap)
> alloc_extra += rt->dst.trailer_len;
>
> if ((flags & MSG_MORE) &&
> @@ -1206,6 +1259,30 @@ static int __ip_append_data(struct sock *sk,
> err = -EFAULT;
> goto error;
> }
> + } else if (flags & MSG_SPLICE_PAGES) {
> + struct msghdr *msg = from;
> + struct page *page = NULL, **pages = &page;
> + size_t off;
> +
> + copy = iov_iter_extract_pages(&msg->msg_iter, &pages,
> + copy, 1, 0, &off);
> + if (copy <= 0) {
> + err = copy ?: -EIO;
> + goto error;
> + }
> +
> + err = skb_append_pagefrags(skb, page, off, copy);
> + if (err < 0)
> + goto error;
> +
> + if (skb->ip_summed == CHECKSUM_NONE) {
> + __wsum csum;
> + csum = csum_page(page, off, copy);
> + skb->csum = csum_block_add(skb->csum, csum, skb->len);
> + }
> +
> + skb_len_add(skb, copy);
> + refcount_add(copy, &sk->sk_wmem_alloc);
> } else if (!zc) {
> int i = skb_shinfo(skb)->nr_frags;
>
>
This does add a lot of code to two functions that are already
unwieldy. It may be unavoidable, but it if can use helpers, that would
be preferable.
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES David Howells
2023-03-30 14:20 ` Willem de Bruijn
@ 2023-03-30 14:39 ` David Howells
2023-03-30 17:46 ` Willem de Bruijn
2023-03-30 15:11 ` David Howells
2 siblings, 1 reply; 77+ messages in thread
From: David Howells @ 2023-03-30 14:39 UTC (permalink / raw)
To: Willem de Bruijn
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm
Willem de Bruijn <willemdebruijn.kernel@gmail.com> wrote:
> > diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
>
> A non-RFC version would require the same for ipv6, of course.
I missed the fact that ipv6 had it's own version of __ip_append_data() despite
sharing tcp_sendmsg(). Could __ip_append_data() and __ip6_append_data() be
shared? I guess that the v6_cork, the flowi6 and the ipcm6_cookie might
prevent that.
David
^ permalink raw reply [flat|nested] 77+ messages in thread
* Re: [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES
2023-03-30 14:39 ` David Howells
@ 2023-03-30 17:46 ` Willem de Bruijn
0 siblings, 0 replies; 77+ messages in thread
From: Willem de Bruijn @ 2023-03-30 17:46 UTC (permalink / raw)
To: David Howells, Willem de Bruijn
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm
David Howells wrote:
> Willem de Bruijn <willemdebruijn.kernel@gmail.com> wrote:
>
> > > diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
> >
> > A non-RFC version would require the same for ipv6, of course.
>
> I missed the fact that ipv6 had it's own version of __ip_append_data() despite
> sharing tcp_sendmsg(). Could __ip_append_data() and __ip6_append_data() be
> shared? I guess that the v6_cork, the flowi6 and the ipcm6_cookie might
> prevent that.
We haven't been able to unify them before. As this series is
complex enough as is, I would not attempt to include it.
If it grows the code, maybe it'll be an incentive to take
another look afterwards.
^ permalink raw reply [flat|nested] 77+ messages in thread
* Re: [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES David Howells
2023-03-30 14:20 ` Willem de Bruijn
2023-03-30 14:39 ` David Howells
@ 2023-03-30 15:11 ` David Howells
2023-03-30 17:55 ` Willem de Bruijn
2023-03-30 19:49 ` David Howells
2 siblings, 2 replies; 77+ messages in thread
From: David Howells @ 2023-03-30 15:11 UTC (permalink / raw)
To: Willem de Bruijn
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm
Willem de Bruijn <willemdebruijn.kernel@gmail.com> wrote:
> > + unsigned int maxfraglen, fragheaderlen, maxnonfragsize, xlength;
>
> Does x here stand for anything?
Yeah... "bad naming". How about if I call it initial_length? I'm trying to
avoid allocating bufferage for the data.
> This does add a lot of code to two functions that are already
> unwieldy. It may be unavoidable, but it if can use helpers, that would
> be preferable.
Something like the attached? (This is on top of patches 16-17, but I would
need to fold it in)
David
---
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index b38dbb2f9c3f..019ed9bb6745 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -956,6 +956,96 @@ csum_page(struct page *page, int offset, int copy)
return csum;
}
+/*
+ * Allocate a packet for MSG_SPLICE_PAGES.
+ */
+static int __ip_splice_alloc(struct sock *sk, struct sk_buff **pskb,
+ unsigned int fragheaderlen, unsigned int maxfraglen,
+ unsigned int hh_len)
+{
+ struct sk_buff *skb_prev = *pskb, *skb;
+ unsigned int fraggap = skb_prev->len - maxfraglen;
+ unsigned int alloclen = fragheaderlen + hh_len + fraggap + 15;
+
+ skb = sock_wmalloc(sk, alloclen, 1, sk->sk_allocation);
+ if (unlikely(!skb))
+ return -ENOBUFS;
+
+ /* Fill in the control structures */
+ skb->ip_summed = CHECKSUM_NONE;
+ skb->csum = 0;
+ skb_reserve(skb, hh_len);
+
+ /* Find where to start putting bytes. */
+ skb_put(skb, fragheaderlen + fraggap);
+ skb_reset_network_header(skb);
+ skb->transport_header = skb->network_header + fragheaderlen;
+ if (fraggap) {
+ skb->csum = skb_copy_and_csum_bits(skb_prev, maxfraglen,
+ skb_transport_header(skb),
+ fraggap);
+ skb_prev->csum = csum_sub(skb_prev->csum, skb->csum);
+ pskb_trim_unique(skb_prev, maxfraglen);
+ }
+
+ /* Put the packet on the pending queue. */
+ __skb_queue_tail(&sk->sk_write_queue, skb);
+ *pskb = skb;
+ return 0;
+}
+
+/*
+ * Add (or copy) data pages for MSG_SPLICE_PAGES.
+ */
+static int __ip_splice_pages(struct sock *sk, struct sk_buff *skb,
+ void *from, size_t *pcopy)
+{
+ struct msghdr *msg = from;
+ struct page *page = NULL, **pages = &page;
+ ssize_t copy = *pcopy;
+ size_t off;
+ bool put = false;
+ int err;
+
+ copy = iov_iter_extract_pages(&msg->msg_iter, &pages, copy, 1, 0, &off);
+ if (copy <= 0)
+ return copy ?: -EIO;
+
+ if (!sendpage_ok(page)) {
+ const void *p = kmap_local_page(page);
+ void *q;
+
+ q = page_frag_memdup(NULL, p + off, copy,
+ sk->sk_allocation, ULONG_MAX);
+ kunmap_local(p);
+ if (!q)
+ return -ENOMEM;
+ page = virt_to_page(q);
+ off = offset_in_page(q);
+ put = true;
+ }
+
+ err = skb_append_pagefrags(skb, page, off, copy);
+ if (put)
+ put_page(page);
+ if (err < 0) {
+ iov_iter_revert(&msg->msg_iter, copy);
+ return err;
+ }
+
+ if (skb->ip_summed == CHECKSUM_NONE) {
+ __wsum csum;
+
+ csum = csum_page(page, off, copy);
+ skb->csum = csum_block_add(skb->csum, csum, skb->len);
+ }
+
+ skb_len_add(skb, copy);
+ refcount_add(copy, &sk->sk_wmem_alloc);
+ *pcopy = copy;
+ return 0;
+}
+
static int __ip_append_data(struct sock *sk,
struct flowi4 *fl4,
struct sk_buff_head *queue,
@@ -977,7 +1067,7 @@ static int __ip_append_data(struct sock *sk,
int err;
int offset = 0;
bool zc = false;
- unsigned int maxfraglen, fragheaderlen, maxnonfragsize, xlength;
+ unsigned int maxfraglen, fragheaderlen, maxnonfragsize, initial_length;
int csummode = CHECKSUM_NONE;
struct rtable *rt = (struct rtable *)cork->dst;
unsigned int wmem_alloc_delta = 0;
@@ -1017,7 +1107,7 @@ static int __ip_append_data(struct sock *sk,
(!exthdrlen || (rt->dst.dev->features & NETIF_F_HW_ESP_TX_CSUM)))
csummode = CHECKSUM_PARTIAL;
- xlength = length;
+ initial_length = length;
if ((flags & MSG_ZEROCOPY) && length) {
struct msghdr *msg = from;
@@ -1053,7 +1143,7 @@ static int __ip_append_data(struct sock *sk,
return -EPERM;
if (!(rt->dst.dev->features & NETIF_F_SG))
return -EOPNOTSUPP;
- xlength = transhdrlen; /* We need an empty buffer to attach stuff to */
+ initial_length = transhdrlen; /* We need an empty buffer to attach stuff to */
}
cork->length += length;
@@ -1083,47 +1173,13 @@ static int __ip_append_data(struct sock *sk,
struct sk_buff *skb_prev;
if (unlikely(flags & MSG_SPLICE_PAGES)) {
- skb_prev = skb;
- fraggap = skb_prev->len - maxfraglen;
-
- alloclen = fragheaderlen + hh_len + fraggap + 15;
- skb = sock_wmalloc(sk, alloclen, 1, sk->sk_allocation);
- if (unlikely(!skb)) {
- err = -ENOBUFS;
+ err = __ip_splice_alloc(sk, &skb, fragheaderlen,
+ maxfraglen, hh_len);
+ if (err < 0)
goto error;
- }
-
- /*
- * Fill in the control structures
- */
- skb->ip_summed = CHECKSUM_NONE;
- skb->csum = 0;
- skb_reserve(skb, hh_len);
-
- /*
- * Find where to start putting bytes.
- */
- skb_put(skb, fragheaderlen + fraggap);
- skb_reset_network_header(skb);
- skb->transport_header = (skb->network_header +
- fragheaderlen);
- if (fraggap) {
- skb->csum = skb_copy_and_csum_bits(
- skb_prev, maxfraglen,
- skb_transport_header(skb),
- fraggap);
- skb_prev->csum = csum_sub(skb_prev->csum,
- skb->csum);
- pskb_trim_unique(skb_prev, maxfraglen);
- }
-
- /*
- * Put the packet on the pending queue.
- */
- __skb_queue_tail(&sk->sk_write_queue, skb);
continue;
}
- xlength = length;
+ initial_length = length;
alloc_new_skb:
skb_prev = skb;
@@ -1136,7 +1192,7 @@ static int __ip_append_data(struct sock *sk,
* If remaining data exceeds the mtu,
* we know we need more fragment(s).
*/
- datalen = xlength + fraggap;
+ datalen = initial_length + fraggap;
if (datalen > mtu - fragheaderlen)
datalen = maxfraglen - fragheaderlen;
fraglen = datalen + fragheaderlen;
@@ -1150,7 +1206,7 @@ static int __ip_append_data(struct sock *sk,
* because we have no idea what fragment will be
* the last.
*/
- if (datalen == xlength + fraggap)
+ if (datalen == initial_length + fraggap)
alloc_extra += rt->dst.trailer_len;
if ((flags & MSG_MORE) &&
@@ -1258,48 +1314,9 @@ static int __ip_append_data(struct sock *sk,
goto error;
}
} else if (flags & MSG_SPLICE_PAGES) {
- struct msghdr *msg = from;
- struct page *page = NULL, **pages = &page;
- size_t off;
- bool put = false;
-
- copy = iov_iter_extract_pages(&msg->msg_iter, &pages,
- copy, 1, 0, &off);
- if (copy <= 0) {
- err = copy ?: -EIO;
- goto error;
- }
-
- if (!sendpage_ok(page)) {
- const void *p = kmap_local_page(page);
- void *q;
-
- q = page_frag_memdup(NULL, p + off, copy,
- sk->sk_allocation, ULONG_MAX);
- kunmap_local(p);
- if (!q) {
- err = copy ?: -ENOMEM;
- goto error;
- }
- page = virt_to_page(q);
- off = offset_in_page(q);
- put = true;
- }
-
- err = skb_append_pagefrags(skb, page, off, copy);
- if (put)
- put_page(page);
+ err = __ip_splice_pages(sk, skb, from, ©);
if (err < 0)
goto error;
-
- if (skb->ip_summed == CHECKSUM_NONE) {
- __wsum csum;
- csum = csum_page(page, off, copy);
- skb->csum = csum_block_add(skb->csum, csum, skb->len);
- }
-
- skb_len_add(skb, copy);
- refcount_add(copy, &sk->sk_wmem_alloc);
} else if (!zc) {
int i = skb_shinfo(skb)->nr_frags;
^ permalink raw reply related [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES
2023-03-30 15:11 ` David Howells
@ 2023-03-30 17:55 ` Willem de Bruijn
2023-03-30 19:49 ` David Howells
1 sibling, 0 replies; 77+ messages in thread
From: Willem de Bruijn @ 2023-03-30 17:55 UTC (permalink / raw)
To: David Howells, Willem de Bruijn
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm
David Howells wrote:
> Willem de Bruijn <willemdebruijn.kernel@gmail.com> wrote:
>
> > > + unsigned int maxfraglen, fragheaderlen, maxnonfragsize, xlength;
> >
> > Does x here stand for anything?
>
> Yeah... "bad naming". How about if I call it initial_length? I'm trying to
> avoid allocating bufferage for the data.
That's more informative, thanks. Let me not bikeshed this further for now.
> > This does add a lot of code to two functions that are already
> > unwieldy. It may be unavoidable, but it if can use helpers, that would
> > be preferable.
>
> Something like the attached? (This is on top of patches 16-17, but I would
> need to fold it in)
Yes exactly. I wasn't sure whether the inner loops required access to
too many function scope variables to pull this off. But seems like it
is doable. Great.
>
> David
> ---
> diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
> index b38dbb2f9c3f..019ed9bb6745 100644
> --- a/net/ipv4/ip_output.c
> +++ b/net/ipv4/ip_output.c
> @@ -956,6 +956,96 @@ csum_page(struct page *page, int offset, int copy)
> return csum;
> }
>
> +/*
> + * Allocate a packet for MSG_SPLICE_PAGES.
> + */
> +static int __ip_splice_alloc(struct sock *sk, struct sk_buff **pskb,
> + unsigned int fragheaderlen, unsigned int maxfraglen,
> + unsigned int hh_len)
> +{
> + struct sk_buff *skb_prev = *pskb, *skb;
> + unsigned int fraggap = skb_prev->len - maxfraglen;
> + unsigned int alloclen = fragheaderlen + hh_len + fraggap + 15;
> +
> + skb = sock_wmalloc(sk, alloclen, 1, sk->sk_allocation);
> + if (unlikely(!skb))
> + return -ENOBUFS;
> +
> + /* Fill in the control structures */
> + skb->ip_summed = CHECKSUM_NONE;
> + skb->csum = 0;
> + skb_reserve(skb, hh_len);
> +
> + /* Find where to start putting bytes. */
> + skb_put(skb, fragheaderlen + fraggap);
> + skb_reset_network_header(skb);
> + skb->transport_header = skb->network_header + fragheaderlen;
> + if (fraggap) {
> + skb->csum = skb_copy_and_csum_bits(skb_prev, maxfraglen,
> + skb_transport_header(skb),
> + fraggap);
> + skb_prev->csum = csum_sub(skb_prev->csum, skb->csum);
> + pskb_trim_unique(skb_prev, maxfraglen);
> + }
> +
> + /* Put the packet on the pending queue. */
> + __skb_queue_tail(&sk->sk_write_queue, skb);
> + *pskb = skb;
> + return 0;
> +}
> +
> +/*
> + * Add (or copy) data pages for MSG_SPLICE_PAGES.
> + */
> +static int __ip_splice_pages(struct sock *sk, struct sk_buff *skb,
> + void *from, size_t *pcopy)
> +{
> + struct msghdr *msg = from;
> + struct page *page = NULL, **pages = &page;
> + ssize_t copy = *pcopy;
> + size_t off;
> + bool put = false;
> + int err;
> +
> + copy = iov_iter_extract_pages(&msg->msg_iter, &pages, copy, 1, 0, &off);
> + if (copy <= 0)
> + return copy ?: -EIO;
> +
> + if (!sendpage_ok(page)) {
> + const void *p = kmap_local_page(page);
> + void *q;
> +
> + q = page_frag_memdup(NULL, p + off, copy,
> + sk->sk_allocation, ULONG_MAX);
> + kunmap_local(p);
> + if (!q)
> + return -ENOMEM;
> + page = virt_to_page(q);
> + off = offset_in_page(q);
> + put = true;
> + }
> +
> + err = skb_append_pagefrags(skb, page, off, copy);
> + if (put)
> + put_page(page);
> + if (err < 0) {
> + iov_iter_revert(&msg->msg_iter, copy);
> + return err;
> + }
> +
> + if (skb->ip_summed == CHECKSUM_NONE) {
> + __wsum csum;
> +
> + csum = csum_page(page, off, copy);
> + skb->csum = csum_block_add(skb->csum, csum, skb->len);
> + }
> +
> + skb_len_add(skb, copy);
> + refcount_add(copy, &sk->sk_wmem_alloc);
> + *pcopy = copy;
> + return 0;
> +}
> +
> static int __ip_append_data(struct sock *sk,
> struct flowi4 *fl4,
> struct sk_buff_head *queue,
> @@ -977,7 +1067,7 @@ static int __ip_append_data(struct sock *sk,
> int err;
> int offset = 0;
> bool zc = false;
> - unsigned int maxfraglen, fragheaderlen, maxnonfragsize, xlength;
> + unsigned int maxfraglen, fragheaderlen, maxnonfragsize, initial_length;
> int csummode = CHECKSUM_NONE;
> struct rtable *rt = (struct rtable *)cork->dst;
> unsigned int wmem_alloc_delta = 0;
> @@ -1017,7 +1107,7 @@ static int __ip_append_data(struct sock *sk,
> (!exthdrlen || (rt->dst.dev->features & NETIF_F_HW_ESP_TX_CSUM)))
> csummode = CHECKSUM_PARTIAL;
>
> - xlength = length;
> + initial_length = length;
> if ((flags & MSG_ZEROCOPY) && length) {
> struct msghdr *msg = from;
>
> @@ -1053,7 +1143,7 @@ static int __ip_append_data(struct sock *sk,
> return -EPERM;
> if (!(rt->dst.dev->features & NETIF_F_SG))
> return -EOPNOTSUPP;
> - xlength = transhdrlen; /* We need an empty buffer to attach stuff to */
> + initial_length = transhdrlen; /* We need an empty buffer to attach stuff to */
> }
>
> cork->length += length;
> @@ -1083,47 +1173,13 @@ static int __ip_append_data(struct sock *sk,
> struct sk_buff *skb_prev;
>
> if (unlikely(flags & MSG_SPLICE_PAGES)) {
> - skb_prev = skb;
> - fraggap = skb_prev->len - maxfraglen;
> -
> - alloclen = fragheaderlen + hh_len + fraggap + 15;
> - skb = sock_wmalloc(sk, alloclen, 1, sk->sk_allocation);
> - if (unlikely(!skb)) {
> - err = -ENOBUFS;
> + err = __ip_splice_alloc(sk, &skb, fragheaderlen,
> + maxfraglen, hh_len);
> + if (err < 0)
> goto error;
> - }
> -
> - /*
> - * Fill in the control structures
> - */
> - skb->ip_summed = CHECKSUM_NONE;
> - skb->csum = 0;
> - skb_reserve(skb, hh_len);
> -
> - /*
> - * Find where to start putting bytes.
> - */
> - skb_put(skb, fragheaderlen + fraggap);
> - skb_reset_network_header(skb);
> - skb->transport_header = (skb->network_header +
> - fragheaderlen);
> - if (fraggap) {
> - skb->csum = skb_copy_and_csum_bits(
> - skb_prev, maxfraglen,
> - skb_transport_header(skb),
> - fraggap);
> - skb_prev->csum = csum_sub(skb_prev->csum,
> - skb->csum);
> - pskb_trim_unique(skb_prev, maxfraglen);
> - }
> -
> - /*
> - * Put the packet on the pending queue.
> - */
> - __skb_queue_tail(&sk->sk_write_queue, skb);
> continue;
> }
> - xlength = length;
> + initial_length = length;
>
> alloc_new_skb:
> skb_prev = skb;
> @@ -1136,7 +1192,7 @@ static int __ip_append_data(struct sock *sk,
> * If remaining data exceeds the mtu,
> * we know we need more fragment(s).
> */
> - datalen = xlength + fraggap;
> + datalen = initial_length + fraggap;
> if (datalen > mtu - fragheaderlen)
> datalen = maxfraglen - fragheaderlen;
> fraglen = datalen + fragheaderlen;
> @@ -1150,7 +1206,7 @@ static int __ip_append_data(struct sock *sk,
> * because we have no idea what fragment will be
> * the last.
> */
> - if (datalen == xlength + fraggap)
> + if (datalen == initial_length + fraggap)
> alloc_extra += rt->dst.trailer_len;
>
> if ((flags & MSG_MORE) &&
> @@ -1258,48 +1314,9 @@ static int __ip_append_data(struct sock *sk,
> goto error;
> }
> } else if (flags & MSG_SPLICE_PAGES) {
> - struct msghdr *msg = from;
> - struct page *page = NULL, **pages = &page;
> - size_t off;
> - bool put = false;
> -
> - copy = iov_iter_extract_pages(&msg->msg_iter, &pages,
> - copy, 1, 0, &off);
> - if (copy <= 0) {
> - err = copy ?: -EIO;
> - goto error;
> - }
> -
> - if (!sendpage_ok(page)) {
> - const void *p = kmap_local_page(page);
> - void *q;
> -
> - q = page_frag_memdup(NULL, p + off, copy,
> - sk->sk_allocation, ULONG_MAX);
> - kunmap_local(p);
> - if (!q) {
> - err = copy ?: -ENOMEM;
> - goto error;
> - }
> - page = virt_to_page(q);
> - off = offset_in_page(q);
> - put = true;
> - }
> -
> - err = skb_append_pagefrags(skb, page, off, copy);
> - if (put)
> - put_page(page);
> + err = __ip_splice_pages(sk, skb, from, ©);
> if (err < 0)
> goto error;
> -
> - if (skb->ip_summed == CHECKSUM_NONE) {
> - __wsum csum;
> - csum = csum_page(page, off, copy);
> - skb->csum = csum_block_add(skb->csum, csum, skb->len);
> - }
> -
> - skb_len_add(skb, copy);
> - refcount_add(copy, &sk->sk_wmem_alloc);
> } else if (!zc) {
> int i = skb_shinfo(skb)->nr_frags;
>
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES
2023-03-30 15:11 ` David Howells
2023-03-30 17:55 ` Willem de Bruijn
@ 2023-03-30 19:49 ` David Howells
1 sibling, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-30 19:49 UTC (permalink / raw)
To: Willem de Bruijn
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm
Willem de Bruijn <willemdebruijn.kernel@gmail.com> wrote:
> Yes exactly. I wasn't sure whether the inner loops required access to
> too many function scope variables to pull this off. But seems like it
> is doable. Great.
The same helpers can be used for both ipv4 and ipv6 as it turns out.
David
^ permalink raw reply [flat|nested] 77+ messages in thread
* [RFC PATCH v2 17/48] ip, udp: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (15 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 16/48] ip, udp: Support MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 18/48] udp: Convert udp_sendpage() to use MSG_SPLICE_PAGES David Howells
` (30 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Willem de Bruijn
If sendmsg() with MSG_SPLICE_PAGES encounters a page that shouldn't be
spliced - a slab page, for instance, or one with a zero count - make
__ip_append_data() copy it.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/ipv4/ip_output.c | 19 +++++++++++++++++++
1 file changed, 19 insertions(+)
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index 07736da70eab..e4aeaab704c8 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -1263,6 +1263,7 @@ static int __ip_append_data(struct sock *sk,
struct msghdr *msg = from;
struct page *page = NULL, **pages = &page;
size_t off;
+ bool put = false;
copy = iov_iter_extract_pages(&msg->msg_iter, &pages,
copy, 1, 0, &off);
@@ -1271,7 +1272,25 @@ static int __ip_append_data(struct sock *sk,
goto error;
}
+ if (!sendpage_ok(page)) {
+ const void *p = kmap_local_page(page);
+ void *q;
+
+ q = page_frag_memdup(NULL, p + off, copy,
+ sk->sk_allocation, ULONG_MAX);
+ kunmap_local(p);
+ if (!q) {
+ err = copy ?: -ENOMEM;
+ goto error;
+ }
+ page = virt_to_page(q);
+ off = offset_in_page(q);
+ put = true;
+ }
+
err = skb_append_pagefrags(skb, page, off, copy);
+ if (put)
+ put_page(page);
if (err < 0)
goto error;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 18/48] udp: Convert udp_sendpage() to use MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (16 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 17/48] ip, udp: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 19/48] af_unix: Support MSG_SPLICE_PAGES David Howells
` (29 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Willem de Bruijn
Convert udp_sendpage() to use sendmsg() with MSG_SPLICE_PAGES rather than
directly splicing in the pages itself.
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/ipv4/udp.c | 50 +++++++++-----------------------------------------
1 file changed, 9 insertions(+), 41 deletions(-)
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index c605d171eb2d..097feb92e215 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -1332,52 +1332,20 @@ EXPORT_SYMBOL(udp_sendmsg);
int udp_sendpage(struct sock *sk, struct page *page, int offset,
size_t size, int flags)
{
- struct inet_sock *inet = inet_sk(sk);
- struct udp_sock *up = udp_sk(sk);
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = flags | MSG_SPLICE_PAGES | MSG_MORE
+ };
int ret;
- if (flags & MSG_SENDPAGE_NOTLAST)
- flags |= MSG_MORE;
+ bvec_set_page(&bvec, page, size, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
- if (!up->pending) {
- struct msghdr msg = { .msg_flags = flags|MSG_MORE };
-
- /* Call udp_sendmsg to specify destination address which
- * sendpage interface can't pass.
- * This will succeed only when the socket is connected.
- */
- ret = udp_sendmsg(sk, &msg, 0);
- if (ret < 0)
- return ret;
- }
+ if (flags & MSG_SENDPAGE_NOTLAST)
+ msg.msg_flags |= MSG_MORE;
lock_sock(sk);
-
- if (unlikely(!up->pending)) {
- release_sock(sk);
-
- net_dbg_ratelimited("cork failed\n");
- return -EINVAL;
- }
-
- ret = ip_append_page(sk, &inet->cork.fl.u.ip4,
- page, offset, size, flags);
- if (ret == -EOPNOTSUPP) {
- release_sock(sk);
- return sock_no_sendpage(sk->sk_socket, page, offset,
- size, flags);
- }
- if (ret < 0) {
- udp_flush_pending_frames(sk);
- goto out;
- }
-
- up->len += size;
- if (!(READ_ONCE(up->corkflag) || (flags&MSG_MORE)))
- ret = udp_push_pending_frames(sk);
- if (!ret)
- ret = size;
-out:
+ ret = udp_sendmsg(sk, &msg, size);
release_sock(sk);
return ret;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 19/48] af_unix: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (17 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 18/48] udp: Convert udp_sendpage() to use MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 20/48] af_unix: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data David Howells
` (28 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Make AF_UNIX sendmsg() support MSG_SPLICE_PAGES, splicing in pages from the
source iterator if given and if ITER_BVEC and copying the data in
otherwise.
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/unix/af_unix.c | 93 ++++++++++++++++++++++++++++++++++++++--------
1 file changed, 77 insertions(+), 16 deletions(-)
diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 347122c3575e..84a0d97f1aa4 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -2151,6 +2151,53 @@ static int queue_oob(struct socket *sock, struct msghdr *msg, struct sock *other
}
#endif
+/*
+ * Extract pages from an iterator and add them to the socket buffer.
+ */
+static ssize_t unix_extract_bvec_to_skb(struct sk_buff *skb,
+ struct iov_iter *iter, ssize_t maxsize)
+{
+ struct page *pages[8], **ppages = pages;
+ unsigned int i, nr;
+ ssize_t ret = 0;
+
+ while (iter->count > 0) {
+ size_t off, len;
+
+ nr = min(MAX_SKB_FRAGS - skb_shinfo(skb)->nr_frags,
+ ARRAY_SIZE(pages));
+ if (nr == 0)
+ break;
+
+ len = iov_iter_extract_pages(iter, &ppages, maxsize, nr, 0, &off);
+ if (len <= 0) {
+ if (!ret)
+ ret = len ?: -EIO;
+ break;
+ }
+
+ i = 0;
+ do {
+ size_t part = min(PAGE_SIZE - off, len);
+
+ if (skb_append_pagefrags(skb, pages[i++], off, part) < 0) {
+ if (!ret)
+ ret = -EMSGSIZE;
+ goto out;
+ }
+ off = 0;
+ ret += part;
+ maxsize -= part;
+ len -= part;
+ } while (len > 0);
+ if (maxsize <= 0)
+ break;
+ }
+
+out:
+ return ret;
+}
+
static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
size_t len)
{
@@ -2194,19 +2241,25 @@ static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
while (sent < len) {
size = len - sent;
- /* Keep two messages in the pipe so it schedules better */
- size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64);
+ if (unlikely(msg->msg_flags & MSG_SPLICE_PAGES)) {
+ skb = sock_alloc_send_pskb(sk, 0, 0,
+ msg->msg_flags & MSG_DONTWAIT,
+ &err, 0);
+ } else {
+ /* Keep two messages in the pipe so it schedules better */
+ size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64);
- /* allow fallback to order-0 allocations */
- size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ);
+ /* allow fallback to order-0 allocations */
+ size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ);
- data_len = max_t(int, 0, size - SKB_MAX_HEAD(0));
+ data_len = max_t(int, 0, size - SKB_MAX_HEAD(0));
- data_len = min_t(size_t, size, PAGE_ALIGN(data_len));
+ data_len = min_t(size_t, size, PAGE_ALIGN(data_len));
- skb = sock_alloc_send_pskb(sk, size - data_len, data_len,
- msg->msg_flags & MSG_DONTWAIT, &err,
- get_order(UNIX_SKB_FRAGS_SZ));
+ skb = sock_alloc_send_pskb(sk, size - data_len, data_len,
+ msg->msg_flags & MSG_DONTWAIT, &err,
+ get_order(UNIX_SKB_FRAGS_SZ));
+ }
if (!skb)
goto out_err;
@@ -2218,13 +2271,21 @@ static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
}
fds_sent = true;
- skb_put(skb, size - data_len);
- skb->data_len = data_len;
- skb->len = size;
- err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
- if (err) {
- kfree_skb(skb);
- goto out_err;
+ if (unlikely(msg->msg_flags & MSG_SPLICE_PAGES)) {
+ size = unix_extract_bvec_to_skb(skb, &msg->msg_iter, size);
+ skb->data_len += size;
+ skb->len += size;
+ skb->truesize += size;
+ refcount_add(size, &sk->sk_wmem_alloc);
+ } else {
+ skb_put(skb, size - data_len);
+ skb->data_len = data_len;
+ skb->len = size;
+ err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
+ if (err) {
+ kfree_skb(skb);
+ goto out_err;
+ }
}
unix_state_lock(other);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 20/48] af_unix: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (18 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 19/48] af_unix: Support MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 21/48] crypto: af_alg: Pin pages rather than ref'ing if appropriate David Howells
` (27 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
If sendmsg() with MSG_SPLICE_PAGES encounters a page that shouldn't be
spliced - a slab page, for instance, or one with a zero count - make
unix_extract_bvec_to_skb() copy it.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/unix/af_unix.c | 44 +++++++++++++++++++++++++++++++++-----------
1 file changed, 33 insertions(+), 11 deletions(-)
diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 84a0d97f1aa4..b4b27a652ef0 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -2154,12 +2154,12 @@ static int queue_oob(struct socket *sock, struct msghdr *msg, struct sock *other
/*
* Extract pages from an iterator and add them to the socket buffer.
*/
-static ssize_t unix_extract_bvec_to_skb(struct sk_buff *skb,
- struct iov_iter *iter, ssize_t maxsize)
+static ssize_t unix_extract_bvec_to_skb(struct sk_buff *skb, struct iov_iter *iter,
+ ssize_t maxsize, gfp_t gfp)
{
struct page *pages[8], **ppages = pages;
unsigned int i, nr;
- ssize_t ret = 0;
+ ssize_t spliced = 0, ret = 0;
while (iter->count > 0) {
size_t off, len;
@@ -2171,31 +2171,52 @@ static ssize_t unix_extract_bvec_to_skb(struct sk_buff *skb,
len = iov_iter_extract_pages(iter, &ppages, maxsize, nr, 0, &off);
if (len <= 0) {
- if (!ret)
- ret = len ?: -EIO;
+ ret = len ?: -EIO;
break;
}
i = 0;
do {
+ struct page *page = pages[i++];
size_t part = min(PAGE_SIZE - off, len);
+ bool put = false;
+
+ if (!sendpage_ok(page)) {
+ const void *p = kmap_local_page(page);
+ void *q;
+
+ q = page_frag_memdup(NULL, p + off, part, gfp,
+ ULONG_MAX);
+ kunmap_local(p);
+ if (!q) {
+ iov_iter_revert(iter, len);
+ ret = -ENOMEM;
+ goto out;
+ }
+ page = virt_to_page(q);
+ off = offset_in_page(q);
+ put = true;
+ }
- if (skb_append_pagefrags(skb, pages[i++], off, part) < 0) {
- if (!ret)
- ret = -EMSGSIZE;
+ ret = skb_append_pagefrags(skb, page, off, part);
+ if (put)
+ put_page(page);
+ if (ret < 0) {
+ iov_iter_revert(iter, len);
goto out;
}
off = 0;
- ret += part;
+ spliced += part;
maxsize -= part;
len -= part;
} while (len > 0);
+
if (maxsize <= 0)
break;
}
out:
- return ret;
+ return spliced ?: ret;
}
static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
@@ -2272,7 +2293,8 @@ static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
fds_sent = true;
if (unlikely(msg->msg_flags & MSG_SPLICE_PAGES)) {
- size = unix_extract_bvec_to_skb(skb, &msg->msg_iter, size);
+ size = unix_extract_bvec_to_skb(skb, &msg->msg_iter, size,
+ sk->sk_allocation);
skb->data_len += size;
skb->len += size;
skb->truesize += size;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 21/48] crypto: af_alg: Pin pages rather than ref'ing if appropriate
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (19 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 20/48] af_unix: Make sendmsg(MSG_SPLICE_PAGES) copy unspliceable data David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 22/48] crypto: af_alg: Use netfs_extract_iter_to_sg() to create scatterlists David Howells
` (26 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Herbert Xu,
linux-crypto
Convert AF_ALG to use iov_iter_extract_pages() instead of
iov_iter_get_pages(). This will pin pages or leave them unaltered rather
than getting a ref on them as appropriate to the iterator.
The pages need to be pinned for DIO-read rather than having refs taken on
them to prevent VM copy-on-write from malfunctioning during a concurrent
fork() (the result of the I/O would otherwise end up only visible to the
child process and not the parent).
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: linux-crypto@vger.kernel.org
---
crypto/af_alg.c | 10 +++++++---
include/crypto/if_alg.h | 1 +
2 files changed, 8 insertions(+), 3 deletions(-)
diff --git a/crypto/af_alg.c b/crypto/af_alg.c
index 5f7252a5b7b4..7caff10df643 100644
--- a/crypto/af_alg.c
+++ b/crypto/af_alg.c
@@ -533,14 +533,17 @@ static const struct net_proto_family alg_family = {
int af_alg_make_sg(struct af_alg_sgl *sgl, struct iov_iter *iter, int len)
{
+ struct page **pages = sgl->pages;
size_t off;
ssize_t n;
int npages, i;
- n = iov_iter_get_pages2(iter, sgl->pages, len, ALG_MAX_PAGES, &off);
+ n = iov_iter_extract_pages(iter, &pages, len, ALG_MAX_PAGES, 0, &off);
if (n < 0)
return n;
+ sgl->need_unpin = iov_iter_extract_will_pin(iter);
+
npages = DIV_ROUND_UP(off + n, PAGE_SIZE);
if (WARN_ON(npages == 0))
return -EINVAL;
@@ -573,8 +576,9 @@ void af_alg_free_sg(struct af_alg_sgl *sgl)
{
int i;
- for (i = 0; i < sgl->npages; i++)
- put_page(sgl->pages[i]);
+ if (sgl->need_unpin)
+ for (i = 0; i < sgl->npages; i++)
+ unpin_user_page(sgl->pages[i]);
}
EXPORT_SYMBOL_GPL(af_alg_free_sg);
diff --git a/include/crypto/if_alg.h b/include/crypto/if_alg.h
index 7e76623f9ec3..46494b33f5bc 100644
--- a/include/crypto/if_alg.h
+++ b/include/crypto/if_alg.h
@@ -59,6 +59,7 @@ struct af_alg_sgl {
struct scatterlist sg[ALG_MAX_PAGES + 1];
struct page *pages[ALG_MAX_PAGES];
unsigned int npages;
+ bool need_unpin;
};
/* TX SGL entry */
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 22/48] crypto: af_alg: Use netfs_extract_iter_to_sg() to create scatterlists
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (20 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 21/48] crypto: af_alg: Pin pages rather than ref'ing if appropriate David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 23/48] crypto: af_alg: Indent the loop in af_alg_sendmsg() David Howells
` (25 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Herbert Xu,
linux-crypto
Use netfs_extract_iter_to_sg() to decant the destination iterator into a
scatterlist in af_alg_get_rsgl(). af_alg_make_sg() can then be removed.
[!] Note that if this fits, netfs_extract_iter_to_sg() should move to core
code.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: linux-crypto@vger.kernel.org
---
crypto/af_alg.c | 55 ++++++++++-------------------------------
crypto/algif_aead.c | 12 ++++-----
crypto/algif_hash.c | 18 ++++++++++----
crypto/algif_skcipher.c | 2 +-
include/crypto/if_alg.h | 6 ++---
5 files changed, 35 insertions(+), 58 deletions(-)
diff --git a/crypto/af_alg.c b/crypto/af_alg.c
index 7caff10df643..1dafd088ad45 100644
--- a/crypto/af_alg.c
+++ b/crypto/af_alg.c
@@ -22,6 +22,7 @@
#include <linux/sched/signal.h>
#include <linux/security.h>
#include <linux/string.h>
+#include <linux/netfs.h>
#include <keys/user-type.h>
#include <keys/trusted-type.h>
#include <keys/encrypted-type.h>
@@ -531,45 +532,11 @@ static const struct net_proto_family alg_family = {
.owner = THIS_MODULE,
};
-int af_alg_make_sg(struct af_alg_sgl *sgl, struct iov_iter *iter, int len)
-{
- struct page **pages = sgl->pages;
- size_t off;
- ssize_t n;
- int npages, i;
-
- n = iov_iter_extract_pages(iter, &pages, len, ALG_MAX_PAGES, 0, &off);
- if (n < 0)
- return n;
-
- sgl->need_unpin = iov_iter_extract_will_pin(iter);
-
- npages = DIV_ROUND_UP(off + n, PAGE_SIZE);
- if (WARN_ON(npages == 0))
- return -EINVAL;
- /* Add one extra for linking */
- sg_init_table(sgl->sg, npages + 1);
-
- for (i = 0, len = n; i < npages; i++) {
- int plen = min_t(int, len, PAGE_SIZE - off);
-
- sg_set_page(sgl->sg + i, sgl->pages[i], plen, off);
-
- off = 0;
- len -= plen;
- }
- sg_mark_end(sgl->sg + npages - 1);
- sgl->npages = npages;
-
- return n;
-}
-EXPORT_SYMBOL_GPL(af_alg_make_sg);
-
static void af_alg_link_sg(struct af_alg_sgl *sgl_prev,
struct af_alg_sgl *sgl_new)
{
- sg_unmark_end(sgl_prev->sg + sgl_prev->npages - 1);
- sg_chain(sgl_prev->sg, sgl_prev->npages + 1, sgl_new->sg);
+ sg_unmark_end(sgl_prev->sgt.sgl + sgl_prev->sgt.nents - 1);
+ sg_chain(sgl_prev->sgt.sgl, sgl_prev->sgt.nents + 1, sgl_new->sgt.sgl);
}
void af_alg_free_sg(struct af_alg_sgl *sgl)
@@ -577,8 +544,8 @@ void af_alg_free_sg(struct af_alg_sgl *sgl)
int i;
if (sgl->need_unpin)
- for (i = 0; i < sgl->npages; i++)
- unpin_user_page(sgl->pages[i]);
+ for (i = 0; i < sgl->sgt.nents; i++)
+ unpin_user_page(sg_page(&sgl->sgt.sgl[i]));
}
EXPORT_SYMBOL_GPL(af_alg_free_sg);
@@ -1292,8 +1259,8 @@ int af_alg_get_rsgl(struct sock *sk, struct msghdr *msg, int flags,
while (maxsize > len && msg_data_left(msg)) {
struct af_alg_rsgl *rsgl;
+ ssize_t err;
size_t seglen;
- int err;
/* limit the amount of readable buffers */
if (!af_alg_readable(sk))
@@ -1310,16 +1277,20 @@ int af_alg_get_rsgl(struct sock *sk, struct msghdr *msg, int flags,
return -ENOMEM;
}
- rsgl->sgl.npages = 0;
+ rsgl->sgl.sgt.sgl = rsgl->sgl.sgl;
+ rsgl->sgl.sgt.nents = 0;
+ rsgl->sgl.sgt.orig_nents = 0;
list_add_tail(&rsgl->list, &areq->rsgl_list);
- /* make one iovec available as scatterlist */
- err = af_alg_make_sg(&rsgl->sgl, &msg->msg_iter, seglen);
+ err = netfs_extract_iter_to_sg(&msg->msg_iter, seglen,
+ &rsgl->sgl.sgt, ALG_MAX_PAGES, 0);
if (err < 0) {
rsgl->sg_num_bytes = 0;
return err;
}
+ rsgl->sgl.need_unpin = iov_iter_extract_will_pin(&msg->msg_iter);
+
/* chain the new scatterlist with previous one */
if (areq->last_rsgl)
af_alg_link_sg(&areq->last_rsgl->sgl, &rsgl->sgl);
diff --git a/crypto/algif_aead.c b/crypto/algif_aead.c
index 42493b4d8ce4..f6aa3856d8d5 100644
--- a/crypto/algif_aead.c
+++ b/crypto/algif_aead.c
@@ -210,7 +210,7 @@ static int _aead_recvmsg(struct socket *sock, struct msghdr *msg,
*/
/* Use the RX SGL as source (and destination) for crypto op. */
- rsgl_src = areq->first_rsgl.sgl.sg;
+ rsgl_src = areq->first_rsgl.sgl.sgt.sgl;
if (ctx->enc) {
/*
@@ -224,7 +224,7 @@ static int _aead_recvmsg(struct socket *sock, struct msghdr *msg,
* RX SGL: AAD || PT || Tag
*/
err = crypto_aead_copy_sgl(null_tfm, tsgl_src,
- areq->first_rsgl.sgl.sg, processed);
+ areq->first_rsgl.sgl.sgt.sgl, processed);
if (err)
goto free;
af_alg_pull_tsgl(sk, processed, NULL, 0);
@@ -242,7 +242,7 @@ static int _aead_recvmsg(struct socket *sock, struct msghdr *msg,
/* Copy AAD || CT to RX SGL buffer for in-place operation. */
err = crypto_aead_copy_sgl(null_tfm, tsgl_src,
- areq->first_rsgl.sgl.sg, outlen);
+ areq->first_rsgl.sgl.sgt.sgl, outlen);
if (err)
goto free;
@@ -268,8 +268,8 @@ static int _aead_recvmsg(struct socket *sock, struct msghdr *msg,
/* RX SGL present */
struct af_alg_sgl *sgl_prev = &areq->last_rsgl->sgl;
- sg_unmark_end(sgl_prev->sg + sgl_prev->npages - 1);
- sg_chain(sgl_prev->sg, sgl_prev->npages + 1,
+ sg_unmark_end(sgl_prev->sgt.sgl + sgl_prev->sgt.nents - 1);
+ sg_chain(sgl_prev->sgt.sgl, sgl_prev->sgt.nents + 1,
areq->tsgl);
} else
/* no RX SGL present (e.g. authentication only) */
@@ -278,7 +278,7 @@ static int _aead_recvmsg(struct socket *sock, struct msghdr *msg,
/* Initialize the crypto operation */
aead_request_set_crypt(&areq->cra_u.aead_req, rsgl_src,
- areq->first_rsgl.sgl.sg, used, ctx->iv);
+ areq->first_rsgl.sgl.sgt.sgl, used, ctx->iv);
aead_request_set_ad(&areq->cra_u.aead_req, ctx->aead_assoclen);
aead_request_set_tfm(&areq->cra_u.aead_req, tfm);
diff --git a/crypto/algif_hash.c b/crypto/algif_hash.c
index 1d017ec5c63c..f051fa624bd7 100644
--- a/crypto/algif_hash.c
+++ b/crypto/algif_hash.c
@@ -14,6 +14,7 @@
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/net.h>
+#include <linux/netfs.h>
#include <net/sock.h>
struct hash_ctx {
@@ -91,13 +92,20 @@ static int hash_sendmsg(struct socket *sock, struct msghdr *msg,
if (len > limit)
len = limit;
- len = af_alg_make_sg(&ctx->sgl, &msg->msg_iter, len);
+ ctx->sgl.sgt.sgl = ctx->sgl.sgl;
+ ctx->sgl.sgt.nents = 0;
+ ctx->sgl.sgt.orig_nents = 0;
+
+ len = netfs_extract_iter_to_sg(&msg->msg_iter, len,
+ &ctx->sgl.sgt, ALG_MAX_PAGES, 0);
if (len < 0) {
err = copied ? 0 : len;
goto unlock;
}
- ahash_request_set_crypt(&ctx->req, ctx->sgl.sg, NULL, len);
+ ctx->sgl.need_unpin = iov_iter_extract_will_pin(&msg->msg_iter);
+
+ ahash_request_set_crypt(&ctx->req, ctx->sgl.sgt.sgl, NULL, len);
err = crypto_wait_req(crypto_ahash_update(&ctx->req),
&ctx->wait);
@@ -141,8 +149,8 @@ static ssize_t hash_sendpage(struct socket *sock, struct page *page,
flags |= MSG_MORE;
lock_sock(sk);
- sg_init_table(ctx->sgl.sg, 1);
- sg_set_page(ctx->sgl.sg, page, size, offset);
+ sg_init_table(ctx->sgl.sgl, 1);
+ sg_set_page(ctx->sgl.sgl, page, size, offset);
if (!(flags & MSG_MORE)) {
err = hash_alloc_result(sk, ctx);
@@ -151,7 +159,7 @@ static ssize_t hash_sendpage(struct socket *sock, struct page *page,
} else if (!ctx->more)
hash_free_result(sk, ctx);
- ahash_request_set_crypt(&ctx->req, ctx->sgl.sg, ctx->result, size);
+ ahash_request_set_crypt(&ctx->req, ctx->sgl.sgl, ctx->result, size);
if (!(flags & MSG_MORE)) {
if (ctx->more)
diff --git a/crypto/algif_skcipher.c b/crypto/algif_skcipher.c
index ee8890ee8f33..a251cd6bd5b9 100644
--- a/crypto/algif_skcipher.c
+++ b/crypto/algif_skcipher.c
@@ -105,7 +105,7 @@ static int _skcipher_recvmsg(struct socket *sock, struct msghdr *msg,
/* Initialize the crypto operation */
skcipher_request_set_tfm(&areq->cra_u.skcipher_req, tfm);
skcipher_request_set_crypt(&areq->cra_u.skcipher_req, areq->tsgl,
- areq->first_rsgl.sgl.sg, len, ctx->iv);
+ areq->first_rsgl.sgl.sgt.sgl, len, ctx->iv);
if (msg->msg_iocb && !is_sync_kiocb(msg->msg_iocb)) {
/* AIO operation */
diff --git a/include/crypto/if_alg.h b/include/crypto/if_alg.h
index 46494b33f5bc..34224e77f5a2 100644
--- a/include/crypto/if_alg.h
+++ b/include/crypto/if_alg.h
@@ -56,9 +56,8 @@ struct af_alg_type {
};
struct af_alg_sgl {
- struct scatterlist sg[ALG_MAX_PAGES + 1];
- struct page *pages[ALG_MAX_PAGES];
- unsigned int npages;
+ struct sg_table sgt;
+ struct scatterlist sgl[ALG_MAX_PAGES + 1];
bool need_unpin;
};
@@ -164,7 +163,6 @@ int af_alg_release(struct socket *sock);
void af_alg_release_parent(struct sock *sk);
int af_alg_accept(struct sock *sk, struct socket *newsock, bool kern);
-int af_alg_make_sg(struct af_alg_sgl *sgl, struct iov_iter *iter, int len);
void af_alg_free_sg(struct af_alg_sgl *sgl);
static inline struct alg_sock *alg_sk(struct sock *sk)
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 23/48] crypto: af_alg: Indent the loop in af_alg_sendmsg()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (21 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 22/48] crypto: af_alg: Use netfs_extract_iter_to_sg() to create scatterlists David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 24/48] crypto: af_alg: Support MSG_SPLICE_PAGES David Howells
` (24 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Herbert Xu,
linux-crypto
Put the loop in af_alg_sendmsg() into an if-statement to indent it to make
the next patch easier to review as that will add another branch to handle
MSG_SPLICE_PAGES to the if-statement.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-crypto@vger.kernel.org
cc: netdev@vger.kernel.org
---
crypto/af_alg.c | 50 +++++++++++++++++++++++++------------------------
1 file changed, 26 insertions(+), 24 deletions(-)
diff --git a/crypto/af_alg.c b/crypto/af_alg.c
index 1dafd088ad45..483821e310e9 100644
--- a/crypto/af_alg.c
+++ b/crypto/af_alg.c
@@ -1031,35 +1031,37 @@ int af_alg_sendmsg(struct socket *sock, struct msghdr *msg, size_t size,
if (sgl->cur)
sg_unmark_end(sg + sgl->cur - 1);
- do {
- struct page *pg;
- unsigned int i = sgl->cur;
+ if (1 /* TODO check MSG_SPLICE_PAGES */) {
+ do {
+ struct page *pg;
+ unsigned int i = sgl->cur;
- plen = min_t(size_t, len, PAGE_SIZE);
+ plen = min_t(size_t, len, PAGE_SIZE);
- pg = alloc_page(GFP_KERNEL);
- if (!pg) {
- err = -ENOMEM;
- goto unlock;
- }
+ pg = alloc_page(GFP_KERNEL);
+ if (!pg) {
+ err = -ENOMEM;
+ goto unlock;
+ }
- sg_assign_page(sg + i, pg);
+ sg_assign_page(sg + i, pg);
- err = memcpy_from_msg(page_address(sg_page(sg + i)),
- msg, plen);
- if (err) {
- __free_page(sg_page(sg + i));
- sg_assign_page(sg + i, NULL);
- goto unlock;
- }
+ err = memcpy_from_msg(page_address(sg_page(sg + i)),
+ msg, plen);
+ if (err) {
+ __free_page(sg_page(sg + i));
+ sg_assign_page(sg + i, NULL);
+ goto unlock;
+ }
- sg[i].length = plen;
- len -= plen;
- ctx->used += plen;
- copied += plen;
- size -= plen;
- sgl->cur++;
- } while (len && sgl->cur < MAX_SGL_ENTS);
+ sg[i].length = plen;
+ len -= plen;
+ ctx->used += plen;
+ copied += plen;
+ size -= plen;
+ sgl->cur++;
+ } while (len && sgl->cur < MAX_SGL_ENTS);
+ }
if (!size)
sg_mark_end(sg + sgl->cur - 1);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 24/48] crypto: af_alg: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (22 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 23/48] crypto: af_alg: Indent the loop in af_alg_sendmsg() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 25/48] crypto: af_alg: Convert af_alg_sendpage() to use MSG_SPLICE_PAGES David Howells
` (23 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Herbert Xu,
linux-crypto
Make AF_ALG sendmsg() support MSG_SPLICE_PAGES. This causes pages to be
spliced from the source iterator if possible (the iterator must be
ITER_BVEC and the pages must be spliceable).
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
[!] Note that this makes use of netfs_extract_iter_to_sg() from netfslib.
This probably needs moving to core code somewhere.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-crypto@vger.kernel.org
cc: netdev@vger.kernel.org
---
crypto/Kconfig | 1 +
crypto/af_alg.c | 28 ++++++++++++++++++++++++++--
crypto/algif_aead.c | 22 +++++++++++-----------
crypto/algif_skcipher.c | 8 ++++----
4 files changed, 42 insertions(+), 17 deletions(-)
diff --git a/crypto/Kconfig b/crypto/Kconfig
index 9c86f7045157..8c04ecbb4395 100644
--- a/crypto/Kconfig
+++ b/crypto/Kconfig
@@ -1297,6 +1297,7 @@ menu "Userspace interface"
config CRYPTO_USER_API
tristate
+ select NETFS_SUPPORT # for netfs_extract_iter_to_sg()
config CRYPTO_USER_API_HASH
tristate "Hash algorithms"
diff --git a/crypto/af_alg.c b/crypto/af_alg.c
index 483821e310e9..3088ab298632 100644
--- a/crypto/af_alg.c
+++ b/crypto/af_alg.c
@@ -941,6 +941,10 @@ int af_alg_sendmsg(struct socket *sock, struct msghdr *msg, size_t size,
bool init = false;
int err = 0;
+ if ((msg->msg_flags & MSG_SPLICE_PAGES) &&
+ !iov_iter_is_bvec(&msg->msg_iter))
+ return -EINVAL;
+
if (msg->msg_controllen) {
err = af_alg_cmsg_send(msg, &con);
if (err)
@@ -986,7 +990,7 @@ int af_alg_sendmsg(struct socket *sock, struct msghdr *msg, size_t size,
while (size) {
struct scatterlist *sg;
size_t len = size;
- size_t plen;
+ ssize_t plen;
/* use the existing memory in an allocated page */
if (ctx->merge) {
@@ -1031,7 +1035,27 @@ int af_alg_sendmsg(struct socket *sock, struct msghdr *msg, size_t size,
if (sgl->cur)
sg_unmark_end(sg + sgl->cur - 1);
- if (1 /* TODO check MSG_SPLICE_PAGES */) {
+ if (msg->msg_flags & MSG_SPLICE_PAGES) {
+ struct sg_table sgtable = {
+ .sgl = sg,
+ .nents = sgl->cur,
+ .orig_nents = sgl->cur,
+ };
+
+ plen = netfs_extract_iter_to_sg(&msg->msg_iter, len,
+ &sgtable, MAX_SGL_ENTS, 0);
+ if (plen < 0) {
+ err = plen;
+ goto unlock;
+ }
+
+ for (; sgl->cur < sgtable.nents; sgl->cur++)
+ get_page(sg_page(&sg[sgl->cur]));
+ len -= plen;
+ ctx->used += plen;
+ copied += plen;
+ size -= plen;
+ } else {
do {
struct page *pg;
unsigned int i = sgl->cur;
diff --git a/crypto/algif_aead.c b/crypto/algif_aead.c
index f6aa3856d8d5..b16111a3025a 100644
--- a/crypto/algif_aead.c
+++ b/crypto/algif_aead.c
@@ -9,8 +9,8 @@
* The following concept of the memory management is used:
*
* The kernel maintains two SGLs, the TX SGL and the RX SGL. The TX SGL is
- * filled by user space with the data submitted via sendpage/sendmsg. Filling
- * up the TX SGL does not cause a crypto operation -- the data will only be
+ * filled by user space with the data submitted via sendpage. Filling up
+ * the TX SGL does not cause a crypto operation -- the data will only be
* tracked by the kernel. Upon receipt of one recvmsg call, the caller must
* provide a buffer which is tracked with the RX SGL.
*
@@ -113,19 +113,19 @@ static int _aead_recvmsg(struct socket *sock, struct msghdr *msg,
}
/*
- * Data length provided by caller via sendmsg/sendpage that has not
- * yet been processed.
+ * Data length provided by caller via sendmsg that has not yet been
+ * processed.
*/
used = ctx->used;
/*
- * Make sure sufficient data is present -- note, the same check is
- * also present in sendmsg/sendpage. The checks in sendpage/sendmsg
- * shall provide an information to the data sender that something is
- * wrong, but they are irrelevant to maintain the kernel integrity.
- * We need this check here too in case user space decides to not honor
- * the error message in sendmsg/sendpage and still call recvmsg. This
- * check here protects the kernel integrity.
+ * Make sure sufficient data is present -- note, the same check is also
+ * present in sendmsg. The checks in sendmsg shall provide an
+ * information to the data sender that something is wrong, but they are
+ * irrelevant to maintain the kernel integrity. We need this check
+ * here too in case user space decides to not honor the error message
+ * in sendmsg and still call recvmsg. This check here protects the
+ * kernel integrity.
*/
if (!aead_sufficient_data(sk))
return -EINVAL;
diff --git a/crypto/algif_skcipher.c b/crypto/algif_skcipher.c
index a251cd6bd5b9..b1f321b9f846 100644
--- a/crypto/algif_skcipher.c
+++ b/crypto/algif_skcipher.c
@@ -9,10 +9,10 @@
* The following concept of the memory management is used:
*
* The kernel maintains two SGLs, the TX SGL and the RX SGL. The TX SGL is
- * filled by user space with the data submitted via sendpage/sendmsg. Filling
- * up the TX SGL does not cause a crypto operation -- the data will only be
- * tracked by the kernel. Upon receipt of one recvmsg call, the caller must
- * provide a buffer which is tracked with the RX SGL.
+ * filled by user space with the data submitted via sendmsg. Filling up the TX
+ * SGL does not cause a crypto operation -- the data will only be tracked by
+ * the kernel. Upon receipt of one recvmsg call, the caller must provide a
+ * buffer which is tracked with the RX SGL.
*
* During the processing of the recvmsg operation, the cipher request is
* allocated and prepared. As part of the recvmsg operation, the processed
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 25/48] crypto: af_alg: Convert af_alg_sendpage() to use MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (23 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 24/48] crypto: af_alg: Support MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 26/48] crypto: af_alg/hash: Support MSG_SPLICE_PAGES David Howells
` (22 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Herbert Xu,
linux-crypto
Convert af_alg_sendpage() to use sendmsg() with MSG_SPLICE_PAGES rather
than directly splicing in the pages itself.
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
[!] Note that this makes use of netfs_extract_iter_to_sg() from netfslib.
This probably needs moving to core code somewhere.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-crypto@vger.kernel.org
cc: netdev@vger.kernel.org
---
crypto/af_alg.c | 53 +++++++++----------------------------------------
1 file changed, 9 insertions(+), 44 deletions(-)
diff --git a/crypto/af_alg.c b/crypto/af_alg.c
index 3088ab298632..7fe8c8db6bb5 100644
--- a/crypto/af_alg.c
+++ b/crypto/af_alg.c
@@ -1118,53 +1118,18 @@ EXPORT_SYMBOL_GPL(af_alg_sendmsg);
ssize_t af_alg_sendpage(struct socket *sock, struct page *page,
int offset, size_t size, int flags)
{
- struct sock *sk = sock->sk;
- struct alg_sock *ask = alg_sk(sk);
- struct af_alg_ctx *ctx = ask->private;
- struct af_alg_tsgl *sgl;
- int err = -EINVAL;
-
- if (flags & MSG_SENDPAGE_NOTLAST)
- flags |= MSG_MORE;
-
- lock_sock(sk);
- if (!ctx->more && ctx->used)
- goto unlock;
-
- if (!size)
- goto done;
-
- if (!af_alg_writable(sk)) {
- err = af_alg_wait_for_wmem(sk, flags);
- if (err)
- goto unlock;
- }
-
- err = af_alg_alloc_tsgl(sk);
- if (err)
- goto unlock;
-
- ctx->merge = 0;
- sgl = list_entry(ctx->tsgl_list.prev, struct af_alg_tsgl, list);
-
- if (sgl->cur)
- sg_unmark_end(sgl->sg + sgl->cur - 1);
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = flags | MSG_SPLICE_PAGES,
+ };
- sg_mark_end(sgl->sg + sgl->cur);
+ bvec_set_page(&bvec, page, size, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
- get_page(page);
- sg_set_page(sgl->sg + sgl->cur, page, size, offset);
- sgl->cur++;
- ctx->used += size;
-
-done:
- ctx->more = flags & MSG_MORE;
-
-unlock:
- af_alg_data_wakeup(sk);
- release_sock(sk);
+ if (flags & MSG_SENDPAGE_NOTLAST)
+ msg.msg_flags |= MSG_MORE;
- return err ?: size;
+ return sock_sendmsg(sock, &msg);
}
EXPORT_SYMBOL_GPL(af_alg_sendpage);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 26/48] crypto: af_alg/hash: Support MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (24 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 25/48] crypto: af_alg: Convert af_alg_sendpage() to use MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 27/48] splice, net: Use sendmsg(MSG_SPLICE_PAGES) rather than ->sendpage() David Howells
` (21 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Herbert Xu,
linux-crypto
Make AF_ALG sendmsg() support MSG_SPLICE_PAGES in the hashing code. This
causes pages to be spliced from the source iterator if possible.
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
[!] Note that this makes use of netfs_extract_iter_to_sg() from netfslib.
This probably needs moving to core code somewhere.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-crypto@vger.kernel.org
cc: netdev@vger.kernel.org
---
crypto/af_alg.c | 11 +++--
crypto/algif_hash.c | 99 ++++++++++++++++++++++++++++-----------------
2 files changed, 70 insertions(+), 40 deletions(-)
diff --git a/crypto/af_alg.c b/crypto/af_alg.c
index 7fe8c8db6bb5..686610a4986f 100644
--- a/crypto/af_alg.c
+++ b/crypto/af_alg.c
@@ -543,9 +543,14 @@ void af_alg_free_sg(struct af_alg_sgl *sgl)
{
int i;
- if (sgl->need_unpin)
- for (i = 0; i < sgl->sgt.nents; i++)
- unpin_user_page(sg_page(&sgl->sgt.sgl[i]));
+ if (sgl->sgt.sgl) {
+ if (sgl->need_unpin)
+ for (i = 0; i < sgl->sgt.nents; i++)
+ unpin_user_page(sg_page(&sgl->sgt.sgl[i]));
+ if (sgl->sgt.sgl != sgl->sgl)
+ kvfree(sgl->sgt.sgl);
+ sgl->sgt.sgl = NULL;
+ }
}
EXPORT_SYMBOL_GPL(af_alg_free_sg);
diff --git a/crypto/algif_hash.c b/crypto/algif_hash.c
index f051fa624bd7..b89c2c50cecc 100644
--- a/crypto/algif_hash.c
+++ b/crypto/algif_hash.c
@@ -64,77 +64,102 @@ static void hash_free_result(struct sock *sk, struct hash_ctx *ctx)
static int hash_sendmsg(struct socket *sock, struct msghdr *msg,
size_t ignored)
{
- int limit = ALG_MAX_PAGES * PAGE_SIZE;
struct sock *sk = sock->sk;
struct alg_sock *ask = alg_sk(sk);
struct hash_ctx *ctx = ask->private;
- long copied = 0;
+ ssize_t copied = 0;
+ size_t len, max_pages = ALG_MAX_PAGES, npages;
+ bool continuing = ctx->more, need_init = false;
int err;
- if (limit > sk->sk_sndbuf)
- limit = sk->sk_sndbuf;
+ /* Don't limit to ALG_MAX_PAGES if the pages are all already pinned. */
+ if (!user_backed_iter(&msg->msg_iter))
+ max_pages = INT_MAX;
+ else
+ max_pages = min_t(size_t, max_pages,
+ DIV_ROUND_UP(sk->sk_sndbuf, PAGE_SIZE));
lock_sock(sk);
- if (!ctx->more) {
+ if (!continuing) {
if ((msg->msg_flags & MSG_MORE))
hash_free_result(sk, ctx);
-
- err = crypto_wait_req(crypto_ahash_init(&ctx->req), &ctx->wait);
- if (err)
- goto unlock;
+ need_init = true;
}
ctx->more = false;
while (msg_data_left(msg)) {
- int len = msg_data_left(msg);
-
- if (len > limit)
- len = limit;
-
ctx->sgl.sgt.sgl = ctx->sgl.sgl;
ctx->sgl.sgt.nents = 0;
ctx->sgl.sgt.orig_nents = 0;
- len = netfs_extract_iter_to_sg(&msg->msg_iter, len,
- &ctx->sgl.sgt, ALG_MAX_PAGES, 0);
- if (len < 0) {
- err = copied ? 0 : len;
- goto unlock;
+ err = -EIO;
+ npages = iov_iter_npages(&msg->msg_iter, max_pages);
+ if (npages == 0)
+ goto unlock_free;
+
+ if (npages > ARRAY_SIZE(ctx->sgl.sgl)) {
+ err = -ENOMEM;
+ ctx->sgl.sgt.sgl =
+ kvmalloc(array_size(npages, sizeof(*ctx->sgl.sgt.sgl)),
+ GFP_KERNEL);
+ if (!ctx->sgl.sgt.sgl)
+ goto unlock_free;
}
+ sg_init_table(ctx->sgl.sgl, npages);
ctx->sgl.need_unpin = iov_iter_extract_will_pin(&msg->msg_iter);
- ahash_request_set_crypt(&ctx->req, ctx->sgl.sgt.sgl, NULL, len);
+ err = netfs_extract_iter_to_sg(&msg->msg_iter, LONG_MAX,
+ &ctx->sgl.sgt, npages, 0);
+ if (err < 0)
+ goto unlock_free;
+ len = err;
+ sg_mark_end(ctx->sgl.sgt.sgl + ctx->sgl.sgt.nents - 1);
- err = crypto_wait_req(crypto_ahash_update(&ctx->req),
- &ctx->wait);
- af_alg_free_sg(&ctx->sgl);
- if (err) {
- iov_iter_revert(&msg->msg_iter, len);
- goto unlock;
+ if (!msg_data_left(msg)) {
+ err = hash_alloc_result(sk, ctx);
+ if (err)
+ goto unlock_free;
}
- copied += len;
- }
+ ahash_request_set_crypt(&ctx->req, ctx->sgl.sgt.sgl, ctx->result, len);
- err = 0;
+ if (!msg_data_left(msg) && !continuing && !(msg->msg_flags & MSG_MORE)) {
+ err = crypto_ahash_digest(&ctx->req);
+ } else {
+ if (need_init) {
+ err = crypto_wait_req(crypto_ahash_init(&ctx->req),
+ &ctx->wait);
+ if (err)
+ goto unlock_free;
+ need_init = false;
+ }
+
+ if (msg_data_left(msg) || (msg->msg_flags & MSG_MORE))
+ err = crypto_ahash_update(&ctx->req);
+ else
+ err = crypto_ahash_finup(&ctx->req);
+ continuing = true;
+ }
- ctx->more = msg->msg_flags & MSG_MORE;
- if (!ctx->more) {
- err = hash_alloc_result(sk, ctx);
+ err = crypto_wait_req(err, &ctx->wait);
if (err)
- goto unlock;
+ goto unlock_free;
- ahash_request_set_crypt(&ctx->req, NULL, ctx->result, 0);
- err = crypto_wait_req(crypto_ahash_final(&ctx->req),
- &ctx->wait);
+ copied += len;
+ af_alg_free_sg(&ctx->sgl);
}
+ ctx->more = msg->msg_flags & MSG_MORE;
+ err = 0;
unlock:
release_sock(sk);
+ return copied ?: err;
- return err ?: copied;
+unlock_free:
+ af_alg_free_sg(&ctx->sgl);
+ goto unlock;
}
static ssize_t hash_sendpage(struct socket *sock, struct page *page,
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 27/48] splice, net: Use sendmsg(MSG_SPLICE_PAGES) rather than ->sendpage()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (25 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 26/48] crypto: af_alg/hash: Support MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 28/48] splice: Reimplement splice_to_socket() to pass multiple bufs to sendmsg() David Howells
` (20 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Use sendmsg(MSG_SPLICE_PAGES) rather than ->sendpage() to splice data from
a pipe to a socket. This paves the way for passing in multiple pages at
once from a pipe and the handling of multipage folios.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
fs/splice.c | 42 +++++++++++++++++++++++-------------------
include/linux/fs.h | 2 --
include/linux/splice.h | 2 ++
net/socket.c | 26 ++------------------------
4 files changed, 27 insertions(+), 45 deletions(-)
diff --git a/fs/splice.c b/fs/splice.c
index f46dd1fb367b..23ead122d631 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -32,6 +32,7 @@
#include <linux/uio.h>
#include <linux/security.h>
#include <linux/gfp.h>
+#include <linux/net.h>
#include <linux/socket.h>
#include <linux/sched/signal.h>
@@ -410,29 +411,32 @@ const struct pipe_buf_operations nosteal_pipe_buf_ops = {
};
EXPORT_SYMBOL(nosteal_pipe_buf_ops);
+#ifdef CONFIG_NET
/*
* Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
* using sendpage(). Return the number of bytes sent.
*/
-static int pipe_to_sendpage(struct pipe_inode_info *pipe,
- struct pipe_buffer *buf, struct splice_desc *sd)
+static int pipe_to_sendmsg(struct pipe_inode_info *pipe,
+ struct pipe_buffer *buf, struct splice_desc *sd)
{
- struct file *file = sd->u.file;
- loff_t pos = sd->pos;
- int more;
-
- if (!likely(file->f_op->sendpage))
- return -EINVAL;
+ struct socket *sock = sock_from_file(sd->u.file);
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = MSG_SPLICE_PAGES,
+ };
- more = (sd->flags & SPLICE_F_MORE) ? MSG_MORE : 0;
+ if (sd->flags & SPLICE_F_MORE)
+ msg.msg_flags |= MSG_MORE;
if (sd->len < sd->total_len &&
pipe_occupancy(pipe->head, pipe->tail) > 1)
- more |= MSG_SENDPAGE_NOTLAST;
+ msg.msg_flags |= MSG_MORE;
- return file->f_op->sendpage(file, buf->page, buf->offset,
- sd->len, &pos, more);
+ bvec_set_page(&bvec, buf->page, sd->len, buf->offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, sd->len);
+ return sock_sendmsg(sock, &msg);
}
+#endif
static void wakeup_pipe_writers(struct pipe_inode_info *pipe)
{
@@ -614,7 +618,7 @@ static void splice_from_pipe_end(struct pipe_inode_info *pipe, struct splice_des
* Description:
* This function does little more than loop over the pipe and call
* @actor to do the actual moving of a single struct pipe_buffer to
- * the desired destination. See pipe_to_file, pipe_to_sendpage, or
+ * the desired destination. See pipe_to_file, pipe_to_sendmsg, or
* pipe_to_user.
*
*/
@@ -795,8 +799,9 @@ iter_file_splice_write(struct pipe_inode_info *pipe, struct file *out,
EXPORT_SYMBOL(iter_file_splice_write);
+#ifdef CONFIG_NET
/**
- * generic_splice_sendpage - splice data from a pipe to a socket
+ * splice_to_socket - splice data from a pipe to a socket
* @pipe: pipe to splice from
* @out: socket to write to
* @ppos: position in @out
@@ -808,13 +813,12 @@ EXPORT_SYMBOL(iter_file_splice_write);
* is involved.
*
*/
-ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe, struct file *out,
- loff_t *ppos, size_t len, unsigned int flags)
+ssize_t splice_to_socket(struct pipe_inode_info *pipe, struct file *out,
+ loff_t *ppos, size_t len, unsigned int flags)
{
- return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendpage);
+ return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendmsg);
}
-
-EXPORT_SYMBOL(generic_splice_sendpage);
+#endif
static int warn_unsupported(struct file *file, const char *op)
{
diff --git a/include/linux/fs.h b/include/linux/fs.h
index c85916e9f7db..f3ccc243851e 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2740,8 +2740,6 @@ extern ssize_t generic_file_splice_read(struct file *, loff_t *,
struct pipe_inode_info *, size_t, unsigned int);
extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
struct file *, loff_t *, size_t, unsigned int);
-extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
- struct file *out, loff_t *, size_t len, unsigned int flags);
extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
loff_t *opos, size_t len, unsigned int flags);
diff --git a/include/linux/splice.h b/include/linux/splice.h
index 8f052c3dae95..e6153feda86c 100644
--- a/include/linux/splice.h
+++ b/include/linux/splice.h
@@ -87,6 +87,8 @@ extern long do_splice(struct file *in, loff_t *off_in,
extern long do_tee(struct file *in, struct file *out, size_t len,
unsigned int flags);
+extern ssize_t splice_to_socket(struct pipe_inode_info *pipe, struct file *out,
+ loff_t *ppos, size_t len, unsigned int flags);
/*
* for dynamic pipe sizing
diff --git a/net/socket.c b/net/socket.c
index dfb912bbed62..2cd5c2bcdde8 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -57,6 +57,7 @@
#include <linux/mm.h>
#include <linux/socket.h>
#include <linux/file.h>
+#include <linux/splice.h>
#include <linux/net.h>
#include <linux/interrupt.h>
#include <linux/thread_info.h>
@@ -126,8 +127,6 @@ static long compat_sock_ioctl(struct file *file,
unsigned int cmd, unsigned long arg);
#endif
static int sock_fasync(int fd, struct file *filp, int on);
-static ssize_t sock_sendpage(struct file *file, struct page *page,
- int offset, size_t size, loff_t *ppos, int more);
static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
struct pipe_inode_info *pipe, size_t len,
unsigned int flags);
@@ -162,8 +161,7 @@ static const struct file_operations socket_file_ops = {
.mmap = sock_mmap,
.release = sock_close,
.fasync = sock_fasync,
- .sendpage = sock_sendpage,
- .splice_write = generic_splice_sendpage,
+ .splice_write = splice_to_socket,
.splice_read = sock_splice_read,
.show_fdinfo = sock_show_fdinfo,
};
@@ -1062,26 +1060,6 @@ int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
}
EXPORT_SYMBOL(kernel_recvmsg);
-static ssize_t sock_sendpage(struct file *file, struct page *page,
- int offset, size_t size, loff_t *ppos, int more)
-{
- struct socket *sock;
- int flags;
- int ret;
-
- sock = file->private_data;
-
- flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
- /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
- flags |= more;
-
- ret = kernel_sendpage(sock, page, offset, size, flags);
-
- if (trace_sock_send_length_enabled())
- call_trace_sock_send_length(sock->sk, ret, 0);
- return ret;
-}
-
static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
struct pipe_inode_info *pipe, size_t len,
unsigned int flags)
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 28/48] splice: Reimplement splice_to_socket() to pass multiple bufs to sendmsg()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (26 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 27/48] splice, net: Use sendmsg(MSG_SPLICE_PAGES) rather than ->sendpage() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 29/48] Remove file->f_op->sendpage David Howells
` (19 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Reimplement splice_to_socket() so that it can pass multiple pipe buffer
pages to sendmsg() in a single go.
Signed-off-by: David Howells <dhowells@redhat.com>
---
fs/splice.c | 148 ++++++++++++++++++++++++++++++++++++++++++----------
1 file changed, 120 insertions(+), 28 deletions(-)
diff --git a/fs/splice.c b/fs/splice.c
index 23ead122d631..d5bc28b59720 100644
--- a/fs/splice.c
+++ b/fs/splice.c
@@ -411,33 +411,6 @@ const struct pipe_buf_operations nosteal_pipe_buf_ops = {
};
EXPORT_SYMBOL(nosteal_pipe_buf_ops);
-#ifdef CONFIG_NET
-/*
- * Send 'sd->len' bytes to socket from 'sd->file' at position 'sd->pos'
- * using sendpage(). Return the number of bytes sent.
- */
-static int pipe_to_sendmsg(struct pipe_inode_info *pipe,
- struct pipe_buffer *buf, struct splice_desc *sd)
-{
- struct socket *sock = sock_from_file(sd->u.file);
- struct bio_vec bvec;
- struct msghdr msg = {
- .msg_flags = MSG_SPLICE_PAGES,
- };
-
- if (sd->flags & SPLICE_F_MORE)
- msg.msg_flags |= MSG_MORE;
-
- if (sd->len < sd->total_len &&
- pipe_occupancy(pipe->head, pipe->tail) > 1)
- msg.msg_flags |= MSG_MORE;
-
- bvec_set_page(&bvec, buf->page, sd->len, buf->offset);
- iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, sd->len);
- return sock_sendmsg(sock, &msg);
-}
-#endif
-
static void wakeup_pipe_writers(struct pipe_inode_info *pipe)
{
smp_mb();
@@ -816,7 +789,126 @@ EXPORT_SYMBOL(iter_file_splice_write);
ssize_t splice_to_socket(struct pipe_inode_info *pipe, struct file *out,
loff_t *ppos, size_t len, unsigned int flags)
{
- return splice_from_pipe(pipe, out, ppos, len, flags, pipe_to_sendmsg);
+ struct socket *sock = sock_from_file(out);
+ struct bio_vec bvec[16];
+ struct msghdr msg = {};
+ ssize_t ret;
+ size_t spliced = 0;
+ bool need_wakeup = false;
+
+ pipe_lock(pipe);
+
+ while (len > 0) {
+ unsigned int head, tail, mask, bc = 0;
+ size_t remain = len;
+
+ /*
+ * Check for signal early to make process killable when there
+ * are always buffers available
+ */
+ ret = -ERESTARTSYS;
+ if (signal_pending(current))
+ break;
+
+ while (pipe_empty(pipe->head, pipe->tail)) {
+ ret = 0;
+ if (!pipe->writers)
+ goto out;
+
+ if (spliced)
+ goto out;
+
+ ret = -EAGAIN;
+ if (flags & SPLICE_F_NONBLOCK)
+ goto out;
+
+ ret = -ERESTARTSYS;
+ if (signal_pending(current))
+ goto out;
+
+ if (need_wakeup) {
+ wakeup_pipe_writers(pipe);
+ need_wakeup = false;
+ }
+
+ pipe_wait_readable(pipe);
+ }
+
+ head = pipe->head;
+ tail = pipe->tail;
+ mask = pipe->ring_size - 1;
+
+ while (!pipe_empty(head, tail)) {
+ struct pipe_buffer *buf = &pipe->bufs[tail & mask];
+ size_t seg;
+
+ if (!buf->len) {
+ tail++;
+ continue;
+ }
+
+ seg = min_t(size_t, remain, buf->len);
+ seg = min_t(size_t, seg, PAGE_SIZE);
+
+ ret = pipe_buf_confirm(pipe, buf);
+ if (unlikely(ret)) {
+ if (ret == -ENODATA)
+ ret = 0;
+ break;
+ }
+
+ bvec_set_page(&bvec[bc++], buf->page, seg, buf->offset);
+ remain -= seg;
+ if (seg >= buf->len)
+ tail++;
+ if (bc >= ARRAY_SIZE(bvec))
+ break;
+ }
+
+ if (!bc)
+ break;
+
+ msg.msg_flags = 0;
+ if (flags & SPLICE_F_MORE)
+ msg.msg_flags = MSG_MORE;
+ if (remain && pipe_occupancy(pipe->head, tail) > 0)
+ msg.msg_flags = MSG_MORE;
+ msg.msg_flags |= MSG_SPLICE_PAGES;
+
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, bvec, bc, len - remain);
+ ret = sock_sendmsg(sock, &msg);
+ if (ret <= 0)
+ break;
+
+ spliced += ret;
+ len -= ret;
+ tail = pipe->tail;
+ while (ret > 0) {
+ struct pipe_buffer *buf = &pipe->bufs[tail & mask];
+ size_t seg = min_t(size_t, ret, buf->len);
+
+ buf->offset += seg;
+ buf->len -= seg;
+ ret -= seg;
+
+ if (!buf->len) {
+ pipe_buf_release(pipe, buf);
+ tail++;
+ }
+ }
+
+ if (tail != pipe->tail) {
+ pipe->tail = tail;
+ if (pipe->files)
+ need_wakeup = true;
+ }
+ }
+
+out:
+ pipe_unlock(pipe);
+ if (need_wakeup)
+ wakeup_pipe_writers(pipe);
+ return spliced ?: ret;
}
#endif
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 29/48] Remove file->f_op->sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (27 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 28/48] splice: Reimplement splice_to_socket() to pass multiple bufs to sendmsg() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage to transmit David Howells
` (18 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Remove file->f_op->sendpage as splicing to a socket now calls sendmsg
rather than sendpage.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
include/linux/fs.h | 1 -
1 file changed, 1 deletion(-)
diff --git a/include/linux/fs.h b/include/linux/fs.h
index f3ccc243851e..a9f1b2543d2c 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1773,7 +1773,6 @@ struct file_operations {
int (*fsync) (struct file *, loff_t, loff_t, int datasync);
int (*fasync) (int, struct file *, int);
int (*lock) (struct file *, int, struct file_lock *);
- ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
int (*check_flags)(int);
int (*flock) (struct file *, int, struct file_lock *);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage to transmit
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (28 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 29/48] Remove file->f_op->sendpage David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 15:18 ` Bernard Metzler
2023-03-29 15:32 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 31/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage David Howells
` (17 subsequent siblings)
47 siblings, 2 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Bernard Metzler,
Tom Talpey, linux-rdma
When transmitting data, call down into TCP using a single sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather than
performing several sendmsg and sendpage calls to transmit header, data
pages and trailer.
To make this work, the data is assembled in a bio_vec array and attached to
a BVEC-type iterator. The header and trailer (if present) are copied into
page fragments that can be freed with put_page().
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Bernard Metzler <bmt@zurich.ibm.com>
cc: Tom Talpey <tom@talpey.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-rdma@vger.kernel.org
cc: netdev@vger.kernel.org
---
drivers/infiniband/sw/siw/siw_qp_tx.c | 234 ++++++--------------------
1 file changed, 48 insertions(+), 186 deletions(-)
diff --git a/drivers/infiniband/sw/siw/siw_qp_tx.c b/drivers/infiniband/sw/siw/siw_qp_tx.c
index fa5de40d85d5..fbe80c06d0ca 100644
--- a/drivers/infiniband/sw/siw/siw_qp_tx.c
+++ b/drivers/infiniband/sw/siw/siw_qp_tx.c
@@ -312,114 +312,8 @@ static int siw_tx_ctrl(struct siw_iwarp_tx *c_tx, struct socket *s,
return rv;
}
-/*
- * 0copy TCP transmit interface: Use MSG_SPLICE_PAGES.
- *
- * Using sendpage to push page by page appears to be less efficient
- * than using sendmsg, even if data are copied.
- *
- * A general performance limitation might be the extra four bytes
- * trailer checksum segment to be pushed after user data.
- */
-static int siw_tcp_sendpages(struct socket *s, struct page **page, int offset,
- size_t size)
-{
- struct bio_vec bvec;
- struct msghdr msg = {
- .msg_flags = (MSG_MORE | MSG_DONTWAIT | MSG_SENDPAGE_NOTLAST |
- MSG_SPLICE_PAGES),
- };
- struct sock *sk = s->sk;
- int i = 0, rv = 0, sent = 0;
-
- while (size) {
- size_t bytes = min_t(size_t, PAGE_SIZE - offset, size);
-
- if (size + offset <= PAGE_SIZE)
- msg.msg_flags = MSG_MORE | MSG_DONTWAIT;
-
- tcp_rate_check_app_limited(sk);
- bvec_set_page(&bvec, page[i], bytes, offset);
- iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
-
-try_page_again:
- lock_sock(sk);
- rv = tcp_sendmsg_locked(sk, &msg, size);
- release_sock(sk);
-
- if (rv > 0) {
- size -= rv;
- sent += rv;
- if (rv != bytes) {
- offset += rv;
- bytes -= rv;
- goto try_page_again;
- }
- offset = 0;
- } else {
- if (rv == -EAGAIN || rv == 0)
- break;
- return rv;
- }
- i++;
- }
- return sent;
-}
-
-/*
- * siw_0copy_tx()
- *
- * Pushes list of pages to TCP socket. If pages from multiple
- * SGE's, all referenced pages of each SGE are pushed in one
- * shot.
- */
-static int siw_0copy_tx(struct socket *s, struct page **page,
- struct siw_sge *sge, unsigned int offset,
- unsigned int size)
-{
- int i = 0, sent = 0, rv;
- int sge_bytes = min(sge->length - offset, size);
-
- offset = (sge->laddr + offset) & ~PAGE_MASK;
-
- while (sent != size) {
- rv = siw_tcp_sendpages(s, &page[i], offset, sge_bytes);
- if (rv >= 0) {
- sent += rv;
- if (size == sent || sge_bytes > rv)
- break;
-
- i += PAGE_ALIGN(sge_bytes + offset) >> PAGE_SHIFT;
- sge++;
- sge_bytes = min(sge->length, size - sent);
- offset = sge->laddr & ~PAGE_MASK;
- } else {
- sent = rv;
- break;
- }
- }
- return sent;
-}
-
#define MAX_TRAILER (MPA_CRC_SIZE + 4)
-static void siw_unmap_pages(struct kvec *iov, unsigned long kmap_mask, int len)
-{
- int i;
-
- /*
- * Work backwards through the array to honor the kmap_local_page()
- * ordering requirements.
- */
- for (i = (len-1); i >= 0; i--) {
- if (kmap_mask & BIT(i)) {
- unsigned long addr = (unsigned long)iov[i].iov_base;
-
- kunmap_local((void *)(addr & PAGE_MASK));
- }
- }
-}
-
/*
* siw_tx_hdt() tries to push a complete packet to TCP where all
* packet fragments are referenced by the elements of one iovec.
@@ -439,15 +333,14 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
{
struct siw_wqe *wqe = &c_tx->wqe_active;
struct siw_sge *sge = &wqe->sqe.sge[c_tx->sge_idx];
- struct kvec iov[MAX_ARRAY];
- struct page *page_array[MAX_ARRAY];
+ struct bio_vec bvec[MAX_ARRAY];
struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_EOR };
+ void *trl, *t;
int seg = 0, do_crc = c_tx->do_crc, is_kva = 0, rv;
unsigned int data_len = c_tx->bytes_unsent, hdr_len = 0, trl_len = 0,
sge_off = c_tx->sge_off, sge_idx = c_tx->sge_idx,
pbl_idx = c_tx->pbl_idx;
- unsigned long kmap_mask = 0L;
if (c_tx->state == SIW_SEND_HDR) {
if (c_tx->use_sendpage) {
@@ -457,10 +350,15 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
c_tx->state = SIW_SEND_DATA;
} else {
- iov[0].iov_base =
- (char *)&c_tx->pkt.ctrl + c_tx->ctrl_sent;
- iov[0].iov_len = hdr_len =
- c_tx->ctrl_len - c_tx->ctrl_sent;
+ const void *hdr = &c_tx->pkt.ctrl + c_tx->ctrl_sent;
+ void *h;
+
+ rv = -ENOMEM;
+ hdr_len = c_tx->ctrl_len - c_tx->ctrl_sent;
+ h = page_frag_memdup(NULL, hdr, hdr_len, GFP_NOFS, ULONG_MAX);
+ if (!h)
+ goto done;
+ bvec_set_virt(&bvec[0], h, hdr_len);
seg = 1;
}
}
@@ -478,28 +376,9 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
} else {
is_kva = 1;
}
- if (is_kva && !c_tx->use_sendpage) {
- /*
- * tx from kernel virtual address: either inline data
- * or memory region with assigned kernel buffer
- */
- iov[seg].iov_base =
- (void *)(uintptr_t)(sge->laddr + sge_off);
- iov[seg].iov_len = sge_len;
-
- if (do_crc)
- crypto_shash_update(c_tx->mpa_crc_hd,
- iov[seg].iov_base,
- sge_len);
- sge_off += sge_len;
- data_len -= sge_len;
- seg++;
- goto sge_done;
- }
while (sge_len) {
size_t plen = min((int)PAGE_SIZE - fp_off, sge_len);
- void *kaddr;
if (!is_kva) {
struct page *p;
@@ -512,33 +391,12 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
p = siw_get_upage(mem->umem,
sge->laddr + sge_off);
if (unlikely(!p)) {
- siw_unmap_pages(iov, kmap_mask, seg);
wqe->processed -= c_tx->bytes_unsent;
rv = -EFAULT;
goto done_crc;
}
- page_array[seg] = p;
-
- if (!c_tx->use_sendpage) {
- void *kaddr = kmap_local_page(p);
-
- /* Remember for later kunmap() */
- kmap_mask |= BIT(seg);
- iov[seg].iov_base = kaddr + fp_off;
- iov[seg].iov_len = plen;
-
- if (do_crc)
- crypto_shash_update(
- c_tx->mpa_crc_hd,
- iov[seg].iov_base,
- plen);
- } else if (do_crc) {
- kaddr = kmap_local_page(p);
- crypto_shash_update(c_tx->mpa_crc_hd,
- kaddr + fp_off,
- plen);
- kunmap_local(kaddr);
- }
+
+ bvec_set_page(&bvec[seg], p, plen, fp_off);
} else {
/*
* Cast to an uintptr_t to preserve all 64 bits
@@ -552,12 +410,15 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
* bits on a 64 bit platform and 32 bits on a
* 32 bit platform.
*/
- page_array[seg] = virt_to_page((void *)(va & PAGE_MASK));
- if (do_crc)
- crypto_shash_update(
- c_tx->mpa_crc_hd,
- (void *)va,
- plen);
+ bvec_set_virt(&bvec[seg], (void *)va, plen);
+ }
+
+ if (do_crc) {
+ void *kaddr = kmap_local_page(bvec[seg].bv_page);
+ crypto_shash_update(c_tx->mpa_crc_hd,
+ kaddr + bvec[seg].bv_offset,
+ bvec[seg].bv_len);
+ kunmap_local(kaddr);
}
sge_len -= plen;
@@ -567,13 +428,12 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
if (++seg > (int)MAX_ARRAY) {
siw_dbg_qp(tx_qp(c_tx), "to many fragments\n");
- siw_unmap_pages(iov, kmap_mask, seg-1);
wqe->processed -= c_tx->bytes_unsent;
rv = -EMSGSIZE;
goto done_crc;
}
}
-sge_done:
+
/* Update SGE variables at end of SGE */
if (sge_off == sge->length &&
(data_len != 0 || wqe->processed < wqe->bytes)) {
@@ -582,15 +442,8 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
sge_off = 0;
}
}
- /* trailer */
- if (likely(c_tx->state != SIW_SEND_TRAILER)) {
- iov[seg].iov_base = &c_tx->trailer.pad[4 - c_tx->pad];
- iov[seg].iov_len = trl_len = MAX_TRAILER - (4 - c_tx->pad);
- } else {
- iov[seg].iov_base = &c_tx->trailer.pad[c_tx->ctrl_sent];
- iov[seg].iov_len = trl_len = MAX_TRAILER - c_tx->ctrl_sent;
- }
+ /* Set the CRC in the trailer */
if (c_tx->pad) {
*(u32 *)c_tx->trailer.pad = 0;
if (do_crc)
@@ -603,23 +456,29 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
else if (do_crc)
crypto_shash_final(c_tx->mpa_crc_hd, (u8 *)&c_tx->trailer.crc);
- data_len = c_tx->bytes_unsent;
-
- if (c_tx->use_sendpage) {
- rv = siw_0copy_tx(s, page_array, &wqe->sqe.sge[c_tx->sge_idx],
- c_tx->sge_off, data_len);
- if (rv == data_len) {
- rv = kernel_sendmsg(s, &msg, &iov[seg], 1, trl_len);
- if (rv > 0)
- rv += data_len;
- else
- rv = data_len;
- }
+ /* Copy the trailer and add it to the output list */
+ if (likely(c_tx->state != SIW_SEND_TRAILER)) {
+ trl = &c_tx->trailer.pad[4 - c_tx->pad];
+ trl_len = MAX_TRAILER - (4 - c_tx->pad);
} else {
- rv = kernel_sendmsg(s, &msg, iov, seg + 1,
- hdr_len + data_len + trl_len);
- siw_unmap_pages(iov, kmap_mask, seg);
+ trl = &c_tx->trailer.pad[c_tx->ctrl_sent];
+ trl_len = MAX_TRAILER - c_tx->ctrl_sent;
}
+
+ rv = -ENOMEM;
+ t = page_frag_memdup(NULL, trl, trl_len, GFP_NOFS, ULONG_MAX);
+ if (!t)
+ goto done_crc;
+ bvec_set_virt(&bvec[seg], t, trl_len);
+
+ data_len = c_tx->bytes_unsent;
+
+ if (c_tx->use_sendpage)
+ msg.msg_flags |= MSG_SPLICE_PAGES;
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, bvec, seg + 1,
+ hdr_len + data_len + trl_len);
+ rv = sock_sendmsg(s, &msg);
+
if (rv < (int)hdr_len) {
/* Not even complete hdr pushed or negative rv */
wqe->processed -= data_len;
@@ -680,6 +539,9 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct socket *s)
}
done_crc:
c_tx->do_crc = 0;
+ if (c_tx->state == SIW_SEND_HDR)
+ folio_put(page_folio(bvec[0].bv_page));
+ folio_put(page_folio(bvec[seg].bv_page));
done:
return rv;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* RE: [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage to transmit
2023-03-29 14:13 ` [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage to transmit David Howells
@ 2023-03-29 15:18 ` Bernard Metzler
2023-03-29 15:32 ` David Howells
1 sibling, 0 replies; 77+ messages in thread
From: Bernard Metzler @ 2023-03-29 15:18 UTC (permalink / raw)
To: David Howells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev@vger.kernel.org, linux-fsdevel@vger.kernel.org,
linux-kernel@vger.kernel.org, linux-mm@kvack.org, Tom Talpey,
linux-rdma@vger.kernel.org
> -----Original Message-----
> From: David Howells <dhowells@redhat.com>
> Sent: Wednesday, 29 March 2023 16:14
> To: Matthew Wilcox <willy@infradead.org>; David S. Miller
> <davem@davemloft.net>; Eric Dumazet <edumazet@google.com>; Jakub Kicinski
> <kuba@kernel.org>; Paolo Abeni <pabeni@redhat.com>
> Cc: David Howells <dhowells@redhat.com>; Al Viro <viro@zeniv.linux.org.uk>;
> Christoph Hellwig <hch@infradead.org>; Jens Axboe <axboe@kernel.dk>; Jeff
> Layton <jlayton@kernel.org>; Christian Brauner <brauner@kernel.org>; Chuck
> Lever III <chuck.lever@oracle.com>; Linus Torvalds <torvalds@linux-
> foundation.org>; netdev@vger.kernel.org; linux-fsdevel@vger.kernel.org;
> linux-kernel@vger.kernel.org; linux-mm@kvack.org; Bernard Metzler
> <BMT@zurich.ibm.com>; Tom Talpey <tom@talpey.com>; linux-
> rdma@vger.kernel.org
> Subject: [EXTERNAL] [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES)
> rather than sendpage to transmit
>
> When transmitting data, call down into TCP using a single sendmsg with
> MSG_SPLICE_PAGES to indicate that content should be spliced rather than
> performing several sendmsg and sendpage calls to transmit header, data
> pages and trailer.
>
> To make this work, the data is assembled in a bio_vec array and attached to
> a BVEC-type iterator. The header and trailer (if present) are copied into
> page fragments that can be freed with put_page().
I like it a lot if it still keeps zero copy sendpage() semantics for
the cases the driver can make use of data transfers w/o copy.
Is 'msg.msg_flags |= MSG_SPLICE_PAGES' doing that magic? 'splicing'
suggest just merging pages to me.
It would simplify the transmit code path substantially, also getting
rid of kmap_local_page()/kunmap_local() sequences for multi-fragment
sendmsg()'s.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Bernard Metzler <bmt@zurich.ibm.com>
> cc: Tom Talpey <tom@talpey.com>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: linux-rdma@vger.kernel.org
> cc: netdev@vger.kernel.org
> ---
> drivers/infiniband/sw/siw/siw_qp_tx.c | 234 ++++++--------------------
> 1 file changed, 48 insertions(+), 186 deletions(-)
>
> diff --git a/drivers/infiniband/sw/siw/siw_qp_tx.c
> b/drivers/infiniband/sw/siw/siw_qp_tx.c
> index fa5de40d85d5..fbe80c06d0ca 100644
> --- a/drivers/infiniband/sw/siw/siw_qp_tx.c
> +++ b/drivers/infiniband/sw/siw/siw_qp_tx.c
> @@ -312,114 +312,8 @@ static int siw_tx_ctrl(struct siw_iwarp_tx *c_tx,
> struct socket *s,
> return rv;
> }
>
> -/*
> - * 0copy TCP transmit interface: Use MSG_SPLICE_PAGES.
> - *
> - * Using sendpage to push page by page appears to be less efficient
> - * than using sendmsg, even if data are copied.
> - *
> - * A general performance limitation might be the extra four bytes
> - * trailer checksum segment to be pushed after user data.
> - */
> -static int siw_tcp_sendpages(struct socket *s, struct page **page, int
> offset,
> - size_t size)
> -{
> - struct bio_vec bvec;
> - struct msghdr msg = {
> - .msg_flags = (MSG_MORE | MSG_DONTWAIT | MSG_SENDPAGE_NOTLAST |
> - MSG_SPLICE_PAGES),
> - };
> - struct sock *sk = s->sk;
> - int i = 0, rv = 0, sent = 0;
> -
> - while (size) {
> - size_t bytes = min_t(size_t, PAGE_SIZE - offset, size);
> -
> - if (size + offset <= PAGE_SIZE)
> - msg.msg_flags = MSG_MORE | MSG_DONTWAIT;
> -
> - tcp_rate_check_app_limited(sk);
> - bvec_set_page(&bvec, page[i], bytes, offset);
> - iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
> -
> -try_page_again:
> - lock_sock(sk);
> - rv = tcp_sendmsg_locked(sk, &msg, size);
> - release_sock(sk);
> -
> - if (rv > 0) {
> - size -= rv;
> - sent += rv;
> - if (rv != bytes) {
> - offset += rv;
> - bytes -= rv;
> - goto try_page_again;
> - }
> - offset = 0;
> - } else {
> - if (rv == -EAGAIN || rv == 0)
> - break;
> - return rv;
> - }
> - i++;
> - }
> - return sent;
> -}
> -
> -/*
> - * siw_0copy_tx()
> - *
> - * Pushes list of pages to TCP socket. If pages from multiple
> - * SGE's, all referenced pages of each SGE are pushed in one
> - * shot.
> - */
> -static int siw_0copy_tx(struct socket *s, struct page **page,
> - struct siw_sge *sge, unsigned int offset,
> - unsigned int size)
> -{
> - int i = 0, sent = 0, rv;
> - int sge_bytes = min(sge->length - offset, size);
> -
> - offset = (sge->laddr + offset) & ~PAGE_MASK;
> -
> - while (sent != size) {
> - rv = siw_tcp_sendpages(s, &page[i], offset, sge_bytes);
> - if (rv >= 0) {
> - sent += rv;
> - if (size == sent || sge_bytes > rv)
> - break;
> -
> - i += PAGE_ALIGN(sge_bytes + offset) >> PAGE_SHIFT;
> - sge++;
> - sge_bytes = min(sge->length, size - sent);
> - offset = sge->laddr & ~PAGE_MASK;
> - } else {
> - sent = rv;
> - break;
> - }
> - }
> - return sent;
> -}
> -
> #define MAX_TRAILER (MPA_CRC_SIZE + 4)
>
> -static void siw_unmap_pages(struct kvec *iov, unsigned long kmap_mask, int
> len)
> -{
> - int i;
> -
> - /*
> - * Work backwards through the array to honor the kmap_local_page()
> - * ordering requirements.
> - */
> - for (i = (len-1); i >= 0; i--) {
> - if (kmap_mask & BIT(i)) {
> - unsigned long addr = (unsigned long)iov[i].iov_base;
> -
> - kunmap_local((void *)(addr & PAGE_MASK));
> - }
> - }
> -}
> -
> /*
> * siw_tx_hdt() tries to push a complete packet to TCP where all
> * packet fragments are referenced by the elements of one iovec.
> @@ -439,15 +333,14 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
> {
> struct siw_wqe *wqe = &c_tx->wqe_active;
> struct siw_sge *sge = &wqe->sqe.sge[c_tx->sge_idx];
> - struct kvec iov[MAX_ARRAY];
> - struct page *page_array[MAX_ARRAY];
> + struct bio_vec bvec[MAX_ARRAY];
> struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_EOR };
> + void *trl, *t;
>
> int seg = 0, do_crc = c_tx->do_crc, is_kva = 0, rv;
> unsigned int data_len = c_tx->bytes_unsent, hdr_len = 0, trl_len = 0,
> sge_off = c_tx->sge_off, sge_idx = c_tx->sge_idx,
> pbl_idx = c_tx->pbl_idx;
> - unsigned long kmap_mask = 0L;
>
> if (c_tx->state == SIW_SEND_HDR) {
> if (c_tx->use_sendpage) {
> @@ -457,10 +350,15 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
>
> c_tx->state = SIW_SEND_DATA;
> } else {
> - iov[0].iov_base =
> - (char *)&c_tx->pkt.ctrl + c_tx->ctrl_sent;
> - iov[0].iov_len = hdr_len =
> - c_tx->ctrl_len - c_tx->ctrl_sent;
> + const void *hdr = &c_tx->pkt.ctrl + c_tx->ctrl_sent;
> + void *h;
> +
> + rv = -ENOMEM;
> + hdr_len = c_tx->ctrl_len - c_tx->ctrl_sent;
> + h = page_frag_memdup(NULL, hdr, hdr_len, GFP_NOFS,
> ULONG_MAX);
> + if (!h)
> + goto done;
> + bvec_set_virt(&bvec[0], h, hdr_len);
> seg = 1;
> }
> }
> @@ -478,28 +376,9 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
> } else {
> is_kva = 1;
> }
> - if (is_kva && !c_tx->use_sendpage) {
> - /*
> - * tx from kernel virtual address: either inline data
> - * or memory region with assigned kernel buffer
> - */
> - iov[seg].iov_base =
> - (void *)(uintptr_t)(sge->laddr + sge_off);
> - iov[seg].iov_len = sge_len;
> -
> - if (do_crc)
> - crypto_shash_update(c_tx->mpa_crc_hd,
> - iov[seg].iov_base,
> - sge_len);
> - sge_off += sge_len;
> - data_len -= sge_len;
> - seg++;
> - goto sge_done;
> - }
>
> while (sge_len) {
> size_t plen = min((int)PAGE_SIZE - fp_off, sge_len);
> - void *kaddr;
>
> if (!is_kva) {
> struct page *p;
> @@ -512,33 +391,12 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
> p = siw_get_upage(mem->umem,
> sge->laddr + sge_off);
> if (unlikely(!p)) {
> - siw_unmap_pages(iov, kmap_mask, seg);
> wqe->processed -= c_tx->bytes_unsent;
> rv = -EFAULT;
> goto done_crc;
> }
> - page_array[seg] = p;
> -
> - if (!c_tx->use_sendpage) {
> - void *kaddr = kmap_local_page(p);
> -
> - /* Remember for later kunmap() */
> - kmap_mask |= BIT(seg);
> - iov[seg].iov_base = kaddr + fp_off;
> - iov[seg].iov_len = plen;
> -
> - if (do_crc)
> - crypto_shash_update(
> - c_tx->mpa_crc_hd,
> - iov[seg].iov_base,
> - plen);
> - } else if (do_crc) {
> - kaddr = kmap_local_page(p);
> - crypto_shash_update(c_tx->mpa_crc_hd,
> - kaddr + fp_off,
> - plen);
> - kunmap_local(kaddr);
> - }
> +
> + bvec_set_page(&bvec[seg], p, plen, fp_off);
> } else {
> /*
> * Cast to an uintptr_t to preserve all 64 bits
> @@ -552,12 +410,15 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
> * bits on a 64 bit platform and 32 bits on a
> * 32 bit platform.
> */
> - page_array[seg] = virt_to_page((void *)(va &
> PAGE_MASK));
> - if (do_crc)
> - crypto_shash_update(
> - c_tx->mpa_crc_hd,
> - (void *)va,
> - plen);
> + bvec_set_virt(&bvec[seg], (void *)va, plen);
> + }
> +
> + if (do_crc) {
> + void *kaddr = kmap_local_page(bvec[seg].bv_page);
> + crypto_shash_update(c_tx->mpa_crc_hd,
> + kaddr + bvec[seg].bv_offset,
> + bvec[seg].bv_len);
> + kunmap_local(kaddr);
> }
>
> sge_len -= plen;
> @@ -567,13 +428,12 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
>
> if (++seg > (int)MAX_ARRAY) {
> siw_dbg_qp(tx_qp(c_tx), "to many fragments\n");
> - siw_unmap_pages(iov, kmap_mask, seg-1);
> wqe->processed -= c_tx->bytes_unsent;
> rv = -EMSGSIZE;
> goto done_crc;
> }
> }
> -sge_done:
> +
> /* Update SGE variables at end of SGE */
> if (sge_off == sge->length &&
> (data_len != 0 || wqe->processed < wqe->bytes)) {
> @@ -582,15 +442,8 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
> sge_off = 0;
> }
> }
> - /* trailer */
> - if (likely(c_tx->state != SIW_SEND_TRAILER)) {
> - iov[seg].iov_base = &c_tx->trailer.pad[4 - c_tx->pad];
> - iov[seg].iov_len = trl_len = MAX_TRAILER - (4 - c_tx->pad);
> - } else {
> - iov[seg].iov_base = &c_tx->trailer.pad[c_tx->ctrl_sent];
> - iov[seg].iov_len = trl_len = MAX_TRAILER - c_tx->ctrl_sent;
> - }
>
> + /* Set the CRC in the trailer */
> if (c_tx->pad) {
> *(u32 *)c_tx->trailer.pad = 0;
> if (do_crc)
> @@ -603,23 +456,29 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx,
> struct socket *s)
> else if (do_crc)
> crypto_shash_final(c_tx->mpa_crc_hd, (u8 *)&c_tx->trailer.crc);
>
> - data_len = c_tx->bytes_unsent;
> -
> - if (c_tx->use_sendpage) {
> - rv = siw_0copy_tx(s, page_array, &wqe->sqe.sge[c_tx->sge_idx],
> - c_tx->sge_off, data_len);
> - if (rv == data_len) {
> - rv = kernel_sendmsg(s, &msg, &iov[seg], 1, trl_len);
> - if (rv > 0)
> - rv += data_len;
> - else
> - rv = data_len;
> - }
> + /* Copy the trailer and add it to the output list */
> + if (likely(c_tx->state != SIW_SEND_TRAILER)) {
> + trl = &c_tx->trailer.pad[4 - c_tx->pad];
> + trl_len = MAX_TRAILER - (4 - c_tx->pad);
> } else {
> - rv = kernel_sendmsg(s, &msg, iov, seg + 1,
> - hdr_len + data_len + trl_len);
> - siw_unmap_pages(iov, kmap_mask, seg);
> + trl = &c_tx->trailer.pad[c_tx->ctrl_sent];
> + trl_len = MAX_TRAILER - c_tx->ctrl_sent;
> }
> +
> + rv = -ENOMEM;
> + t = page_frag_memdup(NULL, trl, trl_len, GFP_NOFS, ULONG_MAX);
> + if (!t)
> + goto done_crc;
> + bvec_set_virt(&bvec[seg], t, trl_len);
> +
> + data_len = c_tx->bytes_unsent;
> +
> + if (c_tx->use_sendpage)
> + msg.msg_flags |= MSG_SPLICE_PAGES;
> + iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, bvec, seg + 1,
> + hdr_len + data_len + trl_len);
> + rv = sock_sendmsg(s, &msg);
> +
> if (rv < (int)hdr_len) {
> /* Not even complete hdr pushed or negative rv */
> wqe->processed -= data_len;
> @@ -680,6 +539,9 @@ static int siw_tx_hdt(struct siw_iwarp_tx *c_tx, struct
> socket *s)
> }
> done_crc:
> c_tx->do_crc = 0;
> + if (c_tx->state == SIW_SEND_HDR)
> + folio_put(page_folio(bvec[0].bv_page));
> + folio_put(page_folio(bvec[seg].bv_page));
> done:
> return rv;
> }
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage to transmit
2023-03-29 14:13 ` [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage to transmit David Howells
2023-03-29 15:18 ` Bernard Metzler
@ 2023-03-29 15:32 ` David Howells
1 sibling, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 15:32 UTC (permalink / raw)
To: Bernard Metzler
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev@vger.kernel.org,
linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
linux-mm@kvack.org, Tom Talpey, linux-rdma@vger.kernel.org
Bernard Metzler <BMT@zurich.ibm.com> wrote:
> > When transmitting data, call down into TCP using a single sendmsg with
> > MSG_SPLICE_PAGES to indicate that content should be spliced rather than
> > performing several sendmsg and sendpage calls to transmit header, data
> > pages and trailer.
> >
> > To make this work, the data is assembled in a bio_vec array and attached to
> > a BVEC-type iterator. The header and trailer (if present) are copied into
> > page fragments that can be freed with put_page().
>
> I like it a lot if it still keeps zero copy sendpage() semantics for
> the cases the driver can make use of data transfers w/o copy.
> Is 'msg.msg_flags |= MSG_SPLICE_PAGES' doing that magic?
Yes. MSG_SPLICE_PAGES indicates that you want the socket to retain your
buffer and pass it directly to the device. Note that it's just a hint,
however, pages that are unspliceable (eg. they belong to the slab) will get
copied into a page fragment instead. Further, if the device cannot support a
vector, then the hint can be ignored and all the data can be copied as normal.
> 'splicing' suggest just merging pages to me.
'splicing' as in what the splice system call does.
Unfortunately, MSG_ZEROCOPY is already a (different) thing.
> It would simplify the transmit code path substantially, also getting
> rid of kmap_local_page()/kunmap_local() sequences for multi-fragment
> sendmsg()'s.
If the ITER_ITERLIST iterator is accepted, then siw would be able to do mix
KVEC and BVEC iterators, e.g. what I did for sunrpc here:
https://lore.kernel.org/linux-fsdevel/20230329141354.516864-42-dhowells@redhat.com/T/#u
This means that in siw_tx_hdt() where I made it copy data into page fragments
using page_frag_memdup() and attach that to a bvec:
hdr_len = c_tx->ctrl_len - c_tx->ctrl_sent;
h = page_frag_memdup(NULL, hdr, hdr_len, GFP_NOFS, ULONG_MAX);
if (!h)
goto done;
bvec_set_virt(&bvec[0], h, hdr_len);
seg = 1;
it can just set up a kvec instead.
Unfortunately, it's not so easy to get rid of all of the kmap'ing as we need
to do some of it to do the hashing.
David
^ permalink raw reply [flat|nested] 77+ messages in thread
* [RFC PATCH v2 31/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (29 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 30/48] siw: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage to transmit David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 32/48] iscsi: " David Howells
` (16 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Ilya Dryomov,
Xiubo Li, ceph-devel
Use sendmsg() and MSG_SPLICE_PAGES rather than sendpage in ceph when
transmitting data. For the moment, this can only transmit one page at a
time because of the architecture of net/ceph/, but if
write_partial_message_data() can be given a bvec[] at a time by the
iteration code, this would allow pages to be sent in a batch.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Ilya Dryomov <idryomov@gmail.com>
cc: Xiubo Li <xiubli@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: ceph-devel@vger.kernel.org
cc: netdev@vger.kernel.org
---
net/ceph/messenger_v1.c | 58 ++++++++++++++---------------------------
1 file changed, 19 insertions(+), 39 deletions(-)
diff --git a/net/ceph/messenger_v1.c b/net/ceph/messenger_v1.c
index d664cb1593a7..b2d801a49122 100644
--- a/net/ceph/messenger_v1.c
+++ b/net/ceph/messenger_v1.c
@@ -74,37 +74,6 @@ static int ceph_tcp_sendmsg(struct socket *sock, struct kvec *iov,
return r;
}
-/*
- * @more: either or both of MSG_MORE and MSG_SENDPAGE_NOTLAST
- */
-static int ceph_tcp_sendpage(struct socket *sock, struct page *page,
- int offset, size_t size, int more)
-{
- ssize_t (*sendpage)(struct socket *sock, struct page *page,
- int offset, size_t size, int flags);
- int flags = MSG_DONTWAIT | MSG_NOSIGNAL | more;
- int ret;
-
- /*
- * sendpage cannot properly handle pages with page_count == 0,
- * we need to fall back to sendmsg if that's the case.
- *
- * Same goes for slab pages: skb_can_coalesce() allows
- * coalescing neighboring slab objects into a single frag which
- * triggers one of hardened usercopy checks.
- */
- if (sendpage_ok(page))
- sendpage = sock->ops->sendpage;
- else
- sendpage = sock_no_sendpage;
-
- ret = sendpage(sock, page, offset, size, flags);
- if (ret == -EAGAIN)
- ret = 0;
-
- return ret;
-}
-
static void con_out_kvec_reset(struct ceph_connection *con)
{
BUG_ON(con->v1.out_skip);
@@ -464,7 +433,6 @@ static int write_partial_message_data(struct ceph_connection *con)
struct ceph_msg *msg = con->out_msg;
struct ceph_msg_data_cursor *cursor = &msg->cursor;
bool do_datacrc = !ceph_test_opt(from_msgr(con->msgr), NOCRC);
- int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
u32 crc;
dout("%s %p msg %p\n", __func__, con, msg);
@@ -482,6 +450,10 @@ static int write_partial_message_data(struct ceph_connection *con)
*/
crc = do_datacrc ? le32_to_cpu(msg->footer.data_crc) : 0;
while (cursor->total_resid) {
+ struct bio_vec bvec;
+ struct msghdr msghdr = {
+ .msg_flags = MSG_SPLICE_PAGES | MSG_SENDPAGE_NOTLAST,
+ };
struct page *page;
size_t page_offset;
size_t length;
@@ -494,9 +466,12 @@ static int write_partial_message_data(struct ceph_connection *con)
page = ceph_msg_data_next(cursor, &page_offset, &length);
if (length == cursor->total_resid)
- more = MSG_MORE;
- ret = ceph_tcp_sendpage(con->sock, page, page_offset, length,
- more);
+ msghdr.msg_flags |= MSG_MORE;
+
+ bvec_set_page(&bvec, page, length, page_offset);
+ iov_iter_bvec(&msghdr.msg_iter, ITER_SOURCE, &bvec, 1, length);
+
+ ret = sock_sendmsg(con->sock, &msghdr);
if (ret <= 0) {
if (do_datacrc)
msg->footer.data_crc = cpu_to_le32(crc);
@@ -526,7 +501,10 @@ static int write_partial_message_data(struct ceph_connection *con)
*/
static int write_partial_skip(struct ceph_connection *con)
{
- int more = MSG_MORE | MSG_SENDPAGE_NOTLAST;
+ struct bio_vec bvec;
+ struct msghdr msghdr = {
+ .msg_flags = MSG_SPLICE_PAGES | MSG_SENDPAGE_NOTLAST | MSG_MORE,
+ };
int ret;
dout("%s %p %d left\n", __func__, con, con->v1.out_skip);
@@ -534,9 +512,11 @@ static int write_partial_skip(struct ceph_connection *con)
size_t size = min(con->v1.out_skip, (int)PAGE_SIZE);
if (size == con->v1.out_skip)
- more = MSG_MORE;
- ret = ceph_tcp_sendpage(con->sock, ceph_zero_page, 0, size,
- more);
+ msghdr.msg_flags &= ~MSG_SENDPAGE_NOTLAST;
+ bvec_set_page(&bvec, ZERO_PAGE(0), size, 0);
+ iov_iter_bvec(&msghdr.msg_iter, ITER_SOURCE, &bvec, 1, size);
+
+ ret = sock_sendmsg(con->sock, &msghdr);
if (ret <= 0)
goto out;
con->v1.out_skip -= ret;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 32/48] iscsi: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (30 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 31/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 33/48] iscsi: Assume "sendpage" is okay in iscsi_tcp_segment_map() David Howells
` (15 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Martin K. Petersen,
linux-scsi, target-devel
Use sendmsg() with MSG_SPLICE_PAGES rather than sendpage. This allows
multiple pages and multipage folios to be passed through.
TODO: iscsit_fe_sendpage_sg() should perhaps set up a bio_vec array for the
entire set of pages it's going to transfer plus two for the header and
trailer and page fragments to hold the header and trailer - and then call
sendmsg once for the entire message.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "Martin K. Petersen" <martin.petersen@oracle.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-scsi@vger.kernel.org
cc: target-devel@vger.kernel.org
cc: netdev@vger.kernel.org
---
drivers/scsi/iscsi_tcp.c | 31 ++++++++++++------------
drivers/scsi/iscsi_tcp.h | 2 +-
drivers/target/iscsi/iscsi_target_util.c | 14 ++++++-----
3 files changed, 24 insertions(+), 23 deletions(-)
diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c
index c76f82fb8b63..cf3eb55d2a76 100644
--- a/drivers/scsi/iscsi_tcp.c
+++ b/drivers/scsi/iscsi_tcp.c
@@ -301,35 +301,37 @@ static int iscsi_sw_tcp_xmit_segment(struct iscsi_tcp_conn *tcp_conn,
while (!iscsi_tcp_segment_done(tcp_conn, segment, 0, r)) {
struct scatterlist *sg;
+ struct msghdr msg = {};
+ union {
+ struct kvec kv;
+ struct bio_vec bv;
+ } vec;
unsigned int offset, copy;
- int flags = 0;
r = 0;
offset = segment->copied;
copy = segment->size - offset;
if (segment->total_copied + segment->size < segment->total_size)
- flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
+ msg.msg_flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
if (tcp_sw_conn->queue_recv)
- flags |= MSG_DONTWAIT;
+ msg.msg_flags |= MSG_DONTWAIT;
- /* Use sendpage if we can; else fall back to sendmsg */
if (!segment->data) {
+ if (tcp_conn->iscsi_conn->datadgst_en)
+ msg.msg_flags |= MSG_SPLICE_PAGES;
sg = segment->sg;
offset += segment->sg_offset + sg->offset;
- r = tcp_sw_conn->sendpage(sk, sg_page(sg), offset,
- copy, flags);
+ bvec_set_page(&vec.bv, sg_page(sg), copy, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &vec.bv, 1, copy);
} else {
- struct msghdr msg = { .msg_flags = flags };
- struct kvec iov = {
- .iov_base = segment->data + offset,
- .iov_len = copy
- };
-
- r = kernel_sendmsg(sk, &msg, &iov, 1, copy);
+ vec.kv.iov_base = segment->data + offset;
+ vec.kv.iov_len = copy;
+ iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, &vec.kv, 1, copy);
}
+ r = sock_sendmsg(sk, &msg);
if (r < 0) {
iscsi_tcp_segment_unmap(segment);
return r;
@@ -746,7 +748,6 @@ iscsi_sw_tcp_conn_bind(struct iscsi_cls_session *cls_session,
sock_no_linger(sk);
iscsi_sw_tcp_conn_set_callbacks(conn);
- tcp_sw_conn->sendpage = tcp_sw_conn->sock->ops->sendpage;
/*
* set receive state machine into initial state
*/
@@ -778,8 +779,6 @@ static int iscsi_sw_tcp_conn_set_param(struct iscsi_cls_conn *cls_conn,
mutex_unlock(&tcp_sw_conn->sock_lock);
return -ENOTCONN;
}
- tcp_sw_conn->sendpage = conn->datadgst_en ?
- sock_no_sendpage : tcp_sw_conn->sock->ops->sendpage;
mutex_unlock(&tcp_sw_conn->sock_lock);
break;
case ISCSI_PARAM_MAX_R2T:
diff --git a/drivers/scsi/iscsi_tcp.h b/drivers/scsi/iscsi_tcp.h
index 68e14a344904..d6ec08d7eb63 100644
--- a/drivers/scsi/iscsi_tcp.h
+++ b/drivers/scsi/iscsi_tcp.h
@@ -48,7 +48,7 @@ struct iscsi_sw_tcp_conn {
uint32_t sendpage_failures_cnt;
uint32_t discontiguous_hdr_cnt;
- ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int);
+ bool can_splice_to_tcp;
};
struct iscsi_sw_tcp_host {
diff --git a/drivers/target/iscsi/iscsi_target_util.c b/drivers/target/iscsi/iscsi_target_util.c
index 26dc8ed3045b..c7d58e41ac3b 100644
--- a/drivers/target/iscsi/iscsi_target_util.c
+++ b/drivers/target/iscsi/iscsi_target_util.c
@@ -1078,6 +1078,8 @@ int iscsit_fe_sendpage_sg(
struct iscsit_conn *conn)
{
struct scatterlist *sg = cmd->first_data_sg;
+ struct bio_vec bvec;
+ struct msghdr msghdr = { .msg_flags = MSG_SPLICE_PAGES, };
struct kvec iov;
u32 tx_hdr_size, data_len;
u32 offset = cmd->first_data_sg_off;
@@ -1121,17 +1123,17 @@ int iscsit_fe_sendpage_sg(
u32 space = (sg->length - offset);
u32 sub_len = min_t(u32, data_len, space);
send_pg:
- tx_sent = conn->sock->ops->sendpage(conn->sock,
- sg_page(sg), sg->offset + offset, sub_len, 0);
+ bvec_set_page(&bvec, sg_page(sg), sub_len, sg->offset + offset);
+ iov_iter_bvec(&msghdr.msg_iter, ITER_SOURCE, &bvec, 1, sub_len);
+
+ tx_sent = conn->sock->ops->sendmsg(conn->sock, &msghdr, sub_len);
if (tx_sent != sub_len) {
if (tx_sent == -EAGAIN) {
- pr_err("tcp_sendpage() returned"
- " -EAGAIN\n");
+ pr_err("sendmsg/splice returned -EAGAIN\n");
goto send_pg;
}
- pr_err("tcp_sendpage() failure: %d\n",
- tx_sent);
+ pr_err("sendmsg/splice failure: %d\n", tx_sent);
return -1;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 33/48] iscsi: Assume "sendpage" is okay in iscsi_tcp_segment_map()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (31 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 32/48] iscsi: " David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 34/48] tcp_bpf: Make tcp_bpf_sendpage() go through tcp_bpf_sendmsg(MSG_SPLICE_PAGES) David Howells
` (14 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Martin K. Petersen,
linux-scsi, target-devel
As iscsi is now using sendmsg() with MSG_SPLICE_PAGES rather than sendpage,
assume that sendpage_ok() will return true in iscsi_tcp_segment_map() and
leave it to TCP to copy the data if not.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "Martin K. Petersen" <martin.petersen@oracle.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-scsi@vger.kernel.org
cc: target-devel@vger.kernel.org
cc: netdev@vger.kernel.org
---
drivers/scsi/libiscsi_tcp.c | 13 +++----------
1 file changed, 3 insertions(+), 10 deletions(-)
diff --git a/drivers/scsi/libiscsi_tcp.c b/drivers/scsi/libiscsi_tcp.c
index c182aa83f2c9..07ba0d864820 100644
--- a/drivers/scsi/libiscsi_tcp.c
+++ b/drivers/scsi/libiscsi_tcp.c
@@ -128,18 +128,11 @@ static void iscsi_tcp_segment_map(struct iscsi_segment *segment, int recv)
* coalescing neighboring slab objects into a single frag which
* triggers one of hardened usercopy checks.
*/
- if (!recv && sendpage_ok(sg_page(sg)))
+ if (!recv)
return;
- if (recv) {
- segment->atomic_mapped = true;
- segment->sg_mapped = kmap_atomic(sg_page(sg));
- } else {
- segment->atomic_mapped = false;
- /* the xmit path can sleep with the page mapped so use kmap */
- segment->sg_mapped = kmap(sg_page(sg));
- }
-
+ segment->atomic_mapped = true;
+ segment->sg_mapped = kmap_atomic(sg_page(sg));
segment->data = segment->sg_mapped + sg->offset + segment->sg_offset;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 34/48] tcp_bpf: Make tcp_bpf_sendpage() go through tcp_bpf_sendmsg(MSG_SPLICE_PAGES)
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (32 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 33/48] iscsi: Assume "sendpage" is okay in iscsi_tcp_segment_map() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 35/48] net: Use sendmsg(MSG_SPLICE_PAGES) not sendpage in skb_send_sock() David Howells
` (13 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, John Fastabend,
Jakub Sitnicki, bpf
Translate tcp_bpf_sendpage() calls to tcp_bpf_sendmsg(MSG_SPLICE_PAGES).
Signed-off-by: David Howells <dhowells@redhat.com>
cc: John Fastabend <john.fastabend@gmail.com>
cc: Jakub Sitnicki <jakub@cloudflare.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: bpf@vger.kernel.org
cc: netdev@vger.kernel.org
---
net/ipv4/tcp_bpf.c | 49 +++++++++-------------------------------------
1 file changed, 9 insertions(+), 40 deletions(-)
diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c
index 7f17134637eb..de37a4372437 100644
--- a/net/ipv4/tcp_bpf.c
+++ b/net/ipv4/tcp_bpf.c
@@ -485,49 +485,18 @@ static int tcp_bpf_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
static int tcp_bpf_sendpage(struct sock *sk, struct page *page, int offset,
size_t size, int flags)
{
- struct sk_msg tmp, *msg = NULL;
- int err = 0, copied = 0;
- struct sk_psock *psock;
- bool enospc = false;
-
- psock = sk_psock_get(sk);
- if (unlikely(!psock))
- return tcp_sendpage(sk, page, offset, size, flags);
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = flags | MSG_SPLICE_PAGES,
+ };
- lock_sock(sk);
- if (psock->cork) {
- msg = psock->cork;
- } else {
- msg = &tmp;
- sk_msg_init(msg);
- }
+ bvec_set_page(&bvec, page, size, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
- /* Catch case where ring is full and sendpage is stalled. */
- if (unlikely(sk_msg_full(msg)))
- goto out_err;
-
- sk_msg_page_add(msg, page, size, offset);
- sk_mem_charge(sk, size);
- copied = size;
- if (sk_msg_full(msg))
- enospc = true;
- if (psock->cork_bytes) {
- if (size > psock->cork_bytes)
- psock->cork_bytes = 0;
- else
- psock->cork_bytes -= size;
- if (psock->cork_bytes && !enospc)
- goto out_err;
- /* All cork bytes are accounted, rerun the prog. */
- psock->eval = __SK_NONE;
- psock->cork_bytes = 0;
- }
+ if (flags & MSG_SENDPAGE_NOTLAST)
+ msg.msg_flags |= MSG_MORE;
- err = tcp_bpf_send_verdict(sk, psock, msg, &copied, flags);
-out_err:
- release_sock(sk);
- sk_psock_put(sk, psock);
- return copied ? copied : err;
+ return tcp_bpf_sendmsg(sk, &msg, size);
}
enum {
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 35/48] net: Use sendmsg(MSG_SPLICE_PAGES) not sendpage in skb_send_sock()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (33 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 34/48] tcp_bpf: Make tcp_bpf_sendpage() go through tcp_bpf_sendmsg(MSG_SPLICE_PAGES) David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 36/48] algif: Remove hash_sendpage*() David Howells
` (12 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm
Use sendmsg() with MSG_SPLICE_PAGES rather than sendpage in
skb_send_sock(). This causes pages to be spliced from the source iterator
if possible (the iterator must be ITER_BVEC and the pages must be
spliceable).
This allows ->sendpage() to be replaced by something that can handle
multiple multipage folios in a single transaction.
Note that this could perhaps be improved to fill out a bvec array with all
the frags and then make a single sendmsg call, possibly sticking the header
on the front also.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/core/skbuff.c | 49 ++++++++++++++++++++++++++---------------------
1 file changed, 27 insertions(+), 22 deletions(-)
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 0506e4cf1ed9..3693b3526d33 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -2919,32 +2919,32 @@ int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
}
EXPORT_SYMBOL_GPL(skb_splice_bits);
-static int sendmsg_unlocked(struct sock *sk, struct msghdr *msg,
- struct kvec *vec, size_t num, size_t size)
+static int sendmsg_locked(struct sock *sk, struct msghdr *msg)
{
struct socket *sock = sk->sk_socket;
+ size_t size = msg_data_left(msg);
if (!sock)
return -EINVAL;
- return kernel_sendmsg(sock, msg, vec, num, size);
+
+ if (!sock->ops->sendmsg_locked)
+ return sock_no_sendmsg_locked(sk, msg, size);
+
+ return sock->ops->sendmsg_locked(sk, msg, size);
}
-static int sendpage_unlocked(struct sock *sk, struct page *page, int offset,
- size_t size, int flags)
+static int sendmsg_unlocked(struct sock *sk, struct msghdr *msg)
{
struct socket *sock = sk->sk_socket;
if (!sock)
return -EINVAL;
- return kernel_sendpage(sock, page, offset, size, flags);
+ return sock_sendmsg(sock, msg);
}
-typedef int (*sendmsg_func)(struct sock *sk, struct msghdr *msg,
- struct kvec *vec, size_t num, size_t size);
-typedef int (*sendpage_func)(struct sock *sk, struct page *page, int offset,
- size_t size, int flags);
+typedef int (*sendmsg_func)(struct sock *sk, struct msghdr *msg);
static int __skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset,
- int len, sendmsg_func sendmsg, sendpage_func sendpage)
+ int len, sendmsg_func sendmsg)
{
unsigned int orig_len = len;
struct sk_buff *head = skb;
@@ -2964,8 +2964,9 @@ static int __skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset,
memset(&msg, 0, sizeof(msg));
msg.msg_flags = MSG_DONTWAIT;
- ret = INDIRECT_CALL_2(sendmsg, kernel_sendmsg_locked,
- sendmsg_unlocked, sk, &msg, &kv, 1, slen);
+ iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, &kv, 1, slen);
+ ret = INDIRECT_CALL_2(sendmsg, sendmsg_locked,
+ sendmsg_unlocked, sk, &msg);
if (ret <= 0)
goto error;
@@ -2996,11 +2997,17 @@ static int __skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset,
slen = min_t(size_t, len, skb_frag_size(frag) - offset);
while (slen) {
- ret = INDIRECT_CALL_2(sendpage, kernel_sendpage_locked,
- sendpage_unlocked, sk,
- skb_frag_page(frag),
- skb_frag_off(frag) + offset,
- slen, MSG_DONTWAIT);
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = MSG_SPLICE_PAGES | MSG_DONTWAIT,
+ };
+
+ bvec_set_page(&bvec, skb_frag_page(frag), slen,
+ skb_frag_off(frag) + offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, slen);
+
+ ret = INDIRECT_CALL_2(sendmsg, sendmsg_locked,
+ sendmsg_unlocked, sk, &msg);
if (ret <= 0)
goto error;
@@ -3037,16 +3044,14 @@ static int __skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset,
int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
int len)
{
- return __skb_send_sock(sk, skb, offset, len, kernel_sendmsg_locked,
- kernel_sendpage_locked);
+ return __skb_send_sock(sk, skb, offset, len, sendmsg_locked);
}
EXPORT_SYMBOL_GPL(skb_send_sock_locked);
/* Send skb data on a socket. Socket must be unlocked. */
int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len)
{
- return __skb_send_sock(sk, skb, offset, len, sendmsg_unlocked,
- sendpage_unlocked);
+ return __skb_send_sock(sk, skb, offset, len, sendmsg_unlocked);
}
/**
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 36/48] algif: Remove hash_sendpage*()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (34 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 35/48] net: Use sendmsg(MSG_SPLICE_PAGES) not sendpage in skb_send_sock() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage() David Howells
` (11 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Herbert Xu,
linux-crypto
Remove hash_sendpage*()..
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Herbert Xu <herbert@gondor.apana.org.au>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-crypto@vger.kernel.org
cc: netdev@vger.kernel.org
---
crypto/algif_hash.c | 66 ---------------------------------------------
1 file changed, 66 deletions(-)
diff --git a/crypto/algif_hash.c b/crypto/algif_hash.c
index b89c2c50cecc..dc6c45637b2d 100644
--- a/crypto/algif_hash.c
+++ b/crypto/algif_hash.c
@@ -162,58 +162,6 @@ static int hash_sendmsg(struct socket *sock, struct msghdr *msg,
goto unlock;
}
-static ssize_t hash_sendpage(struct socket *sock, struct page *page,
- int offset, size_t size, int flags)
-{
- struct sock *sk = sock->sk;
- struct alg_sock *ask = alg_sk(sk);
- struct hash_ctx *ctx = ask->private;
- int err;
-
- if (flags & MSG_SENDPAGE_NOTLAST)
- flags |= MSG_MORE;
-
- lock_sock(sk);
- sg_init_table(ctx->sgl.sgl, 1);
- sg_set_page(ctx->sgl.sgl, page, size, offset);
-
- if (!(flags & MSG_MORE)) {
- err = hash_alloc_result(sk, ctx);
- if (err)
- goto unlock;
- } else if (!ctx->more)
- hash_free_result(sk, ctx);
-
- ahash_request_set_crypt(&ctx->req, ctx->sgl.sgl, ctx->result, size);
-
- if (!(flags & MSG_MORE)) {
- if (ctx->more)
- err = crypto_ahash_finup(&ctx->req);
- else
- err = crypto_ahash_digest(&ctx->req);
- } else {
- if (!ctx->more) {
- err = crypto_ahash_init(&ctx->req);
- err = crypto_wait_req(err, &ctx->wait);
- if (err)
- goto unlock;
- }
-
- err = crypto_ahash_update(&ctx->req);
- }
-
- err = crypto_wait_req(err, &ctx->wait);
- if (err)
- goto unlock;
-
- ctx->more = flags & MSG_MORE;
-
-unlock:
- release_sock(sk);
-
- return err ?: size;
-}
-
static int hash_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
int flags)
{
@@ -318,7 +266,6 @@ static struct proto_ops algif_hash_ops = {
.release = af_alg_release,
.sendmsg = hash_sendmsg,
- .sendpage = hash_sendpage,
.recvmsg = hash_recvmsg,
.accept = hash_accept,
};
@@ -370,18 +317,6 @@ static int hash_sendmsg_nokey(struct socket *sock, struct msghdr *msg,
return hash_sendmsg(sock, msg, size);
}
-static ssize_t hash_sendpage_nokey(struct socket *sock, struct page *page,
- int offset, size_t size, int flags)
-{
- int err;
-
- err = hash_check_key(sock);
- if (err)
- return err;
-
- return hash_sendpage(sock, page, offset, size, flags);
-}
-
static int hash_recvmsg_nokey(struct socket *sock, struct msghdr *msg,
size_t ignored, int flags)
{
@@ -420,7 +355,6 @@ static struct proto_ops algif_hash_ops_nokey = {
.release = af_alg_release,
.sendmsg = hash_sendmsg_nokey,
- .sendpage = hash_sendpage_nokey,
.recvmsg = hash_recvmsg_nokey,
.accept = hash_accept_nokey,
};
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (35 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 36/48] algif: Remove hash_sendpage*() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-30 1:45 ` Xiubo Li
2023-03-30 6:48 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 38/48] rds: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage David Howells
` (10 subsequent siblings)
47 siblings, 2 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Ilya Dryomov,
Xiubo Li, ceph-devel
Use sendmsg() and MSG_SPLICE_PAGES rather than sendpage in ceph when
transmitting data. For the moment, this can only transmit one page at a
time because of the architecture of net/ceph/, but if
write_partial_message_data() can be given a bvec[] at a time by the
iteration code, this would allow pages to be sent in a batch.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Ilya Dryomov <idryomov@gmail.com>
cc: Xiubo Li <xiubli@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: ceph-devel@vger.kernel.org
cc: netdev@vger.kernel.org
---
net/ceph/messenger_v2.c | 89 +++++++++--------------------------------
1 file changed, 18 insertions(+), 71 deletions(-)
diff --git a/net/ceph/messenger_v2.c b/net/ceph/messenger_v2.c
index 301a991dc6a6..1637a0c21126 100644
--- a/net/ceph/messenger_v2.c
+++ b/net/ceph/messenger_v2.c
@@ -117,91 +117,38 @@ static int ceph_tcp_recv(struct ceph_connection *con)
return ret;
}
-static int do_sendmsg(struct socket *sock, struct iov_iter *it)
-{
- struct msghdr msg = { .msg_flags = CEPH_MSG_FLAGS };
- int ret;
-
- msg.msg_iter = *it;
- while (iov_iter_count(it)) {
- ret = sock_sendmsg(sock, &msg);
- if (ret <= 0) {
- if (ret == -EAGAIN)
- ret = 0;
- return ret;
- }
-
- iov_iter_advance(it, ret);
- }
-
- WARN_ON(msg_data_left(&msg));
- return 1;
-}
-
-static int do_try_sendpage(struct socket *sock, struct iov_iter *it)
-{
- struct msghdr msg = { .msg_flags = CEPH_MSG_FLAGS };
- struct bio_vec bv;
- int ret;
-
- if (WARN_ON(!iov_iter_is_bvec(it)))
- return -EINVAL;
-
- while (iov_iter_count(it)) {
- /* iov_iter_iovec() for ITER_BVEC */
- bvec_set_page(&bv, it->bvec->bv_page,
- min(iov_iter_count(it),
- it->bvec->bv_len - it->iov_offset),
- it->bvec->bv_offset + it->iov_offset);
-
- /*
- * sendpage cannot properly handle pages with
- * page_count == 0, we need to fall back to sendmsg if
- * that's the case.
- *
- * Same goes for slab pages: skb_can_coalesce() allows
- * coalescing neighboring slab objects into a single frag
- * which triggers one of hardened usercopy checks.
- */
- if (sendpage_ok(bv.bv_page)) {
- ret = sock->ops->sendpage(sock, bv.bv_page,
- bv.bv_offset, bv.bv_len,
- CEPH_MSG_FLAGS);
- } else {
- iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bv, 1, bv.bv_len);
- ret = sock_sendmsg(sock, &msg);
- }
- if (ret <= 0) {
- if (ret == -EAGAIN)
- ret = 0;
- return ret;
- }
-
- iov_iter_advance(it, ret);
- }
-
- return 1;
-}
-
/*
* Write as much as possible. The socket is expected to be corked,
* so we don't bother with MSG_MORE/MSG_SENDPAGE_NOTLAST here.
*
* Return:
- * 1 - done, nothing (else) to write
+ * >0 - done, nothing (else) to write
* 0 - socket is full, need to wait
* <0 - error
*/
static int ceph_tcp_send(struct ceph_connection *con)
{
+ struct msghdr msg = {
+ .msg_iter = con->v2.out_iter,
+ .msg_flags = CEPH_MSG_FLAGS,
+ };
int ret;
+ if (WARN_ON(!iov_iter_is_bvec(&con->v2.out_iter)))
+ return -EINVAL;
+
+ if (con->v2.out_iter_sendpage)
+ msg.msg_flags |= MSG_SPLICE_PAGES;
+
dout("%s con %p have %zu try_sendpage %d\n", __func__, con,
iov_iter_count(&con->v2.out_iter), con->v2.out_iter_sendpage);
- if (con->v2.out_iter_sendpage)
- ret = do_try_sendpage(con->sock, &con->v2.out_iter);
- else
- ret = do_sendmsg(con->sock, &con->v2.out_iter);
+
+ ret = sock_sendmsg(con->sock, &msg);
+ if (ret > 0)
+ iov_iter_advance(&con->v2.out_iter, ret);
+ else if (ret == -EAGAIN)
+ ret = 0;
+
dout("%s con %p ret %d left %zu\n", __func__, con, ret,
iov_iter_count(&con->v2.out_iter));
return ret;
^ permalink raw reply related [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-03-29 14:13 ` [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage() David Howells
@ 2023-03-30 1:45 ` Xiubo Li
2023-03-30 6:48 ` David Howells
1 sibling, 0 replies; 77+ messages in thread
From: Xiubo Li @ 2023-03-30 1:45 UTC (permalink / raw)
To: David Howells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni
Cc: Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Chuck Lever III, Linus Torvalds, netdev,
linux-fsdevel, linux-kernel, linux-mm, Ilya Dryomov, ceph-devel
David,
BTW, will this two patch depend on the others in this patch series ?
I am planing to run a test with these two later.
Thanks
- Xiubo
On 29/03/2023 22:13, David Howells wrote:
> Use sendmsg() and MSG_SPLICE_PAGES rather than sendpage in ceph when
> transmitting data. For the moment, this can only transmit one page at a
> time because of the architecture of net/ceph/, but if
> write_partial_message_data() can be given a bvec[] at a time by the
> iteration code, this would allow pages to be sent in a batch.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Ilya Dryomov <idryomov@gmail.com>
> cc: Xiubo Li <xiubli@redhat.com>
> cc: Jeff Layton <jlayton@kernel.org>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: ceph-devel@vger.kernel.org
> cc: netdev@vger.kernel.org
> ---
> net/ceph/messenger_v2.c | 89 +++++++++--------------------------------
> 1 file changed, 18 insertions(+), 71 deletions(-)
>
> diff --git a/net/ceph/messenger_v2.c b/net/ceph/messenger_v2.c
> index 301a991dc6a6..1637a0c21126 100644
> --- a/net/ceph/messenger_v2.c
> +++ b/net/ceph/messenger_v2.c
> @@ -117,91 +117,38 @@ static int ceph_tcp_recv(struct ceph_connection *con)
> return ret;
> }
>
> -static int do_sendmsg(struct socket *sock, struct iov_iter *it)
> -{
> - struct msghdr msg = { .msg_flags = CEPH_MSG_FLAGS };
> - int ret;
> -
> - msg.msg_iter = *it;
> - while (iov_iter_count(it)) {
> - ret = sock_sendmsg(sock, &msg);
> - if (ret <= 0) {
> - if (ret == -EAGAIN)
> - ret = 0;
> - return ret;
> - }
> -
> - iov_iter_advance(it, ret);
> - }
> -
> - WARN_ON(msg_data_left(&msg));
> - return 1;
> -}
> -
> -static int do_try_sendpage(struct socket *sock, struct iov_iter *it)
> -{
> - struct msghdr msg = { .msg_flags = CEPH_MSG_FLAGS };
> - struct bio_vec bv;
> - int ret;
> -
> - if (WARN_ON(!iov_iter_is_bvec(it)))
> - return -EINVAL;
> -
> - while (iov_iter_count(it)) {
> - /* iov_iter_iovec() for ITER_BVEC */
> - bvec_set_page(&bv, it->bvec->bv_page,
> - min(iov_iter_count(it),
> - it->bvec->bv_len - it->iov_offset),
> - it->bvec->bv_offset + it->iov_offset);
> -
> - /*
> - * sendpage cannot properly handle pages with
> - * page_count == 0, we need to fall back to sendmsg if
> - * that's the case.
> - *
> - * Same goes for slab pages: skb_can_coalesce() allows
> - * coalescing neighboring slab objects into a single frag
> - * which triggers one of hardened usercopy checks.
> - */
> - if (sendpage_ok(bv.bv_page)) {
> - ret = sock->ops->sendpage(sock, bv.bv_page,
> - bv.bv_offset, bv.bv_len,
> - CEPH_MSG_FLAGS);
> - } else {
> - iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bv, 1, bv.bv_len);
> - ret = sock_sendmsg(sock, &msg);
> - }
> - if (ret <= 0) {
> - if (ret == -EAGAIN)
> - ret = 0;
> - return ret;
> - }
> -
> - iov_iter_advance(it, ret);
> - }
> -
> - return 1;
> -}
> -
> /*
> * Write as much as possible. The socket is expected to be corked,
> * so we don't bother with MSG_MORE/MSG_SENDPAGE_NOTLAST here.
> *
> * Return:
> - * 1 - done, nothing (else) to write
> + * >0 - done, nothing (else) to write
> * 0 - socket is full, need to wait
> * <0 - error
> */
> static int ceph_tcp_send(struct ceph_connection *con)
> {
> + struct msghdr msg = {
> + .msg_iter = con->v2.out_iter,
> + .msg_flags = CEPH_MSG_FLAGS,
> + };
> int ret;
>
> + if (WARN_ON(!iov_iter_is_bvec(&con->v2.out_iter)))
> + return -EINVAL;
> +
> + if (con->v2.out_iter_sendpage)
> + msg.msg_flags |= MSG_SPLICE_PAGES;
> +
> dout("%s con %p have %zu try_sendpage %d\n", __func__, con,
> iov_iter_count(&con->v2.out_iter), con->v2.out_iter_sendpage);
> - if (con->v2.out_iter_sendpage)
> - ret = do_try_sendpage(con->sock, &con->v2.out_iter);
> - else
> - ret = do_sendmsg(con->sock, &con->v2.out_iter);
> +
> + ret = sock_sendmsg(con->sock, &msg);
> + if (ret > 0)
> + iov_iter_advance(&con->v2.out_iter, ret);
> + else if (ret == -EAGAIN)
> + ret = 0;
> +
> dout("%s con %p ret %d left %zu\n", __func__, con, ret,
> iov_iter_count(&con->v2.out_iter));
> return ret;
>
--
Best Regards,
Xiubo Li (李秀波)
Email: xiubli@redhat.com/xiubli@ibm.com
Slack: @Xiubo Li
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-03-29 14:13 ` [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage() David Howells
2023-03-30 1:45 ` Xiubo Li
@ 2023-03-30 6:48 ` David Howells
2023-03-31 13:05 ` Xiubo Li
` (2 more replies)
1 sibling, 3 replies; 77+ messages in thread
From: David Howells @ 2023-03-30 6:48 UTC (permalink / raw)
To: Xiubo Li
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm,
Ilya Dryomov, ceph-devel
Xiubo Li <xiubli@redhat.com> wrote:
> BTW, will this two patch depend on the others in this patch series ?
Yes. You'll need patches that affect TCP at least so that TCP supports
MSG_SPLICE_PAGES, so 04-08 and perhaps 09. It's also on top of the patches
that remove ITER_PIPE on my iov-extract branch, but I don't think that should
affect you.
David
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-03-30 6:48 ` David Howells
@ 2023-03-31 13:05 ` Xiubo Li
2023-04-03 3:27 ` Xiubo Li
2023-04-03 8:32 ` David Howells
2 siblings, 0 replies; 77+ messages in thread
From: Xiubo Li @ 2023-03-31 13:05 UTC (permalink / raw)
To: David Howells
Cc: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Chuck Lever III, Linus Torvalds, netdev,
linux-fsdevel, linux-kernel, linux-mm, Ilya Dryomov, ceph-devel
On 3/30/23 02:48, David Howells wrote:
> Xiubo Li <xiubli@redhat.com> wrote:
>
>> BTW, will this two patch depend on the others in this patch series ?
> Yes. You'll need patches that affect TCP at least so that TCP supports
> MSG_SPLICE_PAGES, so 04-08 and perhaps 09. It's also on top of the patches
> that remove ITER_PIPE on my iov-extract branch, but I don't think that should
> affect you.
Okay, I will check that.
Thanks.
> David
>
^ permalink raw reply [flat|nested] 77+ messages in thread
* Re: [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-03-30 6:48 ` David Howells
2023-03-31 13:05 ` Xiubo Li
@ 2023-04-03 3:27 ` Xiubo Li
2023-04-03 8:32 ` David Howells
2 siblings, 0 replies; 77+ messages in thread
From: Xiubo Li @ 2023-04-03 3:27 UTC (permalink / raw)
To: David Howells
Cc: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Chuck Lever III, Linus Torvalds, netdev,
linux-fsdevel, linux-kernel, linux-mm, Ilya Dryomov, ceph-devel
On 3/30/23 14:48, David Howells wrote:
> Xiubo Li <xiubli@redhat.com> wrote:
>
>> BTW, will this two patch depend on the others in this patch series ?
> Yes. You'll need patches that affect TCP at least so that TCP supports
> MSG_SPLICE_PAGES, so 04-08 and perhaps 09. It's also on top of the patches
> that remove ITER_PIPE on my iov-extract branch, but I don't think that should
> affect you.
Why I asked this is because I only could see these two ceph relevant
patches currently.
Thanks
- Xiubo
> David
>
^ permalink raw reply [flat|nested] 77+ messages in thread
* Re: [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-03-30 6:48 ` David Howells
2023-03-31 13:05 ` Xiubo Li
2023-04-03 3:27 ` Xiubo Li
@ 2023-04-03 8:32 ` David Howells
2023-04-10 0:38 ` Xiubo Li
2 siblings, 1 reply; 77+ messages in thread
From: David Howells @ 2023-04-03 8:32 UTC (permalink / raw)
To: Xiubo Li
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Chuck Lever III,
Linus Torvalds, netdev, linux-fsdevel, linux-kernel, linux-mm,
Ilya Dryomov, ceph-devel
Xiubo Li <xiubli@redhat.com> wrote:
> On 3/30/23 14:48, David Howells wrote:
> > Xiubo Li <xiubli@redhat.com> wrote:
> >
> >> BTW, will this two patch depend on the others in this patch series ?
> > Yes. You'll need patches that affect TCP at least so that TCP supports
> > MSG_SPLICE_PAGES, so 04-08 and perhaps 09. It's also on top of the
> > patches that remove ITER_PIPE on my iov-extract branch, but I don't think
> > that should affect you.
>
> Why I asked this is because I only could see these two ceph relevant patches
> currently.
Depends on how you defined 'relevant', I guess. Only two patches modify ceph
directly, but there's a dependency: to make those work, TCP needs altering
also.
David
^ permalink raw reply [flat|nested] 77+ messages in thread
* Re: [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-04-03 8:32 ` David Howells
@ 2023-04-10 0:38 ` Xiubo Li
0 siblings, 0 replies; 77+ messages in thread
From: Xiubo Li @ 2023-04-10 0:38 UTC (permalink / raw)
To: David Howells
Cc: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Chuck Lever III, Linus Torvalds, netdev,
linux-fsdevel, linux-kernel, linux-mm, Ilya Dryomov, ceph-devel
On 4/3/23 16:32, David Howells wrote:
> Xiubo Li <xiubli@redhat.com> wrote:
>
>> On 3/30/23 14:48, David Howells wrote:
>>> Xiubo Li <xiubli@redhat.com> wrote:
>>>
>>>> BTW, will this two patch depend on the others in this patch series ?
>>> Yes. You'll need patches that affect TCP at least so that TCP supports
>>> MSG_SPLICE_PAGES, so 04-08 and perhaps 09. It's also on top of the
>>> patches that remove ITER_PIPE on my iov-extract branch, but I don't think
>>> that should affect you.
>> Why I asked this is because I only could see these two ceph relevant patches
>> currently.
> Depends on how you defined 'relevant', I guess. Only two patches modify ceph
> directly, but there's a dependency: to make those work, TCP needs altering
> also.
Okay. Will run the test today. I was reinstalling my laptop last week.
Thanks
- Xiubo
> David
>
^ permalink raw reply [flat|nested] 77+ messages in thread
* [RFC PATCH v2 38/48] rds: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (36 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 37/48] ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 39/48] dlm: " David Howells
` (9 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Santosh Shilimkar,
linux-rdma, rds-devel
When transmitting data, call down into TCP using a single sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather than
performing several sendmsg and sendpage calls to transmit header and data
pages.
To make this work, the data is assembled in a bio_vec array and attached to
a BVEC-type iterator. The header are copied into memory acquired from
zcopy_alloc() which just breaks a page up into small pieces that can be
freed with put_page().
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Santosh Shilimkar <santosh.shilimkar@oracle.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-rdma@vger.kernel.org
cc: rds-devel@oss.oracle.com
cc: netdev@vger.kernel.org
---
net/rds/tcp_send.c | 86 +++++++++++++++++++++-------------------------
1 file changed, 40 insertions(+), 46 deletions(-)
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 8c4d1d6e9249..660d9f203d99 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -52,29 +52,24 @@ void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
tcp_sock_set_cork(tc->t_sock->sk, false);
}
-/* the core send_sem serializes this with other xmit and shutdown */
-static int rds_tcp_sendmsg(struct socket *sock, void *data, unsigned int len)
-{
- struct kvec vec = {
- .iov_base = data,
- .iov_len = len,
- };
- struct msghdr msg = {
- .msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL,
- };
-
- return kernel_sendmsg(sock, &msg, &vec, 1, vec.iov_len);
-}
-
/* the core send_sem serializes this with other xmit and shutdown */
int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
unsigned int hdr_off, unsigned int sg, unsigned int off)
{
struct rds_conn_path *cp = rm->m_inc.i_conn_path;
struct rds_tcp_connection *tc = cp->cp_transport_data;
+ struct msghdr msg = {
+ .msg_flags = MSG_SPLICE_PAGES | MSG_DONTWAIT | MSG_NOSIGNAL,
+ };
+ struct bio_vec *bvec;
+ unsigned int i, size = 0, ix = 0;
+ bool free_hdr = false;
int done = 0;
- int ret = 0;
- int more;
+ int ret = -ENOMEM;
+
+ bvec = kmalloc_array(1 + sg, sizeof(struct bio_vec), GFP_KERNEL);
+ if (!bvec)
+ goto out;
if (hdr_off == 0) {
/*
@@ -99,43 +94,37 @@ int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
if (hdr_off < sizeof(struct rds_header)) {
/* see rds_tcp_write_space() */
+ void *p;
+
set_bit(SOCK_NOSPACE, &tc->t_sock->sk->sk_socket->flags);
- ret = rds_tcp_sendmsg(tc->t_sock,
- (void *)&rm->m_inc.i_hdr + hdr_off,
- sizeof(rm->m_inc.i_hdr) - hdr_off);
- if (ret < 0)
- goto out;
- done += ret;
- if (hdr_off + done != sizeof(struct rds_header))
+ ret = -ENOMEM;
+ p = page_frag_memdup(NULL,
+ (void *)&rm->m_inc.i_hdr + hdr_off,
+ sizeof(rm->m_inc.i_hdr) - hdr_off,
+ GFP_KERNEL, ULONG_MAX);
+ if (!p)
goto out;
+ bvec_set_virt(&bvec[ix], p, sizeof(rm->m_inc.i_hdr) - hdr_off);
+ free_hdr = true;
+ size += bvec[ix].bv_len;
+ ix++;
}
- more = rm->data.op_nents > 1 ? (MSG_MORE | MSG_SENDPAGE_NOTLAST) : 0;
- while (sg < rm->data.op_nents) {
- int flags = MSG_DONTWAIT | MSG_NOSIGNAL | more;
-
- ret = tc->t_sock->ops->sendpage(tc->t_sock,
- sg_page(&rm->data.op_sg[sg]),
- rm->data.op_sg[sg].offset + off,
- rm->data.op_sg[sg].length - off,
- flags);
- rdsdebug("tcp sendpage %p:%u:%u ret %d\n", (void *)sg_page(&rm->data.op_sg[sg]),
- rm->data.op_sg[sg].offset + off, rm->data.op_sg[sg].length - off,
- ret);
- if (ret <= 0)
- break;
-
- off += ret;
- done += ret;
- if (off == rm->data.op_sg[sg].length) {
- off = 0;
- sg++;
- }
- if (sg == rm->data.op_nents - 1)
- more = 0;
+ for (i = sg; i < rm->data.op_nents; i++) {
+ bvec_set_page(&bvec[ix],
+ sg_page(&rm->data.op_sg[i]),
+ rm->data.op_sg[i].length - off,
+ rm->data.op_sg[i].offset + off);
+ off = 0;
+ size += bvec[ix].bv_len;
+ ix++;
}
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, bvec, ix, size);
+ ret = sock_sendmsg(tc->t_sock, &msg);
+ rdsdebug("tcp sendmsg-splice %u,%u ret %d\n", ix, size, ret);
+
out:
if (ret <= 0) {
/* write_space will hit after EAGAIN, all else fatal */
@@ -158,6 +147,11 @@ int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
}
if (done == 0)
done = ret;
+ if (bvec) {
+ if (free_hdr)
+ put_page(bvec[0].bv_page);
+ kfree(bvec);
+ }
return done;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 39/48] dlm: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (37 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 38/48] rds: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
` (8 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm,
Christine Caulfield, David Teigland, cluster-devel
When transmitting data, call down a layer using a single sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather using
sendpage. This allows ->sendpage() to be replaced by something that can
handle multiple multipage folios in a single transaction.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Christine Caulfield <ccaulfie@redhat.com>
cc: David Teigland <teigland@redhat.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: cluster-devel@redhat.com
cc: netdev@vger.kernel.org
---
fs/dlm/lowcomms.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index a9b14f81d655..9c0c691b6106 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1394,8 +1394,11 @@ int dlm_lowcomms_resend_msg(struct dlm_msg *msg)
/* Send a message */
static int send_to_sock(struct connection *con)
{
- const int msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
struct writequeue_entry *e;
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = MSG_SPLICE_PAGES | MSG_DONTWAIT | MSG_NOSIGNAL,
+ };
int len, offset, ret;
spin_lock_bh(&con->writequeue_lock);
@@ -1411,8 +1414,9 @@ static int send_to_sock(struct connection *con)
WARN_ON_ONCE(len == 0 && e->users == 0);
spin_unlock_bh(&con->writequeue_lock);
- ret = kernel_sendpage(con->sock, e->page, offset, len,
- msg_flags);
+ bvec_set_page(&bvec, e->page, len, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, len);
+ ret = sock_sendmsg(con->sock, &msg);
trace_dlm_send(con->nodeid, ret);
if (ret == -EAGAIN || ret == 0) {
lock_sock(con->sock->sk);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (38 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 39/48] dlm: " David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 15:28 ` Chuck Lever III
` (5 more replies)
2023-03-29 14:13 ` [RFC PATCH v2 41/48] sunrpc: Rely on TCP sendmsg + MSG_SPLICE_PAGES to copy unspliceable data David Howells
` (7 subsequent siblings)
47 siblings, 6 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Trond Myklebust,
Anna Schumaker, linux-nfs
When transmitting data, call down into TCP using a single sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather than
performing several sendmsg and sendpage calls to transmit header, data
pages and trailer.
To make this work, the data is assembled in a bio_vec array and attached to
a BVEC-type iterator. The header and trailer are copied into page
fragments so that they can be freed with put_page and attached to iterators
of their own. An iterator-of-iterators is then created to bridge all three
iterators (headers, data, trailer) and that is passed to sendmsg to pass
the entire message in a single call.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Trond Myklebust <trond.myklebust@hammerspace.com>
cc: Anna Schumaker <anna@kernel.org>
cc: Chuck Lever <chuck.lever@oracle.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-nfs@vger.kernel.org
cc: netdev@vger.kernel.org
---
include/linux/sunrpc/svc.h | 11 +++--
net/sunrpc/svcsock.c | 89 +++++++++++++++-----------------------
2 files changed, 40 insertions(+), 60 deletions(-)
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index 877891536c2f..456ae554aa11 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -161,16 +161,15 @@ static inline bool svc_put_not_last(struct svc_serv *serv)
extern u32 svc_max_payload(const struct svc_rqst *rqstp);
/*
- * RPC Requsts and replies are stored in one or more pages.
+ * RPC Requests and replies are stored in one or more pages.
* We maintain an array of pages for each server thread.
* Requests are copied into these pages as they arrive. Remaining
* pages are available to write the reply into.
*
- * Pages are sent using ->sendpage so each server thread needs to
- * allocate more to replace those used in sending. To help keep track
- * of these pages we have a receive list where all pages initialy live,
- * and a send list where pages are moved to when there are to be part
- * of a reply.
+ * Pages are sent using ->sendmsg with MSG_SPLICE_PAGES so each server thread
+ * needs to allocate more to replace those used in sending. To help keep track
+ * of these pages we have a receive list where all pages initialy live, and a
+ * send list where pages are moved to when there are to be part of a reply.
*
* We use xdr_buf for holding responses as it fits well with NFS
* read responses (that have a header, and some data pages, and possibly
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 03a4f5615086..f1cc53aad6e0 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1060,16 +1060,8 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
return 0; /* record not complete */
}
-static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec,
- int flags)
-{
- return kernel_sendpage(sock, virt_to_page(vec->iov_base),
- offset_in_page(vec->iov_base),
- vec->iov_len, flags);
-}
-
/*
- * kernel_sendpage() is used exclusively to reduce the number of
+ * MSG_SPLICE_PAGES is used exclusively to reduce the number of
* copy operations in this path. Therefore the caller must ensure
* that the pages backing @xdr are unchanging.
*
@@ -1081,65 +1073,54 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
{
const struct kvec *head = xdr->head;
const struct kvec *tail = xdr->tail;
- struct kvec rm = {
- .iov_base = &marker,
- .iov_len = sizeof(marker),
- };
+ struct iov_iter iters[3];
+ struct bio_vec head_bv, tail_bv;
struct msghdr msg = {
- .msg_flags = 0,
+ .msg_flags = MSG_SPLICE_PAGES,
};
- int ret;
+ void *m, *t;
+ int ret, n = 2, size;
*sentp = 0;
ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
if (ret < 0)
return ret;
- ret = kernel_sendmsg(sock, &msg, &rm, 1, rm.iov_len);
- if (ret < 0)
- return ret;
- *sentp += ret;
- if (ret != rm.iov_len)
- return -EAGAIN;
+ m = page_frag_alloc(NULL, sizeof(marker) + head->iov_len + tail->iov_len,
+ GFP_KERNEL);
+ if (!m)
+ return -ENOMEM;
- ret = svc_tcp_send_kvec(sock, head, 0);
- if (ret < 0)
- return ret;
- *sentp += ret;
- if (ret != head->iov_len)
- goto out;
+ memcpy(m, &marker, sizeof(marker));
+ if (head->iov_len)
+ memcpy(m + sizeof(marker), head->iov_base, head->iov_len);
+ bvec_set_virt(&head_bv, m, sizeof(marker) + head->iov_len);
+ iov_iter_bvec(&iters[0], ITER_SOURCE, &head_bv, 1,
+ sizeof(marker) + head->iov_len);
- if (xdr->page_len) {
- unsigned int offset, len, remaining;
- struct bio_vec *bvec;
-
- bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
- offset = offset_in_page(xdr->page_base);
- remaining = xdr->page_len;
- while (remaining > 0) {
- len = min(remaining, bvec->bv_len - offset);
- ret = kernel_sendpage(sock, bvec->bv_page,
- bvec->bv_offset + offset,
- len, 0);
- if (ret < 0)
- return ret;
- *sentp += ret;
- if (ret != len)
- goto out;
- remaining -= len;
- offset = 0;
- bvec++;
- }
- }
+ iov_iter_bvec(&iters[1], ITER_SOURCE, xdr->bvec,
+ xdr_buf_pagecount(xdr), xdr->page_len);
if (tail->iov_len) {
- ret = svc_tcp_send_kvec(sock, tail, 0);
- if (ret < 0)
- return ret;
- *sentp += ret;
+ t = page_frag_alloc(NULL, tail->iov_len, GFP_KERNEL);
+ if (!t)
+ return -ENOMEM;
+ memcpy(t, tail->iov_base, tail->iov_len);
+ bvec_set_virt(&tail_bv, t, tail->iov_len);
+ iov_iter_bvec(&iters[2], ITER_SOURCE, &tail_bv, 1, tail->iov_len);
+ n++;
}
-out:
+ size = sizeof(marker) + head->iov_len + xdr->page_len + tail->iov_len;
+ iov_iter_iterlist(&msg.msg_iter, ITER_SOURCE, iters, n, size);
+
+ ret = sock_sendmsg(sock, &msg);
+ if (ret < 0)
+ return ret;
+ if (ret > 0)
+ *sentp = ret;
+ if (ret != size)
+ return -EAGAIN;
return 0;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
@ 2023-03-29 15:28 ` Chuck Lever III
2023-03-29 19:58 ` David Howells
` (4 subsequent siblings)
5 siblings, 0 replies; 77+ messages in thread
From: Chuck Lever III @ 2023-03-29 15:28 UTC (permalink / raw)
To: David Howells
Cc: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Linus Torvalds, open list:NETWORKING [GENERAL],
linux-fsdevel, Linux Kernel Mailing List,
Linux Memory Management List, Trond Myklebust, Anna Schumaker,
Linux NFS Mailing List
> On Mar 29, 2023, at 10:13 AM, David Howells <dhowells@redhat.com> wrote:
>
> When transmitting data, call down into TCP using a single sendmsg with
> MSG_SPLICE_PAGES to indicate that content should be spliced rather than
> performing several sendmsg and sendpage calls to transmit header, data
> pages and trailer.
>
> To make this work, the data is assembled in a bio_vec array and attached to
> a BVEC-type iterator. The header and trailer are copied into page
> fragments so that they can be freed with put_page and attached to iterators
> of their own. An iterator-of-iterators is then created to bridge all three
> iterators (headers, data, trailer) and that is passed to sendmsg to pass
> the entire message in a single call.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Trond Myklebust <trond.myklebust@hammerspace.com>
> cc: Anna Schumaker <anna@kernel.org>
> cc: Chuck Lever <chuck.lever@oracle.com>
> cc: Jeff Layton <jlayton@kernel.org>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: linux-nfs@vger.kernel.org
> cc: netdev@vger.kernel.org
> ---
> include/linux/sunrpc/svc.h | 11 +++--
> net/sunrpc/svcsock.c | 89 +++++++++++++++-----------------------
> 2 files changed, 40 insertions(+), 60 deletions(-)
>
> diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
> index 877891536c2f..456ae554aa11 100644
> --- a/include/linux/sunrpc/svc.h
> +++ b/include/linux/sunrpc/svc.h
> @@ -161,16 +161,15 @@ static inline bool svc_put_not_last(struct svc_serv *serv)
> extern u32 svc_max_payload(const struct svc_rqst *rqstp);
>
> /*
> - * RPC Requsts and replies are stored in one or more pages.
> + * RPC Requests and replies are stored in one or more pages.
> * We maintain an array of pages for each server thread.
> * Requests are copied into these pages as they arrive. Remaining
> * pages are available to write the reply into.
> *
> - * Pages are sent using ->sendpage so each server thread needs to
> - * allocate more to replace those used in sending. To help keep track
> - * of these pages we have a receive list where all pages initialy live,
> - * and a send list where pages are moved to when there are to be part
> - * of a reply.
> + * Pages are sent using ->sendmsg with MSG_SPLICE_PAGES so each server thread
> + * needs to allocate more to replace those used in sending. To help keep track
> + * of these pages we have a receive list where all pages initialy live, and a
> + * send list where pages are moved to when there are to be part of a reply.
> *
> * We use xdr_buf for holding responses as it fits well with NFS
> * read responses (that have a header, and some data pages, and possibly
> diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
> index 03a4f5615086..f1cc53aad6e0 100644
> --- a/net/sunrpc/svcsock.c
> +++ b/net/sunrpc/svcsock.c
> @@ -1060,16 +1060,8 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
> return 0; /* record not complete */
> }
>
> -static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec,
> - int flags)
> -{
> - return kernel_sendpage(sock, virt_to_page(vec->iov_base),
> - offset_in_page(vec->iov_base),
> - vec->iov_len, flags);
> -}
> -
> /*
> - * kernel_sendpage() is used exclusively to reduce the number of
> + * MSG_SPLICE_PAGES is used exclusively to reduce the number of
> * copy operations in this path. Therefore the caller must ensure
> * that the pages backing @xdr are unchanging.
> *
> @@ -1081,65 +1073,54 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
> {
> const struct kvec *head = xdr->head;
> const struct kvec *tail = xdr->tail;
> - struct kvec rm = {
> - .iov_base = &marker,
> - .iov_len = sizeof(marker),
> - };
> + struct iov_iter iters[3];
> + struct bio_vec head_bv, tail_bv;
> struct msghdr msg = {
> - .msg_flags = 0,
> + .msg_flags = MSG_SPLICE_PAGES,
> };
> - int ret;
> + void *m, *t;
> + int ret, n = 2, size;
>
> *sentp = 0;
> ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
> if (ret < 0)
> return ret;
>
> - ret = kernel_sendmsg(sock, &msg, &rm, 1, rm.iov_len);
> - if (ret < 0)
> - return ret;
> - *sentp += ret;
> - if (ret != rm.iov_len)
> - return -EAGAIN;
> + m = page_frag_alloc(NULL, sizeof(marker) + head->iov_len + tail->iov_len,
> + GFP_KERNEL);
> + if (!m)
> + return -ENOMEM;
I'm not excited about adding another memory allocation for this
very common case.
It seems to me that you could eliminate the kernel_sendpage()
consumer here in svc_tcp_sendmsg() without also replacing the
kernel_sendmsg() calls. That would be a conservative step-wise
approach which would carry less risk, and would accomplish
your stated goal without more radical surgery.
Later maybe we can find a way to deal with the head, tail, and
record marker without additional memory allocations. I believe
on the server side, head and tail are already in pages, for
example, not in kmalloc'd memory. That would need some code
auditing, but I'm OK with combining these into a single
sock_sendmsg() call once we've worked out the disposition of
the xdr_buf components outside of the bvec. That seems a bit
outside your stated goal.
Simply replacing the kernel_sendpage() loop would be a
straightforward change and easy to evaluate and test, and
I'd welcome that without hesitation.
> - ret = svc_tcp_send_kvec(sock, head, 0);
> - if (ret < 0)
> - return ret;
> - *sentp += ret;
> - if (ret != head->iov_len)
> - goto out;
> + memcpy(m, &marker, sizeof(marker));
> + if (head->iov_len)
> + memcpy(m + sizeof(marker), head->iov_base, head->iov_len);
> + bvec_set_virt(&head_bv, m, sizeof(marker) + head->iov_len);
> + iov_iter_bvec(&iters[0], ITER_SOURCE, &head_bv, 1,
> + sizeof(marker) + head->iov_len);
>
> - if (xdr->page_len) {
> - unsigned int offset, len, remaining;
> - struct bio_vec *bvec;
> -
> - bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
> - offset = offset_in_page(xdr->page_base);
> - remaining = xdr->page_len;
> - while (remaining > 0) {
> - len = min(remaining, bvec->bv_len - offset);
> - ret = kernel_sendpage(sock, bvec->bv_page,
> - bvec->bv_offset + offset,
> - len, 0);
> - if (ret < 0)
> - return ret;
> - *sentp += ret;
> - if (ret != len)
> - goto out;
> - remaining -= len;
> - offset = 0;
> - bvec++;
> - }
> - }
> + iov_iter_bvec(&iters[1], ITER_SOURCE, xdr->bvec,
> + xdr_buf_pagecount(xdr), xdr->page_len);
>
> if (tail->iov_len) {
> - ret = svc_tcp_send_kvec(sock, tail, 0);
> - if (ret < 0)
> - return ret;
> - *sentp += ret;
> + t = page_frag_alloc(NULL, tail->iov_len, GFP_KERNEL);
> + if (!t)
> + return -ENOMEM;
> + memcpy(t, tail->iov_base, tail->iov_len);
> + bvec_set_virt(&tail_bv, t, tail->iov_len);
> + iov_iter_bvec(&iters[2], ITER_SOURCE, &tail_bv, 1, tail->iov_len);
> + n++;
> }
>
> -out:
> + size = sizeof(marker) + head->iov_len + xdr->page_len + tail->iov_len;
size = sizeof(marker) + xdr->len;
If xdr->len != head->iov_len + xdr->page_len + tail->iov_len,
that is a bug these days.
> + iov_iter_iterlist(&msg.msg_iter, ITER_SOURCE, iters, n, size);
> +
> + ret = sock_sendmsg(sock, &msg);
> + if (ret < 0)
> + return ret;
> + if (ret > 0)
> + *sentp = ret;
> + if (ret != size)
> + return -EAGAIN;
> return 0;
> }
>
>
--
Chuck Lever
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
2023-03-29 15:28 ` Chuck Lever III
@ 2023-03-29 19:58 ` David Howells
2023-03-30 9:29 ` David Howells
` (3 subsequent siblings)
5 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 19:58 UTC (permalink / raw)
To: Chuck Lever III
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Linus Torvalds,
open list:NETWORKING [GENERAL], linux-fsdevel,
Linux Kernel Mailing List, Linux Memory Management List,
Trond Myklebust, Anna Schumaker, Linux NFS Mailing List
Hi Chuck,
Do you have a simple AF_TLS test to hand?
David
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
2023-03-29 15:28 ` Chuck Lever III
2023-03-29 19:58 ` David Howells
@ 2023-03-30 9:29 ` David Howells
2023-03-30 9:41 ` David Howells
` (2 subsequent siblings)
5 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-30 9:29 UTC (permalink / raw)
To: Chuck Lever III
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Linus Torvalds,
open list:NETWORKING [GENERAL], linux-fsdevel,
Linux Kernel Mailing List, Linux Memory Management List,
Trond Myklebust, Anna Schumaker, Linux NFS Mailing List
Chuck Lever III <chuck.lever@oracle.com> wrote:
> It seems to me that you could eliminate the kernel_sendpage()
> consumer here in svc_tcp_sendmsg() without also replacing the
> kernel_sendmsg() calls. That would be a conservative step-wise
> approach which would carry less risk, and would accomplish
> your stated goal without more radical surgery.
Note that only the marker is sent with kernel_sendmsg() in the unmodified
code; the head and tail are sent with svc_tcp_send_kvec()... which uses
kernel_sendpage() which needs to be changed in my patchset. I can make it do
individual sendmsg calls for all those for now.
David
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
` (2 preceding siblings ...)
2023-03-30 9:29 ` David Howells
@ 2023-03-30 9:41 ` David Howells
2023-03-30 13:16 ` Chuck Lever III
2023-03-30 13:01 ` David Howells
2023-03-30 13:16 ` David Howells
5 siblings, 1 reply; 77+ messages in thread
From: David Howells @ 2023-03-30 9:41 UTC (permalink / raw)
To: Chuck Lever III
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Linus Torvalds,
open list:NETWORKING [GENERAL], linux-fsdevel,
Linux Kernel Mailing List, Linux Memory Management List,
Trond Myklebust, Anna Schumaker, Linux NFS Mailing List
Chuck Lever III <chuck.lever@oracle.com> wrote:
> > + if (ret > 0)
> > + *sentp = ret;
Should that be:
*sentp = ret - sizeof(marker);
David
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-30 9:41 ` David Howells
@ 2023-03-30 13:16 ` Chuck Lever III
0 siblings, 0 replies; 77+ messages in thread
From: Chuck Lever III @ 2023-03-30 13:16 UTC (permalink / raw)
To: David Howells
Cc: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Linus Torvalds, open list:NETWORKING [GENERAL],
linux-fsdevel, Linux Kernel Mailing List,
Linux Memory Management List, Trond Myklebust, Anna Schumaker,
Linux NFS Mailing List
> On Mar 30, 2023, at 5:41 AM, David Howells <dhowells@redhat.com> wrote:
>
> Chuck Lever III <chuck.lever@oracle.com> wrote:
>
>>> + if (ret > 0)
>>> + *sentp = ret;
>
> Should that be:
>
> *sentp = ret - sizeof(marker);
>
> David
>
That's a bit out of context, but ...
The return value of ->xpo_sendto is effectively ignored. There is
no caller of svc_process that checks its return code.
svc_rdma_sendto(), for example, returns zero or a negative errno.
That should be cleaned up one day.
--
Chuck Lever
^ permalink raw reply [flat|nested] 77+ messages in thread
* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
` (3 preceding siblings ...)
2023-03-30 9:41 ` David Howells
@ 2023-03-30 13:01 ` David Howells
2023-03-30 13:16 ` David Howells
5 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-30 13:01 UTC (permalink / raw)
To: Chuck Lever III
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Linus Torvalds,
open list:NETWORKING [GENERAL], linux-fsdevel,
Linux Kernel Mailing List, Linux Memory Management List,
Trond Myklebust, Anna Schumaker, Linux NFS Mailing List
Chuck Lever III <chuck.lever@oracle.com> wrote:
> Simply replacing the kernel_sendpage() loop would be a
> straightforward change and easy to evaluate and test, and
> I'd welcome that without hesitation.
How about the attached for a first phase?
It does three sendmsgs, one for the marker + header, one for the body and one
for the tail.
David
---
sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
When transmitting data, call down into TCP using sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather than
performing sendpage calls to transmit header, data pages and trailer.
The marker and the header are passed in an array of kvecs. The marker will
get copied and the header will get spliced.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Trond Myklebust <trond.myklebust@hammerspace.com>
cc: Anna Schumaker <anna@kernel.org>
cc: Chuck Lever <chuck.lever@oracle.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-nfs@vger.kernel.org
cc: netdev@vger.kernel.org
---
include/linux/sunrpc/svc.h | 11 +++---
net/sunrpc/svcsock.c | 75 ++++++++++++++-------------------------------
2 files changed, 29 insertions(+), 57 deletions(-)
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index 877891536c2f..456ae554aa11 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -161,16 +161,15 @@ static inline bool svc_put_not_last(struct svc_serv *serv)
extern u32 svc_max_payload(const struct svc_rqst *rqstp);
/*
- * RPC Requsts and replies are stored in one or more pages.
+ * RPC Requests and replies are stored in one or more pages.
* We maintain an array of pages for each server thread.
* Requests are copied into these pages as they arrive. Remaining
* pages are available to write the reply into.
*
- * Pages are sent using ->sendpage so each server thread needs to
- * allocate more to replace those used in sending. To help keep track
- * of these pages we have a receive list where all pages initialy live,
- * and a send list where pages are moved to when there are to be part
- * of a reply.
+ * Pages are sent using ->sendmsg with MSG_SPLICE_PAGES so each server thread
+ * needs to allocate more to replace those used in sending. To help keep track
+ * of these pages we have a receive list where all pages initialy live, and a
+ * send list where pages are moved to when there are to be part of a reply.
*
* We use xdr_buf for holding responses as it fits well with NFS
* read responses (that have a header, and some data pages, and possibly
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 03a4f5615086..14efcc08c6f8 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1060,16 +1060,8 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
return 0; /* record not complete */
}
-static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec,
- int flags)
-{
- return kernel_sendpage(sock, virt_to_page(vec->iov_base),
- offset_in_page(vec->iov_base),
- vec->iov_len, flags);
-}
-
/*
- * kernel_sendpage() is used exclusively to reduce the number of
+ * MSG_SPLICE_PAGES is used exclusively to reduce the number of
* copy operations in this path. Therefore the caller must ensure
* that the pages backing @xdr are unchanging.
*
@@ -1081,13 +1073,9 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
{
const struct kvec *head = xdr->head;
const struct kvec *tail = xdr->tail;
- struct kvec rm = {
- .iov_base = &marker,
- .iov_len = sizeof(marker),
- };
- struct msghdr msg = {
- .msg_flags = 0,
- };
+ struct kvec kv[2];
+ struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES | MSG_MORE, };
+ size_t sent;
int ret;
*sentp = 0;
@@ -1095,51 +1083,36 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
if (ret < 0)
return ret;
- ret = kernel_sendmsg(sock, &msg, &rm, 1, rm.iov_len);
+ kv[0].iov_base = ▮
+ kv[0].iov_len = sizeof(marker);
+ kv[1] = *head;
+ iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, kv, 2, sizeof(marker) + head->iov_len);
+ ret = sock_sendmsg(sock, &msg);
if (ret < 0)
return ret;
- *sentp += ret;
- if (ret != rm.iov_len)
- return -EAGAIN;
+ sent = ret;
- ret = svc_tcp_send_kvec(sock, head, 0);
+ if (!tail->iov_len)
+ msg.msg_flags &= ~MSG_MORE;
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec,
+ xdr_buf_pagecount(xdr), xdr->page_len);
+ ret = sock_sendmsg(sock, &msg);
if (ret < 0)
return ret;
- *sentp += ret;
- if (ret != head->iov_len)
- goto out;
-
- if (xdr->page_len) {
- unsigned int offset, len, remaining;
- struct bio_vec *bvec;
-
- bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
- offset = offset_in_page(xdr->page_base);
- remaining = xdr->page_len;
- while (remaining > 0) {
- len = min(remaining, bvec->bv_len - offset);
- ret = kernel_sendpage(sock, bvec->bv_page,
- bvec->bv_offset + offset,
- len, 0);
- if (ret < 0)
- return ret;
- *sentp += ret;
- if (ret != len)
- goto out;
- remaining -= len;
- offset = 0;
- bvec++;
- }
- }
+ sent += ret;
if (tail->iov_len) {
- ret = svc_tcp_send_kvec(sock, tail, 0);
+ msg.msg_flags &= ~MSG_MORE;
+ iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, tail, 1, tail->iov_len);
+ ret = sock_sendmsg(sock, &msg);
if (ret < 0)
return ret;
- *sentp += ret;
+ sent += ret;
}
-
-out:
+ if (sent > 0)
+ *sentp = sent;
+ if (sent != sizeof(marker) + xdr->len)
+ return -EAGAIN;
return 0;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
` (4 preceding siblings ...)
2023-03-30 13:01 ` David Howells
@ 2023-03-30 13:16 ` David Howells
2023-03-30 13:27 ` Chuck Lever III
2023-03-30 14:26 ` David Howells
5 siblings, 2 replies; 77+ messages in thread
From: David Howells @ 2023-03-30 13:16 UTC (permalink / raw)
To: Chuck Lever III
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Linus Torvalds,
open list:NETWORKING [GENERAL], linux-fsdevel,
Linux Kernel Mailing List, Linux Memory Management List,
Trond Myklebust, Anna Schumaker, Linux NFS Mailing List
David Howells <dhowells@redhat.com> wrote:
> Chuck Lever III <chuck.lever@oracle.com> wrote:
>
> > Simply replacing the kernel_sendpage() loop would be a
> > straightforward change and easy to evaluate and test, and
> > I'd welcome that without hesitation.
>
> How about the attached for a first phase?
>
> It does three sendmsgs, one for the marker + header, one for the body and one
> for the tail.
... And this as a second phase.
David
---
sunrpc: Allow xdr->bvec[] to be extended to do a single sendmsg
Allow xdr->bvec[] to be extended and insert the marker, the header and the
tail into it so that a single sendmsg() can be used to transmit the message.
I wonder if it would be possible to insert the marker at the beginning of the
head buffer.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Trond Myklebust <trond.myklebust@hammerspace.com>
cc: Anna Schumaker <anna@kernel.org>
cc: Chuck Lever <chuck.lever@oracle.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-nfs@vger.kernel.org
cc: netdev@vger.kernel.org
---
include/linux/sunrpc/xdr.h | 2 -
net/sunrpc/svcsock.c | 46 ++++++++++++++-------------------------------
net/sunrpc/xdr.c | 19 ++++++++++--------
net/sunrpc/xprtsock.c | 6 ++---
4 files changed, 30 insertions(+), 43 deletions(-)
diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h
index 72014c9216fc..c74ea483228b 100644
--- a/include/linux/sunrpc/xdr.h
+++ b/include/linux/sunrpc/xdr.h
@@ -137,7 +137,7 @@ void xdr_inline_pages(struct xdr_buf *, unsigned int,
struct page **, unsigned int, unsigned int);
void xdr_terminate_string(const struct xdr_buf *, const u32);
size_t xdr_buf_pagecount(const struct xdr_buf *buf);
-int xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp);
+int xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp, unsigned int head, unsigned int tail);
void xdr_free_bvec(struct xdr_buf *buf);
static inline __be32 *xdr_encode_array(__be32 *p, const void *s, unsigned int len)
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 14efcc08c6f8..e55761fe1ccf 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -569,7 +569,7 @@ static int svc_udp_sendto(struct svc_rqst *rqstp)
if (svc_xprt_is_dead(xprt))
goto out_notconn;
- err = xdr_alloc_bvec(xdr, GFP_KERNEL);
+ err = xdr_alloc_bvec(xdr, GFP_KERNEL, 0, 0);
if (err < 0)
goto out_unlock;
@@ -1073,45 +1073,29 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
{
const struct kvec *head = xdr->head;
const struct kvec *tail = xdr->tail;
- struct kvec kv[2];
- struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES | MSG_MORE, };
- size_t sent;
+ struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES, };
+ size_t n;
int ret;
*sentp = 0;
- ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
+ ret = xdr_alloc_bvec(xdr, GFP_KERNEL, 2, 1);
if (ret < 0)
return ret;
- kv[0].iov_base = ▮
- kv[0].iov_len = sizeof(marker);
- kv[1] = *head;
- iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, kv, 2, sizeof(marker) + head->iov_len);
+ n = 2 + xdr_buf_pagecount(xdr);
+ bvec_set_virt(&xdr->bvec[0], &marker, sizeof(marker));
+ bvec_set_virt(&xdr->bvec[1], head->iov_base, head->iov_len);
+ bvec_set_virt(&xdr->bvec[n], tail->iov_base, tail->iov_len);
+ if (tail->iov_len)
+ n++;
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec, n,
+ sizeof(marker) + xdr->len);
ret = sock_sendmsg(sock, &msg);
if (ret < 0)
return ret;
- sent = ret;
-
- if (!tail->iov_len)
- msg.msg_flags &= ~MSG_MORE;
- iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec,
- xdr_buf_pagecount(xdr), xdr->page_len);
- ret = sock_sendmsg(sock, &msg);
- if (ret < 0)
- return ret;
- sent += ret;
-
- if (tail->iov_len) {
- msg.msg_flags &= ~MSG_MORE;
- iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, tail, 1, tail->iov_len);
- ret = sock_sendmsg(sock, &msg);
- if (ret < 0)
- return ret;
- sent += ret;
- }
- if (sent > 0)
- *sentp = sent;
- if (sent != sizeof(marker) + xdr->len)
+ if (ret > 0)
+ *sentp = ret;
+ if (ret != sizeof(marker) + xdr->len)
return -EAGAIN;
return 0;
}
diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c
index 36835b2f5446..695821963849 100644
--- a/net/sunrpc/xdr.c
+++ b/net/sunrpc/xdr.c
@@ -141,18 +141,21 @@ size_t xdr_buf_pagecount(const struct xdr_buf *buf)
}
int
-xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp)
+xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp, unsigned int head, unsigned int tail)
{
- size_t i, n = xdr_buf_pagecount(buf);
+ size_t i, j = 0, n = xdr_buf_pagecount(buf);
- if (n != 0 && buf->bvec == NULL) {
- buf->bvec = kmalloc_array(n, sizeof(buf->bvec[0]), gfp);
+ if (head + n + tail != 0 && buf->bvec == NULL) {
+ buf->bvec = kmalloc_array(head + n + tail,
+ sizeof(buf->bvec[0]), gfp);
if (!buf->bvec)
return -ENOMEM;
- for (i = 0; i < n; i++) {
- bvec_set_page(&buf->bvec[i], buf->pages[i], PAGE_SIZE,
- 0);
- }
+ for (i = 0; i < head; i++)
+ bvec_set_page(&buf->bvec[j++], NULL, 0, 0);
+ for (i = 0; i < n; i++)
+ bvec_set_page(&buf->bvec[j++], buf->pages[i], PAGE_SIZE, 0);
+ for (i = 0; i < tail; i++)
+ bvec_set_page(&buf->bvec[j++], NULL, 0, 0);
}
return 0;
}
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index adcbedc244d6..fdf67e84b1c7 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -825,7 +825,7 @@ static int xs_stream_nospace(struct rpc_rqst *req, bool vm_wait)
static int xs_stream_prepare_request(struct rpc_rqst *req, struct xdr_buf *buf)
{
- return xdr_alloc_bvec(buf, rpc_task_gfp_mask());
+ return xdr_alloc_bvec(buf, rpc_task_gfp_mask(), 0, 0);
}
/*
@@ -954,7 +954,7 @@ static int xs_udp_send_request(struct rpc_rqst *req)
if (!xprt_request_get_cong(xprt, req))
return -EBADSLT;
- status = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
+ status = xdr_alloc_bvec(xdr, rpc_task_gfp_mask(), 0, 0);
if (status < 0)
return status;
req->rq_xtime = ktime_get();
@@ -2591,7 +2591,7 @@ static int bc_sendto(struct rpc_rqst *req)
int err;
req->rq_xtime = ktime_get();
- err = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
+ err = xdr_alloc_bvec(xdr, rpc_task_gfp_mask(), 0, 0);
if (err < 0)
return err;
err = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, marker, &sent);
^ permalink raw reply related [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-30 13:16 ` David Howells
@ 2023-03-30 13:27 ` Chuck Lever III
2023-03-30 14:26 ` David Howells
1 sibling, 0 replies; 77+ messages in thread
From: Chuck Lever III @ 2023-03-30 13:27 UTC (permalink / raw)
To: David Howells
Cc: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Linus Torvalds, open list:NETWORKING [GENERAL],
linux-fsdevel, Linux Kernel Mailing List,
Linux Memory Management List, Trond Myklebust, Anna Schumaker,
Linux NFS Mailing List
> On Mar 30, 2023, at 9:16 AM, David Howells <dhowells@redhat.com> wrote:
>
> David Howells <dhowells@redhat.com> wrote:
>
>> Chuck Lever III <chuck.lever@oracle.com> wrote:
>>
>>> Simply replacing the kernel_sendpage() loop would be a
>>> straightforward change and easy to evaluate and test, and
>>> I'd welcome that without hesitation.
>>
>> How about the attached for a first phase?
>>
>> It does three sendmsgs, one for the marker + header, one for the body and one
>> for the tail.
>
> ... And this as a second phase.
>
> David
> ---
> sunrpc: Allow xdr->bvec[] to be extended to do a single sendmsg
>
> Allow xdr->bvec[] to be extended and insert the marker, the header and the
> tail into it so that a single sendmsg() can be used to transmit the message.
Don't. Just change svc_tcp_send_kvec() to use sock_sendmsg, and
leave the marker alone for now, please.
Let's focus on replacing kernel_sendpage() in this series and
leave the deeper clean-ups for another time.
> I wonder if it would be possible to insert the marker at the beginning of the
> head buffer.
That's the way it used to work. The reason we don't do that is
because each transport has its own record marking mechanism.
UDP has nothing, since each RPC message is encapsulated in a
single datagram. RDMA has a full XDR-encoded header which
contains the location of data chunks to be moved via RDMA.
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Trond Myklebust <trond.myklebust@hammerspace.com>
> cc: Anna Schumaker <anna@kernel.org>
> cc: Chuck Lever <chuck.lever@oracle.com>
> cc: Jeff Layton <jlayton@kernel.org>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: linux-nfs@vger.kernel.org
> cc: netdev@vger.kernel.org
> ---
> include/linux/sunrpc/xdr.h | 2 -
> net/sunrpc/svcsock.c | 46 ++++++++++++++-------------------------------
> net/sunrpc/xdr.c | 19 ++++++++++--------
> net/sunrpc/xprtsock.c | 6 ++---
> 4 files changed, 30 insertions(+), 43 deletions(-)
>
> diff --git a/include/linux/sunrpc/xdr.h b/include/linux/sunrpc/xdr.h
> index 72014c9216fc..c74ea483228b 100644
> --- a/include/linux/sunrpc/xdr.h
> +++ b/include/linux/sunrpc/xdr.h
> @@ -137,7 +137,7 @@ void xdr_inline_pages(struct xdr_buf *, unsigned int,
> struct page **, unsigned int, unsigned int);
> void xdr_terminate_string(const struct xdr_buf *, const u32);
> size_t xdr_buf_pagecount(const struct xdr_buf *buf);
> -int xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp);
> +int xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp, unsigned int head, unsigned int tail);
> void xdr_free_bvec(struct xdr_buf *buf);
>
> static inline __be32 *xdr_encode_array(__be32 *p, const void *s, unsigned int len)
> diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
> index 14efcc08c6f8..e55761fe1ccf 100644
> --- a/net/sunrpc/svcsock.c
> +++ b/net/sunrpc/svcsock.c
> @@ -569,7 +569,7 @@ static int svc_udp_sendto(struct svc_rqst *rqstp)
> if (svc_xprt_is_dead(xprt))
> goto out_notconn;
>
> - err = xdr_alloc_bvec(xdr, GFP_KERNEL);
> + err = xdr_alloc_bvec(xdr, GFP_KERNEL, 0, 0);
> if (err < 0)
> goto out_unlock;
>
> @@ -1073,45 +1073,29 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
> {
> const struct kvec *head = xdr->head;
> const struct kvec *tail = xdr->tail;
> - struct kvec kv[2];
> - struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES | MSG_MORE, };
> - size_t sent;
> + struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES, };
> + size_t n;
> int ret;
>
> *sentp = 0;
> - ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
> + ret = xdr_alloc_bvec(xdr, GFP_KERNEL, 2, 1);
> if (ret < 0)
> return ret;
>
> - kv[0].iov_base = ▮
> - kv[0].iov_len = sizeof(marker);
> - kv[1] = *head;
> - iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, kv, 2, sizeof(marker) + head->iov_len);
> + n = 2 + xdr_buf_pagecount(xdr);
> + bvec_set_virt(&xdr->bvec[0], &marker, sizeof(marker));
> + bvec_set_virt(&xdr->bvec[1], head->iov_base, head->iov_len);
> + bvec_set_virt(&xdr->bvec[n], tail->iov_base, tail->iov_len);
> + if (tail->iov_len)
> + n++;
> + iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec, n,
> + sizeof(marker) + xdr->len);
> ret = sock_sendmsg(sock, &msg);
> if (ret < 0)
> return ret;
> - sent = ret;
> -
> - if (!tail->iov_len)
> - msg.msg_flags &= ~MSG_MORE;
> - iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec,
> - xdr_buf_pagecount(xdr), xdr->page_len);
> - ret = sock_sendmsg(sock, &msg);
> - if (ret < 0)
> - return ret;
> - sent += ret;
> -
> - if (tail->iov_len) {
> - msg.msg_flags &= ~MSG_MORE;
> - iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, tail, 1, tail->iov_len);
> - ret = sock_sendmsg(sock, &msg);
> - if (ret < 0)
> - return ret;
> - sent += ret;
> - }
> - if (sent > 0)
> - *sentp = sent;
> - if (sent != sizeof(marker) + xdr->len)
> + if (ret > 0)
> + *sentp = ret;
> + if (ret != sizeof(marker) + xdr->len)
> return -EAGAIN;
> return 0;
> }
> diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c
> index 36835b2f5446..695821963849 100644
> --- a/net/sunrpc/xdr.c
> +++ b/net/sunrpc/xdr.c
> @@ -141,18 +141,21 @@ size_t xdr_buf_pagecount(const struct xdr_buf *buf)
> }
>
> int
> -xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp)
> +xdr_alloc_bvec(struct xdr_buf *buf, gfp_t gfp, unsigned int head, unsigned int tail)
> {
> - size_t i, n = xdr_buf_pagecount(buf);
> + size_t i, j = 0, n = xdr_buf_pagecount(buf);
>
> - if (n != 0 && buf->bvec == NULL) {
> - buf->bvec = kmalloc_array(n, sizeof(buf->bvec[0]), gfp);
> + if (head + n + tail != 0 && buf->bvec == NULL) {
> + buf->bvec = kmalloc_array(head + n + tail,
> + sizeof(buf->bvec[0]), gfp);
> if (!buf->bvec)
> return -ENOMEM;
> - for (i = 0; i < n; i++) {
> - bvec_set_page(&buf->bvec[i], buf->pages[i], PAGE_SIZE,
> - 0);
> - }
> + for (i = 0; i < head; i++)
> + bvec_set_page(&buf->bvec[j++], NULL, 0, 0);
> + for (i = 0; i < n; i++)
> + bvec_set_page(&buf->bvec[j++], buf->pages[i], PAGE_SIZE, 0);
> + for (i = 0; i < tail; i++)
> + bvec_set_page(&buf->bvec[j++], NULL, 0, 0);
> }
> return 0;
> }
> diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
> index adcbedc244d6..fdf67e84b1c7 100644
> --- a/net/sunrpc/xprtsock.c
> +++ b/net/sunrpc/xprtsock.c
> @@ -825,7 +825,7 @@ static int xs_stream_nospace(struct rpc_rqst *req, bool vm_wait)
>
> static int xs_stream_prepare_request(struct rpc_rqst *req, struct xdr_buf *buf)
> {
> - return xdr_alloc_bvec(buf, rpc_task_gfp_mask());
> + return xdr_alloc_bvec(buf, rpc_task_gfp_mask(), 0, 0);
> }
>
> /*
> @@ -954,7 +954,7 @@ static int xs_udp_send_request(struct rpc_rqst *req)
> if (!xprt_request_get_cong(xprt, req))
> return -EBADSLT;
>
> - status = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
> + status = xdr_alloc_bvec(xdr, rpc_task_gfp_mask(), 0, 0);
> if (status < 0)
> return status;
> req->rq_xtime = ktime_get();
> @@ -2591,7 +2591,7 @@ static int bc_sendto(struct rpc_rqst *req)
> int err;
>
> req->rq_xtime = ktime_get();
> - err = xdr_alloc_bvec(xdr, rpc_task_gfp_mask());
> + err = xdr_alloc_bvec(xdr, rpc_task_gfp_mask(), 0, 0);
> if (err < 0)
> return err;
> err = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, marker, &sent);
>
--
Chuck Lever
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-30 13:16 ` David Howells
2023-03-30 13:27 ` Chuck Lever III
@ 2023-03-30 14:26 ` David Howells
2023-03-30 16:36 ` Chuck Lever III
1 sibling, 1 reply; 77+ messages in thread
From: David Howells @ 2023-03-30 14:26 UTC (permalink / raw)
To: Chuck Lever III
Cc: dhowells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Linus Torvalds,
open list:NETWORKING [GENERAL], linux-fsdevel,
Linux Kernel Mailing List, Linux Memory Management List,
Trond Myklebust, Anna Schumaker, Linux NFS Mailing List
Chuck Lever III <chuck.lever@oracle.com> wrote:
> Don't. Just change svc_tcp_send_kvec() to use sock_sendmsg, and
> leave the marker alone for now, please.
If you insist. See attached.
David
---
sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
When transmitting data, call down into TCP using sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather than
performing sendpage calls to transmit header, data pages and trailer.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Trond Myklebust <trond.myklebust@hammerspace.com>
cc: Anna Schumaker <anna@kernel.org>
cc: Chuck Lever <chuck.lever@oracle.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-nfs@vger.kernel.org
cc: netdev@vger.kernel.org
---
include/linux/sunrpc/svc.h | 11 +++++------
net/sunrpc/svcsock.c | 40 +++++++++++++---------------------------
2 files changed, 18 insertions(+), 33 deletions(-)
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index 877891536c2f..456ae554aa11 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -161,16 +161,15 @@ static inline bool svc_put_not_last(struct svc_serv *serv)
extern u32 svc_max_payload(const struct svc_rqst *rqstp);
/*
- * RPC Requsts and replies are stored in one or more pages.
+ * RPC Requests and replies are stored in one or more pages.
* We maintain an array of pages for each server thread.
* Requests are copied into these pages as they arrive. Remaining
* pages are available to write the reply into.
*
- * Pages are sent using ->sendpage so each server thread needs to
- * allocate more to replace those used in sending. To help keep track
- * of these pages we have a receive list where all pages initialy live,
- * and a send list where pages are moved to when there are to be part
- * of a reply.
+ * Pages are sent using ->sendmsg with MSG_SPLICE_PAGES so each server thread
+ * needs to allocate more to replace those used in sending. To help keep track
+ * of these pages we have a receive list where all pages initialy live, and a
+ * send list where pages are moved to when there are to be part of a reply.
*
* We use xdr_buf for holding responses as it fits well with NFS
* read responses (that have a header, and some data pages, and possibly
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 03a4f5615086..af146e053dfc 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1059,17 +1059,18 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
svc_xprt_received(rqstp->rq_xprt);
return 0; /* record not complete */
}
-
+
static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec,
int flags)
{
- return kernel_sendpage(sock, virt_to_page(vec->iov_base),
- offset_in_page(vec->iov_base),
- vec->iov_len, flags);
+ struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES | flags, };
+
+ iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, vec, 1, vec->iov_len);
+ return sock_sendmsg(sock, &msg);
}
/*
- * kernel_sendpage() is used exclusively to reduce the number of
+ * MSG_SPLICE_PAGES is used exclusively to reduce the number of
* copy operations in this path. Therefore the caller must ensure
* that the pages backing @xdr are unchanging.
*
@@ -1109,28 +1110,13 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
if (ret != head->iov_len)
goto out;
- if (xdr->page_len) {
- unsigned int offset, len, remaining;
- struct bio_vec *bvec;
-
- bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
- offset = offset_in_page(xdr->page_base);
- remaining = xdr->page_len;
- while (remaining > 0) {
- len = min(remaining, bvec->bv_len - offset);
- ret = kernel_sendpage(sock, bvec->bv_page,
- bvec->bv_offset + offset,
- len, 0);
- if (ret < 0)
- return ret;
- *sentp += ret;
- if (ret != len)
- goto out;
- remaining -= len;
- offset = 0;
- bvec++;
- }
- }
+ msg.msg_flags = MSG_SPLICE_PAGES;
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec,
+ xdr_buf_pagecount(xdr), xdr->page_len);
+ ret = sock_sendmsg(sock, &msg);
+ if (ret < 0)
+ return ret;
+ *sentp += ret;
if (tail->iov_len) {
ret = svc_tcp_send_kvec(sock, tail, 0);
^ permalink raw reply related [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-30 14:26 ` David Howells
@ 2023-03-30 16:36 ` Chuck Lever III
2023-04-14 14:41 ` Daire Byrne
0 siblings, 1 reply; 77+ messages in thread
From: Chuck Lever III @ 2023-03-30 16:36 UTC (permalink / raw)
To: David Howells
Cc: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, Al Viro, Christoph Hellwig, Jens Axboe, Jeff Layton,
Christian Brauner, Linus Torvalds, open list:NETWORKING [GENERAL],
linux-fsdevel, Linux Kernel Mailing List,
Linux Memory Management List, Trond Myklebust, Anna Schumaker,
Linux NFS Mailing List
> On Mar 30, 2023, at 10:26 AM, David Howells <dhowells@redhat.com> wrote:
>
> Chuck Lever III <chuck.lever@oracle.com> wrote:
>
>> Don't. Just change svc_tcp_send_kvec() to use sock_sendmsg, and
>> leave the marker alone for now, please.
>
> If you insist. See attached.
Very good, thank you for accommodating my regression paranoia.
Acked-by: Chuck Lever <chuck.lever@oracle.com>
>
> David
> ---
> sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
>
> When transmitting data, call down into TCP using sendmsg with
> MSG_SPLICE_PAGES to indicate that content should be spliced rather than
> performing sendpage calls to transmit header, data pages and trailer.
>
> Signed-off-by: David Howells <dhowells@redhat.com>
> cc: Trond Myklebust <trond.myklebust@hammerspace.com>
> cc: Anna Schumaker <anna@kernel.org>
> cc: Chuck Lever <chuck.lever@oracle.com>
> cc: Jeff Layton <jlayton@kernel.org>
> cc: "David S. Miller" <davem@davemloft.net>
> cc: Eric Dumazet <edumazet@google.com>
> cc: Jakub Kicinski <kuba@kernel.org>
> cc: Paolo Abeni <pabeni@redhat.com>
> cc: Jens Axboe <axboe@kernel.dk>
> cc: Matthew Wilcox <willy@infradead.org>
> cc: linux-nfs@vger.kernel.org
> cc: netdev@vger.kernel.org
> ---
> include/linux/sunrpc/svc.h | 11 +++++------
> net/sunrpc/svcsock.c | 40 +++++++++++++---------------------------
> 2 files changed, 18 insertions(+), 33 deletions(-)
>
> diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
> index 877891536c2f..456ae554aa11 100644
> --- a/include/linux/sunrpc/svc.h
> +++ b/include/linux/sunrpc/svc.h
> @@ -161,16 +161,15 @@ static inline bool svc_put_not_last(struct svc_serv *serv)
> extern u32 svc_max_payload(const struct svc_rqst *rqstp);
>
> /*
> - * RPC Requsts and replies are stored in one or more pages.
> + * RPC Requests and replies are stored in one or more pages.
> * We maintain an array of pages for each server thread.
> * Requests are copied into these pages as they arrive. Remaining
> * pages are available to write the reply into.
> *
> - * Pages are sent using ->sendpage so each server thread needs to
> - * allocate more to replace those used in sending. To help keep track
> - * of these pages we have a receive list where all pages initialy live,
> - * and a send list where pages are moved to when there are to be part
> - * of a reply.
> + * Pages are sent using ->sendmsg with MSG_SPLICE_PAGES so each server thread
> + * needs to allocate more to replace those used in sending. To help keep track
> + * of these pages we have a receive list where all pages initialy live, and a
> + * send list where pages are moved to when there are to be part of a reply.
> *
> * We use xdr_buf for holding responses as it fits well with NFS
> * read responses (that have a header, and some data pages, and possibly
> diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
> index 03a4f5615086..af146e053dfc 100644
> --- a/net/sunrpc/svcsock.c
> +++ b/net/sunrpc/svcsock.c
> @@ -1059,17 +1059,18 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
> svc_xprt_received(rqstp->rq_xprt);
> return 0; /* record not complete */
> }
> -
> +
> static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec,
> int flags)
> {
> - return kernel_sendpage(sock, virt_to_page(vec->iov_base),
> - offset_in_page(vec->iov_base),
> - vec->iov_len, flags);
> + struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES | flags, };
> +
> + iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, vec, 1, vec->iov_len);
> + return sock_sendmsg(sock, &msg);
> }
>
> /*
> - * kernel_sendpage() is used exclusively to reduce the number of
> + * MSG_SPLICE_PAGES is used exclusively to reduce the number of
> * copy operations in this path. Therefore the caller must ensure
> * that the pages backing @xdr are unchanging.
> *
> @@ -1109,28 +1110,13 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
> if (ret != head->iov_len)
> goto out;
>
> - if (xdr->page_len) {
> - unsigned int offset, len, remaining;
> - struct bio_vec *bvec;
> -
> - bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
> - offset = offset_in_page(xdr->page_base);
> - remaining = xdr->page_len;
> - while (remaining > 0) {
> - len = min(remaining, bvec->bv_len - offset);
> - ret = kernel_sendpage(sock, bvec->bv_page,
> - bvec->bv_offset + offset,
> - len, 0);
> - if (ret < 0)
> - return ret;
> - *sentp += ret;
> - if (ret != len)
> - goto out;
> - remaining -= len;
> - offset = 0;
> - bvec++;
> - }
> - }
> + msg.msg_flags = MSG_SPLICE_PAGES;
> + iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec,
> + xdr_buf_pagecount(xdr), xdr->page_len);
> + ret = sock_sendmsg(sock, &msg);
> + if (ret < 0)
> + return ret;
> + *sentp += ret;
>
> if (tail->iov_len) {
> ret = svc_tcp_send_kvec(sock, tail, 0);
>
--
Chuck Lever
^ permalink raw reply [flat|nested] 77+ messages in thread* Re: [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-30 16:36 ` Chuck Lever III
@ 2023-04-14 14:41 ` Daire Byrne
0 siblings, 0 replies; 77+ messages in thread
From: Daire Byrne @ 2023-04-14 14:41 UTC (permalink / raw)
To: Chuck Lever III
Cc: David Howells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Al Viro, Christoph Hellwig,
Jens Axboe, Jeff Layton, Christian Brauner, Linus Torvalds,
open list:NETWORKING [GENERAL], linux-fsdevel,
Linux Kernel Mailing List, Linux Memory Management List,
Trond Myklebust, Anna Schumaker, Linux NFS Mailing List
I gave this a spin because I had noticed a previous regression around
the 5.7 time frame in sendpage/sendmsg code changes:
https://bugzilla.kernel.org/show_bug.cgi?id=209439
In that case there was a noticeable regression in performance for high
performance servers (100gbit).
I see no such performance problems with David's iov-sendpage branch
and it all looks good to me with simple benchmarks (100gbit server,
100 x 1gbit clients reading data).
Tested-by: Daire Byrne <daire@dneg.com>
Cheers,
Daire
On Thu, 30 Mar 2023 at 17:37, Chuck Lever III <chuck.lever@oracle.com> wrote:
>
>
>
> > On Mar 30, 2023, at 10:26 AM, David Howells <dhowells@redhat.com> wrote:
> >
> > Chuck Lever III <chuck.lever@oracle.com> wrote:
> >
> >> Don't. Just change svc_tcp_send_kvec() to use sock_sendmsg, and
> >> leave the marker alone for now, please.
> >
> > If you insist. See attached.
>
> Very good, thank you for accommodating my regression paranoia.
>
> Acked-by: Chuck Lever <chuck.lever@oracle.com>
>
>
> >
> > David
> > ---
> > sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
> >
> > When transmitting data, call down into TCP using sendmsg with
> > MSG_SPLICE_PAGES to indicate that content should be spliced rather than
> > performing sendpage calls to transmit header, data pages and trailer.
> >
> > Signed-off-by: David Howells <dhowells@redhat.com>
> > cc: Trond Myklebust <trond.myklebust@hammerspace.com>
> > cc: Anna Schumaker <anna@kernel.org>
> > cc: Chuck Lever <chuck.lever@oracle.com>
> > cc: Jeff Layton <jlayton@kernel.org>
> > cc: "David S. Miller" <davem@davemloft.net>
> > cc: Eric Dumazet <edumazet@google.com>
> > cc: Jakub Kicinski <kuba@kernel.org>
> > cc: Paolo Abeni <pabeni@redhat.com>
> > cc: Jens Axboe <axboe@kernel.dk>
> > cc: Matthew Wilcox <willy@infradead.org>
> > cc: linux-nfs@vger.kernel.org
> > cc: netdev@vger.kernel.org
> > ---
> > include/linux/sunrpc/svc.h | 11 +++++------
> > net/sunrpc/svcsock.c | 40 +++++++++++++---------------------------
> > 2 files changed, 18 insertions(+), 33 deletions(-)
> >
> > diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
> > index 877891536c2f..456ae554aa11 100644
> > --- a/include/linux/sunrpc/svc.h
> > +++ b/include/linux/sunrpc/svc.h
> > @@ -161,16 +161,15 @@ static inline bool svc_put_not_last(struct svc_serv *serv)
> > extern u32 svc_max_payload(const struct svc_rqst *rqstp);
> >
> > /*
> > - * RPC Requsts and replies are stored in one or more pages.
> > + * RPC Requests and replies are stored in one or more pages.
> > * We maintain an array of pages for each server thread.
> > * Requests are copied into these pages as they arrive. Remaining
> > * pages are available to write the reply into.
> > *
> > - * Pages are sent using ->sendpage so each server thread needs to
> > - * allocate more to replace those used in sending. To help keep track
> > - * of these pages we have a receive list where all pages initialy live,
> > - * and a send list where pages are moved to when there are to be part
> > - * of a reply.
> > + * Pages are sent using ->sendmsg with MSG_SPLICE_PAGES so each server thread
> > + * needs to allocate more to replace those used in sending. To help keep track
> > + * of these pages we have a receive list where all pages initialy live, and a
> > + * send list where pages are moved to when there are to be part of a reply.
> > *
> > * We use xdr_buf for holding responses as it fits well with NFS
> > * read responses (that have a header, and some data pages, and possibly
> > diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
> > index 03a4f5615086..af146e053dfc 100644
> > --- a/net/sunrpc/svcsock.c
> > +++ b/net/sunrpc/svcsock.c
> > @@ -1059,17 +1059,18 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
> > svc_xprt_received(rqstp->rq_xprt);
> > return 0; /* record not complete */
> > }
> > -
> > +
> > static int svc_tcp_send_kvec(struct socket *sock, const struct kvec *vec,
> > int flags)
> > {
> > - return kernel_sendpage(sock, virt_to_page(vec->iov_base),
> > - offset_in_page(vec->iov_base),
> > - vec->iov_len, flags);
> > + struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES | flags, };
> > +
> > + iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, vec, 1, vec->iov_len);
> > + return sock_sendmsg(sock, &msg);
> > }
> >
> > /*
> > - * kernel_sendpage() is used exclusively to reduce the number of
> > + * MSG_SPLICE_PAGES is used exclusively to reduce the number of
> > * copy operations in this path. Therefore the caller must ensure
> > * that the pages backing @xdr are unchanging.
> > *
> > @@ -1109,28 +1110,13 @@ static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
> > if (ret != head->iov_len)
> > goto out;
> >
> > - if (xdr->page_len) {
> > - unsigned int offset, len, remaining;
> > - struct bio_vec *bvec;
> > -
> > - bvec = xdr->bvec + (xdr->page_base >> PAGE_SHIFT);
> > - offset = offset_in_page(xdr->page_base);
> > - remaining = xdr->page_len;
> > - while (remaining > 0) {
> > - len = min(remaining, bvec->bv_len - offset);
> > - ret = kernel_sendpage(sock, bvec->bv_page,
> > - bvec->bv_offset + offset,
> > - len, 0);
> > - if (ret < 0)
> > - return ret;
> > - *sentp += ret;
> > - if (ret != len)
> > - goto out;
> > - remaining -= len;
> > - offset = 0;
> > - bvec++;
> > - }
> > - }
> > + msg.msg_flags = MSG_SPLICE_PAGES;
> > + iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, xdr->bvec,
> > + xdr_buf_pagecount(xdr), xdr->page_len);
> > + ret = sock_sendmsg(sock, &msg);
> > + if (ret < 0)
> > + return ret;
> > + *sentp += ret;
> >
> > if (tail->iov_len) {
> > ret = svc_tcp_send_kvec(sock, tail, 0);
> >
>
> --
> Chuck Lever
>
>
^ permalink raw reply [flat|nested] 77+ messages in thread
* [RFC PATCH v2 41/48] sunrpc: Rely on TCP sendmsg + MSG_SPLICE_PAGES to copy unspliceable data
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (39 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 40/48] sunrpc: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 42/48] nvme: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
` (6 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Trond Myklebust,
Anna Schumaker, linux-nfs
Rather than copying data in svc_tcp_sendmsg() into page fragments, just
hand in ITER_KVEC iterators as part of the ITER_ITERLIST and rely on TCP to
copy them if the pages they're residing on are belong to the slab or have a
zero refcount.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Trond Myklebust <trond.myklebust@hammerspace.com>
cc: Anna Schumaker <anna@kernel.org>
cc: Chuck Lever <chuck.lever@oracle.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-nfs@vger.kernel.org
cc: netdev@vger.kernel.org
---
net/sunrpc/svcsock.c | 44 ++++++++++++--------------------------------
1 file changed, 12 insertions(+), 32 deletions(-)
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index f1cc53aad6e0..c1421f6fe57a 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1071,47 +1071,27 @@ static int svc_tcp_recvfrom(struct svc_rqst *rqstp)
static int svc_tcp_sendmsg(struct socket *sock, struct xdr_buf *xdr,
rpc_fraghdr marker, unsigned int *sentp)
{
- const struct kvec *head = xdr->head;
- const struct kvec *tail = xdr->tail;
- struct iov_iter iters[3];
- struct bio_vec head_bv, tail_bv;
- struct msghdr msg = {
- .msg_flags = MSG_SPLICE_PAGES,
- };
- void *m, *t;
- int ret, n = 2, size;
+ struct iov_iter iters[4];
+ struct kvec marker_kv;
+ struct msghdr msg = { .msg_flags = MSG_SPLICE_PAGES, };
+ int ret, n = 0, size;
*sentp = 0;
ret = xdr_alloc_bvec(xdr, GFP_KERNEL);
if (ret < 0)
return ret;
- m = page_frag_alloc(NULL, sizeof(marker) + head->iov_len + tail->iov_len,
- GFP_KERNEL);
- if (!m)
- return -ENOMEM;
-
- memcpy(m, &marker, sizeof(marker));
- if (head->iov_len)
- memcpy(m + sizeof(marker), head->iov_base, head->iov_len);
- bvec_set_virt(&head_bv, m, sizeof(marker) + head->iov_len);
- iov_iter_bvec(&iters[0], ITER_SOURCE, &head_bv, 1,
- sizeof(marker) + head->iov_len);
-
- iov_iter_bvec(&iters[1], ITER_SOURCE, xdr->bvec,
+ marker_kv.iov_base = ▮
+ marker_kv.iov_len = sizeof(marker);
+ iov_iter_kvec(&iters[n++], ITER_SOURCE, &marker_kv, 1, sizeof(marker));
+ iov_iter_kvec(&iters[n++], ITER_SOURCE, xdr->head, 1, xdr->head->iov_len);
+ iov_iter_bvec(&iters[n++], ITER_SOURCE, xdr->bvec,
xdr_buf_pagecount(xdr), xdr->page_len);
- if (tail->iov_len) {
- t = page_frag_alloc(NULL, tail->iov_len, GFP_KERNEL);
- if (!t)
- return -ENOMEM;
- memcpy(t, tail->iov_base, tail->iov_len);
- bvec_set_virt(&tail_bv, t, tail->iov_len);
- iov_iter_bvec(&iters[2], ITER_SOURCE, &tail_bv, 1, tail->iov_len);
- n++;
- }
+ if (xdr->tail->iov_len)
+ iov_iter_kvec(&iters[n++], ITER_SOURCE, xdr->tail, 1, xdr->tail->iov_len);
- size = sizeof(marker) + head->iov_len + xdr->page_len + tail->iov_len;
+ size = sizeof(marker) + xdr->head->iov_len + xdr->page_len + xdr->tail->iov_len;
iov_iter_iterlist(&msg.msg_iter, ITER_SOURCE, iters, n, size);
ret = sock_sendmsg(sock, &msg);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 42/48] nvme: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (40 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 41/48] sunrpc: Rely on TCP sendmsg + MSG_SPLICE_PAGES to copy unspliceable data David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 43/48] kcm: " David Howells
` (5 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Keith Busch,
Jens Axboe, Christoph Hellwig, Sagi Grimberg, Chaitanya Kulkarni,
linux-nvme
When transmitting data, call down into TCP using a single sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather than
performing several sendmsg and sendpage calls to transmit header, data
pages and trailer.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Keith Busch <kbusch@kernel.org>
cc: Jens Axboe <axboe@fb.com>
cc: Christoph Hellwig <hch@lst.de>
cc: Sagi Grimberg <sagi@grimberg.me>
cc: Chaitanya Kulkarni <kch@nvidia.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-nvme@lists.infradead.org
cc: netdev@vger.kernel.org
---
drivers/nvme/host/tcp.c | 44 ++++++++++++++++++------------------
drivers/nvme/target/tcp.c | 47 +++++++++++++++++++++++++--------------
2 files changed, 52 insertions(+), 39 deletions(-)
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index fa32969b532f..cc617692702d 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -979,25 +979,23 @@ static int nvme_tcp_try_send_data(struct nvme_tcp_request *req)
u32 h2cdata_left = req->h2cdata_left;
while (true) {
+ struct bio_vec bvec;
+ struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_SPLICE_PAGES, };
struct page *page = nvme_tcp_req_cur_page(req);
size_t offset = nvme_tcp_req_cur_offset(req);
size_t len = nvme_tcp_req_cur_length(req);
bool last = nvme_tcp_pdu_last_send(req, len);
int req_data_sent = req->data_sent;
- int ret, flags = MSG_DONTWAIT;
+ int ret;
if (last && !queue->data_digest && !nvme_tcp_queue_more(queue))
- flags |= MSG_EOR;
+ msg.msg_flags |= MSG_EOR;
else
- flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
+ msg.msg_flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
- if (sendpage_ok(page)) {
- ret = kernel_sendpage(queue->sock, page, offset, len,
- flags);
- } else {
- ret = sock_no_sendpage(queue->sock, page, offset, len,
- flags);
- }
+ bvec_set_page(&bvec, page, len, offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, len);
+ ret = sock_sendmsg(queue->sock, &msg);
if (ret <= 0)
return ret;
@@ -1036,22 +1034,24 @@ static int nvme_tcp_try_send_cmd_pdu(struct nvme_tcp_request *req)
{
struct nvme_tcp_queue *queue = req->queue;
struct nvme_tcp_cmd_pdu *pdu = req->pdu;
+ struct bio_vec bvec;
+ struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_SPLICE_PAGES, };
bool inline_data = nvme_tcp_has_inline_data(req);
u8 hdgst = nvme_tcp_hdgst_len(queue);
int len = sizeof(*pdu) + hdgst - req->offset;
- int flags = MSG_DONTWAIT;
int ret;
if (inline_data || nvme_tcp_queue_more(queue))
- flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
+ msg.msg_flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
else
- flags |= MSG_EOR;
+ msg.msg_flags |= MSG_EOR;
if (queue->hdr_digest && !req->offset)
nvme_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
- ret = kernel_sendpage(queue->sock, virt_to_page(pdu),
- offset_in_page(pdu) + req->offset, len, flags);
+ bvec_set_virt(&bvec, (void *)pdu + req->offset, len);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, len);
+ ret = sock_sendmsg(queue->sock, &msg);
if (unlikely(ret <= 0))
return ret;
@@ -1075,6 +1075,8 @@ static int nvme_tcp_try_send_data_pdu(struct nvme_tcp_request *req)
{
struct nvme_tcp_queue *queue = req->queue;
struct nvme_tcp_data_pdu *pdu = req->pdu;
+ struct bio_vec bvec;
+ struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_MORE, };
u8 hdgst = nvme_tcp_hdgst_len(queue);
int len = sizeof(*pdu) - req->offset + hdgst;
int ret;
@@ -1083,13 +1085,11 @@ static int nvme_tcp_try_send_data_pdu(struct nvme_tcp_request *req)
nvme_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
if (!req->h2cdata_left)
- ret = kernel_sendpage(queue->sock, virt_to_page(pdu),
- offset_in_page(pdu) + req->offset, len,
- MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST);
- else
- ret = sock_no_sendpage(queue->sock, virt_to_page(pdu),
- offset_in_page(pdu) + req->offset, len,
- MSG_DONTWAIT | MSG_MORE);
+ msg.msg_flags |= MSG_SPLICE_PAGES | MSG_SENDPAGE_NOTLAST;
+
+ bvec_set_virt(&bvec, (void *)pdu + req->offset, len);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, len);
+ ret = sock_sendmsg(queue->sock, &msg);
if (unlikely(ret <= 0))
return ret;
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index d6cc557cc539..00b491abf50f 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -548,13 +548,18 @@ static void nvmet_tcp_execute_request(struct nvmet_tcp_cmd *cmd)
static int nvmet_try_send_data_pdu(struct nvmet_tcp_cmd *cmd)
{
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = (MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST |
+ MSG_SPLICE_PAGES),
+ };
u8 hdgst = nvmet_tcp_hdgst_len(cmd->queue);
int left = sizeof(*cmd->data_pdu) - cmd->offset + hdgst;
int ret;
- ret = kernel_sendpage(cmd->queue->sock, virt_to_page(cmd->data_pdu),
- offset_in_page(cmd->data_pdu) + cmd->offset,
- left, MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST);
+ bvec_set_virt(&bvec, (void *)cmd->data_pdu + cmd->offset, left);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, left);
+ ret = sock_sendmsg(cmd->queue->sock, &msg);
if (ret <= 0)
return ret;
@@ -575,17 +580,21 @@ static int nvmet_try_send_data(struct nvmet_tcp_cmd *cmd, bool last_in_batch)
int ret;
while (cmd->cur_sg) {
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = MSG_DONTWAIT | MSG_SPLICE_PAGES,
+ };
struct page *page = sg_page(cmd->cur_sg);
u32 left = cmd->cur_sg->length - cmd->offset;
- int flags = MSG_DONTWAIT;
if ((!last_in_batch && cmd->queue->send_list_len) ||
cmd->wbytes_done + left < cmd->req.transfer_len ||
queue->data_digest || !queue->nvme_sq.sqhd_disabled)
- flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
+ msg.msg_flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
- ret = kernel_sendpage(cmd->queue->sock, page, cmd->offset,
- left, flags);
+ bvec_set_page(&bvec, page, left, cmd->offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, left);
+ ret = sock_sendmsg(cmd->queue->sock, &msg);
if (ret <= 0)
return ret;
@@ -621,18 +630,20 @@ static int nvmet_try_send_data(struct nvmet_tcp_cmd *cmd, bool last_in_batch)
static int nvmet_try_send_response(struct nvmet_tcp_cmd *cmd,
bool last_in_batch)
{
+ struct bio_vec bvec;
+ struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_SPLICE_PAGES, };
u8 hdgst = nvmet_tcp_hdgst_len(cmd->queue);
int left = sizeof(*cmd->rsp_pdu) - cmd->offset + hdgst;
- int flags = MSG_DONTWAIT;
int ret;
if (!last_in_batch && cmd->queue->send_list_len)
- flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
+ msg.msg_flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
else
- flags |= MSG_EOR;
+ msg.msg_flags |= MSG_EOR;
- ret = kernel_sendpage(cmd->queue->sock, virt_to_page(cmd->rsp_pdu),
- offset_in_page(cmd->rsp_pdu) + cmd->offset, left, flags);
+ bvec_set_virt(&bvec, (void *)cmd->rsp_pdu + cmd->offset, left);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, left);
+ ret = sock_sendmsg(cmd->queue->sock, &msg);
if (ret <= 0)
return ret;
cmd->offset += ret;
@@ -649,18 +660,20 @@ static int nvmet_try_send_response(struct nvmet_tcp_cmd *cmd,
static int nvmet_try_send_r2t(struct nvmet_tcp_cmd *cmd, bool last_in_batch)
{
+ struct bio_vec bvec;
+ struct msghdr msg = { .msg_flags = MSG_DONTWAIT | MSG_SPLICE_PAGES, };
u8 hdgst = nvmet_tcp_hdgst_len(cmd->queue);
int left = sizeof(*cmd->r2t_pdu) - cmd->offset + hdgst;
- int flags = MSG_DONTWAIT;
int ret;
if (!last_in_batch && cmd->queue->send_list_len)
- flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
+ msg.msg_flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
else
- flags |= MSG_EOR;
+ msg.msg_flags |= MSG_EOR;
- ret = kernel_sendpage(cmd->queue->sock, virt_to_page(cmd->r2t_pdu),
- offset_in_page(cmd->r2t_pdu) + cmd->offset, left, flags);
+ bvec_set_virt(&bvec, (void *)cmd->r2t_pdu + cmd->offset, left);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, left);
+ ret = sock_sendmsg(cmd->queue->sock, &msg);
if (ret <= 0)
return ret;
cmd->offset += ret;
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 43/48] kcm: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (41 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 42/48] nvme: Use sendmsg(MSG_SPLICE_PAGES) rather then sendpage David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 44/48] smc: Drop smc_sendpage() in favour of smc_sendmsg() + MSG_SPLICE_PAGES David Howells
` (4 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Tom Herbert,
Tom Herbert
When transmitting data, call down into TCP using a single sendmsg with
MSG_SPLICE_PAGES to indicate that content should be spliced rather than
performing several sendmsg and sendpage calls to transmit header, data
pages and trailer.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Tom Herbert <tom@herbertland.com>
cc: Tom Herbert <tom@quantonium.net>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: netdev@vger.kernel.org
---
net/kcm/kcmsock.c | 15 ++++++++++-----
1 file changed, 10 insertions(+), 5 deletions(-)
diff --git a/net/kcm/kcmsock.c b/net/kcm/kcmsock.c
index cfe828bd7fc6..819259149952 100644
--- a/net/kcm/kcmsock.c
+++ b/net/kcm/kcmsock.c
@@ -641,6 +641,10 @@ static int kcm_write_msgs(struct kcm_sock *kcm)
for (fragidx = 0; fragidx < skb_shinfo(skb)->nr_frags;
fragidx++) {
+ struct bio_vec bvec;
+ struct msghdr msg = {
+ .msg_flags = MSG_DONTWAIT | MSG_SPLICE_PAGES,
+ };
skb_frag_t *frag;
frag_offset = 0;
@@ -651,11 +655,12 @@ static int kcm_write_msgs(struct kcm_sock *kcm)
goto out;
}
- ret = kernel_sendpage(psock->sk->sk_socket,
- skb_frag_page(frag),
- skb_frag_off(frag) + frag_offset,
- skb_frag_size(frag) - frag_offset,
- MSG_DONTWAIT);
+ bvec_set_page(&bvec,
+ skb_frag_page(frag),
+ skb_frag_size(frag) - frag_offset,
+ skb_frag_off(frag) + frag_offset);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, bvec.bv_len);
+ ret = sock_sendmsg(psock->sk->sk_socket, &msg);
if (ret <= 0) {
if (ret == -EAGAIN) {
/* Save state to try again when there's
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 44/48] smc: Drop smc_sendpage() in favour of smc_sendmsg() + MSG_SPLICE_PAGES
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (42 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 43/48] kcm: " David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 45/48] ocfs2: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage() David Howells
` (3 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Karsten Graul,
Wenjia Zhang, Jan Karcher, linux-s390
Drop the smc_sendpage() code as smc_sendmsg() just passes the call down to
the underlying TCP socket and smc_tx_sendpage() is just a wrapper around
its sendmsg implementation.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Karsten Graul <kgraul@linux.ibm.com>
cc: Wenjia Zhang <wenjia@linux.ibm.com>
cc: Jan Karcher <jaka@linux.ibm.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
cc: linux-s390@vger.kernel.org
cc: netdev@vger.kernel.org
---
net/smc/af_smc.c | 29 -----------------------------
net/smc/smc_stats.c | 2 +-
net/smc/smc_stats.h | 1 -
net/smc/smc_tx.c | 16 ----------------
net/smc/smc_tx.h | 2 --
5 files changed, 1 insertion(+), 49 deletions(-)
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index a4cccdfdc00a..d4113c8a7cda 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -3125,34 +3125,6 @@ static int smc_ioctl(struct socket *sock, unsigned int cmd,
return put_user(answ, (int __user *)arg);
}
-static ssize_t smc_sendpage(struct socket *sock, struct page *page,
- int offset, size_t size, int flags)
-{
- struct sock *sk = sock->sk;
- struct smc_sock *smc;
- int rc = -EPIPE;
-
- smc = smc_sk(sk);
- lock_sock(sk);
- if (sk->sk_state != SMC_ACTIVE) {
- release_sock(sk);
- goto out;
- }
- release_sock(sk);
- if (smc->use_fallback) {
- rc = kernel_sendpage(smc->clcsock, page, offset,
- size, flags);
- } else {
- lock_sock(sk);
- rc = smc_tx_sendpage(smc, page, offset, size, flags);
- release_sock(sk);
- SMC_STAT_INC(smc, sendpage_cnt);
- }
-
-out:
- return rc;
-}
-
/* Map the affected portions of the rmbe into an spd, note the number of bytes
* to splice in conn->splice_pending, and press 'go'. Delays consumer cursor
* updates till whenever a respective page has been fully processed.
@@ -3224,7 +3196,6 @@ static const struct proto_ops smc_sock_ops = {
.sendmsg = smc_sendmsg,
.recvmsg = smc_recvmsg,
.mmap = sock_no_mmap,
- .sendpage = smc_sendpage,
.splice_read = smc_splice_read,
};
diff --git a/net/smc/smc_stats.c b/net/smc/smc_stats.c
index e80e34f7ac15..ca14c0f3a07d 100644
--- a/net/smc/smc_stats.c
+++ b/net/smc/smc_stats.c
@@ -227,7 +227,7 @@ static int smc_nl_fill_stats_tech_data(struct sk_buff *skb,
SMC_NLA_STATS_PAD))
goto errattr;
if (nla_put_u64_64bit(skb, SMC_NLA_STATS_T_SENDPAGE_CNT,
- smc_tech->sendpage_cnt,
+ 0,
SMC_NLA_STATS_PAD))
goto errattr;
if (nla_put_u64_64bit(skb, SMC_NLA_STATS_T_CORK_CNT,
diff --git a/net/smc/smc_stats.h b/net/smc/smc_stats.h
index 84b7ecd8c05c..b60fe1eb37ab 100644
--- a/net/smc/smc_stats.h
+++ b/net/smc/smc_stats.h
@@ -71,7 +71,6 @@ struct smc_stats_tech {
u64 clnt_v2_succ_cnt;
u64 srv_v1_succ_cnt;
u64 srv_v2_succ_cnt;
- u64 sendpage_cnt;
u64 urg_data_cnt;
u64 splice_cnt;
u64 cork_cnt;
diff --git a/net/smc/smc_tx.c b/net/smc/smc_tx.c
index f4b6a71ac488..d31ce8209fa2 100644
--- a/net/smc/smc_tx.c
+++ b/net/smc/smc_tx.c
@@ -298,22 +298,6 @@ int smc_tx_sendmsg(struct smc_sock *smc, struct msghdr *msg, size_t len)
return rc;
}
-int smc_tx_sendpage(struct smc_sock *smc, struct page *page, int offset,
- size_t size, int flags)
-{
- struct msghdr msg = {.msg_flags = flags};
- char *kaddr = kmap(page);
- struct kvec iov;
- int rc;
-
- iov.iov_base = kaddr + offset;
- iov.iov_len = size;
- iov_iter_kvec(&msg.msg_iter, ITER_SOURCE, &iov, 1, size);
- rc = smc_tx_sendmsg(smc, &msg, size);
- kunmap(page);
- return rc;
-}
-
/***************************** sndbuf consumer *******************************/
/* sndbuf consumer: actual data transfer of one target chunk with ISM write */
diff --git a/net/smc/smc_tx.h b/net/smc/smc_tx.h
index 34b578498b1f..a59f370b8b43 100644
--- a/net/smc/smc_tx.h
+++ b/net/smc/smc_tx.h
@@ -31,8 +31,6 @@ void smc_tx_pending(struct smc_connection *conn);
void smc_tx_work(struct work_struct *work);
void smc_tx_init(struct smc_sock *smc);
int smc_tx_sendmsg(struct smc_sock *smc, struct msghdr *msg, size_t len);
-int smc_tx_sendpage(struct smc_sock *smc, struct page *page, int offset,
- size_t size, int flags);
int smc_tx_sndbuf_nonempty(struct smc_connection *conn);
void smc_tx_sndbuf_nonfull(struct smc_sock *smc);
void smc_tx_consumer_update(struct smc_connection *conn, bool force);
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 45/48] ocfs2: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (43 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 44/48] smc: Drop smc_sendpage() in favour of smc_sendmsg() + MSG_SPLICE_PAGES David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 46/48] drbd: Use sendmsg(MSG_SPLICE_PAGES) rather than sendmsg() David Howells
` (2 subsequent siblings)
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Mark Fasheh,
Joel Becker, Joseph Qi, ocfs2-devel
Fix ocfs2 to use the page fragment allocator rather than kzalloc in order
to allocate the buffers for the handshake message and keepalive request and
reply messages. Slab pages should not be given to sendpage, but fragments
can be.
Switch from using sendpage() to using sendmsg() + MSG_SPLICE_PAGES so that
sendpage can be phased out.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Mark Fasheh <mark@fasheh.com>
cc: Joel Becker <jlbec@evilplan.org>
cc: Joseph Qi <joseph.qi@linux.alibaba.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: ocfs2-devel@oss.oracle.com
cc: netdev@vger.kernel.org
---
fs/ocfs2/cluster/tcp.c | 107 ++++++++++++++++++++++-------------------
1 file changed, 58 insertions(+), 49 deletions(-)
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index aecbd712a00c..e568ad2f34bf 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -110,9 +110,6 @@ static struct work_struct o2net_listen_work;
static struct o2hb_callback_func o2net_hb_up, o2net_hb_down;
#define O2NET_HB_PRI 0x1
-static struct o2net_handshake *o2net_hand;
-static struct o2net_msg *o2net_keep_req, *o2net_keep_resp;
-
static int o2net_sys_err_translations[O2NET_ERR_MAX] =
{[O2NET_ERR_NONE] = 0,
[O2NET_ERR_NO_HNDLR] = -ENOPROTOOPT,
@@ -930,19 +927,22 @@ static int o2net_send_tcp_msg(struct socket *sock, struct kvec *vec,
}
static void o2net_sendpage(struct o2net_sock_container *sc,
- void *kmalloced_virt,
- size_t size)
+ void *virt, size_t size)
{
struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
+ struct msghdr msg = {};
+ struct bio_vec bv;
ssize_t ret;
+ bvec_set_virt(&bv, virt, size);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bv, 1, size);
+
while (1) {
+ msg.msg_flags = MSG_DONTWAIT | MSG_SPLICE_PAGES;
mutex_lock(&sc->sc_send_lock);
- ret = sc->sc_sock->ops->sendpage(sc->sc_sock,
- virt_to_page(kmalloced_virt),
- offset_in_page(kmalloced_virt),
- size, MSG_DONTWAIT);
+ ret = sock_sendmsg(sc->sc_sock, &msg);
mutex_unlock(&sc->sc_send_lock);
+
if (ret == size)
break;
if (ret == (ssize_t)-EAGAIN) {
@@ -1168,6 +1168,7 @@ static int o2net_process_message(struct o2net_sock_container *sc,
struct o2net_msg *hdr)
{
struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num);
+ struct o2net_msg *keep_resp;
int ret = 0, handler_status;
enum o2net_system_error syserr;
struct o2net_msg_handler *nmh = NULL;
@@ -1186,8 +1187,16 @@ static int o2net_process_message(struct o2net_sock_container *sc,
be32_to_cpu(hdr->status));
goto out;
case O2NET_MSG_KEEP_REQ_MAGIC:
- o2net_sendpage(sc, o2net_keep_resp,
- sizeof(*o2net_keep_resp));
+ keep_resp = page_frag_alloc(NULL, sizeof(*keep_resp),
+ GFP_KERNEL);
+ if (!keep_resp) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ memset(keep_resp, 0, sizeof(*keep_resp));
+ keep_resp->magic = cpu_to_be16(O2NET_MSG_KEEP_RESP_MAGIC);
+ o2net_sendpage(sc, keep_resp, sizeof(*keep_resp));
+ folio_put(virt_to_folio(keep_resp));
goto out;
case O2NET_MSG_KEEP_RESP_MAGIC:
goto out;
@@ -1439,15 +1448,22 @@ static void o2net_rx_until_empty(struct work_struct *work)
sc_put(sc);
}
-static void o2net_initialize_handshake(void)
+static struct o2net_handshake *o2net_initialize_handshake(void)
{
- o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
- O2HB_MAX_WRITE_TIMEOUT_MS);
- o2net_hand->o2net_idle_timeout_ms = cpu_to_be32(o2net_idle_timeout());
- o2net_hand->o2net_keepalive_delay_ms = cpu_to_be32(
- o2net_keepalive_delay());
- o2net_hand->o2net_reconnect_delay_ms = cpu_to_be32(
- o2net_reconnect_delay());
+ struct o2net_handshake *hand;
+
+ hand = page_frag_alloc(NULL, sizeof(*hand), GFP_KERNEL);
+ if (!hand)
+ return NULL;
+
+ memset(hand, 0, sizeof(*hand));
+ hand->protocol_version = cpu_to_be64(O2NET_PROTOCOL_VERSION);
+ hand->connector_id = cpu_to_be64(1);
+ hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(O2HB_MAX_WRITE_TIMEOUT_MS);
+ hand->o2net_idle_timeout_ms = cpu_to_be32(o2net_idle_timeout());
+ hand->o2net_keepalive_delay_ms = cpu_to_be32(o2net_keepalive_delay());
+ hand->o2net_reconnect_delay_ms = cpu_to_be32(o2net_reconnect_delay());
+ return hand;
}
/* ------------------------------------------------------------ */
@@ -1456,16 +1472,22 @@ static void o2net_initialize_handshake(void)
* rx path will see the response and mark the sc valid */
static void o2net_sc_connect_completed(struct work_struct *work)
{
+ struct o2net_handshake *hand;
struct o2net_sock_container *sc =
container_of(work, struct o2net_sock_container,
sc_connect_work);
+ hand = o2net_initialize_handshake();
+ if (!hand)
+ goto out;
+
mlog(ML_MSG, "sc sending handshake with ver %llu id %llx\n",
(unsigned long long)O2NET_PROTOCOL_VERSION,
- (unsigned long long)be64_to_cpu(o2net_hand->connector_id));
+ (unsigned long long)be64_to_cpu(hand->connector_id));
- o2net_initialize_handshake();
- o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand));
+ o2net_sendpage(sc, hand, sizeof(*hand));
+ folio_put(virt_to_folio(hand));
+out:
sc_put(sc);
}
@@ -1475,8 +1497,15 @@ static void o2net_sc_send_keep_req(struct work_struct *work)
struct o2net_sock_container *sc =
container_of(work, struct o2net_sock_container,
sc_keepalive_work.work);
+ struct o2net_msg *keep_req;
- o2net_sendpage(sc, o2net_keep_req, sizeof(*o2net_keep_req));
+ keep_req = page_frag_alloc(NULL, sizeof(*keep_req), GFP_KERNEL);
+ if (keep_req) {
+ memset(keep_req, 0, sizeof(*keep_req));
+ keep_req->magic = cpu_to_be16(O2NET_MSG_KEEP_REQ_MAGIC);
+ o2net_sendpage(sc, keep_req, sizeof(*keep_req));
+ folio_put(virt_to_folio(keep_req));
+ }
sc_put(sc);
}
@@ -1780,6 +1809,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
struct socket *new_sock = NULL;
struct o2nm_node *node = NULL;
struct o2nm_node *local_node = NULL;
+ struct o2net_handshake *hand;
struct o2net_sock_container *sc = NULL;
struct o2net_node *nn;
unsigned int nofs_flag;
@@ -1882,8 +1912,11 @@ static int o2net_accept_one(struct socket *sock, int *more)
o2net_register_callbacks(sc->sc_sock->sk, sc);
o2net_sc_queue_work(sc, &sc->sc_rx_work);
- o2net_initialize_handshake();
- o2net_sendpage(sc, o2net_hand, sizeof(*o2net_hand));
+ hand = o2net_initialize_handshake();
+ if (hand) {
+ o2net_sendpage(sc, hand, sizeof(*hand));
+ folio_put(virt_to_folio(hand));
+ }
out:
if (new_sock)
@@ -2090,21 +2123,8 @@ int o2net_init(void)
unsigned long i;
o2quo_init();
-
o2net_debugfs_init();
- o2net_hand = kzalloc(sizeof(struct o2net_handshake), GFP_KERNEL);
- o2net_keep_req = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL);
- o2net_keep_resp = kzalloc(sizeof(struct o2net_msg), GFP_KERNEL);
- if (!o2net_hand || !o2net_keep_req || !o2net_keep_resp)
- goto out;
-
- o2net_hand->protocol_version = cpu_to_be64(O2NET_PROTOCOL_VERSION);
- o2net_hand->connector_id = cpu_to_be64(1);
-
- o2net_keep_req->magic = cpu_to_be16(O2NET_MSG_KEEP_REQ_MAGIC);
- o2net_keep_resp->magic = cpu_to_be16(O2NET_MSG_KEEP_RESP_MAGIC);
-
for (i = 0; i < ARRAY_SIZE(o2net_nodes); i++) {
struct o2net_node *nn = o2net_nn_from_num(i);
@@ -2122,21 +2142,10 @@ int o2net_init(void)
}
return 0;
-
-out:
- kfree(o2net_hand);
- kfree(o2net_keep_req);
- kfree(o2net_keep_resp);
- o2net_debugfs_exit();
- o2quo_exit();
- return -ENOMEM;
}
void o2net_exit(void)
{
o2quo_exit();
- kfree(o2net_hand);
- kfree(o2net_keep_req);
- kfree(o2net_keep_resp);
o2net_debugfs_exit();
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 46/48] drbd: Use sendmsg(MSG_SPLICE_PAGES) rather than sendmsg()
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (44 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 45/48] ocfs2: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage() David Howells
@ 2023-03-29 14:13 ` David Howells
2023-03-29 14:13 ` [RFC PATCH v2 47/48] drdb: Send an entire bio in a single sendmsg David Howells
[not found] ` <20230329141354.516864-49-dhowells@redhat.com>
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Philipp Reisner,
Lars Ellenberg, Christoph Böhmwalder, drbd-dev, linux-block
Use sendmsg() conditionally with MSG_SPLICE_PAGES in _drbd_send_page()
rather than calling sendpage() or _drbd_no_send_page().
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Philipp Reisner <philipp.reisner@linbit.com>
cc: Lars Ellenberg <lars.ellenberg@linbit.com>
cc: "Christoph Böhmwalder" <christoph.boehmwalder@linbit.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: drbd-dev@lists.linbit.com
cc: linux-block@vger.kernel.org
cc: netdev@vger.kernel.org
---
drivers/block/drbd/drbd_main.c | 23 ++++++++++++-----------
1 file changed, 12 insertions(+), 11 deletions(-)
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index 2c764f7ee4a7..e5f90abd29b6 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -1532,7 +1532,8 @@ static int _drbd_send_page(struct drbd_peer_device *peer_device, struct page *pa
int offset, size_t size, unsigned msg_flags)
{
struct socket *socket = peer_device->connection->data.socket;
- int len = size;
+ struct bio_vec bvec;
+ struct msghdr msg = { .msg_flags = msg_flags, };
int err = -EIO;
/* e.g. XFS meta- & log-data is in slab pages, which have a
@@ -1541,33 +1542,33 @@ static int _drbd_send_page(struct drbd_peer_device *peer_device, struct page *pa
* put_page(); and would cause either a VM_BUG directly, or
* __page_cache_release a page that would actually still be referenced
* by someone, leading to some obscure delayed Oops somewhere else. */
- if (drbd_disable_sendpage || !sendpage_ok(page))
- return _drbd_no_send_page(peer_device, page, offset, size, msg_flags);
+ if (!drbd_disable_sendpage && sendpage_ok(page))
+ msg.msg_flags |= MSG_NOSIGNAL | MSG_SPLICE_PAGES;
+
+ bvec_set_page(&bvec, page, offset, size);
+ iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
- msg_flags |= MSG_NOSIGNAL;
drbd_update_congested(peer_device->connection);
do {
int sent;
- sent = socket->ops->sendpage(socket, page, offset, len, msg_flags);
+ sent = sock_sendmsg(socket, &msg);
if (sent <= 0) {
if (sent == -EAGAIN) {
if (we_should_drop_the_connection(peer_device->connection, socket))
break;
continue;
}
- drbd_warn(peer_device->device, "%s: size=%d len=%d sent=%d\n",
- __func__, (int)size, len, sent);
+ drbd_warn(peer_device->device, "%s: size=%d len=%zu sent=%d\n",
+ __func__, (int)size, msg_data_left(&msg), sent);
if (sent < 0)
err = sent;
break;
}
- len -= sent;
- offset += sent;
- } while (len > 0 /* THINK && device->cstate >= C_CONNECTED*/);
+ } while (msg_data_left(&msg) /* THINK && device->cstate >= C_CONNECTED*/);
clear_bit(NET_CONGESTED, &peer_device->connection->flags);
- if (len == 0) {
+ if (!msg_data_left(&msg)) {
err = 0;
peer_device->device->send_cnt += size >> 9;
}
^ permalink raw reply related [flat|nested] 77+ messages in thread* [RFC PATCH v2 47/48] drdb: Send an entire bio in a single sendmsg
2023-03-29 14:13 ` [RFC PATCH v2 00/48] splice, net: Replace sendpage with sendmsg(MSG_SPLICE_PAGES) David Howells
` (45 preceding siblings ...)
2023-03-29 14:13 ` [RFC PATCH v2 46/48] drbd: Use sendmsg(MSG_SPLICE_PAGES) rather than sendmsg() David Howells
@ 2023-03-29 14:13 ` David Howells
[not found] ` <20230329141354.516864-49-dhowells@redhat.com>
47 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:13 UTC (permalink / raw)
To: Matthew Wilcox, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni
Cc: David Howells, Al Viro, Christoph Hellwig, Jens Axboe,
Jeff Layton, Christian Brauner, Chuck Lever III, Linus Torvalds,
netdev, linux-fsdevel, linux-kernel, linux-mm, Philipp Reisner,
Lars Ellenberg, Christoph Böhmwalder, drbd-dev, linux-block
Since _drdb_sendpage() is now using sendmsg to send the pages rather
sendpage, pass the entire bio in one go using a bvec iterator instead of
doing it piecemeal.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Philipp Reisner <philipp.reisner@linbit.com>
cc: Lars Ellenberg <lars.ellenberg@linbit.com>
cc: "Christoph Böhmwalder" <christoph.boehmwalder@linbit.com>
cc: Jens Axboe <axboe@kernel.dk>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: drbd-dev@lists.linbit.com
cc: linux-block@vger.kernel.org
cc: netdev@vger.kernel.org
---
drivers/block/drbd/drbd_main.c | 77 +++++++++++-----------------------
1 file changed, 25 insertions(+), 52 deletions(-)
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index e5f90abd29b6..ab63d6138407 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -1512,28 +1512,15 @@ static void drbd_update_congested(struct drbd_connection *connection)
* As a workaround, we disable sendpage on pages
* with page_count == 0 or PageSlab.
*/
-static int _drbd_no_send_page(struct drbd_peer_device *peer_device, struct page *page,
- int offset, size_t size, unsigned msg_flags)
-{
- struct socket *socket;
- void *addr;
- int err;
-
- socket = peer_device->connection->data.socket;
- addr = kmap(page) + offset;
- err = drbd_send_all(peer_device->connection, socket, addr, size, msg_flags);
- kunmap(page);
- if (!err)
- peer_device->device->send_cnt += size >> 9;
- return err;
-}
-
-static int _drbd_send_page(struct drbd_peer_device *peer_device, struct page *page,
- int offset, size_t size, unsigned msg_flags)
+static int _drbd_send_pages(struct drbd_peer_device *peer_device,
+ struct iov_iter *iter, unsigned msg_flags)
{
struct socket *socket = peer_device->connection->data.socket;
- struct bio_vec bvec;
- struct msghdr msg = { .msg_flags = msg_flags, };
+ struct msghdr msg = {
+ .msg_flags = msg_flags | MSG_NOSIGNAL,
+ .msg_iter = *iter,
+ };
+ size_t size = iov_iter_count(iter);
int err = -EIO;
/* e.g. XFS meta- & log-data is in slab pages, which have a
@@ -1542,11 +1529,8 @@ static int _drbd_send_page(struct drbd_peer_device *peer_device, struct page *pa
* put_page(); and would cause either a VM_BUG directly, or
* __page_cache_release a page that would actually still be referenced
* by someone, leading to some obscure delayed Oops somewhere else. */
- if (!drbd_disable_sendpage && sendpage_ok(page))
- msg.msg_flags |= MSG_NOSIGNAL | MSG_SPLICE_PAGES;
-
- bvec_set_page(&bvec, page, offset, size);
- iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
+ if (drbd_disable_sendpage)
+ msg.msg_flags &= ~(MSG_NOSIGNAL | MSG_SPLICE_PAGES);
drbd_update_congested(peer_device->connection);
do {
@@ -1577,39 +1561,22 @@ static int _drbd_send_page(struct drbd_peer_device *peer_device, struct page *pa
static int _drbd_send_bio(struct drbd_peer_device *peer_device, struct bio *bio)
{
- struct bio_vec bvec;
- struct bvec_iter iter;
+ struct iov_iter iter;
- /* hint all but last page with MSG_MORE */
- bio_for_each_segment(bvec, bio, iter) {
- int err;
+ iov_iter_bvec(&iter, ITER_SOURCE, bio->bi_io_vec, bio->bi_vcnt,
+ bio->bi_iter.bi_size);
- err = _drbd_no_send_page(peer_device, bvec.bv_page,
- bvec.bv_offset, bvec.bv_len,
- bio_iter_last(bvec, iter)
- ? 0 : MSG_MORE);
- if (err)
- return err;
- }
- return 0;
+ return _drbd_send_pages(peer_device, &iter, 0);
}
static int _drbd_send_zc_bio(struct drbd_peer_device *peer_device, struct bio *bio)
{
- struct bio_vec bvec;
- struct bvec_iter iter;
+ struct iov_iter iter;
- /* hint all but last page with MSG_MORE */
- bio_for_each_segment(bvec, bio, iter) {
- int err;
+ iov_iter_bvec(&iter, ITER_SOURCE, bio->bi_io_vec, bio->bi_vcnt,
+ bio->bi_iter.bi_size);
- err = _drbd_send_page(peer_device, bvec.bv_page,
- bvec.bv_offset, bvec.bv_len,
- bio_iter_last(bvec, iter) ? 0 : MSG_MORE);
- if (err)
- return err;
- }
- return 0;
+ return _drbd_send_pages(peer_device, &iter, MSG_SPLICE_PAGES);
}
static int _drbd_send_zc_ee(struct drbd_peer_device *peer_device,
@@ -1621,10 +1588,16 @@ static int _drbd_send_zc_ee(struct drbd_peer_device *peer_device,
/* hint all but last page with MSG_MORE */
page_chain_for_each(page) {
+ struct iov_iter iter;
+ struct bio_vec bvec;
unsigned l = min_t(unsigned, len, PAGE_SIZE);
- err = _drbd_send_page(peer_device, page, 0, l,
- page_chain_next(page) ? MSG_MORE : 0);
+ bvec_set_page(&bvec, page, 0, l);
+ iov_iter_bvec(&iter, ITER_SOURCE, &bvec, 1, l);
+
+ err = _drbd_send_pages(peer_device, &iter,
+ MSG_SPLICE_PAGES |
+ (page_chain_next(page) ? MSG_MORE : 0));
if (err)
return err;
len -= l;
^ permalink raw reply related [flat|nested] 77+ messages in thread[parent not found: <20230329141354.516864-49-dhowells@redhat.com>]
* Re: [RFC PATCH v2 48/48] sock: Remove ->sendpage*() in favour of sendmsg(MSG_SPLICE_PAGES)
[not found] ` <20230329141354.516864-49-dhowells@redhat.com>
@ 2023-03-29 14:39 ` David Howells
0 siblings, 0 replies; 77+ messages in thread
From: David Howells @ 2023-03-29 14:39 UTC (permalink / raw)
To: Hannes Reinecke
Cc: David Howells, Matthew Wilcox, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, linux-doc, virtualization, linux-mm,
linux-sctp, linux-afs, rds-devel, linux-x25, dccp, linux-rdma,
Christoph Hellwig, linux-wpan, linux-arm-msm, linux-can,
Marc Kleine-Budde, Al Viro, linux-hams, mptcp, Jens Axboe,
Christian Brauner, netdev, Jeff Layton, linux-kernel,
Chuck Lever III, tipc-discussion, linux-crypto, linux-fsdevel,
bpf, Linus Torvalds
Hannes Reinecke <hare@suse.de> wrote:
> > [!] Note: This is a work in progress. At the moment, some things won't
> > build if this patch is applied. nvme, kcm, smc, tls.
Actually, that needs updating. nvme and smc now build.
> Weelll ... what happens to consumers of kernel_sendpage()?
> (Let's call them nvme ...)
> Should they be moved over, too?
Patch 42 should address NVMe, I think. I can't test it, though, as I don't
have hardware.
There should be no callers of kernel_sendmsg() by the end of this patchset,
and the only remaining implementors of sendpage are Chelsio-TLS, AF_TLS and
AF_KCM, which as stated in the cover, aren't yet converted and won't build.
> Or what is the general consensus here?
>
> (And what do we do with TLS? It does have a ->sendpage() version, too ...)
I know. There are three things left that I need to tackle, but I'd like to
get opinions on some of the other bits and I might need some help with AF_TLS
and AF_KCM.
That said, should I just remove tls_sw_do_sendpage() since presumably the data
is going to get copied(?) and encrypted and the source pages aren't going to
be held onto?
David
^ permalink raw reply [flat|nested] 77+ messages in thread