[PATCH v4 00/10] netfs, cachefiles: Miscellaneous fixes

[PATCH v4 00/10] netfs, cachefiles: Miscellaneous fixes

[David Howells]

Hi Christian,

Here are some miscellaneous fixes for netfslib and one for cachefiles, if
you could pick them up?

The first six are all in unbuffered/DIO write:

 (1) Fix an uninitialised return value from netfs_unbuffered_write().

 (2) Fix the normal error return, preferring partial transfer size over
     request error over immediate error (e.g. EINTR).

 (3) Fix the async error return, preferring partial transfer size over
     error.

 (4) Fix the update of i_size on partial transfer ending in an error.

 (5) Fix a subrequest leak in an error path.

 (6) Fix the code to handle subrequest allocation failure.

Then there's:

 (7) Fix synchronisation issues with using a progressive rolling buffer for
     readahead by fetching everything into it upfront so that the issues
     don't arise.  Dropping the refs so acquired is deferred until after
     I/O is begun.

 (8) Change the marking of folios to be copied to the cache to be done
     whilst subreqs are being issued rather than at the time they're
     collected.  Whilst this is not strictly a fix, it means that the
     collector thread doesn't need to try and keep track of that, which
     means that delayed progress reporting will not be a problem in patch
     (9).

 (9) Fix read progress reporting to avoid 64-bit tearing on a 32-bit
     machine.  This has been modified from the previous submission as part
     of a different series to take account of a sashiko reported issue[1].

(10) Fix a potential UAF/KASAN warning reported by sashiko[1] in cachefiles
     in which the coherency data buffer is cast to a __be64* and
     dereferenced in a tracepoint - even though it might not be at least
     that large (or aligned).

The patches can also be found here:

	https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/log/?h=netfs-fixes

Thanks,
David

Changes
=======
ver #4)
- Fixed a number of issues reported by Sashiko[4]:
  - Fixed netfs_mark_copy_to_cache() to decrement len to end the loop.
  - Fixed netfs_read_to_pagecache() to init offset.
  - Fixed netfs_unlock_abandoned_read_pages() to clean up copy mark.
  - Fixed netfs_unlock_read_folio() to clean up copy mark.
  - Fixed copy-to-cache cancellation handling in netfs_unlock_read_folio().
  - Removed duplicate var init in netfs_read_unlock_folios().
  - Fixed setting of rreq->progress_at to ULONG_MAX (set to rreq->len
    instead).

ver #3)
- Fixed a number of issues reported by Sashiko[3]:
  - Fixed netfs_unbuffered_write() to return -ENOMEM to async writes too.
  - Added a patch to preload the entire readahead folio collection up front
    to avoid producer-consumer synchronisation issues.
  - Added a patch to premark buffered read folios with whether or not
    they're going to require caching so that the read-progress fix works
    correctly.

ver #2)
- Imported a patch to add subreq alloc failure handling in unbuffered/DIO
  write.
- Fixed a number of issues reported by Sashiko[2]:
  - Added four patches to fix bugs in unbuffered/DIO write.
  - Added a patch to fix readahead rolling buffer issues.
  - Fixed the cachefiles xattr stuff to use max() not min() when
    allocating.
  - Fixed the cachefiles trace comment with regard to alignment.
  - Fixed netfs_read_set_unlock_at() to use folioq_count(), not
    folio_nr_slots().

[1] https://sashiko.dev/#/patchset/20260810144746.574036-1-dhowells%40redhat.com
[2] https://sashiko.dev/#/patchset/20260824120224.504575-1-dhowells%40redhat.com
[3] https://sashiko.dev/#/patchset/20260825132045.1000787-1-dhowells%40redhat.com
[4] https://sashiko.dev/#/patchset/20260827093828.1956211-1-dhowells%40redhat.com

David Howells (8):
  netfs: Fix unbuffered/DIO write partial transfer error return
  netfs: Fix error vs transferred passed to ->ki_complete()
  netfs: Fix i_size update for partial transfer
  netfs: Fix subreq ref leak
  netfs: Fix readahead synchronisation issues by loading all folios
    upfront
  netfs: Mark folios with COPY_TO_CACHE whilst issuing subreqs
  netfs: Fix read progress reporting
  cachefiles: Fix potential UAF/KASAN warning

Edward Adam Davis (1):
  netfs: break unbuffered write when netfs_alloc_subrequest() fails

Karl Mehltretter (1):
  netfs: Fix uninitialized return value in netfs_unbuffered_write()

 fs/cachefiles/xattr.c             |  16 +--
 fs/netfs/buffered_read.c          | 164 ++++++++++++++++++++++--------
 fs/netfs/direct_write.c           |  31 ++++--
 fs/netfs/internal.h               |   3 +
 fs/netfs/misc.c                   |  19 ++++
 fs/netfs/objects.c                |  32 +++---
 fs/netfs/read_collect.c           | 128 ++++++++++++++++-------
 fs/netfs/read_pgpriv2.c           |  15 +--
 fs/netfs/read_retry.c             |  13 ++-
 fs/netfs/read_single.c            |   2 +
 fs/netfs/rolling_buffer.c         |  81 +++++++++------
 fs/netfs/write_issue.c            |   2 +
 include/linux/netfs.h             |   5 +-
 include/linux/rolling_buffer.h    |   6 +-
 include/trace/events/cachefiles.h |  19 +++-
 include/trace/events/netfs.h      |  30 +++++-
 16 files changed, 411 insertions(+), 155 deletions(-)

[PATCH v4 01/10] netfs: Fix uninitialized return value in netfs_unbuffered_write()

[David Howells]

From: Karl Mehltretter <kmehltretter@gmail.com>

If preparation of the first subrequest fails,
netfs_unbuffered_write() exits its loop before ret is initialized. The
empty-iterator check can do the same.

For synchronous writes, netfs_unbuffered_write_iter_locked() may then
return an unrelated error instead of wreq->error. This is reachable
through CIFS if cifs_prepare_write() fails to reopen the file or obtain
credits.

Initialize ret to 0 so the caller returns wreq->error if no data was
written, or the number of bytes already written otherwise.

Found with Clang's -Wconditional-uninitialized.

Fixes: a0b4c7a49137e ("netfs: Fix unbuffered/DIO writes to dispatch subrequests in strict sequence")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-fable-5
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
---
 fs/netfs/direct_write.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c
index c16fbad286a1..b04019097ab8 100644
--- a/fs/netfs/direct_write.c
+++ b/fs/netfs/direct_write.c
@@ -95,7 +95,7 @@ static int netfs_unbuffered_write(struct netfs_io_request *wreq)
 {
 	struct netfs_io_subrequest *subreq = NULL;
 	struct netfs_io_stream *stream = &wreq->io_streams[0];
-	int ret;
+	int ret = 0;
 
 	_enter("%llx", wreq->len);
 

[PATCH v4 02/10] netfs: Fix unbuffered/DIO write partial transfer error return

[David Howells]

Fix unbuffered/DIO write to return the amount of data transferred in
preference to an error if a partial transfer has been achieved, and to
prefer an error stashed in the request over the one returned by
netfs_unbuffered_write() (likely -EINTR or -ERESTARTSYS).

Fixes: a0b4c7a49137e ("netfs: Fix unbuffered/DIO writes to dispatch subrequests in strict sequence")
Link: https://sashiko.dev/#/patchset/20260824120224.504575-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara <pc@manguebit.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/direct_write.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c
index b04019097ab8..544a4243fc59 100644
--- a/fs/netfs/direct_write.c
+++ b/fs/netfs/direct_write.c
@@ -139,13 +139,11 @@ static int netfs_unbuffered_write(struct netfs_io_request *wreq)
 		if (test_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags)) {
 			retry = true;
 		} else if (test_bit(NETFS_SREQ_FAILED, &subreq->flags)) {
-			ret = subreq->error;
-			wreq->error = ret;
+			wreq->error = subreq->error;
 			netfs_see_subrequest(subreq, netfs_sreq_trace_see_failed);
 			subreq = NULL;
 			break;
 		}
-		ret = 0;
 
 		if (!retry) {
 			netfs_unbuffered_write_collect(wreq, stream, subreq);
@@ -288,11 +286,11 @@ ssize_t netfs_unbuffered_write_iter_locked(struct kiocb *iocb, struct iov_iter *
 		ret = -EIOCBQUEUED;
 	} else {
 		ret = netfs_unbuffered_write(wreq);
-		if (ret < 0) {
-			_debug("begin = %zd", ret);
-		} else {
+		if (wreq->transferred) {
 			iocb->ki_pos += wreq->transferred;
-			ret = wreq->transferred ?: wreq->error;
+			ret = wreq->transferred;
+		} else if (wreq->error) {
+			ret = wreq->error;
 		}
 
 		netfs_put_request(wreq, netfs_rreq_trace_put_complete);

[PATCH v4 03/10] netfs: Fix error vs transferred passed to ->ki_complete()

[David Howells]

Fix netfs_unbuffered_write_done() to pass the amount written to
->ki_complete() rather than the error in the event of a partially complete
transfer.

Fixes: a0b4c7a49137e ("netfs: Fix unbuffered/DIO writes to dispatch subrequests in strict sequence")
Link: https://sashiko.dev/#/patchset/20260824120224.504575-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara <pc@manguebit.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/direct_write.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c
index 544a4243fc59..f7d7e1b54653 100644
--- a/fs/netfs/direct_write.c
+++ b/fs/netfs/direct_write.c
@@ -51,7 +51,7 @@ static void netfs_unbuffered_write_done(struct netfs_io_request *wreq)
 		wreq->iocb->ki_pos += written;
 		if (wreq->iocb->ki_complete) {
 			trace_netfs_rreq(wreq, netfs_rreq_trace_ki_complete);
-			wreq->iocb->ki_complete(wreq->iocb, wreq->error ?: written);
+			wreq->iocb->ki_complete(wreq->iocb, written ?: wreq->error);
 		}
 		wreq->iocb = VFS_PTR_POISON;
 	}

[PATCH v4 04/10] netfs: Fix i_size update for partial transfer

[David Howells]

Fix netfs_unbuffered_write_done() to pass the amount written to
netfs_update_i_size() in the event of a partial transfer that ends in an
error.

That said, it might be better for the filesystem to mark the inode data as
invalid and recheck it in case something like a network error occurred that
prevented the reply from the server from being received.

Fixes: a0b4c7a49137e ("netfs: Fix unbuffered/DIO writes to dispatch subrequests in strict sequence")
Link: https://sashiko.dev/#/patchset/20260824120224.504575-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara <pc@manguebit.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/direct_write.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c
index f7d7e1b54653..f33ccddaa826 100644
--- a/fs/netfs/direct_write.c
+++ b/fs/netfs/direct_write.c
@@ -21,7 +21,7 @@ static void netfs_unbuffered_write_done(struct netfs_io_request *wreq)
 	/* Okay, declare that all I/O is complete. */
 	trace_netfs_rreq(wreq, netfs_rreq_trace_write_done);
 
-	if (!wreq->error)
+	if (wreq->transferred)
 		netfs_update_i_size(ictx, &ictx->inode, wreq->start, wreq->transferred);
 
 	if (wreq->origin == NETFS_DIO_WRITE &&

[PATCH v4 05/10] netfs: Fix subreq ref leak

[David Howells]

Fix a subrequest ref leak in netfs_unbuffered_write() in the event that
subreq->io_iter ends up zero length during preparation.

Fixes: a0b4c7a49137e ("netfs: Fix unbuffered/DIO writes to dispatch subrequests in strict sequence")
Link: https://sashiko.dev/#/patchset/20260824120224.504575-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara <pc@manguebit.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/direct_write.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c
index f33ccddaa826..fbcfadb232ee 100644
--- a/fs/netfs/direct_write.c
+++ b/fs/netfs/direct_write.c
@@ -121,8 +121,14 @@ static int netfs_unbuffered_write(struct netfs_io_request *wreq)
 		}
 
 		iov_iter_truncate(&subreq->io_iter, wreq->len - wreq->transferred);
-		if (!iov_iter_count(&subreq->io_iter))
+		if (!iov_iter_count(&subreq->io_iter)) {
+			pr_warn("netfs: Unexpected zero-length iterator R=%08x\n",
+				wreq->debug_id);
+			__set_bit(NETFS_SREQ_FAILED, &subreq->flags);
+			netfs_write_subrequest_terminated(subreq, -EIO);
+			wreq->error = -EIO;
 			break;
+		}
 
 		subreq->len = netfs_limit_iter(&subreq->io_iter, 0,
 					       stream->sreq_max_len,

[PATCH v4 06/10] netfs: break unbuffered write when netfs_alloc_subrequest() fails

[David Howells]

From: Edward Adam Davis <eadavis@qq.com>

syzbot reported a null-ptr-deref below [1] following a fault injection in
netfs_alloc_subrequest(). [0]

When netfs_alloc_subrequest() fails, subreq is NULL.
Later, netfs_prepare_write() tries to initialize members of
subreq(e.g., source), the issue in [1] is triggered.

Let's handle the error of netfs_prepare_write() properly.

[0]
FAULT_INJECTION: forcing a failure.
name failslab, interval 1, probability 0, space 0, times 0
Call Trace:
 netfs_alloc_subrequest+0x116/0x3f0
 netfs_prepare_write+0x76/0x7b0
 netfs_unbuffered_write+0x75c/0x2020
 netfs_unbuffered_write_iter_locked+0x7d6/0xa80
 netfs_unbuffered_write_iter+0x442/0x720
 v9fs_file_write_iter+0xbf/0x100
 vfs_write+0x6ac/0x1050

[1]
KASAN: null-ptr-deref in range [0x00000000000000a8-0x00000000000000af]
RIP: 0010:netfs_prepare_write+0xbc/0x7b0 fs/netfs/write_issue.c:173
Call Trace:
 netfs_unbuffered_write+0x75c/0x2020 fs/netfs/direct_write.c:111
 netfs_unbuffered_write_iter_locked+0x7d6/0xa80 fs/netfs/direct_write.c:290
 netfs_unbuffered_write_iter+0x442/0x720 fs/netfs/direct_write.c:382
 v9fs_file_write_iter+0xbf/0x100 fs/9p/vfs_file.c:409
 new_sync_write fs/read_write.c:595 [inline]

Fixes: 288ace2f57c9 ("netfs: New writeback implementation")
Reported-by: syzbot+6a13fc77eb6f0802be2d@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=6a13fc77eb6f0802be2d
Tested-by: syzbot+6a13fc77eb6f0802be2d@syzkaller.appspotmail.com
Signed-off-by: Edward Adam Davis <eadavis@qq.com>
Signed-off-by: David Howells <dhowells@redhat.com>
---
 fs/netfs/direct_write.c | 5 +++++
 fs/netfs/write_issue.c  | 2 ++
 2 files changed, 7 insertions(+)

diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c
index fbcfadb232ee..2361277416c7 100644
--- a/fs/netfs/direct_write.c
+++ b/fs/netfs/direct_write.c
@@ -110,6 +110,11 @@ static int netfs_unbuffered_write(struct netfs_io_request *wreq)
 		if (!subreq) {
 			netfs_prepare_write(wreq, stream, wreq->start + wreq->transferred);
 			subreq = stream->construct;
+			if (!subreq) {
+				wreq->error = -ENOMEM;
+				ret = -ENOMEM;
+				break;
+			}
 			stream->construct = NULL;
 		}
 
diff --git a/fs/netfs/write_issue.c b/fs/netfs/write_issue.c
index 2d9cfcd43658..851f6f93ad45 100644
--- a/fs/netfs/write_issue.c
+++ b/fs/netfs/write_issue.c
@@ -170,6 +170,8 @@ void netfs_prepare_write(struct netfs_io_request *wreq,
 		rolling_buffer_make_space(&wreq->buffer, wreq->gfp);
 
 	subreq = netfs_alloc_subrequest(wreq);
+	if (!subreq)
+		return;
 	subreq->source		= stream->source;
 	subreq->start		= start;
 	subreq->stream_nr	= stream->stream_nr;

[PATCH v4 07/10] netfs: Fix readahead synchronisation issues by loading all folios upfront

[David Howells]

There are some synchronisation issues that derive from the app thread
adding more folios to the rolling buffer whilst the collector thread is
looking at them or trying to clear them, such as determining the setting of
front_folio_order when the next folio hasn't been added yet,

The reason for the rolling buffer approach is that loading the buffer
upfront and then dropping all the refs just acquired is quite a slow
operation, and loading progressively allows some of the cost to be deferred
until after at least some of the I/O is started.

Instead, a better way is to load all the folios into the rolling buffer
upfront - and then drop the refs later, once the I/O is in progress.  (Even
better would be for the refs not to be there at all.)

Fix this by changing the rolling buffer loader to load all the folios
selected by the VM for readahead upfront into the folio queue.  The folio
queue is allocated a batch worth at a time as we don't know how many folios
are involved (the readahead_control struct, alas, has a page count, not a
folio count).

The folio refs acquired from readahead are then dropped in bulk once the
first subrequest is dispatched as it's quite a slow operation.  The
collector waits for NETFS_RREQ_NEED_PUT_RA_REFS to be cleared so that it
doesn't unlock folios before the xarray has been scanned for them.

This simplifies the buffer handling later and isn't noticeably slower as
the xarray doesn't need to be modified and the folios are all already
pre-locked.

Fixes: ee4cdf7ba857 ("netfs: Speed up buffered reading")
Link: https://sashiko.dev/#/patchset/20260824120224.504575-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara (Red Hat) <pc@manguebit.org>
cc: Matthew Wilcox <willy@infradead.org>
cc: netfs@lists.linux.dev
cc: linux-mm@kvack.org
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/buffered_read.c       | 101 ++++++++++++++++++++-------------
 fs/netfs/internal.h            |   1 +
 fs/netfs/misc.c                |  19 +++++++
 fs/netfs/read_collect.c        |   7 +++
 fs/netfs/read_retry.c          |   7 +++
 fs/netfs/rolling_buffer.c      |  81 ++++++++++++++++----------
 include/linux/netfs.h          |   1 +
 include/linux/rolling_buffer.h |   6 +-
 include/trace/events/netfs.h   |   3 +
 9 files changed, 154 insertions(+), 72 deletions(-)

diff --git a/fs/netfs/buffered_read.c b/fs/netfs/buffered_read.c
index 7fdfa4f27e34..303fdce54fba 100644
--- a/fs/netfs/buffered_read.c
+++ b/fs/netfs/buffered_read.c
@@ -54,6 +54,42 @@ static void netfs_rreq_expand(struct netfs_io_request *rreq,
 	}
 }
 
+/*
+ * Drop the folio refs acquired from the readahead API.
+ */
+static void netfs_bulk_drop_ra_refs(struct netfs_io_request *rreq)
+{
+	struct folio_batch fbatch;
+	struct folio *folio;
+	pgoff_t nr_pages = DIV_ROUND_UP(rreq->len, PAGE_SIZE);
+	pgoff_t first = rreq->start / PAGE_SIZE;
+	XA_STATE(xas, &rreq->mapping->i_pages, first);
+
+	folio_batch_init(&fbatch);
+
+	rcu_read_lock();
+
+	xas_for_each(&xas, folio,  first + nr_pages - 1) {
+		if (xas_retry(&xas, folio))
+			continue;
+
+		if (!folio_batch_add(&fbatch, folio))
+			folio_batch_release(&fbatch);
+	}
+
+	rcu_read_unlock();
+	folio_batch_release(&fbatch);
+	trace_netfs_rreq(rreq, netfs_rreq_trace_ra_put_ref);
+	clear_bit_unlock(NETFS_RREQ_NEED_PUT_RA_REFS, &rreq->flags);
+	wake_up(&rreq->waitq);
+}
+
+static void netfs_maybe_bulk_drop_ra_refs(struct netfs_io_request *rreq)
+{
+	if (test_bit(NETFS_RREQ_NEED_PUT_RA_REFS, &rreq->flags))
+		netfs_bulk_drop_ra_refs(rreq);
+}
+
 /*
  * Begin an operation, and fetch the stored zero point value from the cookie if
  * available.
@@ -74,12 +110,8 @@ static int netfs_begin_cache_read(struct netfs_io_request *rreq, struct netfs_in
  *
  * Returns the limited size if successful and -ENOMEM if insufficient memory
  * available.
- *
- * [!] NOTE: This must be run in the same thread as ->issue_read() was called
- * in as we access the readahead_control struct.
  */
-static ssize_t netfs_prepare_read_iterator(struct netfs_io_subrequest *subreq,
-					   struct readahead_control *ractl)
+static ssize_t netfs_prepare_read_iterator(struct netfs_io_subrequest *subreq)
 {
 	struct netfs_io_request *rreq = subreq->rreq;
 	size_t rsize = subreq->len;
@@ -87,30 +119,6 @@ static ssize_t netfs_prepare_read_iterator(struct netfs_io_subrequest *subreq,
 	if (subreq->source == NETFS_DOWNLOAD_FROM_SERVER)
 		rsize = umin(rsize, rreq->io_streams[0].sreq_max_len);
 
-	if (ractl) {
-		/* If we don't have sufficient folios in the rolling buffer,
-		 * extract a folioq's worth from the readahead region at a time
-		 * into the buffer.  Note that this acquires a ref on each page
-		 * that we will need to release later - but we don't want to do
-		 * that until after we've started the I/O.
-		 */
-		struct folio_batch put_batch;
-
-		folio_batch_init(&put_batch);
-		while (rreq->submitted < subreq->start + rsize) {
-			ssize_t added;
-
-			added = rolling_buffer_load_from_ra(&rreq->buffer, ractl,
-							    &put_batch);
-			if (added < 0) {
-				folio_batch_release(&put_batch);
-				return added;
-			}
-			rreq->submitted += added;
-		}
-		folio_batch_release(&put_batch);
-	}
-
 	subreq->len = rsize;
 	if (unlikely(rreq->io_streams[0].sreq_max_segs)) {
 		size_t limit = netfs_limit_iter(&rreq->buffer.iter, 0, rsize,
@@ -208,8 +216,7 @@ static void netfs_issue_read(struct netfs_io_request *rreq,
  * slicing up the region to be read according to available cache blocks and
  * network rsize.
  */
-static void netfs_read_to_pagecache(struct netfs_io_request *rreq,
-				    struct readahead_control *ractl)
+static void netfs_read_to_pagecache(struct netfs_io_request *rreq)
 {
 	unsigned long long start = rreq->start;
 	ssize_t size = rreq->len;
@@ -288,7 +295,7 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq,
 		break;
 
 	issue:
-		slice = netfs_prepare_read_iterator(subreq, ractl);
+		slice = netfs_prepare_read_iterator(subreq);
 		if (slice < 0) {
 			ret = slice;
 			netfs_cancel_read(subreq, ret);
@@ -302,6 +309,7 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq,
 		}
 
 		netfs_issue_read(rreq, subreq);
+		netfs_maybe_bulk_drop_ra_refs(rreq);
 
 		if (test_bit(NETFS_RREQ_PAUSE, &rreq->flags))
 			netfs_wait_for_paused_read(rreq);
@@ -339,7 +347,8 @@ void netfs_readahead(struct readahead_control *ractl)
 {
 	struct netfs_io_request *rreq;
 	struct netfs_inode *ictx = netfs_inode(ractl->mapping->host);
-	unsigned long long start = readahead_pos(ractl);
+	ssize_t added;
+	uoff_t start = readahead_pos(ractl);
 	size_t size = readahead_length(ractl);
 	int ret;
 
@@ -360,11 +369,23 @@ void netfs_readahead(struct readahead_control *ractl)
 
 	netfs_rreq_expand(rreq, ractl);
 
-	rreq->submitted = rreq->start;
-	if (rolling_buffer_init(&rreq->buffer, rreq->debug_id, ITER_DEST, rreq->gfp) < 0)
+	/* Load the folios to be read into a bvecq chain.  Note that this
+	 * acquires a ref on each folio that we will need to release later -
+	 * but we don't want to do that until after we've started the I/O.
+	 */
+	added = rolling_buffer_bulk_load_from_ra(&rreq->buffer, ractl,
+						 rreq->debug_id, rreq->gfp);
+	if (added < 0) {
+		ret = added;
 		goto cleanup_free;
-	netfs_read_to_pagecache(rreq, ractl);
+	}
+	__set_bit(NETFS_RREQ_NEED_PUT_RA_REFS, &rreq->flags);
+
+	rreq->submitted = rreq->start + added;
+	rreq->cleaned_to = rreq->start;
 
+	netfs_read_to_pagecache(rreq);
+	netfs_maybe_bulk_drop_ra_refs(rreq);
 	return netfs_put_request(rreq, netfs_rreq_trace_put_return);
 
 cleanup_free:
@@ -457,7 +478,7 @@ static int netfs_read_gaps(struct file *file, struct folio *folio)
 	iov_iter_bvec(&rreq->buffer.iter, ITER_DEST, bvec, i, rreq->len);
 	rreq->submitted = rreq->start + flen;
 
-	netfs_read_to_pagecache(rreq, NULL);
+	netfs_read_to_pagecache(rreq);
 
 	ret = netfs_wait_for_read(rreq);
 	if (ret >= 0) {
@@ -532,7 +553,7 @@ int netfs_read_folio(struct file *file, struct folio *folio)
 	if (ret < 0)
 		goto discard;
 
-	netfs_read_to_pagecache(rreq, NULL);
+	netfs_read_to_pagecache(rreq);
 	ret = netfs_wait_for_read(rreq);
 	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 	return ret < 0 ? ret : 0;
@@ -689,7 +710,7 @@ int netfs_write_begin(struct netfs_inode *ctx,
 	if (ret < 0)
 		goto error_put;
 
-	netfs_read_to_pagecache(rreq, NULL);
+	netfs_read_to_pagecache(rreq);
 	ret = netfs_wait_for_read(rreq);
 	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 	if (ret < 0)
@@ -754,7 +775,7 @@ int netfs_prefetch_for_write(struct file *file, struct folio *folio,
 	if (ret < 0)
 		goto error_put;
 
-	netfs_read_to_pagecache(rreq, NULL);
+	netfs_read_to_pagecache(rreq);
 	ret = netfs_wait_for_read(rreq);
 	netfs_put_request(rreq, netfs_rreq_trace_put_return);
 	return ret < 0 ? ret : 0;
diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h
index 420ee7b26580..bd8b2d633f96 100644
--- a/fs/netfs/internal.h
+++ b/fs/netfs/internal.h
@@ -79,6 +79,7 @@ ssize_t netfs_wait_for_read(struct netfs_io_request *rreq);
 ssize_t netfs_wait_for_write(struct netfs_io_request *rreq);
 void netfs_wait_for_paused_read(struct netfs_io_request *rreq);
 void netfs_wait_for_paused_write(struct netfs_io_request *rreq);
+void netfs_wait_for_put_ra_refs(struct netfs_io_request *rreq);
 
 /*
  * objects.c
diff --git a/fs/netfs/misc.c b/fs/netfs/misc.c
index 5d554512ed23..f5c1c463f4ff 100644
--- a/fs/netfs/misc.c
+++ b/fs/netfs/misc.c
@@ -563,3 +563,22 @@ void netfs_wait_for_paused_write(struct netfs_io_request *rreq)
 {
 	return netfs_wait_for_pause(rreq, netfs_write_collection);
 }
+
+/*
+ * Wait for the readahead-acquired refs to be put.
+ */
+void netfs_wait_for_put_ra_refs(struct netfs_io_request *rreq)
+{
+	DEFINE_WAIT(myself);
+
+	for (;;) {
+		trace_netfs_rreq(rreq, netfs_rreq_trace_wait_put_ra_refs);
+		prepare_to_wait(&rreq->waitq, &myself, TASK_UNINTERRUPTIBLE);
+		if (!test_bit(NETFS_RREQ_NEED_PUT_RA_REFS, &rreq->flags))
+			break;
+		schedule();
+	}
+
+	trace_netfs_rreq(rreq, netfs_rreq_trace_waited_put_ra_refs);
+	finish_wait(&rreq->waitq, &myself);
+}
diff --git a/fs/netfs/read_collect.c b/fs/netfs/read_collect.c
index 23660a590124..edf7cea7e2f9 100644
--- a/fs/netfs/read_collect.c
+++ b/fs/netfs/read_collect.c
@@ -118,6 +118,13 @@ static void netfs_read_unlock_folios(struct netfs_io_request *rreq,
 		slot = 0;
 	}
 
+	/* We have to wait for readahead refs to have been released before we
+	 * can unlock any folios as the ref-dropper walks i_pages and the only
+	 * thing preventing these folios from being removed is the folio lock.
+	 */
+	if (test_bit(NETFS_RREQ_NEED_PUT_RA_REFS, &rreq->flags))
+		netfs_wait_for_put_ra_refs(rreq);
+
 	for (;;) {
 		struct folio *folio;
 		unsigned long long fpos, fend;
diff --git a/fs/netfs/read_retry.c b/fs/netfs/read_retry.c
index 2b42758e01ec..dd463a485139 100644
--- a/fs/netfs/read_retry.c
+++ b/fs/netfs/read_retry.c
@@ -292,6 +292,13 @@ void netfs_unlock_abandoned_read_pages(struct netfs_io_request *rreq)
 {
 	struct folio_queue *p;
 
+	/* We have to wait for readahead refs to have been released before we
+	 * can unlock any folios as the ref-dropper walks i_pages and the only
+	 * thing preventing these folios from being removed is the folio lock.
+	 */
+	if (test_bit(NETFS_RREQ_NEED_PUT_RA_REFS, &rreq->flags))
+		netfs_wait_for_put_ra_refs(rreq);
+
 	for (p = rreq->buffer.tail; p; p = p->next) {
 		for (int slot = 0; slot < folioq_count(p); slot++) {
 			struct folio *folio = folioq_folio(p, slot);
diff --git a/fs/netfs/rolling_buffer.c b/fs/netfs/rolling_buffer.c
index 8c0026836f9c..424e77a9a109 100644
--- a/fs/netfs/rolling_buffer.c
+++ b/fs/netfs/rolling_buffer.c
@@ -115,42 +115,65 @@ int rolling_buffer_make_space(struct rolling_buffer *roll, gfp_t gfp)
 }
 
 /*
- * Decant the list of folios to read into a rolling buffer.
+ * Decant the entire list of folios to read into a rolling buffer.
  */
-ssize_t rolling_buffer_load_from_ra(struct rolling_buffer *roll,
-				    struct readahead_control *ractl,
-				    struct folio_batch *put_batch)
+ssize_t rolling_buffer_bulk_load_from_ra(struct rolling_buffer *roll,
+					 struct readahead_control *ractl,
+					 unsigned int rreq_id, gfp_t gfp)
 {
 	struct folio_queue *fq;
-	struct page **vec;
-	int nr, ix, to;
-	ssize_t size = 0;
+	ssize_t loaded = 0;
 
-	if (rolling_buffer_make_space(roll, GFP_KERNEL) < 0)
-		return -ENOMEM;
+	while (ractl->_nr_pages - ractl->_batch_count > 0) {
+		unsigned int nr;
 
-	fq = roll->head;
-	vec = (struct page **)fq->vec.folios;
-	nr = __readahead_batch(ractl, vec + folio_batch_count(&fq->vec),
-			       folio_batch_space(&fq->vec));
-	ix = fq->vec.nr;
-	to = ix + nr;
-	fq->vec.nr = to;
-	for (; ix < to; ix++) {
-		struct folio *folio = folioq_folio(fq, ix);
-		unsigned int order = folio_order(folio);
-
-		fq->orders[ix] = order;
-		size += PAGE_SIZE << order;
-		trace_netfs_folio(folio, netfs_folio_trace_read);
-		if (!folio_batch_add(put_batch, folio))
-			folio_batch_release(put_batch);
+		/* Allocate a folioq to put some folios into and attach it to
+		 * the rolling buffer.
+		 */
+		fq = netfs_folioq_alloc(rreq_id, gfp,
+					netfs_trace_folioq_make_space);
+		if (!fq)
+			goto nomem_unlock;
+		fq->prev = roll->head;
+		if (!roll->tail)
+			roll->tail = fq;
+		else
+			roll->head->next = fq;
+		roll->head = fq;
+
+		/* Get a batch of folios and note their orders. */
+		nr = __readahead_batch(ractl, (struct page **)fq->vec.folios,
+				       folioq_nr_slots(fq));
+		if (WARN_ON_ONCE(!nr))
+			break;
+		fq->vec.nr = nr;
+
+		for (int slot = 0; slot < nr; slot++) {
+			struct folio *folio = folioq_folio(fq, slot);
+			unsigned int order;
+
+			order = folio_order(folio);
+			fq->orders[slot] = order;
+			loaded += PAGE_SIZE << order;
+			trace_netfs_folio(folio, netfs_folio_trace_read);
+		}
 	}
-	WRITE_ONCE(roll->iter.count, roll->iter.count + size);
 
-	/* Store the counter after setting the slot. */
-	smp_store_release(&roll->next_head_slot, to);
-	return size;
+	WRITE_ONCE(roll->iter.count, loaded);
+	iov_iter_folio_queue(&roll->iter, ITER_DEST, roll->tail, 0, 0, loaded);
+	return loaded;
+
+nomem_unlock:
+	for (fq = roll->tail; fq; fq = fq->next) {
+		for (int slot = 0; slot < folioq_count(fq); slot++) {
+			folio_unlock(fq->vec.folios[slot]);
+			folioq_mark(fq, slot);
+		}
+	}
+	rolling_buffer_clear(roll);
+	roll->head = NULL;
+	roll->tail = NULL;
+	return -ENOMEM;
 }
 
 /*
diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index f837a501008c..5c538d0c5d79 100644
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -278,6 +278,7 @@ struct netfs_io_request {
 #define NETFS_RREQ_FOLIO_COPY_TO_CACHE	10	/* Copy current folio to cache from read */
 #define NETFS_RREQ_UPLOAD_TO_SERVER	11	/* Need to write to the server */
 #define NETFS_RREQ_USE_IO_ITER		12	/* Use ->io_iter rather than ->i_pages */
+#define NETFS_RREQ_NEED_PUT_RA_REFS	17	/* Need to put the folio refs RA gave us */
 #define NETFS_RREQ_USE_PGPRIV2		31	/* [DEPRECATED] Use PG_private_2 to mark
 						 * write to cache on read */
 	const struct netfs_request_ops *netfs_ops;
diff --git a/include/linux/rolling_buffer.h b/include/linux/rolling_buffer.h
index 9e5dad29669c..a97f7cfaacaa 100644
--- a/include/linux/rolling_buffer.h
+++ b/include/linux/rolling_buffer.h
@@ -45,9 +45,9 @@ struct rolling_buffer_snapshot {
 int rolling_buffer_init(struct rolling_buffer *roll, unsigned int rreq_id,
 			unsigned int direction, gfp_t gfp);
 int rolling_buffer_make_space(struct rolling_buffer *roll, gfp_t gfp);
-ssize_t rolling_buffer_load_from_ra(struct rolling_buffer *roll,
-				    struct readahead_control *ractl,
-				    struct folio_batch *put_batch);
+ssize_t rolling_buffer_bulk_load_from_ra(struct rolling_buffer *roll,
+					 struct readahead_control *ractl,
+					 unsigned int rreq_id, gfp_t gfp);
 ssize_t rolling_buffer_append(struct rolling_buffer *roll, struct folio *folio,
 			      unsigned int flags, gfp_t gfp);
 struct folio_queue *rolling_buffer_delete_spent(struct rolling_buffer *roll);
diff --git a/include/trace/events/netfs.h b/include/trace/events/netfs.h
index 082cb03c6131..9bda9302be90 100644
--- a/include/trace/events/netfs.h
+++ b/include/trace/events/netfs.h
@@ -59,6 +59,7 @@
 	EM(netfs_rreq_trace_free,		"FREE   ")	\
 	EM(netfs_rreq_trace_intr,		"INTR   ")	\
 	EM(netfs_rreq_trace_ki_complete,	"KI-CMPL")	\
+	EM(netfs_rreq_trace_ra_put_ref,		"RA-PUT ")	\
 	EM(netfs_rreq_trace_recollect,		"RECLLCT")	\
 	EM(netfs_rreq_trace_redirty,		"REDIRTY")	\
 	EM(netfs_rreq_trace_resubmit,		"RESUBMT")	\
@@ -70,9 +71,11 @@
 	EM(netfs_rreq_trace_unpause,		"UNPAUSE")	\
 	EM(netfs_rreq_trace_wait_ip,		"WAIT-IP")	\
 	EM(netfs_rreq_trace_wait_pause,		"--PAUSED--")	\
+	EM(netfs_rreq_trace_wait_put_ra_refs,	"WAIT-P-RA")	\
 	EM(netfs_rreq_trace_wait_quiesce,	"WAIT-QUIESCE")	\
 	EM(netfs_rreq_trace_waited_ip,		"DONE-IP")	\
 	EM(netfs_rreq_trace_waited_pause,	"--UNPAUSED--")	\
+	EM(netfs_rreq_trace_waited_put_ra_refs,	"DONE-P-RA")	\
 	EM(netfs_rreq_trace_waited_quiesce,	"DONE-QUIESCE")	\
 	EM(netfs_rreq_trace_wake_ip,		"WAKE-IP")	\
 	EM(netfs_rreq_trace_wake_queue,		"WAKE-Q ")	\

[PATCH v4 08/10] netfs: Mark folios with COPY_TO_CACHE whilst issuing subreqs

[David Howells]

Mark folios with NETFS_FOLIO_COPY_TO_CACHE whilst issuing subreqs rather than
when collecting them.  This means that the collector thread doesn't have to
try and keep track of which subreqs contribute to which folios - and thus
which folios will need to be copied to the cache because at least one byte
wasn't in the cache.  Instead, this is marked on the folios up front and the
collector need only consider the folios.

For PG_private_2-using filesystems, PG_private_2 is set instead of
NETFS_FOLIO_COPY_TO_CACHE, but otherwise it works the same.

The NETFS_RREQ_COPY_TO_CACHE is replaced with NETFS_RREQ_CANCEL_CACHING, which
is now set if caching fails somewhere, thereby causing the collection thread
to cancel the copy-to-cache marks on the remaining folios.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara (Red Hat) <pc@manguebit.org>
cc: Matthew Wilcox <willy@infradead.org>
cc: netfs@lists.linux.dev
cc: linux-mm@kvack.org
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/buffered_read.c     | 61 +++++++++++++++++++++++++++++++-
 fs/netfs/internal.h          |  1 +
 fs/netfs/read_collect.c      | 67 ++++++++++++++++++++++--------------
 fs/netfs/read_pgpriv2.c      | 15 ++++----
 fs/netfs/read_retry.c        |  6 +++-
 include/linux/netfs.h        |  2 +-
 include/trace/events/netfs.h |  6 ++--
 7 files changed, 121 insertions(+), 37 deletions(-)

diff --git a/fs/netfs/buffered_read.c b/fs/netfs/buffered_read.c
index 303fdce54fba..16d4db776f6a 100644
--- a/fs/netfs/buffered_read.c
+++ b/fs/netfs/buffered_read.c
@@ -211,6 +211,56 @@ static void netfs_issue_read(struct netfs_io_request *rreq,
 	}
 }
 
+/*
+ * Mark folios that we want to copy to the cache.  For filesystems that use
+ * netfslib fully, we set folio->private to NETFS_FOLIO_COPY_TO_CACHE;
+ * otherwise we set the deprecated PG_private_2.
+ */
+static void netfs_mark_copy_to_cache(struct netfs_io_request *rreq,
+				     struct folio_queue **fq,
+				     unsigned int *offset,
+				     int *slot,
+				     size_t len,
+				     bool copy)
+{
+	while (len > 0) {
+		struct folio *folio;
+		size_t fsize, overlap;
+
+		if (!*fq)
+			break;
+		if (*slot >= folioq_count(*fq)) {
+			*fq = (*fq)->next;
+			*slot = 0;
+			*offset = 0;
+			continue;
+		}
+
+		/* Determine how much the subreq overlaps the folio, if at all. */
+		fsize = folioq_folio_size(*fq, *slot);
+		overlap = min(len, fsize - *offset);
+
+		if (overlap > 0 && copy) {
+			folio = folioq_folio(*fq, *slot);
+			if (unlikely(test_bit(NETFS_RREQ_USE_PGPRIV2, &rreq->flags))) {
+				if (!folio_test_private_2(folio))
+					folio_start_private_2(folio);
+			} else {
+				if (!folio_get_private(folio))
+					folio_attach_private(folio, NETFS_FOLIO_COPY_TO_CACHE);
+			}
+			trace_netfs_folio(folio, netfs_folio_trace_mark_copy);
+		}
+
+		len -= overlap;
+		*offset += overlap;
+		if (*offset >= fsize) {
+			*slot += 1;
+			*offset = 0;
+		}
+	}
+}
+
 /*
  * Perform a read to the pagecache from a series of sources of different types,
  * slicing up the region to be read according to available cache blocks and
@@ -218,9 +268,11 @@ static void netfs_issue_read(struct netfs_io_request *rreq,
  */
 static void netfs_read_to_pagecache(struct netfs_io_request *rreq)
 {
+	struct folio_queue *fq = rreq->buffer.tail;
 	unsigned long long start = rreq->start;
+	unsigned int offset = 0;
 	ssize_t size = rreq->len;
-	int ret = 0;
+	int ret = 0, slot = 0;
 
 	do {
 		struct netfs_io_subrequest *subreq;
@@ -308,6 +360,13 @@ static void netfs_read_to_pagecache(struct netfs_io_request *rreq)
 			set_bit(NETFS_RREQ_ALL_QUEUED, &rreq->flags);
 		}
 
+		if (fq) {
+			/* See if the cache indicated this should be cached. */
+			bool copy = test_bit(NETFS_SREQ_COPY_TO_CACHE, &subreq->flags);
+
+			netfs_mark_copy_to_cache(rreq, &fq, &slot, &offset, slice, copy);
+		}
+
 		netfs_issue_read(rreq, subreq);
 		netfs_maybe_bulk_drop_ra_refs(rreq);
 
diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h
index bd8b2d633f96..dfe7939f35f3 100644
--- a/fs/netfs/internal.h
+++ b/fs/netfs/internal.h
@@ -110,6 +110,7 @@ static inline void netfs_see_subrequest(struct netfs_io_subrequest *subreq,
 /*
  * read_collect.c
  */
+void netfs_cancel_copy_to_cache(struct netfs_io_request *rreq, struct folio *folio);
 bool netfs_read_collection(struct netfs_io_request *rreq);
 void netfs_read_collection_worker(struct work_struct *work);
 void netfs_cancel_read(struct netfs_io_subrequest *subreq, int error);
diff --git a/fs/netfs/read_collect.c b/fs/netfs/read_collect.c
index edf7cea7e2f9..12a786be1ea2 100644
--- a/fs/netfs/read_collect.c
+++ b/fs/netfs/read_collect.c
@@ -19,7 +19,6 @@
 #define MADE_PROGRESS	0x04	/* Made progress cleaning up a stream or the folio set */
 #define BUFFERED	0x08	/* The pagecache needs cleaning up */
 #define NEED_RETRY	0x10	/* A front op requests retrying */
-#define COPY_TO_CACHE	0x40	/* Need to copy subrequest to cache */
 #define ABANDON_SREQ	0x80	/* Need to abandon untransferred part of subrequest */
 
 /*
@@ -34,6 +33,30 @@ static void netfs_clear_unread(struct netfs_io_subrequest *subreq)
 		__set_bit(NETFS_SREQ_HIT_EOF, &subreq->flags);
 }
 
+/*
+ * Cancel the copy-to-cache mark on a folio.
+ */
+void netfs_cancel_copy_to_cache(struct netfs_io_request *rreq, struct folio *folio)
+{
+	if (!test_bit(NETFS_RREQ_USE_PGPRIV2, &rreq->flags)) {
+		if (folio_get_private(folio) == NETFS_FOLIO_COPY_TO_CACHE) {
+			folio_detach_private(folio);
+			trace_netfs_folio(folio, netfs_folio_trace_cancel_copy);
+		} else if (netfs_folio_group(folio) == NETFS_FOLIO_COPY_TO_CACHE)  {
+			struct netfs_folio *finfo = netfs_folio_info(folio);
+
+			finfo->netfs_group = NULL;
+			trace_netfs_folio(folio, netfs_folio_trace_cancel_copy);
+		}
+	} else {
+		// TODO: Use of PG_private_2 is deprecated.
+		if (folio_test_private_2(folio)) {
+			folio_end_private_2(folio);
+			trace_netfs_folio(folio, netfs_folio_trace_cancel_copy);
+		}
+	}
+}
+
 /*
  * Flush, mark and unlock a folio that's now completely read.  If we want to
  * cache the folio, we set the group to NETFS_FOLIO_COPY_TO_CACHE, mark it
@@ -48,37 +71,37 @@ static void netfs_unlock_read_folio(struct netfs_io_request *rreq,
 
 	if (unlikely(folio_pos(folio) < rreq->abandon_to)) {
 		trace_netfs_folio(folio, netfs_folio_trace_abandon);
+		netfs_cancel_copy_to_cache(rreq, folio);
 		goto just_unlock;
 	}
 
 	flush_dcache_folio(folio);
 	folio_mark_uptodate(folio);
 
-	if (!test_bit(NETFS_RREQ_USE_PGPRIV2, &rreq->flags)) {
-		finfo = netfs_folio_info(folio);
-		if (finfo) {
-			trace_netfs_folio(folio, netfs_folio_trace_filled_gaps);
-			if (finfo->netfs_group)
-				folio_change_private(folio, finfo->netfs_group);
-			else
-				folio_detach_private(folio);
-			kfree(finfo);
-		}
+	if (unlikely(test_bit(NETFS_RREQ_CANCEL_CACHING, &rreq->flags)))
+		netfs_cancel_copy_to_cache(rreq, folio);
 
-		if (test_bit(NETFS_RREQ_FOLIO_COPY_TO_CACHE, &rreq->flags)) {
-			if (!WARN_ON_ONCE(folio_get_private(folio) != NULL)) {
-				trace_netfs_folio(folio, netfs_folio_trace_copy_to_cache);
-				folio_attach_private(folio, NETFS_FOLIO_COPY_TO_CACHE);
-				folio_mark_dirty(folio);
-			}
+	if (!test_bit(NETFS_RREQ_USE_PGPRIV2, &rreq->flags)) {
+		if (netfs_folio_group(folio) == NETFS_FOLIO_COPY_TO_CACHE)  {
+			trace_netfs_folio(folio, netfs_folio_trace_sched_copy);
+			folio_mark_dirty(folio);
 		} else {
+			finfo = netfs_folio_info(folio);
+			if (finfo) {
+				trace_netfs_folio(folio, netfs_folio_trace_filled_gaps);
+				if (finfo->netfs_group)
+					folio_change_private(folio, finfo->netfs_group);
+				else
+					folio_detach_private(folio);
+				kfree(finfo);
+			}
 			trace_netfs_folio(folio, netfs_folio_trace_read_done);
 		}
 
 		folioq_clear(folioq, slot);
 	} else {
 		// TODO: Use of PG_private_2 is deprecated.
-		if (test_bit(NETFS_RREQ_FOLIO_COPY_TO_CACHE, &rreq->flags))
+		if (folio_test_private_2(folio))
 			netfs_pgpriv2_copy_to_cache(rreq, folio);
 	}
 
@@ -131,9 +154,6 @@ static void netfs_read_unlock_folios(struct netfs_io_request *rreq,
 		unsigned int order;
 		size_t fsize;
 
-		if (*notes & COPY_TO_CACHE)
-			set_bit(NETFS_RREQ_FOLIO_COPY_TO_CACHE, &rreq->flags);
-
 		folio = folioq_folio(folioq, slot);
 		if (WARN_ONCE(!folio_test_locked(folio),
 			      "R=%08x: folio %lx is not locked\n",
@@ -156,8 +176,6 @@ static void netfs_read_unlock_folios(struct netfs_io_request *rreq,
 		WRITE_ONCE(rreq->cleaned_to, fpos + fsize);
 		*notes |= MADE_PROGRESS;
 
-		clear_bit(NETFS_RREQ_FOLIO_COPY_TO_CACHE, &rreq->flags);
-
 		/* Clean up the head folioq.  If we clear an entire folioq, then
 		 * we can get rid of it provided it's not also the tail folioq
 		 * being filled by the issuer.
@@ -255,9 +273,6 @@ static void netfs_collect_read_results(struct netfs_io_request *rreq)
 			stream->collected_to = front->start + transferred;
 			rreq->collected_to = stream->collected_to;
 
-			if (test_bit(NETFS_SREQ_COPY_TO_CACHE, &front->flags))
-				notes |= COPY_TO_CACHE;
-
 			if (test_bit(NETFS_SREQ_FAILED, &front->flags)) {
 				rreq->abandon_to = front->start + front->len;
 				front->transferred = front->len;
diff --git a/fs/netfs/read_pgpriv2.c b/fs/netfs/read_pgpriv2.c
index c31190993b76..a4b7bb88cbdb 100644
--- a/fs/netfs/read_pgpriv2.c
+++ b/fs/netfs/read_pgpriv2.c
@@ -54,8 +54,8 @@ static void netfs_pgpriv2_copy_folio(struct netfs_io_request *creq, struct folio
 
 	/* Attach the folio to the rolling buffer. */
 	if (rolling_buffer_append(&creq->buffer, folio, 0, creq->gfp) < 0) {
+		set_bit(NETFS_RREQ_CANCEL_CACHING, &creq->flags);
 		folio_end_private_2(folio);
-		clear_bit(NETFS_RREQ_FOLIO_COPY_TO_CACHE, &creq->flags);
 		return;
 	}
 
@@ -122,13 +122,14 @@ static struct netfs_io_request *netfs_pgpriv2_begin_copy_to_cache(
 	netfs_put_failed_request(creq);
 cancel:
 	rreq->copy_to_cache = ERR_PTR(-ENOBUFS);
-	clear_bit(NETFS_RREQ_FOLIO_COPY_TO_CACHE, &rreq->flags);
+	set_bit(NETFS_RREQ_CANCEL_CACHING, &rreq->flags);
 	return ERR_PTR(-ENOBUFS);
 }
 
 /*
  * [DEPRECATED] Mark page as requiring copy-to-cache using PG_private_2 and add
- * it to the copy write request.
+ * it to the copy write request.  PG_private_2 should already be set on the
+ * folio.
  */
 void netfs_pgpriv2_copy_to_cache(struct netfs_io_request *rreq, struct folio *folio)
 {
@@ -136,11 +137,13 @@ void netfs_pgpriv2_copy_to_cache(struct netfs_io_request *rreq, struct folio *fo
 
 	if (!creq)
 		creq = netfs_pgpriv2_begin_copy_to_cache(rreq, folio);
-	if (IS_ERR(creq))
+	if (IS_ERR(creq)) {
+		set_bit(NETFS_RREQ_CANCEL_CACHING, &rreq->flags);
+		netfs_cancel_copy_to_cache(rreq, folio);
 		return;
+	}
 
-	trace_netfs_folio(folio, netfs_folio_trace_copy_to_cache);
-	folio_start_private_2(folio);
+	trace_netfs_folio(folio, netfs_folio_trace_pgpriv2_copy);
 	netfs_pgpriv2_copy_folio(creq, folio);
 }
 
diff --git a/fs/netfs/read_retry.c b/fs/netfs/read_retry.c
index dd463a485139..4f6a36c6e214 100644
--- a/fs/netfs/read_retry.c
+++ b/fs/netfs/read_retry.c
@@ -303,7 +303,11 @@ void netfs_unlock_abandoned_read_pages(struct netfs_io_request *rreq)
 		for (int slot = 0; slot < folioq_count(p); slot++) {
 			struct folio *folio = folioq_folio(p, slot);
 
-			if (folio && !folioq_is_marked2(p, slot)) {
+			if (!folio)
+				continue;
+			netfs_cancel_copy_to_cache(rreq, folio);
+
+			if (!folioq_is_marked2(p, slot)) {
 				if (folio == rreq->no_unlock_folio &&
 				    test_bit(NETFS_RREQ_NO_UNLOCK_FOLIO,
 					     &rreq->flags)) {
diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index 5c538d0c5d79..9881f4afdc0c 100644
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -275,7 +275,7 @@ struct netfs_io_request {
 #define NETFS_RREQ_SHORT_TRANSFER	5	/* Set if we have a short transfer */
 #define NETFS_RREQ_OFFLOAD_COLLECTION	8	/* Offload collection to workqueue */
 #define NETFS_RREQ_NO_UNLOCK_FOLIO	9	/* Don't unlock no_unlock_folio on completion */
-#define NETFS_RREQ_FOLIO_COPY_TO_CACHE	10	/* Copy current folio to cache from read */
+#define NETFS_RREQ_CANCEL_CACHING	10	/* Set to cancel caching */
 #define NETFS_RREQ_UPLOAD_TO_SERVER	11	/* Need to write to the server */
 #define NETFS_RREQ_USE_IO_ITER		12	/* Use ->io_iter rather than ->i_pages */
 #define NETFS_RREQ_NEED_PUT_RA_REFS	17	/* Need to put the folio refs RA gave us */
diff --git a/include/trace/events/netfs.h b/include/trace/events/netfs.h
index 9bda9302be90..a22084813cb5 100644
--- a/include/trace/events/netfs.h
+++ b/include/trace/events/netfs.h
@@ -198,7 +198,6 @@
 	EM(netfs_folio_trace_clear_cc,		"clear-cc")	\
 	EM(netfs_folio_trace_clear_g,		"clear-g")	\
 	EM(netfs_folio_trace_clear_s,		"clear-s")	\
-	EM(netfs_folio_trace_copy_to_cache,	"mark-copy")	\
 	EM(netfs_folio_trace_end_copy,		"end-copy")	\
 	EM(netfs_folio_trace_filled_gaps,	"filled-gaps")	\
 	EM(netfs_folio_trace_invalidate_all,	"inval-all")	\
@@ -209,16 +208,19 @@
 	EM(netfs_folio_trace_kill_cc,		"kill-cc")	\
 	EM(netfs_folio_trace_kill_g,		"kill-g")	\
 	EM(netfs_folio_trace_kill_s,		"kill-s")	\
+	EM(netfs_folio_trace_mark_copy,		"mark-copy")	\
 	EM(netfs_folio_trace_mkwrite,		"mkwrite")	\
 	EM(netfs_folio_trace_mkwrite_plus,	"mkwrite+")	\
-	EM(netfs_folio_trace_not_under_wback,	"!wback")	\
 	EM(netfs_folio_trace_not_locked,	"!locked")	\
+	EM(netfs_folio_trace_not_under_wback,	"!wback")	\
+	EM(netfs_folio_trace_pgpriv2_copy,	"pgpriv2-copy")	\
 	EM(netfs_folio_trace_put,		"put")		\
 	EM(netfs_folio_trace_read,		"read")		\
 	EM(netfs_folio_trace_read_done,		"read-done")	\
 	EM(netfs_folio_trace_read_gaps,		"read-gaps")	\
 	EM(netfs_folio_trace_read_unlock,	"read-unlock")	\
 	EM(netfs_folio_trace_redirtied,		"redirtied")	\
+	EM(netfs_folio_trace_sched_copy,	"sched-copy")	\
 	EM(netfs_folio_trace_store,		"store")	\
 	EM(netfs_folio_trace_store_copy,	"store-copy")	\
 	EM(netfs_folio_trace_store_plus,	"store+")	\

[PATCH v4 09/10] netfs: Fix read progress reporting

[David Howells]

For really big read RPC ops that span multiple folios, netfslib allows the
filesystem to give progress notifications to wake up the collector thread
to do a collection of folios that have now been fetched, even if the RPC is
still ongoing, thereby allowing the application to make progress.

This works by taking the current rreq->cleaned_to value (which indicates
which folios have been unlocked) and adding the stashed size of the next
folio to it.  cleaned_to, however, is subject to 64-bit tearing on a 32-bit
arch.

Fix this by stashing the next progress notification point as a size_t
(which won't tear) to be added to rreq->start (which won't change), with
the collector thread calculating that from cleaned_to plus the next folio
size.

Further, however, if the folios are small, the collector thread gets
constantly woken up - which has a negative performance impact on the
system.

Fix that too by setting a minimum trigger of 256KiB or the size of the
folio at the front of the queue, whichever is larger.  Note that this has
an issue that different subreqs have different need-to-be-cached
properties; this is solved by a preceding patch that marks the property on
the folios whilst issuing subreqs rather than when collecting them.

Also, make sure rreq->cleaned_to is initialised up front, along with
rreq->collected_to and stream->collected_to.

Fixes: e2d46f2ec332 ("netfs: Change the read result collector to only use one work item")
Link: https://sashiko.dev/#/patchset/20260804100224.2748935-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara <pc@manguebit.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
---
 fs/netfs/buffered_read.c     |  2 ++
 fs/netfs/internal.h          |  1 +
 fs/netfs/objects.c           | 32 +++++++++++++--------
 fs/netfs/read_collect.c      | 54 ++++++++++++++++++++++++++++--------
 fs/netfs/read_single.c       |  2 ++
 include/linux/netfs.h        |  2 +-
 include/trace/events/netfs.h | 21 ++++++++++++++
 7 files changed, 89 insertions(+), 25 deletions(-)

diff --git a/fs/netfs/buffered_read.c b/fs/netfs/buffered_read.c
index 16d4db776f6a..424df70a5c30 100644
--- a/fs/netfs/buffered_read.c
+++ b/fs/netfs/buffered_read.c
@@ -442,6 +442,7 @@ void netfs_readahead(struct readahead_control *ractl)
 
 	rreq->submitted = rreq->start + added;
 	rreq->cleaned_to = rreq->start;
+	netfs_read_set_unlock_at(rreq);
 
 	netfs_read_to_pagecache(rreq);
 	netfs_maybe_bulk_drop_ra_refs(rreq);
@@ -467,6 +468,7 @@ static int netfs_create_singular_buffer(struct netfs_io_request *rreq, struct fo
 	if (added < 0)
 		return added;
 	rreq->submitted = rreq->start + added;
+	rreq->progress_at = added;
 	return 0;
 }
 
diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h
index dfe7939f35f3..c79c8e69d60c 100644
--- a/fs/netfs/internal.h
+++ b/fs/netfs/internal.h
@@ -111,6 +111,7 @@ static inline void netfs_see_subrequest(struct netfs_io_subrequest *subreq,
  * read_collect.c
  */
 void netfs_cancel_copy_to_cache(struct netfs_io_request *rreq, struct folio *folio);
+void netfs_read_set_unlock_at(struct netfs_io_request *rreq);
 bool netfs_read_collection(struct netfs_io_request *rreq);
 void netfs_read_collection_worker(struct work_struct *work);
 void netfs_cancel_read(struct netfs_io_subrequest *subreq, int error);
diff --git a/fs/netfs/objects.c b/fs/netfs/objects.c
index 01461a74642d..7f6a3e912602 100644
--- a/fs/netfs/objects.c
+++ b/fs/netfs/objects.c
@@ -41,24 +41,32 @@ struct netfs_io_request *netfs_alloc_request(struct address_space *mapping,
 
 	memset(rreq, 0, kmem_cache_size(cache));
 	INIT_WORK(&rreq->cleanup_work, netfs_free_request);
-	rreq->gfp	= gfp;
-	rreq->start	= start;
-	rreq->len	= len;
-	rreq->origin	= origin;
-	rreq->netfs_ops	= ctx->ops;
-	rreq->mapping	= mapping;
-	rreq->inode	= inode;
-	rreq->i_size	= i_size_read(inode);
-	rreq->debug_id	= atomic_inc_return(&debug_ids);
-	rreq->wsize	= INT_MAX;
+	rreq->gfp		= gfp;
+	rreq->start		= start;
+	rreq->collected_to	= start;
+	rreq->cleaned_to	= start;
+	rreq->len		= len;
+	rreq->progress_at	= 0;
+	rreq->origin		= origin;
+	rreq->netfs_ops		= ctx->ops;
+	rreq->mapping		= mapping;
+	rreq->inode		= inode;
+	rreq->i_size		= i_size_read(inode);
+	rreq->debug_id		= atomic_inc_return(&debug_ids);
+	rreq->wsize		= INT_MAX;
 	rreq->io_streams[0].sreq_max_len = ULONG_MAX;
 	rreq->io_streams[0].sreq_max_segs = 0;
 	spin_lock_init(&rreq->lock);
-	INIT_LIST_HEAD(&rreq->io_streams[0].subrequests);
-	INIT_LIST_HEAD(&rreq->io_streams[1].subrequests);
 	init_waitqueue_head(&rreq->waitq);
 	refcount_set(&rreq->ref, 2);
 
+	for (int s = 0; s < NR_IO_STREAMS; s++) {
+		struct netfs_io_stream *stream = &rreq->io_streams[s];
+
+		INIT_LIST_HEAD(&stream->subrequests);
+		stream->collected_to = rreq->start;
+	}
+
 	if (origin == NETFS_READAHEAD ||
 	    origin == NETFS_READPAGE ||
 	    origin == NETFS_READ_GAPS ||
diff --git a/fs/netfs/read_collect.c b/fs/netfs/read_collect.c
index 12a786be1ea2..5cf22087d243 100644
--- a/fs/netfs/read_collect.c
+++ b/fs/netfs/read_collect.c
@@ -117,6 +117,35 @@ static void netfs_unlock_read_folio(struct netfs_io_request *rreq,
 	folioq_clear(folioq, slot);
 }
 
+/*
+ * Determine how much to gather before unlocking more folios.
+ */
+void netfs_read_set_unlock_at(struct netfs_io_request *rreq)
+{
+	struct folio_queue *folioq = rreq->buffer.tail;
+	unsigned int slot = rreq->buffer.first_tail_slot;
+	size_t cleaned_to = rreq->cleaned_to - rreq->start;
+	size_t progress_at = cleaned_to;
+	size_t minimum = 256 * 1024;
+
+	while (progress_at < rreq->len) {
+		if (slot >= folioq_count(folioq)) {
+			folioq = folioq->next;
+			if (!folioq)
+				break;
+			slot = 0;
+		}
+
+		progress_at += folioq_folio_size(folioq, slot);
+		if (progress_at - cleaned_to >= minimum)
+			break;
+		slot++;
+	}
+
+	WRITE_ONCE(rreq->progress_at, progress_at);
+	trace_netfs_read_progress_at(rreq);
+}
+
 /*
  * Unlock any folios we've finished with.
  */
@@ -135,7 +164,7 @@ static void netfs_read_unlock_folios(struct netfs_io_request *rreq,
 	if (slot >= folioq_nr_slots(folioq)) {
 		folioq = rolling_buffer_delete_spent(&rreq->buffer);
 		if (!folioq) {
-			rreq->front_folio_order = 0;
+			WRITE_ONCE(rreq->progress_at, rreq->len);
 			return;
 		}
 		slot = 0;
@@ -151,7 +180,6 @@ static void netfs_read_unlock_folios(struct netfs_io_request *rreq,
 	for (;;) {
 		struct folio *folio;
 		unsigned long long fpos, fend;
-		unsigned int order;
 		size_t fsize;
 
 		folio = folioq_folio(folioq, slot);
@@ -160,9 +188,7 @@ static void netfs_read_unlock_folios(struct netfs_io_request *rreq,
 			      rreq->debug_id, folio->index))
 			trace_netfs_folio(folio, netfs_folio_trace_not_locked);
 
-		order = folioq_folio_order(folioq, slot);
-		rreq->front_folio_order = order;
-		fsize = PAGE_SIZE << order;
+		fsize = folioq_folio_size(folioq, slot);
 		fpos = folio_pos(folio);
 		fend = fpos + fsize;
 
@@ -197,6 +223,8 @@ static void netfs_read_unlock_folios(struct netfs_io_request *rreq,
 	rreq->buffer.tail = folioq;
 done:
 	rreq->buffer.first_tail_slot = slot;
+
+	netfs_read_set_unlock_at(rreq);
 }
 
 /*
@@ -257,7 +285,7 @@ static void netfs_collect_read_results(struct netfs_io_request *rreq)
 		 * subreqs.
 		 */
 		if (notes & BUFFERED) {
-			size_t fsize = PAGE_SIZE << rreq->front_folio_order;
+			uoff_t unlock_at = rreq->start + rreq->progress_at;
 
 			/* Clear the tail of a short read. */
 			if (!(notes & HIT_PENDING) &&
@@ -279,7 +307,7 @@ static void netfs_collect_read_results(struct netfs_io_request *rreq)
 				transferred = front->len;
 				trace_netfs_rreq(rreq, netfs_rreq_trace_set_abandon);
 			}
-			if (front->start + transferred >= rreq->cleaned_to + fsize ||
+			if (front->start + transferred >= unlock_at ||
 			    test_bit(NETFS_SREQ_HIT_EOF, &front->flags))
 				netfs_read_unlock_folios(rreq, &notes);
 		} else {
@@ -499,20 +527,22 @@ void netfs_read_collection_worker(struct work_struct *work)
 void netfs_read_subreq_progress(struct netfs_io_subrequest *subreq)
 {
 	struct netfs_io_request *rreq = subreq->rreq;
-	struct netfs_io_stream *stream = &rreq->io_streams[0];
-	size_t fsize = PAGE_SIZE << rreq->front_folio_order;
-
-	trace_netfs_sreq(subreq, netfs_sreq_trace_progress);
+	struct netfs_io_stream *stream = &rreq->io_streams[subreq->stream_nr];
+	size_t progress_at = READ_ONCE(rreq->progress_at);
+	uoff_t update_at = rreq->start + progress_at;
+	uoff_t transferred_to = subreq->start + subreq->transferred;
 
 	/* If we are at the head of the queue, wake up the collector,
 	 * getting a ref to it if we were the ones to do so.
 	 */
-	if (subreq->start + subreq->transferred > rreq->cleaned_to + fsize &&
+	if (progress_at < rreq->len &&
+	    transferred_to >= update_at &&
 	    (rreq->origin == NETFS_READAHEAD ||
 	     rreq->origin == NETFS_READPAGE ||
 	     rreq->origin == NETFS_READ_FOR_WRITE) &&
 	    list_is_first(&subreq->rreq_link, &stream->subrequests)
 	    ) {
+		trace_netfs_sreq(subreq, netfs_sreq_trace_progress);
 		__set_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
 		netfs_wake_collector(rreq);
 	}
diff --git a/fs/netfs/read_single.c b/fs/netfs/read_single.c
index 8833550d2eb6..de67ac41548d 100644
--- a/fs/netfs/read_single.c
+++ b/fs/netfs/read_single.c
@@ -170,6 +170,8 @@ ssize_t netfs_read_single(struct inode *inode, struct file *file, struct iov_ite
 	if (IS_ERR(rreq))
 		return PTR_ERR(rreq);
 
+	rreq->progress_at = rreq->len;
+
 	ret = netfs_single_begin_cache_read(rreq, ictx);
 	if (ret == -ENOMEM || ret == -EINTR || ret == -ERESTARTSYS)
 		goto cleanup_free;
diff --git a/include/linux/netfs.h b/include/linux/netfs.h
index 9881f4afdc0c..b4dd32863dd4 100644
--- a/include/linux/netfs.h
+++ b/include/linux/netfs.h
@@ -246,6 +246,7 @@ struct netfs_io_request {
 	unsigned long long	submitted;	/* Amount submitted for I/O so far */
 	unsigned long long	len;		/* Length of the request */
 	size_t			transferred;	/* Amount to be indicated as transferred */
+	size_t			progress_at;	/* Report read progress when hit this much read */
 	long			error;		/* 0 or error that occurred */
 	unsigned long long	i_size;		/* Size of the file */
 	unsigned long long	start;		/* Start position */
@@ -262,7 +263,6 @@ struct netfs_io_request {
 	atomic_t		subreq_counter;	/* Next subreq->debug_index */
 	unsigned int		nr_group_rel;	/* Number of refs to release on ->group */
 	spinlock_t		lock;		/* Lock for queuing subreqs */
-	unsigned char		front_folio_order; /* Order (size) of front folio */
 	enum netfs_io_origin	origin;		/* Origin of the request */
 	bool			direct_bv_unpin; /* T if direct_bv[] must be unpinned */
 	refcount_t		ref;
diff --git a/include/trace/events/netfs.h b/include/trace/events/netfs.h
index a22084813cb5..3fec3e8f91c8 100644
--- a/include/trace/events/netfs.h
+++ b/include/trace/events/netfs.h
@@ -791,6 +791,27 @@ TRACE_EVENT(netfs_folioq,
 		      __print_symbolic(__entry->trace, netfs_folioq_traces))
 	    );
 
+TRACE_EVENT(netfs_read_progress_at,
+	    TP_PROTO(const struct netfs_io_request *rreq),
+
+	    TP_ARGS(rreq),
+
+	    TP_STRUCT__entry(
+		    __field(unsigned int,	rreq)
+		    __field(size_t,		progress_at)
+		    __field(size_t,		cleaned_to)
+			     ),
+
+	    TP_fast_assign(
+		    __entry->rreq	= rreq->debug_id;
+		    __entry->cleaned_to	= rreq->cleaned_to - rreq->start;
+		    __entry->progress_at = rreq->progress_at;
+			   ),
+
+	    TP_printk("R=%08x cln=%zx prg=%zx",
+		      __entry->rreq, __entry->cleaned_to, __entry->progress_at)
+	    );
+
 #undef EM
 #undef E_
 #endif /* _TRACE_NETFS_H */

[PATCH v4 10/10] cachefiles: Fix potential UAF/KASAN warning

[David Howells]

Currently, trace_cachefiles_coherency() is being passed a pointer to a
__be64 lain over the coherency data in struct cachefiles_xattr so that it
can display the first 8 bytes.  However, the data is of variable length and
could even be 0 bytes.  This could lead to a UAF or KASAN warning.

Fix this by making sure the buffer has room for at least 8 bytes and that
those 8 bytes are pre-cleared.

Further, those bytes are not 8-byte aligned, so fix the tracepoint to
extract the data as four 2-byte words (they are 2-byte aligned) and
reassemble the __be64.  The compiler will convert this into a single 8-byte
load where the CPU supports it.

Fixes: 229105e5cfd9 ("cachefiles: Add auxiliary data trace")
Link: https://sashiko.dev/#/patchset/20260810144746.574036-1-dhowells%40redhat.com
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Paulo Alcantara <pc@manguebit.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
---
 fs/cachefiles/xattr.c             | 16 ++++++++--------
 include/trace/events/cachefiles.h | 19 +++++++++++++++++--
 2 files changed, 25 insertions(+), 10 deletions(-)

diff --git a/fs/cachefiles/xattr.c b/fs/cachefiles/xattr.c
index f8ae78b3f7b6..c70bf67e52b0 100644
--- a/fs/cachefiles/xattr.c
+++ b/fs/cachefiles/xattr.c
@@ -13,6 +13,7 @@
 #include <linux/quotaops.h>
 #include <linux/xattr.h>
 #include <linux/slab.h>
+#include <linux/unaligned.h>
 #include "internal.h"
 
 #define CACHEFILES_COOKIE_TYPE_DATA 1
@@ -50,7 +51,7 @@ int cachefiles_set_object_xattr(struct cachefiles_object *object)
 
 	_enter("%x,#%d", object->debug_id, len);
 
-	buf = kmalloc(sizeof(struct cachefiles_xattr) + len, GFP_KERNEL);
+	buf = kmalloc(sizeof(struct cachefiles_xattr) + max(len, sizeof(__be64)), GFP_KERNEL);
 	if (!buf)
 		return -ENOMEM;
 
@@ -60,6 +61,7 @@ int cachefiles_set_object_xattr(struct cachefiles_object *object)
 	buf->content		= object->content_info;
 	if (test_bit(FSCACHE_COOKIE_LOCAL_WRITE, &object->cookie->flags))
 		buf->content	= CACHEFILES_CONTENT_DIRTY;
+	put_unaligned_be64(0, (__be64 *)buf->data);
 	if (len > 0)
 		memcpy(buf->data, fscache_get_aux(object->cookie), len);
 
@@ -77,8 +79,7 @@ int cachefiles_set_object_xattr(struct cachefiles_object *object)
 		trace_cachefiles_vfs_error(object, file_inode(file), ret,
 					   cachefiles_trace_setxattr_error);
 		trace_cachefiles_coherency(object, file_inode(file)->i_ino,
-					   be64_to_cpup((__be64 *)buf->data),
-					   buf->content,
+					   buf->data, buf->content,
 					   cachefiles_coherency_set_fail);
 		if (ret != -ENOMEM)
 			cachefiles_io_error_obj(
@@ -86,8 +87,7 @@ int cachefiles_set_object_xattr(struct cachefiles_object *object)
 				"Failed to set xattr with error %d", ret);
 	} else {
 		trace_cachefiles_coherency(object, file_inode(file)->i_ino,
-					   be64_to_cpup((__be64 *)buf->data),
-					   buf->content,
+					   buf->data, buf->content,
 					   cachefiles_coherency_set_ok);
 	}
 
@@ -110,9 +110,10 @@ int cachefiles_check_auxdata(struct cachefiles_object *object, struct file *file
 	int ret = -ESTALE;
 
 	tlen = sizeof(struct cachefiles_xattr) + len;
-	buf = kmalloc(tlen, GFP_KERNEL);
+	buf = kmalloc(sizeof(struct cachefiles_xattr) + max(len, sizeof(__be64)), GFP_KERNEL);
 	if (!buf)
 		return -ENOMEM;
+	put_unaligned_be64(0, (__be64 *)buf->data);
 
 	xlen = cachefiles_inject_read_error();
 	if (xlen == 0)
@@ -148,8 +149,7 @@ int cachefiles_check_auxdata(struct cachefiles_object *object, struct file *file
 
 out:
 	trace_cachefiles_coherency(object, file_inode(file)->i_ino,
-				   be64_to_cpup((__be64 *)buf->data),
-				   buf->content, why);
+				   buf->data, buf->content, why);
 	kfree(buf);
 	return ret;
 }
diff --git a/include/trace/events/cachefiles.h b/include/trace/events/cachefiles.h
index 9259bc71049e..e3101410e8b2 100644
--- a/include/trace/events/cachefiles.h
+++ b/include/trace/events/cachefiles.h
@@ -372,7 +372,7 @@ TRACE_EVENT(cachefiles_rename,
 TRACE_EVENT(cachefiles_coherency,
 	    TP_PROTO(struct cachefiles_object *obj,
 		     ino_t ino,
-		     u64 disk_aux,
+		     const void *disk_aux,
 		     enum cachefiles_content content,
 		     enum cachefiles_coherency_trace why),
 
@@ -389,12 +389,27 @@ TRACE_EVENT(cachefiles_coherency,
 			     ),
 
 	    TP_fast_assign(
+		    union {
+			    __be16 s[4];
+			    __be64 ll;
+		    } x;
+
 		    __entry->obj	= obj->debug_id;
 		    __entry->why	= why;
 		    __entry->content	= content;
 		    __entry->ino	= ino;
 		    __entry->aux	= be64_to_cpup((__be64 *)obj->cookie->inline_aux);
-		    __entry->disk_aux	= disk_aux;
+
+		    /* cachefiles_xattr::data is 2-byte aligned but not 8-byte aligned. */
+		    if (disk_aux) {
+			    x.s[0] = ((__be16 *)disk_aux)[0];
+			    x.s[1] = ((__be16 *)disk_aux)[1];
+			    x.s[2] = ((__be16 *)disk_aux)[2];
+			    x.s[3] = ((__be16 *)disk_aux)[3];
+			    __entry->disk_aux = be64_to_cpu(x.ll);
+		    } else {
+			    __entry->disk_aux = 0;
+		    }
 			   ),
 
 	    TP_printk("o=%08x %s B=%llx c=%u aux=%llx dsk=%llx",

Re: [PATCH v4 00/10] netfs, cachefiles: Miscellaneous fixes

[Paulo Alcantara]

Looks good.

Acked-by: Paulo Alcantara <pc@manguebit.org>