* [PATCH 1/8] Btrfs: rework outstanding_extents
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
@ 2017-10-19 18:15 ` Josef Bacik
2017-10-19 18:15 ` [PATCH 2/8] btrfs: add tracepoints for outstanding extents mods Josef Bacik
` (7 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:15 UTC (permalink / raw)
To: kernel-team, linux-btrfs
Right now we do a lot of weird hoops around outstanding_extents in order
to keep the extent count consistent. This is because we logically
transfer the outstanding_extent count from the initial reservation
through the set_delalloc_bits. This makes it pretty difficult to get a
handle on how and when we need to mess with outstanding_extents.
Fix this by revamping the rules of how we deal with outstanding_extents.
Now instead everybody that is holding on to a delalloc extent is
required to increase the outstanding extents count for itself. This
means we'll have something like this
btrfs_delalloc_reserve_metadata - outstanding_extents = 1
btrfs_set_extent_delalloc - outstanding_extents = 2
btrfs_release_delalloc_extents - outstanding_extents = 1
for an initial file write. Now take the append write where we extend an
existing delalloc range but still under the maximum extent size
btrfs_delalloc_reserve_metadata - outstanding_extents = 2
btrfs_set_extent_delalloc
btrfs_set_bit_hook - outstanding_extents = 3
btrfs_merge_extent_hook - outstanding_extents = 2
btrfs_delalloc_release_extents - outstanding_extnets = 1
In order to make the ordered extent transition we of course must now
make ordered extents carry their own outstanding_extent reservation, so
for cow_file_range we end up with
btrfs_add_ordered_extent - outstanding_extents = 2
clear_extent_bit - outstanding_extents = 1
btrfs_remove_ordered_extent - outstanding_extents = 0
This makes all manipulations of outstanding_extents much more explicit.
Every successful call to btrfs_delalloc_reserve_metadata _must_ now be
combined with btrfs_release_delalloc_extents, even in the error case, as
that is the only function that actually modifies the
outstanding_extents counter.
The drawback to this is now we are much more likely to have transient
cases where outstanding_extents is much larger than it actually should
be. This could happen before as we manipulated the delalloc bits, but
now it happens basically at every write. This may put more pressure on
the ENOSPC flushing code, but I think making this code simpler is worth
the cost. I have another change coming to mitigate this side-effect
somewhat.
I also added trace points for the counter manipulation. These were used
by a bpf script I wrote to help track down leak issues.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/btrfs_inode.h | 18 ++++++
fs/btrfs/ctree.h | 2 +
fs/btrfs/extent-tree.c | 139 ++++++++++++++++++++++++++++---------------
fs/btrfs/file.c | 22 +++----
fs/btrfs/inode-map.c | 3 +-
fs/btrfs/inode.c | 114 +++++++++++------------------------
fs/btrfs/ioctl.c | 2 +
fs/btrfs/ordered-data.c | 21 ++++++-
fs/btrfs/relocation.c | 3 +
fs/btrfs/tests/inode-tests.c | 18 ++----
10 files changed, 186 insertions(+), 156 deletions(-)
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index eccadb5f62a5..e3ac29e72714 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -267,6 +267,24 @@ static inline bool btrfs_is_free_space_inode(struct btrfs_inode *inode)
return false;
}
+static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
+ int mod)
+{
+ lockdep_assert_held(&inode->lock);
+ inode->outstanding_extents += mod;
+ if (btrfs_is_free_space_inode(inode))
+ return;
+}
+
+static inline void btrfs_mod_reserved_extents(struct btrfs_inode *inode,
+ int mod)
+{
+ lockdep_assert_held(&inode->lock);
+ inode->reserved_extents += mod;
+ if (btrfs_is_free_space_inode(inode))
+ return;
+}
+
static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
{
int ret = 0;
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 7bda8429e93f..9d950c2dd53f 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -2747,6 +2747,8 @@ int btrfs_subvolume_reserve_metadata(struct btrfs_root *root,
u64 *qgroup_reserved, bool use_global_rsv);
void btrfs_subvolume_release_metadata(struct btrfs_fs_info *fs_info,
struct btrfs_block_rsv *rsv);
+void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes);
+
int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes);
void btrfs_delalloc_release_metadata(struct btrfs_inode *inode, u64 num_bytes);
int btrfs_delalloc_reserve_space(struct inode *inode,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 4f874d02f310..aaa346562df6 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -5954,42 +5954,31 @@ void btrfs_subvolume_release_metadata(struct btrfs_fs_info *fs_info,
}
/**
- * drop_outstanding_extent - drop an outstanding extent
+ * drop_over_reserved_extents - drop our extra extent reservations
* @inode: the inode we're dropping the extent for
- * @num_bytes: the number of bytes we're releasing.
*
- * This is called when we are freeing up an outstanding extent, either called
- * after an error or after an extent is written. This will return the number of
- * reserved extents that need to be freed. This must be called with
- * BTRFS_I(inode)->lock held.
+ * We reserve extents we may use, but they may have been merged with other
+ * extents and we may not need the extra reservation.
+ *
+ * We also call this when we've completed io to an extent or had an error and
+ * cleared the outstanding extent, in either case we no longer need our
+ * reservation and can drop the excess.
*/
-static unsigned drop_outstanding_extent(struct btrfs_inode *inode,
- u64 num_bytes)
+static unsigned drop_over_reserved_extents(struct btrfs_inode *inode)
{
- unsigned drop_inode_space = 0;
- unsigned dropped_extents = 0;
- unsigned num_extents;
+ unsigned num_extents = 0;
- num_extents = count_max_extents(num_bytes);
- ASSERT(num_extents);
- ASSERT(inode->outstanding_extents >= num_extents);
- inode->outstanding_extents -= num_extents;
+ if (inode->reserved_extents > inode->outstanding_extents) {
+ num_extents = inode->reserved_extents -
+ inode->outstanding_extents;
+ btrfs_mod_reserved_extents(inode, -num_extents);
+ }
if (inode->outstanding_extents == 0 &&
test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
&inode->runtime_flags))
- drop_inode_space = 1;
-
- /*
- * If we have more or the same amount of outstanding extents than we have
- * reserved then we need to leave the reserved extents count alone.
- */
- if (inode->outstanding_extents >= inode->reserved_extents)
- return drop_inode_space;
-
- dropped_extents = inode->reserved_extents - inode->outstanding_extents;
- inode->reserved_extents -= dropped_extents;
- return dropped_extents + drop_inode_space;
+ num_extents++;
+ return num_extents;
}
/**
@@ -6044,13 +6033,15 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
struct btrfs_block_rsv *block_rsv = &fs_info->delalloc_block_rsv;
u64 to_reserve = 0;
u64 csum_bytes;
- unsigned nr_extents;
+ unsigned nr_extents, reserve_extents;
enum btrfs_reserve_flush_enum flush = BTRFS_RESERVE_FLUSH_ALL;
int ret = 0;
bool delalloc_lock = true;
u64 to_free = 0;
unsigned dropped;
bool release_extra = false;
+ bool underflow = false;
+ bool did_retry = false;
/* If we are a free space inode we need to not flush since we will be in
* the middle of a transaction commit. We also don't need the delalloc
@@ -6075,18 +6066,31 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
mutex_lock(&inode->delalloc_mutex);
num_bytes = ALIGN(num_bytes, fs_info->sectorsize);
-
+retry:
spin_lock(&inode->lock);
- nr_extents = count_max_extents(num_bytes);
- inode->outstanding_extents += nr_extents;
+ reserve_extents = nr_extents = count_max_extents(num_bytes);
+ btrfs_mod_outstanding_extents(inode, nr_extents);
- nr_extents = 0;
- if (inode->outstanding_extents > inode->reserved_extents)
- nr_extents += inode->outstanding_extents -
+ /*
+ * Because we add an outstanding extent for ordered before we clear
+ * delalloc we will double count our outstanding extents slightly. This
+ * could mean that we transiently over-reserve, which could result in an
+ * early ENOSPC if our timing is unlucky. Keep track of the case that
+ * we had a reservation underflow so we can retry if we fail.
+ *
+ * Keep in mind we can legitimately have more outstanding extents than
+ * reserved because of fragmentation, so only allow a retry once.
+ */
+ if (inode->outstanding_extents >
+ inode->reserved_extents + nr_extents) {
+ reserve_extents = inode->outstanding_extents -
inode->reserved_extents;
+ underflow = true;
+ }
/* We always want to reserve a slot for updating the inode. */
- to_reserve = btrfs_calc_trans_metadata_size(fs_info, nr_extents + 1);
+ to_reserve = btrfs_calc_trans_metadata_size(fs_info,
+ reserve_extents + 1);
to_reserve += calc_csum_metadata_size(inode, num_bytes, 1);
csum_bytes = inode->csum_bytes;
spin_unlock(&inode->lock);
@@ -6111,7 +6115,7 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
to_reserve -= btrfs_calc_trans_metadata_size(fs_info, 1);
release_extra = true;
}
- inode->reserved_extents += nr_extents;
+ btrfs_mod_reserved_extents(inode, reserve_extents);
spin_unlock(&inode->lock);
if (delalloc_lock)
@@ -6127,7 +6131,10 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
out_fail:
spin_lock(&inode->lock);
- dropped = drop_outstanding_extent(inode, num_bytes);
+ nr_extents = count_max_extents(num_bytes);
+ btrfs_mod_outstanding_extents(inode, -nr_extents);
+
+ dropped = drop_over_reserved_extents(inode);
/*
* If the inodes csum_bytes is the same as the original
* csum_bytes then we know we haven't raced with any free()ers
@@ -6184,6 +6191,11 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
trace_btrfs_space_reservation(fs_info, "delalloc",
btrfs_ino(inode), to_free, 0);
}
+ if (underflow && !did_retry) {
+ did_retry = true;
+ underflow = false;
+ goto retry;
+ }
if (delalloc_lock)
mutex_unlock(&inode->delalloc_mutex);
return ret;
@@ -6191,12 +6203,12 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
/**
* btrfs_delalloc_release_metadata - release a metadata reservation for an inode
- * @inode: the inode to release the reservation for
- * @num_bytes: the number of bytes we're releasing
+ * @inode: the inode to release the reservation for.
+ * @num_bytes: the number of bytes we are releasing.
*
* This will release the metadata reservation for an inode. This can be called
* once we complete IO for a given set of bytes to release their metadata
- * reservations.
+ * reservations, or on error for the same reason.
*/
void btrfs_delalloc_release_metadata(struct btrfs_inode *inode, u64 num_bytes)
{
@@ -6206,8 +6218,7 @@ void btrfs_delalloc_release_metadata(struct btrfs_inode *inode, u64 num_bytes)
num_bytes = ALIGN(num_bytes, fs_info->sectorsize);
spin_lock(&inode->lock);
- dropped = drop_outstanding_extent(inode, num_bytes);
-
+ dropped = drop_over_reserved_extents(inode);
if (num_bytes)
to_free = calc_csum_metadata_size(inode, num_bytes, 0);
spin_unlock(&inode->lock);
@@ -6224,6 +6235,42 @@ void btrfs_delalloc_release_metadata(struct btrfs_inode *inode, u64 num_bytes)
}
/**
+ * btrfs_delalloc_release_extents - release our outstanding_extents
+ * @inode: the inode to balance the reservation for.
+ * @num_bytes: the number of bytes we originally reserved with
+ *
+ * When we reserve space we increase outstanding_extents for the extents we may
+ * add. Once we've set the range as delalloc or created our ordered extents we
+ * have outstanding_extents to track the real usage, so we use this to free our
+ * temporarily tracked outstanding_extents. This _must_ be used in conjunction
+ * with btrfs_delalloc_reserve_metadata.
+ */
+void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes)
+{
+ struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
+ unsigned num_extents;
+ u64 to_free;
+ unsigned dropped;
+
+ spin_lock(&inode->lock);
+ num_extents = count_max_extents(num_bytes);
+ btrfs_mod_outstanding_extents(inode, -num_extents);
+ dropped = drop_over_reserved_extents(inode);
+ spin_unlock(&inode->lock);
+
+ if (!dropped)
+ return;
+
+ if (btrfs_is_testing(fs_info))
+ return;
+
+ to_free = btrfs_calc_trans_metadata_size(fs_info, dropped);
+ trace_btrfs_space_reservation(fs_info, "delalloc", btrfs_ino(inode),
+ to_free, 0);
+ btrfs_block_rsv_release(fs_info, &fs_info->delalloc_block_rsv, to_free);
+}
+
+/**
* btrfs_delalloc_reserve_space - reserve data and metadata space for
* delalloc
* @inode: inode we're writing to
@@ -6267,10 +6314,7 @@ int btrfs_delalloc_reserve_space(struct inode *inode,
* @inode: inode we're releasing space for
* @start: start position of the space already reserved
* @len: the len of the space already reserved
- *
- * This must be matched with a call to btrfs_delalloc_reserve_space. This is
- * called in the case that we don't need the metadata AND data reservations
- * anymore. So if there is an error or we insert an inline extent.
+ * @release_bytes: the len of the space we consumed or didn't use
*
* This function will release the metadata space that was not used and will
* decrement ->delalloc_bytes and remove it from the fs_info delalloc_inodes
@@ -6278,7 +6322,8 @@ int btrfs_delalloc_reserve_space(struct inode *inode,
* Also it will handle the qgroup reserved space.
*/
void btrfs_delalloc_release_space(struct inode *inode,
- struct extent_changeset *reserved, u64 start, u64 len)
+ struct extent_changeset *reserved,
+ u64 start, u64 len)
{
btrfs_delalloc_release_metadata(BTRFS_I(inode), len);
btrfs_free_reserved_data_space(inode, reserved, start, len);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 4de174b664ff..f80254d82f40 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1656,6 +1656,7 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
}
}
+ WARN_ON(reserve_bytes == 0);
ret = btrfs_delalloc_reserve_metadata(BTRFS_I(inode),
reserve_bytes);
if (ret) {
@@ -1678,8 +1679,11 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
ret = prepare_pages(inode, pages, num_pages,
pos, write_bytes,
force_page_uptodate);
- if (ret)
+ if (ret) {
+ btrfs_delalloc_release_extents(BTRFS_I(inode),
+ reserve_bytes);
break;
+ }
extents_locked = lock_and_cleanup_extent_if_need(
BTRFS_I(inode), pages,
@@ -1688,6 +1692,8 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
if (extents_locked < 0) {
if (extents_locked == -EAGAIN)
goto again;
+ btrfs_delalloc_release_extents(BTRFS_I(inode),
+ reserve_bytes);
ret = extents_locked;
break;
}
@@ -1716,23 +1722,10 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
PAGE_SIZE);
}
- /*
- * If we had a short copy we need to release the excess delaloc
- * bytes we reserved. We need to increment outstanding_extents
- * because btrfs_delalloc_release_space and
- * btrfs_delalloc_release_metadata will decrement it, but
- * we still have an outstanding extent for the chunk we actually
- * managed to copy.
- */
if (num_sectors > dirty_sectors) {
/* release everything except the sectors we dirtied */
release_bytes -= dirty_sectors <<
fs_info->sb->s_blocksize_bits;
- if (copied > 0) {
- spin_lock(&BTRFS_I(inode)->lock);
- BTRFS_I(inode)->outstanding_extents++;
- spin_unlock(&BTRFS_I(inode)->lock);
- }
if (only_release_metadata) {
btrfs_delalloc_release_metadata(BTRFS_I(inode),
release_bytes);
@@ -1758,6 +1751,7 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
unlock_extent_cached(&BTRFS_I(inode)->io_tree,
lockstart, lockend, &cached_state,
GFP_NOFS);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes);
if (ret) {
btrfs_drop_pages(pages, num_pages);
break;
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index d02019747d00..022b19336fee 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -500,11 +500,12 @@ int btrfs_save_ino_cache(struct btrfs_root *root,
ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc,
prealloc, prealloc, &alloc_hint);
if (ret) {
- btrfs_delalloc_release_metadata(BTRFS_I(inode), prealloc);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), prealloc);
goto out_put;
}
ret = btrfs_write_out_ino_cache(root, trans, path, inode);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), prealloc);
out_put:
iput(inode);
out_release:
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index f2787cab6f3b..741852511d77 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -67,7 +67,6 @@ struct btrfs_iget_args {
};
struct btrfs_dio_data {
- u64 outstanding_extents;
u64 reserve;
u64 unsubmitted_oe_range_start;
u64 unsubmitted_oe_range_end;
@@ -348,7 +347,6 @@ static noinline int cow_file_range_inline(struct btrfs_root *root,
}
set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
- btrfs_delalloc_release_metadata(BTRFS_I(inode), end + 1 - start);
btrfs_drop_extent_cache(BTRFS_I(inode), start, aligned_end - 1, 0);
out:
/*
@@ -581,16 +579,21 @@ static noinline void compress_file_range(struct inode *inode,
}
if (ret <= 0) {
unsigned long clear_flags = EXTENT_DELALLOC |
- EXTENT_DELALLOC_NEW | EXTENT_DEFRAG;
+ EXTENT_DELALLOC_NEW | EXTENT_DEFRAG |
+ EXTENT_DO_ACCOUNTING;
unsigned long page_error_op;
- clear_flags |= (ret < 0) ? EXTENT_DO_ACCOUNTING : 0;
page_error_op = ret < 0 ? PAGE_SET_ERROR : 0;
/*
* inline extent creation worked or returned error,
* we don't need to create any more async work items.
* Unlock and free up our temp pages.
+ *
+ * We use DO_ACCOUNTING here because we need the
+ * delalloc_release_metadata to be done _after_ we drop
+ * our outstanding extent for clearing delalloc for this
+ * range.
*/
extent_clear_unlock_delalloc(inode, start, end, end,
NULL, clear_flags,
@@ -599,10 +602,6 @@ static noinline void compress_file_range(struct inode *inode,
PAGE_SET_WRITEBACK |
page_error_op |
PAGE_END_WRITEBACK);
- if (ret == 0)
- btrfs_free_reserved_data_space_noquota(inode,
- start,
- end - start + 1);
goto free_pages_out;
}
}
@@ -978,15 +977,19 @@ static noinline int cow_file_range(struct inode *inode,
ret = cow_file_range_inline(root, inode, start, end, 0,
BTRFS_COMPRESS_NONE, NULL);
if (ret == 0) {
+ /*
+ * We use DO_ACCOUNTING here because we need the
+ * delalloc_release_metadata to be run _after_ we drop
+ * our outstanding extent for clearing delalloc for this
+ * range.
+ */
extent_clear_unlock_delalloc(inode, start, end,
delalloc_end, NULL,
EXTENT_LOCKED | EXTENT_DELALLOC |
- EXTENT_DELALLOC_NEW |
- EXTENT_DEFRAG, PAGE_UNLOCK |
+ EXTENT_DELALLOC_NEW | EXTENT_DEFRAG |
+ EXTENT_DO_ACCOUNTING, PAGE_UNLOCK |
PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK |
PAGE_END_WRITEBACK);
- btrfs_free_reserved_data_space_noquota(inode, start,
- end - start + 1);
*nr_written = *nr_written +
(end - start + PAGE_SIZE) / PAGE_SIZE;
*page_started = 1;
@@ -1624,7 +1627,7 @@ static void btrfs_split_extent_hook(void *private_data,
}
spin_lock(&BTRFS_I(inode)->lock);
- BTRFS_I(inode)->outstanding_extents++;
+ btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
spin_unlock(&BTRFS_I(inode)->lock);
}
@@ -1654,7 +1657,7 @@ static void btrfs_merge_extent_hook(void *private_data,
/* we're not bigger than the max, unreserve the space and go */
if (new_size <= BTRFS_MAX_EXTENT_SIZE) {
spin_lock(&BTRFS_I(inode)->lock);
- BTRFS_I(inode)->outstanding_extents--;
+ btrfs_mod_outstanding_extents(BTRFS_I(inode), -1);
spin_unlock(&BTRFS_I(inode)->lock);
return;
}
@@ -1685,7 +1688,7 @@ static void btrfs_merge_extent_hook(void *private_data,
return;
spin_lock(&BTRFS_I(inode)->lock);
- BTRFS_I(inode)->outstanding_extents--;
+ btrfs_mod_outstanding_extents(BTRFS_I(inode), -1);
spin_unlock(&BTRFS_I(inode)->lock);
}
@@ -1755,15 +1758,12 @@ static void btrfs_set_bit_hook(void *private_data,
if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
struct btrfs_root *root = BTRFS_I(inode)->root;
u64 len = state->end + 1 - state->start;
+ u32 num_extents = count_max_extents(len);
bool do_list = !btrfs_is_free_space_inode(BTRFS_I(inode));
- if (*bits & EXTENT_FIRST_DELALLOC) {
- *bits &= ~EXTENT_FIRST_DELALLOC;
- } else {
- spin_lock(&BTRFS_I(inode)->lock);
- BTRFS_I(inode)->outstanding_extents++;
- spin_unlock(&BTRFS_I(inode)->lock);
- }
+ spin_lock(&BTRFS_I(inode)->lock);
+ btrfs_mod_outstanding_extents(BTRFS_I(inode), num_extents);
+ spin_unlock(&BTRFS_I(inode)->lock);
/* For sanity tests */
if (btrfs_is_testing(fs_info))
@@ -1817,13 +1817,9 @@ static void btrfs_clear_bit_hook(void *private_data,
struct btrfs_root *root = inode->root;
bool do_list = !btrfs_is_free_space_inode(inode);
- if (*bits & EXTENT_FIRST_DELALLOC) {
- *bits &= ~EXTENT_FIRST_DELALLOC;
- } else if (!(*bits & EXTENT_CLEAR_META_RESV)) {
- spin_lock(&inode->lock);
- inode->outstanding_extents -= num_extents;
- spin_unlock(&inode->lock);
- }
+ spin_lock(&inode->lock);
+ btrfs_mod_outstanding_extents(inode, -num_extents);
+ spin_unlock(&inode->lock);
/*
* We don't reserve metadata space for space cache inodes so we
@@ -2094,6 +2090,7 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
0);
ClearPageChecked(page);
set_page_dirty(page);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
out:
unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
&cached_state, GFP_NOFS);
@@ -3048,9 +3045,6 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
0, &cached_state, GFP_NOFS);
}
- if (root != fs_info->tree_root)
- btrfs_delalloc_release_metadata(BTRFS_I(inode),
- ordered_extent->len);
if (trans)
btrfs_end_transaction(trans);
@@ -4791,8 +4785,11 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
(!len || ((len & (blocksize - 1)) == 0)))
goto out;
+ block_start = round_down(from, blocksize);
+ block_end = block_start + blocksize - 1;
+
ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
- round_down(from, blocksize), blocksize);
+ block_start, blocksize);
if (ret)
goto out;
@@ -4800,15 +4797,12 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
page = find_or_create_page(mapping, index, mask);
if (!page) {
btrfs_delalloc_release_space(inode, data_reserved,
- round_down(from, blocksize),
- blocksize);
+ block_start, blocksize);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize);
ret = -ENOMEM;
goto out;
}
- block_start = round_down(from, blocksize);
- block_end = block_start + blocksize - 1;
-
if (!PageUptodate(page)) {
ret = btrfs_readpage(NULL, page);
lock_page(page);
@@ -4873,6 +4867,7 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
if (ret)
btrfs_delalloc_release_space(inode, data_reserved, block_start,
blocksize);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize);
unlock_page(page);
put_page(page);
out:
@@ -7787,33 +7782,6 @@ static struct extent_map *create_io_em(struct inode *inode, u64 start, u64 len,
return em;
}
-static void adjust_dio_outstanding_extents(struct inode *inode,
- struct btrfs_dio_data *dio_data,
- const u64 len)
-{
- unsigned num_extents = count_max_extents(len);
-
- /*
- * If we have an outstanding_extents count still set then we're
- * within our reservation, otherwise we need to adjust our inode
- * counter appropriately.
- */
- if (dio_data->outstanding_extents >= num_extents) {
- dio_data->outstanding_extents -= num_extents;
- } else {
- /*
- * If dio write length has been split due to no large enough
- * contiguous space, we need to compensate our inode counter
- * appropriately.
- */
- u64 num_needed = num_extents - dio_data->outstanding_extents;
-
- spin_lock(&BTRFS_I(inode)->lock);
- BTRFS_I(inode)->outstanding_extents += num_needed;
- spin_unlock(&BTRFS_I(inode)->lock);
- }
-}
-
static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
struct buffer_head *bh_result, int create)
{
@@ -7975,7 +7943,6 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
if (!dio_data->overwrite && start + len > i_size_read(inode))
i_size_write(inode, start + len);
- adjust_dio_outstanding_extents(inode, dio_data, len);
WARN_ON(dio_data->reserve < len);
dio_data->reserve -= len;
dio_data->unsubmitted_oe_range_end = start + len;
@@ -8005,14 +7972,6 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
err:
if (dio_data)
current->journal_info = dio_data;
- /*
- * Compensate the delalloc release we do in btrfs_direct_IO() when we
- * write less data then expected, so that we don't underflow our inode's
- * outstanding extents counter.
- */
- if (create && dio_data)
- adjust_dio_outstanding_extents(inode, dio_data, len);
-
return ret;
}
@@ -8857,7 +8816,6 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
offset, count);
if (ret)
goto out;
- dio_data.outstanding_extents = count_max_extents(count);
/*
* We need to know how many extents we reserved so that we can
@@ -8884,6 +8842,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
if (iov_iter_rw(iter) == WRITE) {
up_read(&BTRFS_I(inode)->dio_sem);
current->journal_info = NULL;
+ btrfs_delalloc_release_extents(BTRFS_I(inode), count);
if (ret < 0 && ret != -EIOCBQUEUED) {
if (dio_data.reserve)
btrfs_delalloc_release_space(inode, data_reserved,
@@ -9221,9 +9180,6 @@ int btrfs_page_mkwrite(struct vm_fault *vmf)
fs_info->sectorsize);
if (reserved_space < PAGE_SIZE) {
end = page_start + reserved_space - 1;
- spin_lock(&BTRFS_I(inode)->lock);
- BTRFS_I(inode)->outstanding_extents++;
- spin_unlock(&BTRFS_I(inode)->lock);
btrfs_delalloc_release_space(inode, data_reserved,
page_start, PAGE_SIZE - reserved_space);
}
@@ -9275,12 +9231,14 @@ int btrfs_page_mkwrite(struct vm_fault *vmf)
out_unlock:
if (!ret) {
+ btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
sb_end_pagefault(inode->i_sb);
extent_changeset_free(data_reserved);
return VM_FAULT_LOCKED;
}
unlock_page(page);
out:
+ btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
btrfs_delalloc_release_space(inode, data_reserved, page_start,
reserved_space);
out_noreserve:
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 847d318756d4..9b0bb448fae7 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -1217,6 +1217,7 @@ static int cluster_pages_for_defrag(struct inode *inode,
unlock_page(pages[i]);
put_page(pages[i]);
}
+ btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
extent_changeset_free(data_reserved);
return i_done;
out:
@@ -1227,6 +1228,7 @@ static int cluster_pages_for_defrag(struct inode *inode,
btrfs_delalloc_release_space(inode, data_reserved,
start_index << PAGE_SHIFT,
page_cnt << PAGE_SHIFT);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
extent_changeset_free(data_reserved);
return ret;
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index a3aca495e33e..5b311aeddcc8 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -242,6 +242,15 @@ static int __btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
}
spin_unlock(&root->ordered_extent_lock);
+ /*
+ * We don't need the count_max_extents here, we can assume that all of
+ * that work has been done at higher layers, so this is truly the
+ * smallest the extent is going to get.
+ */
+ spin_lock(&BTRFS_I(inode)->lock);
+ btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
+ spin_unlock(&BTRFS_I(inode)->lock);
+
return 0;
}
@@ -591,11 +600,19 @@ void btrfs_remove_ordered_extent(struct inode *inode,
{
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
struct btrfs_ordered_inode_tree *tree;
- struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
+ struct btrfs_root *root = btrfs_inode->root;
struct rb_node *node;
bool dec_pending_ordered = false;
- tree = &BTRFS_I(inode)->ordered_tree;
+ /* This is paired with btrfs_add_ordered_extent. */
+ spin_lock(&btrfs_inode->lock);
+ btrfs_mod_outstanding_extents(btrfs_inode, -1);
+ spin_unlock(&btrfs_inode->lock);
+ if (root != fs_info->tree_root)
+ btrfs_delalloc_release_metadata(btrfs_inode, entry->len);
+
+ tree = &btrfs_inode->ordered_tree;
spin_lock_irq(&tree->lock);
node = &entry->rb_node;
rb_erase(node, &tree->tree);
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 7d506a3e46dd..4cf2eb67eba6 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -3246,6 +3246,8 @@ static int relocate_file_extent_cluster(struct inode *inode,
put_page(page);
btrfs_delalloc_release_metadata(BTRFS_I(inode),
PAGE_SIZE);
+ btrfs_delalloc_release_extents(BTRFS_I(inode),
+ PAGE_SIZE);
ret = -EIO;
goto out;
}
@@ -3275,6 +3277,7 @@ static int relocate_file_extent_cluster(struct inode *inode,
put_page(page);
index++;
+ btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
balance_dirty_pages_ratelimited(inode->i_mapping);
btrfs_throttle(fs_info);
}
diff --git a/fs/btrfs/tests/inode-tests.c b/fs/btrfs/tests/inode-tests.c
index 330815eb07b4..f797642c013d 100644
--- a/fs/btrfs/tests/inode-tests.c
+++ b/fs/btrfs/tests/inode-tests.c
@@ -968,7 +968,6 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
btrfs_test_inode_set_ops(inode);
/* [BTRFS_MAX_EXTENT_SIZE] */
- BTRFS_I(inode)->outstanding_extents++;
ret = btrfs_set_extent_delalloc(inode, 0, BTRFS_MAX_EXTENT_SIZE - 1,
NULL, 0);
if (ret) {
@@ -983,7 +982,6 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
}
/* [BTRFS_MAX_EXTENT_SIZE][sectorsize] */
- BTRFS_I(inode)->outstanding_extents++;
ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE,
BTRFS_MAX_EXTENT_SIZE + sectorsize - 1,
NULL, 0);
@@ -1003,7 +1001,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
BTRFS_MAX_EXTENT_SIZE >> 1,
(BTRFS_MAX_EXTENT_SIZE >> 1) + sectorsize - 1,
EXTENT_DELALLOC | EXTENT_DIRTY |
- EXTENT_UPTODATE | EXTENT_DO_ACCOUNTING, 0, 0,
+ EXTENT_UPTODATE, 0, 0,
NULL, GFP_KERNEL);
if (ret) {
test_msg("clear_extent_bit returned %d\n", ret);
@@ -1017,7 +1015,6 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
}
/* [BTRFS_MAX_EXTENT_SIZE][sectorsize] */
- BTRFS_I(inode)->outstanding_extents++;
ret = btrfs_set_extent_delalloc(inode, BTRFS_MAX_EXTENT_SIZE >> 1,
(BTRFS_MAX_EXTENT_SIZE >> 1)
+ sectorsize - 1,
@@ -1035,12 +1032,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
/*
* [BTRFS_MAX_EXTENT_SIZE+sectorsize][sectorsize HOLE][BTRFS_MAX_EXTENT_SIZE+sectorsize]
- *
- * I'm artificially adding 2 to outstanding_extents because in the
- * buffered IO case we'd add things up as we go, but I don't feel like
- * doing that here, this isn't the interesting case we want to test.
*/
- BTRFS_I(inode)->outstanding_extents += 2;
ret = btrfs_set_extent_delalloc(inode,
BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize,
(BTRFS_MAX_EXTENT_SIZE << 1) + 3 * sectorsize - 1,
@@ -1059,7 +1051,6 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
/*
* [BTRFS_MAX_EXTENT_SIZE+sectorsize][sectorsize][BTRFS_MAX_EXTENT_SIZE+sectorsize]
*/
- BTRFS_I(inode)->outstanding_extents++;
ret = btrfs_set_extent_delalloc(inode,
BTRFS_MAX_EXTENT_SIZE + sectorsize,
BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, NULL, 0);
@@ -1079,7 +1070,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
BTRFS_MAX_EXTENT_SIZE + sectorsize,
BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1,
EXTENT_DIRTY | EXTENT_DELALLOC |
- EXTENT_DO_ACCOUNTING | EXTENT_UPTODATE, 0, 0,
+ EXTENT_UPTODATE, 0, 0,
NULL, GFP_KERNEL);
if (ret) {
test_msg("clear_extent_bit returned %d\n", ret);
@@ -1096,7 +1087,6 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
* Refill the hole again just for good measure, because I thought it
* might fail and I'd rather satisfy my paranoia at this point.
*/
- BTRFS_I(inode)->outstanding_extents++;
ret = btrfs_set_extent_delalloc(inode,
BTRFS_MAX_EXTENT_SIZE + sectorsize,
BTRFS_MAX_EXTENT_SIZE + 2 * sectorsize - 1, NULL, 0);
@@ -1114,7 +1104,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
/* Empty */
ret = clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, (u64)-1,
EXTENT_DIRTY | EXTENT_DELALLOC |
- EXTENT_DO_ACCOUNTING | EXTENT_UPTODATE, 0, 0,
+ EXTENT_UPTODATE, 0, 0,
NULL, GFP_KERNEL);
if (ret) {
test_msg("clear_extent_bit returned %d\n", ret);
@@ -1131,7 +1121,7 @@ static int test_extent_accounting(u32 sectorsize, u32 nodesize)
if (ret)
clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, (u64)-1,
EXTENT_DIRTY | EXTENT_DELALLOC |
- EXTENT_DO_ACCOUNTING | EXTENT_UPTODATE, 0, 0,
+ EXTENT_UPTODATE, 0, 0,
NULL, GFP_KERNEL);
iput(inode);
btrfs_free_dummy_root(root);
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* [PATCH 2/8] btrfs: add tracepoints for outstanding extents mods
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
2017-10-19 18:15 ` [PATCH 1/8] Btrfs: rework outstanding_extents Josef Bacik
@ 2017-10-19 18:15 ` Josef Bacik
2017-10-19 18:15 ` [PATCH 3/8] btrfs: make the delalloc block rsv per inode Josef Bacik
` (6 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:15 UTC (permalink / raw)
To: kernel-team, linux-btrfs
This is handy for tracing problems with modifying the outstanding
extents counters.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/btrfs_inode.h | 2 ++
include/trace/events/btrfs.h | 21 +++++++++++++++++++++
2 files changed, 23 insertions(+)
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index e3ac29e72714..5ebeafc19936 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -274,6 +274,8 @@ static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
inode->outstanding_extents += mod;
if (btrfs_is_free_space_inode(inode))
return;
+ trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
+ mod);
}
static inline void btrfs_mod_reserved_extents(struct btrfs_inode *inode,
diff --git a/include/trace/events/btrfs.h b/include/trace/events/btrfs.h
index bfe2f23b578c..567dcf2022bb 100644
--- a/include/trace/events/btrfs.h
+++ b/include/trace/events/btrfs.h
@@ -1695,6 +1695,27 @@ DEFINE_EVENT(btrfs__prelim_ref, btrfs_prelim_ref_insert,
TP_ARGS(fs_info, oldref, newref, tree_size)
);
+TRACE_EVENT(btrfs_inode_mod_outstanding_extents,
+ TP_PROTO(struct btrfs_root *root, u64 ino, int mod),
+
+ TP_ARGS(root, ino, mod),
+
+ TP_STRUCT__entry_btrfs(
+ __field( u64, root_objectid )
+ __field( u64, ino )
+ __field( int, mod )
+ ),
+
+ TP_fast_assign_btrfs(root->fs_info,
+ __entry->root_objectid = root->objectid;
+ __entry->ino = ino;
+ __entry->mod = mod;
+ ),
+
+ TP_printk_btrfs("root = %llu(%s) ino = %llu mod = %d",
+ show_root_type(__entry->root_objectid),
+ (unsigned long long)__entry->ino, __entry->mod)
+);
#endif /* _TRACE_BTRFS_H */
/* This part must be outside protection */
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* [PATCH 3/8] btrfs: make the delalloc block rsv per inode
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
2017-10-19 18:15 ` [PATCH 1/8] Btrfs: rework outstanding_extents Josef Bacik
2017-10-19 18:15 ` [PATCH 2/8] btrfs: add tracepoints for outstanding extents mods Josef Bacik
@ 2017-10-19 18:15 ` Josef Bacik
2017-10-19 18:15 ` [PATCH 4/8] btrfs: switch args for comp_*_refs Josef Bacik
` (5 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:15 UTC (permalink / raw)
To: kernel-team, linux-btrfs
The way we handle delalloc metadata reservations has gotten
progressively more complicated over the years. There is so much cruft
and weirdness around keeping the reserved count and outstanding counters
consistent and handling the error cases that it's impossible to
understand.
Fix this by making the delalloc block rsv per-inode. This way we can
calculate the actual size of the outstanding metadata reservations every
time we make a change, and then reserve the delta based on that amount.
This greatly simplifies the code everywhere, and makes the error
handling in btrfs_delalloc_reserve_metadata far less terrifying.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/btrfs_inode.h | 27 ++--
fs/btrfs/ctree.h | 5 +-
fs/btrfs/delayed-inode.c | 46 +------
fs/btrfs/disk-io.c | 18 ++-
fs/btrfs/extent-tree.c | 320 ++++++++++++++++-------------------------------
fs/btrfs/inode.c | 18 +--
6 files changed, 141 insertions(+), 293 deletions(-)
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 5ebeafc19936..63f0ccc92a71 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -36,14 +36,13 @@
#define BTRFS_INODE_ORPHAN_META_RESERVED 1
#define BTRFS_INODE_DUMMY 2
#define BTRFS_INODE_IN_DEFRAG 3
-#define BTRFS_INODE_DELALLOC_META_RESERVED 4
-#define BTRFS_INODE_HAS_ORPHAN_ITEM 5
-#define BTRFS_INODE_HAS_ASYNC_EXTENT 6
-#define BTRFS_INODE_NEEDS_FULL_SYNC 7
-#define BTRFS_INODE_COPY_EVERYTHING 8
-#define BTRFS_INODE_IN_DELALLOC_LIST 9
-#define BTRFS_INODE_READDIO_NEED_LOCK 10
-#define BTRFS_INODE_HAS_PROPS 11
+#define BTRFS_INODE_HAS_ORPHAN_ITEM 4
+#define BTRFS_INODE_HAS_ASYNC_EXTENT 5
+#define BTRFS_INODE_NEEDS_FULL_SYNC 6
+#define BTRFS_INODE_COPY_EVERYTHING 7
+#define BTRFS_INODE_IN_DELALLOC_LIST 8
+#define BTRFS_INODE_READDIO_NEED_LOCK 9
+#define BTRFS_INODE_HAS_PROPS 10
/* in memory btrfs inode */
struct btrfs_inode {
@@ -176,7 +175,8 @@ struct btrfs_inode {
* of extent items we've reserved metadata for.
*/
unsigned outstanding_extents;
- unsigned reserved_extents;
+
+ struct btrfs_block_rsv block_rsv;
/*
* Cached values of inode properties
@@ -278,15 +278,6 @@ static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
mod);
}
-static inline void btrfs_mod_reserved_extents(struct btrfs_inode *inode,
- int mod)
-{
- lockdep_assert_held(&inode->lock);
- inode->reserved_extents += mod;
- if (btrfs_is_free_space_inode(inode))
- return;
-}
-
static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
{
int ret = 0;
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9d950c2dd53f..0685ec774d72 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -763,8 +763,6 @@ struct btrfs_fs_info {
* delayed dir index item
*/
struct btrfs_block_rsv global_block_rsv;
- /* block reservation for delay allocation */
- struct btrfs_block_rsv delalloc_block_rsv;
/* block reservation for metadata operations */
struct btrfs_block_rsv trans_block_rsv;
/* block reservation for chunk tree */
@@ -2756,6 +2754,9 @@ int btrfs_delalloc_reserve_space(struct inode *inode,
void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv, unsigned short type);
struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_fs_info *fs_info,
unsigned short type);
+void btrfs_init_metadata_block_rsv(struct btrfs_fs_info *fs_info,
+ struct btrfs_block_rsv *rsv,
+ unsigned short type);
void btrfs_free_block_rsv(struct btrfs_fs_info *fs_info,
struct btrfs_block_rsv *rsv);
void __btrfs_free_block_rsv(struct btrfs_block_rsv *rsv);
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index 19e4ad2f3f2e..5d73f79ded8b 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -581,7 +581,6 @@ static int btrfs_delayed_inode_reserve_metadata(
struct btrfs_block_rsv *dst_rsv;
u64 num_bytes;
int ret;
- bool release = false;
src_rsv = trans->block_rsv;
dst_rsv = &fs_info->delayed_block_rsv;
@@ -589,36 +588,13 @@ static int btrfs_delayed_inode_reserve_metadata(
num_bytes = btrfs_calc_trans_metadata_size(fs_info, 1);
/*
- * If our block_rsv is the delalloc block reserve then check and see if
- * we have our extra reservation for updating the inode. If not fall
- * through and try to reserve space quickly.
- *
- * We used to try and steal from the delalloc block rsv or the global
- * reserve, but we'd steal a full reservation, which isn't kind. We are
- * here through delalloc which means we've likely just cowed down close
- * to the leaf that contains the inode, so we would steal less just
- * doing the fallback inode update, so if we do end up having to steal
- * from the global block rsv we hopefully only steal one or two blocks
- * worth which is less likely to hurt us.
- */
- if (src_rsv && src_rsv->type == BTRFS_BLOCK_RSV_DELALLOC) {
- spin_lock(&inode->lock);
- if (test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
- &inode->runtime_flags))
- release = true;
- else
- src_rsv = NULL;
- spin_unlock(&inode->lock);
- }
-
- /*
* btrfs_dirty_inode will update the inode under btrfs_join_transaction
* which doesn't reserve space for speed. This is a problem since we
* still need to reserve space for this update, so try to reserve the
* space.
*
* Now if src_rsv == delalloc_block_rsv we'll let it just steal since
- * we're accounted for.
+ * we always reserve enough to update the inode item.
*/
if (!src_rsv || (!trans->bytes_reserved &&
src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC)) {
@@ -643,32 +619,12 @@ static int btrfs_delayed_inode_reserve_metadata(
}
ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes, 1);
-
- /*
- * Migrate only takes a reservation, it doesn't touch the size of the
- * block_rsv. This is to simplify people who don't normally have things
- * migrated from their block rsv. If they go to release their
- * reservation, that will decrease the size as well, so if migrate
- * reduced size we'd end up with a negative size. But for the
- * delalloc_meta_reserved stuff we will only know to drop 1 reservation,
- * but we could in fact do this reserve/migrate dance several times
- * between the time we did the original reservation and we'd clean it
- * up. So to take care of this, release the space for the meta
- * reservation here. I think it may be time for a documentation page on
- * how block rsvs. work.
- */
if (!ret) {
trace_btrfs_space_reservation(fs_info, "delayed_inode",
btrfs_ino(inode), num_bytes, 1);
node->bytes_reserved = num_bytes;
}
- if (release) {
- trace_btrfs_space_reservation(fs_info, "delalloc",
- btrfs_ino(inode), num_bytes, 0);
- btrfs_block_rsv_release(fs_info, src_rsv, num_bytes);
- }
-
return ret;
}
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 484cf8fc952c..d1f396f72979 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -2447,14 +2447,6 @@ int open_ctree(struct super_block *sb,
goto fail_delalloc_bytes;
}
- fs_info->btree_inode = new_inode(sb);
- if (!fs_info->btree_inode) {
- err = -ENOMEM;
- goto fail_bio_counter;
- }
-
- mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
-
INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
INIT_LIST_HEAD(&fs_info->trans_list);
@@ -2487,8 +2479,6 @@ int open_ctree(struct super_block *sb,
btrfs_mapping_init(&fs_info->mapping_tree);
btrfs_init_block_rsv(&fs_info->global_block_rsv,
BTRFS_BLOCK_RSV_GLOBAL);
- btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
- BTRFS_BLOCK_RSV_DELALLOC);
btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
@@ -2517,6 +2507,14 @@ int open_ctree(struct super_block *sb,
INIT_LIST_HEAD(&fs_info->ordered_roots);
spin_lock_init(&fs_info->ordered_root_lock);
+
+ fs_info->btree_inode = new_inode(sb);
+ if (!fs_info->btree_inode) {
+ err = -ENOMEM;
+ goto fail_bio_counter;
+ }
+ mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
+
fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
GFP_KERNEL);
if (!fs_info->delayed_root) {
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index aaa346562df6..fc9720e28005 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -26,6 +26,7 @@
#include <linux/slab.h>
#include <linux/ratelimit.h>
#include <linux/percpu_counter.h>
+#include <linux/lockdep.h>
#include "hash.h"
#include "tree-log.h"
#include "disk-io.h"
@@ -4811,7 +4812,6 @@ static inline u64 calc_reclaim_items_nr(struct btrfs_fs_info *fs_info,
static void shrink_delalloc(struct btrfs_fs_info *fs_info, u64 to_reclaim,
u64 orig, bool wait_ordered)
{
- struct btrfs_block_rsv *block_rsv;
struct btrfs_space_info *space_info;
struct btrfs_trans_handle *trans;
u64 delalloc_bytes;
@@ -4827,8 +4827,7 @@ static void shrink_delalloc(struct btrfs_fs_info *fs_info, u64 to_reclaim,
to_reclaim = items * EXTENT_SIZE_PER_ITEM;
trans = (struct btrfs_trans_handle *)current->journal_info;
- block_rsv = &fs_info->delalloc_block_rsv;
- space_info = block_rsv->space_info;
+ space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
delalloc_bytes = percpu_counter_sum_positive(
&fs_info->delalloc_bytes);
@@ -5564,11 +5563,12 @@ static void space_info_add_new_bytes(struct btrfs_fs_info *fs_info,
}
}
-static void block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
+static u64 block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
struct btrfs_block_rsv *block_rsv,
struct btrfs_block_rsv *dest, u64 num_bytes)
{
struct btrfs_space_info *space_info = block_rsv->space_info;
+ u64 ret;
spin_lock(&block_rsv->lock);
if (num_bytes == (u64)-1)
@@ -5583,6 +5583,7 @@ static void block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
}
spin_unlock(&block_rsv->lock);
+ ret = num_bytes;
if (num_bytes > 0) {
if (dest) {
spin_lock(&dest->lock);
@@ -5602,6 +5603,7 @@ static void block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
space_info_add_old_bytes(fs_info, space_info,
num_bytes);
}
+ return ret;
}
int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src,
@@ -5625,6 +5627,15 @@ void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv, unsigned short type)
rsv->type = type;
}
+void btrfs_init_metadata_block_rsv(struct btrfs_fs_info *fs_info,
+ struct btrfs_block_rsv *rsv,
+ unsigned short type)
+{
+ btrfs_init_block_rsv(rsv, type);
+ rsv->space_info = __find_space_info(fs_info,
+ BTRFS_BLOCK_GROUP_METADATA);
+}
+
struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_fs_info *fs_info,
unsigned short type)
{
@@ -5634,9 +5645,7 @@ struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_fs_info *fs_info,
if (!block_rsv)
return NULL;
- btrfs_init_block_rsv(block_rsv, type);
- block_rsv->space_info = __find_space_info(fs_info,
- BTRFS_BLOCK_GROUP_METADATA);
+ btrfs_init_metadata_block_rsv(fs_info, block_rsv, type);
return block_rsv;
}
@@ -5719,6 +5728,66 @@ int btrfs_block_rsv_refill(struct btrfs_root *root,
return ret;
}
+/**
+ * btrfs_inode_rsv_refill - refill the inode block rsv.
+ * @inode - the inode we are refilling.
+ * @flush - the flusing restriction.
+ *
+ * Essentially the same as btrfs_block_rsv_refill, except it uses the
+ * block_rsv->size as the minimum size. We'll either refill the missing amount
+ * or return if we already have enough space. This will also handle the resreve
+ * tracepoint for the reserved amount.
+ */
+int btrfs_inode_rsv_refill(struct btrfs_inode *inode,
+ enum btrfs_reserve_flush_enum flush)
+{
+ struct btrfs_root *root = inode->root;
+ struct btrfs_block_rsv *block_rsv = &inode->block_rsv;
+ u64 num_bytes = 0;
+ int ret = -ENOSPC;
+
+ spin_lock(&block_rsv->lock);
+ if (block_rsv->reserved < block_rsv->size)
+ num_bytes = block_rsv->size - block_rsv->reserved;
+ spin_unlock(&block_rsv->lock);
+
+ if (num_bytes == 0)
+ return 0;
+
+ ret = reserve_metadata_bytes(root, block_rsv, num_bytes, flush);
+ if (!ret) {
+ block_rsv_add_bytes(block_rsv, num_bytes, 0);
+ trace_btrfs_space_reservation(root->fs_info, "delalloc",
+ btrfs_ino(inode), num_bytes, 1);
+ }
+ return ret;
+}
+
+/**
+ * btrfs_inode_rsv_release - release any excessive reservation.
+ * @inode - the inode we need to release from.
+ *
+ * This is the same as btrfs_block_rsv_release, except that it handles the
+ * tracepoint for the reservation.
+ */
+void btrfs_inode_rsv_release(struct btrfs_inode *inode)
+{
+ struct btrfs_fs_info *fs_info = inode->root->fs_info;
+ struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
+ struct btrfs_block_rsv *block_rsv = &inode->block_rsv;
+ u64 released = 0;
+
+ /*
+ * Since we statically set the block_rsv->size we just want to say we
+ * are releasing 0 bytes, and then we'll just get the reservation over
+ * the size free'd.
+ */
+ released = block_rsv_release_bytes(fs_info, block_rsv, global_rsv, 0);
+ if (released > 0)
+ trace_btrfs_space_reservation(fs_info, "delalloc",
+ btrfs_ino(inode), released, 0);
+}
+
void btrfs_block_rsv_release(struct btrfs_fs_info *fs_info,
struct btrfs_block_rsv *block_rsv,
u64 num_bytes)
@@ -5790,7 +5859,6 @@ static void init_global_block_rsv(struct btrfs_fs_info *fs_info)
space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
fs_info->global_block_rsv.space_info = space_info;
- fs_info->delalloc_block_rsv.space_info = space_info;
fs_info->trans_block_rsv.space_info = space_info;
fs_info->empty_block_rsv.space_info = space_info;
fs_info->delayed_block_rsv.space_info = space_info;
@@ -5810,8 +5878,6 @@ static void release_global_block_rsv(struct btrfs_fs_info *fs_info)
{
block_rsv_release_bytes(fs_info, &fs_info->global_block_rsv, NULL,
(u64)-1);
- WARN_ON(fs_info->delalloc_block_rsv.size > 0);
- WARN_ON(fs_info->delalloc_block_rsv.reserved > 0);
WARN_ON(fs_info->trans_block_rsv.size > 0);
WARN_ON(fs_info->trans_block_rsv.reserved > 0);
WARN_ON(fs_info->chunk_block_rsv.size > 0);
@@ -5953,95 +6019,37 @@ void btrfs_subvolume_release_metadata(struct btrfs_fs_info *fs_info,
btrfs_block_rsv_release(fs_info, rsv, (u64)-1);
}
-/**
- * drop_over_reserved_extents - drop our extra extent reservations
- * @inode: the inode we're dropping the extent for
- *
- * We reserve extents we may use, but they may have been merged with other
- * extents and we may not need the extra reservation.
- *
- * We also call this when we've completed io to an extent or had an error and
- * cleared the outstanding extent, in either case we no longer need our
- * reservation and can drop the excess.
- */
-static unsigned drop_over_reserved_extents(struct btrfs_inode *inode)
-{
- unsigned num_extents = 0;
-
- if (inode->reserved_extents > inode->outstanding_extents) {
- num_extents = inode->reserved_extents -
- inode->outstanding_extents;
- btrfs_mod_reserved_extents(inode, -num_extents);
- }
-
- if (inode->outstanding_extents == 0 &&
- test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
- &inode->runtime_flags))
- num_extents++;
- return num_extents;
-}
-
-/**
- * calc_csum_metadata_size - return the amount of metadata space that must be
- * reserved/freed for the given bytes.
- * @inode: the inode we're manipulating
- * @num_bytes: the number of bytes in question
- * @reserve: 1 if we are reserving space, 0 if we are freeing space
- *
- * This adjusts the number of csum_bytes in the inode and then returns the
- * correct amount of metadata that must either be reserved or freed. We
- * calculate how many checksums we can fit into one leaf and then divide the
- * number of bytes that will need to be checksumed by this value to figure out
- * how many checksums will be required. If we are adding bytes then the number
- * may go up and we will return the number of additional bytes that must be
- * reserved. If it is going down we will return the number of bytes that must
- * be freed.
- *
- * This must be called with BTRFS_I(inode)->lock held.
- */
-static u64 calc_csum_metadata_size(struct btrfs_inode *inode, u64 num_bytes,
- int reserve)
+static void btrfs_calculate_inode_block_rsv_size(struct btrfs_fs_info *fs_info,
+ struct btrfs_inode *inode)
{
- struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
- u64 old_csums, num_csums;
-
- if (inode->flags & BTRFS_INODE_NODATASUM && inode->csum_bytes == 0)
- return 0;
-
- old_csums = btrfs_csum_bytes_to_leaves(fs_info, inode->csum_bytes);
- if (reserve)
- inode->csum_bytes += num_bytes;
- else
- inode->csum_bytes -= num_bytes;
- num_csums = btrfs_csum_bytes_to_leaves(fs_info, inode->csum_bytes);
-
- /* No change, no need to reserve more */
- if (old_csums == num_csums)
- return 0;
+ struct btrfs_block_rsv *block_rsv = &inode->block_rsv;
+ u64 reserve_size = 0;
+ u64 csum_leaves;
+ unsigned outstanding_extents;
- if (reserve)
- return btrfs_calc_trans_metadata_size(fs_info,
- num_csums - old_csums);
+ lockdep_assert_held(&inode->lock);
+ outstanding_extents = inode->outstanding_extents;
+ if (outstanding_extents)
+ reserve_size = btrfs_calc_trans_metadata_size(fs_info,
+ outstanding_extents + 1);
+ csum_leaves = btrfs_csum_bytes_to_leaves(fs_info,
+ inode->csum_bytes);
+ reserve_size += btrfs_calc_trans_metadata_size(fs_info,
+ csum_leaves);
- return btrfs_calc_trans_metadata_size(fs_info, old_csums - num_csums);
+ spin_lock(&block_rsv->lock);
+ block_rsv->size = reserve_size;
+ spin_unlock(&block_rsv->lock);
}
int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
{
struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
struct btrfs_root *root = inode->root;
- struct btrfs_block_rsv *block_rsv = &fs_info->delalloc_block_rsv;
- u64 to_reserve = 0;
- u64 csum_bytes;
- unsigned nr_extents, reserve_extents;
+ unsigned nr_extents;
enum btrfs_reserve_flush_enum flush = BTRFS_RESERVE_FLUSH_ALL;
int ret = 0;
bool delalloc_lock = true;
- u64 to_free = 0;
- unsigned dropped;
- bool release_extra = false;
- bool underflow = false;
- bool did_retry = false;
/* If we are a free space inode we need to not flush since we will be in
* the middle of a transaction commit. We also don't need the delalloc
@@ -6066,33 +6074,13 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
mutex_lock(&inode->delalloc_mutex);
num_bytes = ALIGN(num_bytes, fs_info->sectorsize);
-retry:
+
+ /* Add our new extents and calculate the new rsv size. */
spin_lock(&inode->lock);
- reserve_extents = nr_extents = count_max_extents(num_bytes);
+ nr_extents = count_max_extents(num_bytes);
btrfs_mod_outstanding_extents(inode, nr_extents);
-
- /*
- * Because we add an outstanding extent for ordered before we clear
- * delalloc we will double count our outstanding extents slightly. This
- * could mean that we transiently over-reserve, which could result in an
- * early ENOSPC if our timing is unlucky. Keep track of the case that
- * we had a reservation underflow so we can retry if we fail.
- *
- * Keep in mind we can legitimately have more outstanding extents than
- * reserved because of fragmentation, so only allow a retry once.
- */
- if (inode->outstanding_extents >
- inode->reserved_extents + nr_extents) {
- reserve_extents = inode->outstanding_extents -
- inode->reserved_extents;
- underflow = true;
- }
-
- /* We always want to reserve a slot for updating the inode. */
- to_reserve = btrfs_calc_trans_metadata_size(fs_info,
- reserve_extents + 1);
- to_reserve += calc_csum_metadata_size(inode, num_bytes, 1);
- csum_bytes = inode->csum_bytes;
+ inode->csum_bytes += num_bytes;
+ btrfs_calculate_inode_block_rsv_size(fs_info, inode);
spin_unlock(&inode->lock);
if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags)) {
@@ -6102,100 +6090,26 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
goto out_fail;
}
- ret = btrfs_block_rsv_add(root, block_rsv, to_reserve, flush);
+ ret = btrfs_inode_rsv_refill(inode, flush);
if (unlikely(ret)) {
btrfs_qgroup_free_meta(root,
nr_extents * fs_info->nodesize);
goto out_fail;
}
- spin_lock(&inode->lock);
- if (test_and_set_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
- &inode->runtime_flags)) {
- to_reserve -= btrfs_calc_trans_metadata_size(fs_info, 1);
- release_extra = true;
- }
- btrfs_mod_reserved_extents(inode, reserve_extents);
- spin_unlock(&inode->lock);
-
if (delalloc_lock)
mutex_unlock(&inode->delalloc_mutex);
-
- if (to_reserve)
- trace_btrfs_space_reservation(fs_info, "delalloc",
- btrfs_ino(inode), to_reserve, 1);
- if (release_extra)
- btrfs_block_rsv_release(fs_info, block_rsv,
- btrfs_calc_trans_metadata_size(fs_info, 1));
return 0;
out_fail:
spin_lock(&inode->lock);
nr_extents = count_max_extents(num_bytes);
btrfs_mod_outstanding_extents(inode, -nr_extents);
-
- dropped = drop_over_reserved_extents(inode);
- /*
- * If the inodes csum_bytes is the same as the original
- * csum_bytes then we know we haven't raced with any free()ers
- * so we can just reduce our inodes csum bytes and carry on.
- */
- if (inode->csum_bytes == csum_bytes) {
- calc_csum_metadata_size(inode, num_bytes, 0);
- } else {
- u64 orig_csum_bytes = inode->csum_bytes;
- u64 bytes;
-
- /*
- * This is tricky, but first we need to figure out how much we
- * freed from any free-ers that occurred during this
- * reservation, so we reset ->csum_bytes to the csum_bytes
- * before we dropped our lock, and then call the free for the
- * number of bytes that were freed while we were trying our
- * reservation.
- */
- bytes = csum_bytes - inode->csum_bytes;
- inode->csum_bytes = csum_bytes;
- to_free = calc_csum_metadata_size(inode, bytes, 0);
-
-
- /*
- * Now we need to see how much we would have freed had we not
- * been making this reservation and our ->csum_bytes were not
- * artificially inflated.
- */
- inode->csum_bytes = csum_bytes - num_bytes;
- bytes = csum_bytes - orig_csum_bytes;
- bytes = calc_csum_metadata_size(inode, bytes, 0);
-
- /*
- * Now reset ->csum_bytes to what it should be. If bytes is
- * more than to_free then we would have freed more space had we
- * not had an artificially high ->csum_bytes, so we need to free
- * the remainder. If bytes is the same or less then we don't
- * need to do anything, the other free-ers did the correct
- * thing.
- */
- inode->csum_bytes = orig_csum_bytes - num_bytes;
- if (bytes > to_free)
- to_free = bytes - to_free;
- else
- to_free = 0;
- }
+ inode->csum_bytes -= num_bytes;
+ btrfs_calculate_inode_block_rsv_size(fs_info, inode);
spin_unlock(&inode->lock);
- if (dropped)
- to_free += btrfs_calc_trans_metadata_size(fs_info, dropped);
- if (to_free) {
- btrfs_block_rsv_release(fs_info, block_rsv, to_free);
- trace_btrfs_space_reservation(fs_info, "delalloc",
- btrfs_ino(inode), to_free, 0);
- }
- if (underflow && !did_retry) {
- did_retry = true;
- underflow = false;
- goto retry;
- }
+ btrfs_inode_rsv_release(inode);
if (delalloc_lock)
mutex_unlock(&inode->delalloc_mutex);
return ret;
@@ -6213,25 +6127,17 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes)
void btrfs_delalloc_release_metadata(struct btrfs_inode *inode, u64 num_bytes)
{
struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
- u64 to_free = 0;
- unsigned dropped;
num_bytes = ALIGN(num_bytes, fs_info->sectorsize);
spin_lock(&inode->lock);
- dropped = drop_over_reserved_extents(inode);
- if (num_bytes)
- to_free = calc_csum_metadata_size(inode, num_bytes, 0);
+ inode->csum_bytes -= num_bytes;
+ btrfs_calculate_inode_block_rsv_size(fs_info, inode);
spin_unlock(&inode->lock);
- if (dropped > 0)
- to_free += btrfs_calc_trans_metadata_size(fs_info, dropped);
if (btrfs_is_testing(fs_info))
return;
- trace_btrfs_space_reservation(fs_info, "delalloc", btrfs_ino(inode),
- to_free, 0);
-
- btrfs_block_rsv_release(fs_info, &fs_info->delalloc_block_rsv, to_free);
+ btrfs_inode_rsv_release(inode);
}
/**
@@ -6249,25 +6155,17 @@ void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes)
{
struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
unsigned num_extents;
- u64 to_free;
- unsigned dropped;
spin_lock(&inode->lock);
num_extents = count_max_extents(num_bytes);
btrfs_mod_outstanding_extents(inode, -num_extents);
- dropped = drop_over_reserved_extents(inode);
+ btrfs_calculate_inode_block_rsv_size(fs_info, inode);
spin_unlock(&inode->lock);
- if (!dropped)
- return;
-
if (btrfs_is_testing(fs_info))
return;
- to_free = btrfs_calc_trans_metadata_size(fs_info, dropped);
- trace_btrfs_space_reservation(fs_info, "delalloc", btrfs_ino(inode),
- to_free, 0);
- btrfs_block_rsv_release(fs_info, &fs_info->delalloc_block_rsv, to_free);
+ btrfs_inode_rsv_release(inode);
}
/**
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 741852511d77..68e28375e159 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -42,6 +42,7 @@
#include <linux/blkdev.h>
#include <linux/posix_acl_xattr.h>
#include <linux/uio.h>
+#include <linux/magic.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
@@ -315,7 +316,7 @@ static noinline int cow_file_range_inline(struct btrfs_root *root,
btrfs_free_path(path);
return PTR_ERR(trans);
}
- trans->block_rsv = &fs_info->delalloc_block_rsv;
+ trans->block_rsv = &BTRFS_I(inode)->block_rsv;
if (compressed_size && compressed_pages)
extent_item_size = btrfs_file_extent_calc_inline_size(
@@ -2951,7 +2952,7 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
trans = NULL;
goto out;
}
- trans->block_rsv = &fs_info->delalloc_block_rsv;
+ trans->block_rsv = &BTRFS_I(inode)->block_rsv;
ret = btrfs_update_inode_fallback(trans, root, inode);
if (ret) /* -ENOMEM or corruption */
btrfs_abort_transaction(trans, ret);
@@ -2987,7 +2988,7 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
goto out;
}
- trans->block_rsv = &fs_info->delalloc_block_rsv;
+ trans->block_rsv = &BTRFS_I(inode)->block_rsv;
if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
compress_type = ordered_extent->compress_type;
@@ -8842,7 +8843,6 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
if (iov_iter_rw(iter) == WRITE) {
up_read(&BTRFS_I(inode)->dio_sem);
current->journal_info = NULL;
- btrfs_delalloc_release_extents(BTRFS_I(inode), count);
if (ret < 0 && ret != -EIOCBQUEUED) {
if (dio_data.reserve)
btrfs_delalloc_release_space(inode, data_reserved,
@@ -8863,6 +8863,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
} else if (ret >= 0 && (size_t)ret < count)
btrfs_delalloc_release_space(inode, data_reserved,
offset, count - (size_t)ret);
+ btrfs_delalloc_release_extents(BTRFS_I(inode), count);
}
out:
if (wakeup)
@@ -9427,6 +9428,7 @@ int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
struct inode *btrfs_alloc_inode(struct super_block *sb)
{
+ struct btrfs_fs_info *fs_info = btrfs_sb(sb);
struct btrfs_inode *ei;
struct inode *inode;
@@ -9453,8 +9455,9 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
spin_lock_init(&ei->lock);
ei->outstanding_extents = 0;
- ei->reserved_extents = 0;
-
+ if (sb->s_magic != BTRFS_TEST_MAGIC)
+ btrfs_init_metadata_block_rsv(fs_info, &ei->block_rsv,
+ BTRFS_BLOCK_RSV_DELALLOC);
ei->runtime_flags = 0;
ei->prop_compress = BTRFS_COMPRESS_NONE;
ei->defrag_compress = BTRFS_COMPRESS_NONE;
@@ -9504,8 +9507,9 @@ void btrfs_destroy_inode(struct inode *inode)
WARN_ON(!hlist_empty(&inode->i_dentry));
WARN_ON(inode->i_data.nrpages);
+ WARN_ON(BTRFS_I(inode)->block_rsv.reserved);
+ WARN_ON(BTRFS_I(inode)->block_rsv.size);
WARN_ON(BTRFS_I(inode)->outstanding_extents);
- WARN_ON(BTRFS_I(inode)->reserved_extents);
WARN_ON(BTRFS_I(inode)->delalloc_bytes);
WARN_ON(BTRFS_I(inode)->new_delalloc_bytes);
WARN_ON(BTRFS_I(inode)->csum_bytes);
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* [PATCH 4/8] btrfs: switch args for comp_*_refs
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
` (2 preceding siblings ...)
2017-10-19 18:15 ` [PATCH 3/8] btrfs: make the delalloc block rsv per inode Josef Bacik
@ 2017-10-19 18:15 ` Josef Bacik
2017-10-19 18:15 ` [PATCH 5/8] btrfs: add a comp_refs() helper Josef Bacik
` (4 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:15 UTC (permalink / raw)
To: kernel-team, linux-btrfs
Make it more consistent, we want the inserted ref to be compared against
what's already in there. This will make the order go from lowest seq ->
highest seq, which will make us more likely to make forward progress if
there's a seqlock currently held.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/delayed-ref.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index a2973340a94f..bc940bb374cf 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -40,8 +40,8 @@ struct kmem_cache *btrfs_delayed_extent_op_cachep;
/*
* compare two delayed tree backrefs with same bytenr and type
*/
-static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref2,
- struct btrfs_delayed_tree_ref *ref1)
+static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref1,
+ struct btrfs_delayed_tree_ref *ref2)
{
if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) {
if (ref1->root < ref2->root)
@@ -60,8 +60,8 @@ static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref2,
/*
* compare two delayed data backrefs with same bytenr and type
*/
-static int comp_data_refs(struct btrfs_delayed_data_ref *ref2,
- struct btrfs_delayed_data_ref *ref1)
+static int comp_data_refs(struct btrfs_delayed_data_ref *ref1,
+ struct btrfs_delayed_data_ref *ref2)
{
if (ref1->node.type == BTRFS_EXTENT_DATA_REF_KEY) {
if (ref1->root < ref2->root)
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* [PATCH 5/8] btrfs: add a comp_refs() helper
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
` (3 preceding siblings ...)
2017-10-19 18:15 ` [PATCH 4/8] btrfs: switch args for comp_*_refs Josef Bacik
@ 2017-10-19 18:15 ` Josef Bacik
2017-10-19 18:16 ` [PATCH 6/8] btrfs: track refs in a rb_tree instead of a list Josef Bacik
` (3 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:15 UTC (permalink / raw)
To: kernel-team, linux-btrfs
Instead of open-coding the delayed ref comparisons, add a helper to do
the comparisons generically and use that everywhere. We compare
sequence numbers last for following patches.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/delayed-ref.c | 54 ++++++++++++++++++++++++++++----------------------
1 file changed, 30 insertions(+), 24 deletions(-)
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index bc940bb374cf..c4cfadb9768c 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -85,6 +85,34 @@ static int comp_data_refs(struct btrfs_delayed_data_ref *ref1,
return 0;
}
+static int comp_refs(struct btrfs_delayed_ref_node *ref1,
+ struct btrfs_delayed_ref_node *ref2,
+ bool check_seq)
+{
+ int ret = 0;
+ if (ref1->type < ref2->type)
+ return -1;
+ if (ref1->type > ref2->type)
+ return 1;
+ if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY ||
+ ref1->type == BTRFS_SHARED_BLOCK_REF_KEY)
+ ret = comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref1),
+ btrfs_delayed_node_to_tree_ref(ref2));
+ else
+ ret = comp_data_refs(btrfs_delayed_node_to_data_ref(ref1),
+ btrfs_delayed_node_to_data_ref(ref2));
+ if (ret)
+ return ret;
+ if (check_seq) {
+ if (ref1->seq < ref2->seq)
+ return -1;
+ if (ref1->seq > ref2->seq)
+ return 1;
+ }
+ return 0;
+}
+
+
/* insert a new ref to head ref rbtree */
static struct btrfs_delayed_ref_head *htree_insert(struct rb_root *root,
struct rb_node *node)
@@ -217,18 +245,7 @@ static bool merge_ref(struct btrfs_trans_handle *trans,
if (seq && next->seq >= seq)
goto next;
- if (next->type != ref->type)
- goto next;
-
- if ((ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
- ref->type == BTRFS_SHARED_BLOCK_REF_KEY) &&
- comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref),
- btrfs_delayed_node_to_tree_ref(next)))
- goto next;
- if ((ref->type == BTRFS_EXTENT_DATA_REF_KEY ||
- ref->type == BTRFS_SHARED_DATA_REF_KEY) &&
- comp_data_refs(btrfs_delayed_node_to_data_ref(ref),
- btrfs_delayed_node_to_data_ref(next)))
+ if (comp_refs(ref, next, false))
goto next;
if (ref->action == next->action) {
@@ -402,18 +419,7 @@ add_delayed_ref_tail_merge(struct btrfs_trans_handle *trans,
exist = list_entry(href->ref_list.prev, struct btrfs_delayed_ref_node,
list);
/* No need to compare bytenr nor is_head */
- if (exist->type != ref->type || exist->seq != ref->seq)
- goto add_tail;
-
- if ((exist->type == BTRFS_TREE_BLOCK_REF_KEY ||
- exist->type == BTRFS_SHARED_BLOCK_REF_KEY) &&
- comp_tree_refs(btrfs_delayed_node_to_tree_ref(exist),
- btrfs_delayed_node_to_tree_ref(ref)))
- goto add_tail;
- if ((exist->type == BTRFS_EXTENT_DATA_REF_KEY ||
- exist->type == BTRFS_SHARED_DATA_REF_KEY) &&
- comp_data_refs(btrfs_delayed_node_to_data_ref(exist),
- btrfs_delayed_node_to_data_ref(ref)))
+ if (comp_refs(exist, ref, true))
goto add_tail;
/* Now we are sure we can merge */
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* [PATCH 6/8] btrfs: track refs in a rb_tree instead of a list
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
` (4 preceding siblings ...)
2017-10-19 18:15 ` [PATCH 5/8] btrfs: add a comp_refs() helper Josef Bacik
@ 2017-10-19 18:16 ` Josef Bacik
2017-10-19 18:16 ` [PATCH 7/8] btrfs: don't call btrfs_start_delalloc_roots in flushoncommit Josef Bacik
` (2 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:16 UTC (permalink / raw)
To: kernel-team, linux-btrfs
If we get a significant amount of delayed refs for a single block (think
modifying multiple snapshots) we can end up spending an ungodly amount
of time looping through all of the entries trying to see if they can be
merged. This is because we only add them to a list, so we have O(2n)
for every ref head. This doesn't make any sense as we likely have refs
for different roots, and so they cannot be merged. Tracking in a tree
will allow us to break as soon as we hit an entry that doesn't match,
making our worst case O(n).
With this we can also merge entries more easily. Before we had to hope
that matching refs were on the ends of our list, but with the tree we
can search down to exact matches and merge them at insert time.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/backref.c | 5 ++-
fs/btrfs/delayed-ref.c | 107 +++++++++++++++++++++++++------------------------
fs/btrfs/delayed-ref.h | 5 +--
fs/btrfs/disk-io.c | 10 +++--
fs/btrfs/extent-tree.c | 21 ++++++----
5 files changed, 81 insertions(+), 67 deletions(-)
diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 33cba1abf8b6..9b627b895806 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -769,6 +769,7 @@ static int add_delayed_refs(const struct btrfs_fs_info *fs_info,
struct btrfs_key key;
struct btrfs_key tmp_op_key;
struct btrfs_key *op_key = NULL;
+ struct rb_node *n;
int count;
int ret = 0;
@@ -778,7 +779,9 @@ static int add_delayed_refs(const struct btrfs_fs_info *fs_info,
}
spin_lock(&head->lock);
- list_for_each_entry(node, &head->ref_list, list) {
+ for (n = rb_first(&head->ref_tree); n; n = rb_next(n)) {
+ node = rb_entry(n, struct btrfs_delayed_ref_node,
+ ref_node);
if (node->seq > seq)
continue;
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index c4cfadb9768c..48a9b23774e6 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -143,6 +143,33 @@ static struct btrfs_delayed_ref_head *htree_insert(struct rb_root *root,
return NULL;
}
+static struct btrfs_delayed_ref_node *
+tree_insert(struct rb_root *root, struct btrfs_delayed_ref_node *ins)
+{
+ struct rb_node **p = &root->rb_node;
+ struct rb_node *node = &ins->ref_node;
+ struct rb_node *parent_node = NULL;
+ struct btrfs_delayed_ref_node *entry;
+
+ while (*p) {
+ int comp;
+ parent_node = *p;
+ entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,
+ ref_node);
+ comp = comp_refs(ins, entry, true);
+ if (comp < 0)
+ p = &(*p)->rb_left;
+ else if (comp > 0)
+ p = &(*p)->rb_right;
+ else
+ return entry;
+ }
+
+ rb_link_node(node, parent_node, p);
+ rb_insert_color(node, root);
+ return NULL;
+}
+
/*
* find an head entry based on bytenr. This returns the delayed ref
* head if it was able to find one, or NULL if nothing was in that spot.
@@ -212,7 +239,8 @@ static inline void drop_delayed_ref(struct btrfs_trans_handle *trans,
struct btrfs_delayed_ref_node *ref)
{
assert_spin_locked(&head->lock);
- list_del(&ref->list);
+ rb_erase(&ref->ref_node, &head->ref_tree);
+ RB_CLEAR_NODE(&ref->ref_node);
if (!list_empty(&ref->add_list))
list_del(&ref->add_list);
ref->in_tree = 0;
@@ -229,24 +257,18 @@ static bool merge_ref(struct btrfs_trans_handle *trans,
u64 seq)
{
struct btrfs_delayed_ref_node *next;
+ struct rb_node *node = rb_next(&ref->ref_node);
bool done = false;
- next = list_first_entry(&head->ref_list, struct btrfs_delayed_ref_node,
- list);
- while (!done && &next->list != &head->ref_list) {
+ while (!done && node) {
int mod;
- struct btrfs_delayed_ref_node *next2;
-
- next2 = list_next_entry(next, list);
-
- if (next == ref)
- goto next;
+ next = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
+ node = rb_next(node);
if (seq && next->seq >= seq)
- goto next;
-
+ break;
if (comp_refs(ref, next, false))
- goto next;
+ break;
if (ref->action == next->action) {
mod = next->ref_mod;
@@ -270,8 +292,6 @@ static bool merge_ref(struct btrfs_trans_handle *trans,
WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
ref->type == BTRFS_SHARED_BLOCK_REF_KEY);
}
-next:
- next = next2;
}
return done;
@@ -283,11 +303,12 @@ void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
struct btrfs_delayed_ref_head *head)
{
struct btrfs_delayed_ref_node *ref;
+ struct rb_node *node;
u64 seq = 0;
assert_spin_locked(&head->lock);
- if (list_empty(&head->ref_list))
+ if (RB_EMPTY_ROOT(&head->ref_tree))
return;
/* We don't have too many refs to merge for data. */
@@ -304,22 +325,13 @@ void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
}
spin_unlock(&fs_info->tree_mod_seq_lock);
- ref = list_first_entry(&head->ref_list, struct btrfs_delayed_ref_node,
- list);
- while (&ref->list != &head->ref_list) {
+again:
+ for (node = rb_first(&head->ref_tree); node; node = rb_next(node)) {
+ ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
if (seq && ref->seq >= seq)
- goto next;
-
- if (merge_ref(trans, delayed_refs, head, ref, seq)) {
- if (list_empty(&head->ref_list))
- break;
- ref = list_first_entry(&head->ref_list,
- struct btrfs_delayed_ref_node,
- list);
continue;
- }
-next:
- ref = list_next_entry(ref, list);
+ if (merge_ref(trans, delayed_refs, head, ref, seq))
+ goto again;
}
}
@@ -402,25 +414,19 @@ btrfs_select_ref_head(struct btrfs_trans_handle *trans)
* Return 0 for insert.
* Return >0 for merge.
*/
-static int
-add_delayed_ref_tail_merge(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_ref_root *root,
- struct btrfs_delayed_ref_head *href,
- struct btrfs_delayed_ref_node *ref)
+static int insert_delayed_ref(struct btrfs_trans_handle *trans,
+ struct btrfs_delayed_ref_root *root,
+ struct btrfs_delayed_ref_head *href,
+ struct btrfs_delayed_ref_node *ref)
{
struct btrfs_delayed_ref_node *exist;
int mod;
int ret = 0;
spin_lock(&href->lock);
- /* Check whether we can merge the tail node with ref */
- if (list_empty(&href->ref_list))
- goto add_tail;
- exist = list_entry(href->ref_list.prev, struct btrfs_delayed_ref_node,
- list);
- /* No need to compare bytenr nor is_head */
- if (comp_refs(exist, ref, true))
- goto add_tail;
+ exist = tree_insert(&href->ref_tree, ref);
+ if (!exist)
+ goto inserted;
/* Now we are sure we can merge */
ret = 1;
@@ -451,9 +457,7 @@ add_delayed_ref_tail_merge(struct btrfs_trans_handle *trans,
drop_delayed_ref(trans, root, href, exist);
spin_unlock(&href->lock);
return ret;
-
-add_tail:
- list_add_tail(&ref->list, &href->ref_list);
+inserted:
if (ref->action == BTRFS_ADD_DELAYED_REF)
list_add_tail(&ref->add_list, &href->ref_add_list);
atomic_inc(&root->num_entries);
@@ -593,7 +597,7 @@ add_delayed_ref_head(struct btrfs_fs_info *fs_info,
head_ref->ref_mod = count_mod;
head_ref->must_insert_reserved = must_insert_reserved;
head_ref->is_data = is_data;
- INIT_LIST_HEAD(&head_ref->ref_list);
+ head_ref->ref_tree = RB_ROOT;
INIT_LIST_HEAD(&head_ref->ref_add_list);
RB_CLEAR_NODE(&head_ref->href_node);
head_ref->processing = 0;
@@ -685,7 +689,7 @@ add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
ref->is_head = 0;
ref->in_tree = 1;
ref->seq = seq;
- INIT_LIST_HEAD(&ref->list);
+ RB_CLEAR_NODE(&ref->ref_node);
INIT_LIST_HEAD(&ref->add_list);
full_ref = btrfs_delayed_node_to_tree_ref(ref);
@@ -699,7 +703,7 @@ add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
trace_add_delayed_tree_ref(fs_info, ref, full_ref, action);
- ret = add_delayed_ref_tail_merge(trans, delayed_refs, head_ref, ref);
+ ret = insert_delayed_ref(trans, delayed_refs, head_ref, ref);
/*
* XXX: memory should be freed at the same level allocated.
@@ -742,7 +746,7 @@ add_delayed_data_ref(struct btrfs_fs_info *fs_info,
ref->is_head = 0;
ref->in_tree = 1;
ref->seq = seq;
- INIT_LIST_HEAD(&ref->list);
+ RB_CLEAR_NODE(&ref->ref_node);
INIT_LIST_HEAD(&ref->add_list);
full_ref = btrfs_delayed_node_to_data_ref(ref);
@@ -758,8 +762,7 @@ add_delayed_data_ref(struct btrfs_fs_info *fs_info,
trace_add_delayed_data_ref(fs_info, ref, full_ref, action);
- ret = add_delayed_ref_tail_merge(trans, delayed_refs, head_ref, ref);
-
+ ret = insert_delayed_ref(trans, delayed_refs, head_ref, ref);
if (ret > 0)
kmem_cache_free(btrfs_delayed_data_ref_cachep, full_ref);
}
diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index 1ce11858d727..a43af432f859 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -27,8 +27,7 @@
#define BTRFS_UPDATE_DELAYED_HEAD 4 /* not changing ref count on head ref */
struct btrfs_delayed_ref_node {
- /*data/tree ref use list, stored in ref_head->ref_list. */
- struct list_head list;
+ struct rb_node ref_node;
/*
* If action is BTRFS_ADD_DELAYED_REF, also link this node to
* ref_head->ref_add_list, then we do not need to iterate the
@@ -92,7 +91,7 @@ struct btrfs_delayed_ref_head {
struct mutex mutex;
spinlock_t lock;
- struct list_head ref_list;
+ struct rb_root ref_tree;
/* accumulate add BTRFS_ADD_DELAYED_REF nodes to this ref_add_list. */
struct list_head ref_add_list;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index d1f396f72979..efce9a2fa9be 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -4113,7 +4113,7 @@ static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
struct btrfs_delayed_ref_head *head;
- struct btrfs_delayed_ref_node *tmp;
+ struct rb_node *n;
bool pin_bytes = false;
head = rb_entry(node, struct btrfs_delayed_ref_head,
@@ -4129,10 +4129,12 @@ static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
continue;
}
spin_lock(&head->lock);
- list_for_each_entry_safe_reverse(ref, tmp, &head->ref_list,
- list) {
+ while ((n = rb_first(&head->ref_tree)) != NULL) {
+ ref = rb_entry(n, struct btrfs_delayed_ref_node,
+ ref_node);
ref->in_tree = 0;
- list_del(&ref->list);
+ rb_erase(&ref->ref_node, &head->ref_tree);
+ RB_CLEAR_NODE(&ref->ref_node);
if (!list_empty(&ref->add_list))
list_del(&ref->add_list);
atomic_dec(&delayed_refs->num_entries);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index fc9720e28005..673ac4e01dd0 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2519,7 +2519,7 @@ select_delayed_ref(struct btrfs_delayed_ref_head *head)
{
struct btrfs_delayed_ref_node *ref;
- if (list_empty(&head->ref_list))
+ if (RB_EMPTY_ROOT(&head->ref_tree))
return NULL;
/*
@@ -2532,8 +2532,8 @@ select_delayed_ref(struct btrfs_delayed_ref_head *head)
return list_first_entry(&head->ref_add_list,
struct btrfs_delayed_ref_node, add_list);
- ref = list_first_entry(&head->ref_list, struct btrfs_delayed_ref_node,
- list);
+ ref = rb_entry(rb_first(&head->ref_tree),
+ struct btrfs_delayed_ref_node, ref_node);
ASSERT(list_empty(&ref->add_list));
return ref;
}
@@ -2593,7 +2593,7 @@ static int cleanup_ref_head(struct btrfs_trans_handle *trans,
spin_unlock(&head->lock);
spin_lock(&delayed_refs->lock);
spin_lock(&head->lock);
- if (!list_empty(&head->ref_list) || head->extent_op) {
+ if (!RB_EMPTY_ROOT(&head->ref_tree) || head->extent_op) {
spin_unlock(&head->lock);
spin_unlock(&delayed_refs->lock);
return 1;
@@ -2740,7 +2740,8 @@ static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
actual_count++;
ref->in_tree = 0;
- list_del(&ref->list);
+ rb_erase(&ref->ref_node, &locked_ref->ref_tree);
+ RB_CLEAR_NODE(&ref->ref_node);
if (!list_empty(&ref->add_list))
list_del(&ref->add_list);
/*
@@ -3138,6 +3139,7 @@ static noinline int check_delayed_ref(struct btrfs_root *root,
struct btrfs_delayed_data_ref *data_ref;
struct btrfs_delayed_ref_root *delayed_refs;
struct btrfs_transaction *cur_trans;
+ struct rb_node *node;
int ret = 0;
cur_trans = root->fs_info->running_transaction;
@@ -3170,7 +3172,12 @@ static noinline int check_delayed_ref(struct btrfs_root *root,
spin_unlock(&delayed_refs->lock);
spin_lock(&head->lock);
- list_for_each_entry(ref, &head->ref_list, list) {
+ /*
+ * XXX: We should replace this with a proper search function in the
+ * future.
+ */
+ for (node = rb_first(&head->ref_tree); node; node = rb_next(node)) {
+ ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
/* If it's a shared ref we know a cross reference exists */
if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
ret = 1;
@@ -7141,7 +7148,7 @@ static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
goto out_delayed_unlock;
spin_lock(&head->lock);
- if (!list_empty(&head->ref_list))
+ if (!RB_EMPTY_ROOT(&head->ref_tree))
goto out;
if (head->extent_op) {
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* [PATCH 7/8] btrfs: don't call btrfs_start_delalloc_roots in flushoncommit
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
` (5 preceding siblings ...)
2017-10-19 18:16 ` [PATCH 6/8] btrfs: track refs in a rb_tree instead of a list Josef Bacik
@ 2017-10-19 18:16 ` Josef Bacik
2017-10-19 18:16 ` [PATCH 8/8] btrfs: move btrfs_truncate_block out of trans handle Josef Bacik
2017-10-20 15:35 ` [PATCH 0/8] Remaining queue David Sterba
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:16 UTC (permalink / raw)
To: kernel-team, linux-btrfs
We're holding the sb_start_intwrite lock at this point, and doing async
filemap_flush of the inodes will result in a deadlock if we freeze the
fs during this operation. This is because we could do a
btrfs_join_transaction() in the thread we are waiting on which would
block at sb_start_intwrite, and thus deadlock. Using
writeback_inodes_sb() side steps the problem by not introducing all of
these extra locking dependencies.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/transaction.c | 11 ++++++++++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 68c3e1c04bca..5a8c2649af2f 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -1916,8 +1916,17 @@ static void cleanup_transaction(struct btrfs_trans_handle *trans,
static inline int btrfs_start_delalloc_flush(struct btrfs_fs_info *fs_info)
{
+ /*
+ * We use writeback_inodes_sb here because if we used
+ * btrfs_start_delalloc_roots we would deadlock with fs freeze.
+ * Currently are holding the fs freeze lock, if we do an async flush
+ * we'll do btrfs_join_transaction() and deadlock because we need to
+ * wait for the fs freeze lock. Using the direct flushing we benefit
+ * from already being in a transaction and our join_transaction doesn't
+ * have to re-take the fs freeze lock.
+ */
if (btrfs_test_opt(fs_info, FLUSHONCOMMIT))
- return btrfs_start_delalloc_roots(fs_info, 1, -1);
+ writeback_inodes_sb(fs_info->sb, WB_REASON_SYNC);
return 0;
}
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* [PATCH 8/8] btrfs: move btrfs_truncate_block out of trans handle
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
` (6 preceding siblings ...)
2017-10-19 18:16 ` [PATCH 7/8] btrfs: don't call btrfs_start_delalloc_roots in flushoncommit Josef Bacik
@ 2017-10-19 18:16 ` Josef Bacik
2017-10-20 15:35 ` [PATCH 0/8] Remaining queue David Sterba
8 siblings, 0 replies; 10+ messages in thread
From: Josef Bacik @ 2017-10-19 18:16 UTC (permalink / raw)
To: kernel-team, linux-btrfs
Since we do a delalloc reserve in btrfs_truncate_block we can deadlock
with freeze. If somebody else is trying to allocate metadata for this
inode and it gets stuck in start_delalloc_inodes because of freeze we
will deadlock. Be safe and move this outside of a trans handle. This
also has a side-effect of making sure that we're not leaving stale data
behind in the other_encoding or encryption case. Not an issue now since
nobody uses it, but it would be a problem in the future.
Signed-off-by: Josef Bacik <jbacik@fb.com>
---
fs/btrfs/inode.c | 119 ++++++++++++++++++++-----------------------------------
1 file changed, 44 insertions(+), 75 deletions(-)
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 68e28375e159..c94e8938b574 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4357,47 +4357,11 @@ static int truncate_space_check(struct btrfs_trans_handle *trans,
}
-static int truncate_inline_extent(struct inode *inode,
- struct btrfs_path *path,
- struct btrfs_key *found_key,
- const u64 item_end,
- const u64 new_size)
-{
- struct extent_buffer *leaf = path->nodes[0];
- int slot = path->slots[0];
- struct btrfs_file_extent_item *fi;
- u32 size = (u32)(new_size - found_key->offset);
- struct btrfs_root *root = BTRFS_I(inode)->root;
-
- fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
-
- if (btrfs_file_extent_compression(leaf, fi) != BTRFS_COMPRESS_NONE) {
- loff_t offset = new_size;
- loff_t page_end = ALIGN(offset, PAGE_SIZE);
-
- /*
- * Zero out the remaining of the last page of our inline extent,
- * instead of directly truncating our inline extent here - that
- * would be much more complex (decompressing all the data, then
- * compressing the truncated data, which might be bigger than
- * the size of the inline extent, resize the extent, etc).
- * We release the path because to get the page we might need to
- * read the extent item from disk (data not in the page cache).
- */
- btrfs_release_path(path);
- return btrfs_truncate_block(inode, offset, page_end - offset,
- 0);
- }
-
- btrfs_set_file_extent_ram_bytes(leaf, fi, size);
- size = btrfs_file_extent_calc_inline_size(size);
- btrfs_truncate_item(root->fs_info, path, size, 1);
-
- if (test_bit(BTRFS_ROOT_REF_COWS, &root->state))
- inode_sub_bytes(inode, item_end + 1 - new_size);
-
- return 0;
-}
+/*
+ * Return this if we need to call truncate_block for the last bit of the
+ * truncate.
+ */
+#define NEED_TRUNCATE_BLOCK 1
/*
* this can truncate away extent items, csum items and directory items.
@@ -4558,11 +4522,6 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
if (found_type != BTRFS_EXTENT_DATA_KEY)
goto delete;
- if (del_item)
- last_size = found_key.offset;
- else
- last_size = new_size;
-
if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
u64 num_dec;
extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
@@ -4604,40 +4563,29 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
*/
if (!del_item &&
btrfs_file_extent_encryption(leaf, fi) == 0 &&
- btrfs_file_extent_other_encoding(leaf, fi) == 0) {
-
+ btrfs_file_extent_other_encoding(leaf, fi) == 0 &&
+ btrfs_file_extent_compression(leaf, fi) == 0) {
+ u32 size = (u32)(new_size - found_key.offset);
+ btrfs_set_file_extent_ram_bytes(leaf, fi, size);
+ size = btrfs_file_extent_calc_inline_size(size);
+ btrfs_truncate_item(root->fs_info, path, size, 1);
+ } else if (!del_item) {
/*
- * Need to release path in order to truncate a
- * compressed extent. So delete any accumulated
- * extent items so far.
+ * We have to bail so the last_size is set to
+ * just before this extent.
*/
- if (btrfs_file_extent_compression(leaf, fi) !=
- BTRFS_COMPRESS_NONE && pending_del_nr) {
- err = btrfs_del_items(trans, root, path,
- pending_del_slot,
- pending_del_nr);
- if (err) {
- btrfs_abort_transaction(trans,
- err);
- goto error;
- }
- pending_del_nr = 0;
- }
+ err = NEED_TRUNCATE_BLOCK;
+ break;
+ }
- err = truncate_inline_extent(inode, path,
- &found_key,
- item_end,
- new_size);
- if (err) {
- btrfs_abort_transaction(trans, err);
- goto error;
- }
- } else if (test_bit(BTRFS_ROOT_REF_COWS,
- &root->state)) {
+ if (test_bit(BTRFS_ROOT_REF_COWS, &root->state))
inode_sub_bytes(inode, item_end + 1 - new_size);
- }
}
delete:
+ if (del_item)
+ last_size = found_key.offset;
+ else
+ last_size = new_size;
if (del_item) {
if (!pending_del_nr) {
/* no pending yet, add ourselves */
@@ -9335,12 +9283,12 @@ static int btrfs_truncate(struct inode *inode)
ret = btrfs_truncate_inode_items(trans, root, inode,
inode->i_size,
BTRFS_EXTENT_DATA_KEY);
+ trans->block_rsv = &fs_info->trans_block_rsv;
if (ret != -ENOSPC && ret != -EAGAIN) {
err = ret;
break;
}
- trans->block_rsv = &fs_info->trans_block_rsv;
ret = btrfs_update_inode(trans, root, inode);
if (ret) {
err = ret;
@@ -9364,6 +9312,27 @@ static int btrfs_truncate(struct inode *inode)
trans->block_rsv = rsv;
}
+ /*
+ * We can't call btrfs_truncate_block inside a trans handle as we could
+ * deadlock with freeze, if we got NEED_TRUNCATE_BLOCK then we know
+ * we've truncated everything except the last little bit, and can do
+ * btrfs_truncate_block and then update the disk_i_size.
+ */
+ if (ret == NEED_TRUNCATE_BLOCK) {
+ btrfs_end_transaction(trans);
+ btrfs_btree_balance_dirty(fs_info);
+
+ ret = btrfs_truncate_block(inode, inode->i_size, 0, 0);
+ if (ret)
+ goto out;
+ trans = btrfs_start_transaction(root, 1);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ goto out;
+ }
+ btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
+ }
+
if (ret == 0 && inode->i_nlink > 0) {
trans->block_rsv = root->orphan_block_rsv;
ret = btrfs_orphan_del(trans, BTRFS_I(inode));
--
2.7.5
^ permalink raw reply related [flat|nested] 10+ messages in thread* Re: [PATCH 0/8] Remaining queue
2017-10-19 18:15 [PATCH 0/8] Remaining queue Josef Bacik
` (7 preceding siblings ...)
2017-10-19 18:16 ` [PATCH 8/8] btrfs: move btrfs_truncate_block out of trans handle Josef Bacik
@ 2017-10-20 15:35 ` David Sterba
8 siblings, 0 replies; 10+ messages in thread
From: David Sterba @ 2017-10-20 15:35 UTC (permalink / raw)
To: Josef Bacik; +Cc: kernel-team, linux-btrfs
On Thu, Oct 19, 2017 at 02:15:54PM -0400, Josef Bacik wrote:
> Here's the updated batch of the remaining queue of patches from me. I've
> addressed all of the outstanding review feedback for everything and they've been
> pretty thoroughly tested. Most of the changes are around changelogs and adding
> comments, as well as switching to lockdep_assert_held from whatever crap I was
> using before. Thanks,
Thanks. I've replaced the patches and moved them to the main 4.15 pile.
If anybody wants to do a final review and have the tag added, there's
still time.
^ permalink raw reply [flat|nested] 10+ messages in thread