Before we split the ordered CIL up into per cpu lists, we need a
mechanism to track the order of the items in the CIL. We need to do
this because there are rules around the order in which related items
must physically appear in the log even inside a single checkpoint
transaction.
An example of this is intents - an intent must appear in the log
before it's intent done record so that log recovery can cancel the
intent correctly. If we have these two records misordered in the
CIL, then they will not be recovered correctly by journal replay.
We also will not be able to move items to the tail of
the CIL list when they are relogged, hence the log items will need
some mechanism to allow the correct log item order to be recreated
before we write log items to the hournal.
Hence we need to have a mechanism for recording global order of
transactions in the log items so that we can recover that order
from un-ordered per-cpu lists.
Do this with a simple monotonic increasing commit counter in the CIL
context. Each log item in the transaction gets stamped with the
current commit order ID before it is added to the CIL. If the item
is already in the CIL, leave it where it is instead of moving it to
the tail of the list and instead sort the list before we start the
push work.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
To get them out from under the CIL lock.
This is an unordered list, so we can simply punt it to per-cpu lists
during transaction commits and reaggregate it back into a single
list during the CIL push work.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Now that we have the CIL percpu structures in place, implement the
space used counter as a per-cpu counter.
We have to be really careful now about ensuring that the checks and
updates run without arbitrary delays, which means they need to run
with pre-emption disabled. We do this by careful placement of
the get_cpu_ptr/put_cpu_ptr calls to access the per-cpu structures
for that CPU.
We need to be able to reliably detect that the CIL has reached
the hard limit threshold so we can take extra reservations for the
iclog headers when the space used overruns the original reservation.
hence we factor out xlog_cil_over_hard_limit() from
xlog_cil_push_background().
The global CIL space used is an atomic variable that is backed by
per-cpu aggregation to minimise the number of atomic updates we do
to the global state in the fast path. While we are under the soft
limit, we aggregate only when the per-cpu aggregation is over the
proportion of the soft limit assigned to that CPU. This means that
all CPUs can use all but one byte of their aggregation threshold
and we will not go over the soft limit.
Hence once we detect that we've gone over both a per-cpu aggregation
threshold and the soft limit, we know that we have only
exceeded the soft limit by one per-cpu aggregation threshold. Even
if all CPUs hit this at the same time, we can't be over the hard
limit, so we can run an aggregation back into the atomic counter
at this point and still be under the hard limit.
At this point, we will be over the soft limit and hence we'll
aggregate into the global atomic used space directly rather than the
per-cpu counters, hence providing accurate detection of hard limit
excursion for accounting and reservation purposes.
Hence we get the best of both worlds - lockless, scalable per-cpu
fast path plus accurate, atomic detection of hard limit excursion.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
The CIL push lock is highly contended on larger machines, becoming a
hard bottleneck that about 700,000 transaction commits/s on >16p
machines. To address this, start moving the CIL tracking
infrastructure to utilise per-CPU structures.
We need to track the space used, the amount of log reservation space
reserved to write the CIL, the log items in the CIL and the busy
extents that need to be completed by the CIL commit. This requires
a couple of per-cpu counters, an unordered per-cpu list and a
globally ordered per-cpu list.
Create a per-cpu structure to hold these and all the management
interfaces needed, as well as the hooks to handle hotplug CPUs.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
For every iclog that a CIL push will use up, we need to ensure we
have space reserved for the iclog header in each iclog. It is
extremely difficult to do this accurately with a per-cpu counter
without expensive summing of the counter in every commit. However,
we know what the maximum CIL size is going to be because of the
hard space limit we have, and hence we know exactly how many iclogs
we are going to need to write out the CIL.
We are constrained by the requirement that small transactions only
have reservation space for a single iclog header built into them.
At commit time we don't know how much of the current transaction
reservation is made up of iclog header reservations as calculated by
xfs_log_calc_unit_res() when the ticket was reserved. As larger
reservations have multiple header spaces reserved, we can steal
more than one iclog header reservation at a time, but we only steal
the exact number needed for the given log vector size delta.
As a result, we don't know exactly when we are going to steal iclog
header reservations, nor do we know exactly how many we are going to
need for a given CIL.
To make things simple, start by calculating the worst case number of
iclog headers a full CIL push will require. Record this into an
atomic variable in the CIL. Then add a byte counter to the log
ticket that records exactly how much iclog header space has been
reserved in this ticket by xfs_log_calc_unit_res(). This tells us
exactly how much space we can steal from the ticket at transaction
commit time.
Now, at transaction commit time, we can check if the CIL has a full
iclog header reservation and, if not, steal the entire reservation
the current ticket holds for iclog headers. This minimises the
number of times we need to do atomic operations in the fast path,
but still guarantees we get all the reservations we need.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
The xc_cil_lock is the most highly contended lock in XFS now. To
start the process of getting rid of it, lift the initial reservation
of the CIL log space out from under the xc_cil_lock.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
In the next patches we are going to make the CIL list itself
per-cpu, and so we cannot use list_empty() to check is the list is
empty. Replace the list_empty() checks with a flag in the CIL to
indicate we have committed at least one transaction to the CIL and
hence the CIL is not empty.
We need this flag to be an atomic so that we can clear it without
holding any locks in the commit fast path, but we also need to be
careful to avoid atomic operations in the fast path. Hence we use
the fact that test_bit() is not an atomic op to first check if the
flag is set and then run the atomic test_and_clear_bit() operation
to clear it and steal the initial unit reservation for the CIL
context checkpoint.
When we are switching to a new context in a push, we place the
setting of the XLOG_CIL_EMPTY flag under the xc_push_lock. THis
allows all the other places that need to check whether the CIL is
empty to use test_bit() and still be serialised correctly with the
CIL context swaps that set the bit.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
This update includes:
- support for printk message indexing.
- large extent counts to provide support for up to 2^47 data extents and 2^32
attribute extents, allowing us to scale beyond 4 billion data extents
to billions of xattrs per inode.
- conversion of various flags fields to be consistently declared as
unsigned bit fields.
- improvements to realtime extent accounting and converts them to per-cpu
counters to match all the other block and inode accounting.
- reworks core log formatting code to reduce iterations, have a shorter, cleaner
fast path and generally be easier to understand and maintain.
- improvements to rmap btree searches that reduce overhead by up
to 30% resulting in xfs_scrub runtime reductions of 15%.
- improvements to reflink that remove the size limitations in remapping operations
and greatly reduce the size of transaction reservations.
- reworks the minimum log size calculations to allow us to change transaction
reservations without changing the minimum supported log size.
- removal of quota warning support as it has never been used on Linux.
- intent whiteouts to allow us to cancel intents that are completed entirely
in memory rather than having use CPU and disk bandwidth formatting and writing
them into the journal when it is not necessary. This makes rmap, reflink and
extent freeing slightly more efficient, but provides massive improvements
for....
- Logged Attribute Replay feature support. This is a fundamental change to the
way we modify attributes, laying the foundation for future integration of
attribute modifications as part of other atomic transactional operations the
filesystem performs.
- Lots of cleanups and fixes for the logged attribute replay functionality.
-----BEGIN PGP SIGNATURE-----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=yUEM
-----END PGP SIGNATURE-----
Merge tag 'xfs-5.19-for-linus' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull xfs updates from Dave Chinner:
"This is a big update with lots of new code. The summary below them
all, so I'll just touch on teh higlights. The two main new features
are Large Extent Counts and Logged Attribute Replay - these are two
new foundational features that we are building more complex future
features on top of.
For upcoming functionality, we need to be able to store hundreds of
millions of xattrs per inode. The Large Extent Count feature removes
the limits that prevent this scale of xattr storage, and while we were
modifying the on disk extent count format we also increased the number
of data extents we support per inode from 2^32 to 2^47.
We also need to be able to modify xattrs as part of larger atomic
transactions rather than as standalone transactions. The Logged
Attribute Replay feature introduces the infrastructure that allows us
to use intents to record the attribute modifications in the journal
before we start them, hence allowing other atomic transactions to log
attribute modification intents and then defer the actual modification
to later. If we then crash, log recovery then guarantees that the
attribute is replayed in the context of the atomic transaction that
logged the intent.
A significant chunk of the commits in this merge are for the base
attribute replay functionality along with fixes, improvements and
cleanups related to this new functioanlity. Allison deserves a big
round of thanks for her ongoing work to get this functionality into
XFS.
There are also many other smaller changes and improvements, so overall
this is one of the bigger XFS merge requests in some time.
I will be following up next week with another smaller pull request -
we already have another round of fixes and improvements to the logged
attribute replay functionality just about ready to go. They'll soak
and test over the next week, and I'll send a pull request for them
near the end of the merge window.
Summary:
- support for printk message indexing.
- large extent counts to provide support for up to 2^47 data extents
and 2^32 attribute extents, allowing us to scale beyond 4 billion
data extents to billions of xattrs per inode.
- conversion of various flags fields to be consistently declared as
unsigned bit fields.
- improvements to realtime extent accounting and converts them to
per-cpu counters to match all the other block and inode accounting.
- reworks core log formatting code to reduce iterations, have a
shorter, cleaner fast path and generally be easier to understand
and maintain.
- improvements to rmap btree searches that reduce overhead by up to
30% resulting in xfs_scrub runtime reductions of 15%.
- improvements to reflink that remove the size limitations in
remapping operations and greatly reduce the size of transaction
reservations.
- reworks the minimum log size calculations to allow us to change
transaction reservations without changing the minimum supported log
size.
- removal of quota warning support as it has never been used on
Linux.
- intent whiteouts to allow us to cancel intents that are completed
entirely in memory rather than having use CPU and disk bandwidth
formatting and writing them into the journal when it is not
necessary. This makes rmap, reflink and extent freeing slightly
more efficient, but provides massive improvements for....
- Logged Attribute Replay feature support. This is a fundamental
change to the way we modify attributes, laying the foundation for
future integration of attribute modifications as part of other
atomic transactional operations the filesystem performs.
- Lots of cleanups and fixes for the logged attribute replay
functionality"
* tag 'xfs-5.19-for-linus' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (124 commits)
xfs: can't use kmem_zalloc() for attribute buffers
xfs: detect empty attr leaf blocks in xfs_attr3_leaf_verify
xfs: ATTR_REPLACE algorithm with LARP enabled needs rework
xfs: use XFS_DA_OP flags in deferred attr ops
xfs: remove xfs_attri_remove_iter
xfs: switch attr remove to xfs_attri_set_iter
xfs: introduce attr remove initial states into xfs_attr_set_iter
xfs: xfs_attr_set_iter() does not need to return EAGAIN
xfs: clean up final attr removal in xfs_attr_set_iter
xfs: remote xattr removal in xfs_attr_set_iter() is conditional
xfs: XFS_DAS_LEAF_REPLACE state only needed if !LARP
xfs: split remote attr setting out from replace path
xfs: consolidate leaf/node states in xfs_attr_set_iter
xfs: kill XFS_DAC_LEAF_ADDNAME_INIT
xfs: separate out initial attr_set states
xfs: don't set quota warning values
xfs: remove warning counters from struct xfs_dquot_res
xfs: remove quota warning limit from struct xfs_quota_limits
xfs: rework deferred attribute operation setup
xfs: make xattri_leaf_bp more useful
...
Because heap allocation of 64kB buffers will fail:
....
XFS: fs_mark(8414) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8417) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8409) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8428) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8430) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8437) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8433) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8406) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8412) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8432) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
XFS: fs_mark(8424) possible memory allocation deadlock size 65768 in kmem_alloc (mode:0x2d40)
....
I'd use kvmalloc() instead, but....
- 48.19% xfs_attr_create_intent
- 46.89% xfs_attri_init
- kvmalloc_node
- 46.04% __kmalloc_node
- kmalloc_large_node
- 45.99% __alloc_pages
- 39.39% __alloc_pages_slowpath.constprop.0
- 38.89% __alloc_pages_direct_compact
- 38.71% try_to_compact_pages
- compact_zone_order
- compact_zone
- 21.09% isolate_migratepages_block
10.31% PageHuge
5.82% set_pfnblock_flags_mask
0.86% get_pfnblock_flags_mask
- 4.48% __reset_isolation_suitable
4.44% __reset_isolation_pfn
- 3.56% __pageblock_pfn_to_page
1.33% pfn_to_online_page
2.83% get_pfnblock_flags_mask
- 0.87% migrate_pages
0.86% compaction_alloc
0.84% find_suitable_fallback
- 6.60% get_page_from_freelist
4.99% clear_page_erms
- 1.19% _raw_spin_lock_irqsave
- do_raw_spin_lock
__pv_queued_spin_lock_slowpath
- 0.86% __vmalloc_node_range
0.65% __alloc_pages_bulk
.... this is just yet another reminder of how much kvmalloc() sucks.
So lift xlog_cil_kvmalloc(), rename it to xlog_kvmalloc() and use
that instead....
We also clean up the attribute name and value lengths as they no
longer need to be rounded out to sizes compatible with log vectors.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Dave Chinner <david@fromorbit.com>
When we log modifications based on intents, we add both intent
and intent done items to the modification being made. These get
written to the log to ensure that the operation is re-run if the
intent done is not found in the log.
However, for operations that complete wholly within a single
checkpoint, the change in the checkpoint is atomic and will never
need replay. In this case, we don't need to actually write the
intent and intent done items to the journal because log recovery
will never need to manually restart this modification.
Log recovery currently handles intent/intent done matching by
inserting the intent into the AIL, then removing it when a matching
intent done item is found. Hence for all the intent-based operations
that complete within a checkpoint, we spend all that time parsing
the intent/intent done items just to cancel them and do nothing with
them.
Hence it follows that the only time we actually need intents in the
log is when the modification crosses checkpoint boundaries in the
log and so may only be partially complete in the journal. Hence if
we commit and intent done item to the CIL and the intent item is in
the same checkpoint, we don't actually have to write them to the
journal because log recovery will always cancel the intents.
We've never really worried about the overhead of logging intents
unnecessarily like this because the intents we log are generally
very much smaller than the change being made. e.g. freeing an extent
involves modifying at lease two freespace btree blocks and the AGF,
so the EFI/EFD overhead is only a small increase in space and
processing time compared to the overall cost of freeing an extent.
However, delayed attributes change this cost equation dramatically,
especially for inline attributes. In the case of adding an inline
attribute, we only log the inode core and attribute fork at present.
With delayed attributes, we now log the attr intent which includes
the name and value, the inode core adn attr fork, and finally the
attr intent done item. We increase the number of items we log from 1
to 3, and the number of log vectors (regions) goes up from 3 to 7.
Hence we tripple the number of objects that the CIL has to process,
and more than double the number of log vectors that need to be
written to the journal.
At scale, this means delayed attributes cause a non-pipelined CIL to
become CPU bound processing all the extra items, resulting in a > 40%
performance degradation on 16-way file+xattr create worklaods.
Pipelining the CIL (as per 5.15) reduces the performance degradation
to 20%, but now the limitation is the rate at which the log items
can be written to the iclogs and iclogs be dispatched for IO and
completed.
Even log IO completion is slowed down by these intents, because it
now has to process 3x the number of items in the checkpoint.
Processing completed intents is especially inefficient here, because
we first insert the intent into the AIL, then remove it from the AIL
when the intent done is processed. IOWs, we are also doing expensive
operations in log IO completion we could completely avoid if we
didn't log completed intent/intent done pairs.
Enter log item whiteouts.
When an intent done is committed, we can check to see if the
associated intent is in the same checkpoint as we are currently
committing the intent done to. If so, we can mark the intent log
item with a whiteout and immediately free the intent done item
rather than committing it to the CIL. We can basically skip the
entire formatting and CIL insertion steps for the intent done item.
However, we cannot remove the intent item from the CIL at this point
because the unlocked per-cpu CIL item lists do not permit removal
without holding the CIL context lock exclusively. Transaction commit
only holds the context lock shared, hence the best we can do is mark
the intent item with a whiteout so that the CIL push can release it
rather than writing it to the log.
This means we never write the intent to the log if the intent done
has also been committed to the same checkpoint, but we'll always
write the intent if the intent done has not been committed or has
been committed to a different checkpoint. This will result in
correct log recovery behaviour in all cases, without the overhead of
logging unnecessary intents.
This intent whiteout concept is generic - we can apply it to all
intent/intent done pairs that have a direct 1:1 relationship. The
way deferred ops iterate and relog intents mean that all intents
currently have a 1:1 relationship with their done intent, and hence
we can apply this cancellation to all existing intent/intent done
implementations.
For delayed attributes with a 16-way 64kB xattr create workload,
whiteouts reduce the amount of journalled metadata from ~2.5GB/s
down to ~600MB/s and improve the creation rate from 9000/s to
14000/s.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
In preparation for adding support for intent item whiteouts.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Now that we account for log opheaders in the log item formatting
code, we don't actually use the aggregated count of log iovecs in
the CIL for anything. Remove it and the tracking code that
calculates it.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
So remove it from the interface and callers.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The caller of xlog_write() usually has a close accounting of the
aggregated vector length contained in the log vector chain passed to
xlog_write(). There is no need to iterate the chain to calculate he
length of the data in xlog_write_calculate_len() if the caller is
already iterating that chain to build it.
Passing in the vector length avoids doing an extra chain iteration,
which can be a significant amount of work given that large CIL
commits can have hundreds of thousands of vectors attached to the
chain.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Current xlog_write() adds op headers to the log manually for every
log item region that is in the vector passed to it. While
xlog_write() needs to stamp the transaction ID into the ophdr, we
already know it's length, flags, clientid, etc at CIL commit time.
This means the only time that xlog write really needs to format and
reserve space for a new ophdr is when a region is split across two
iclogs. Adding the opheader and accounting for it as part of the
normal formatted item region means we simplify the accounting
of space used by a transaction and we don't have to special case
reserving of space in for the ophdrs in xlog_write(). It also means
we can largely initialise the ophdr in transaction commit instead
of xlog_write, making the xlog_write formatting inner loop much
tighter.
xlog_prepare_iovec() is now too large to stay as an inline function,
so we move it out of line and into xfs_log.c.
Object sizes:
text data bss dec hex filename
1125934 305951 484 1432369 15db31 fs/xfs/built-in.a.before
1123360 305951 484 1429795 15d123 fs/xfs/built-in.a.after
So the code is a roughly 2.5kB smaller with xlog_prepare_iovec() now
out of line, even though it grew in size itself.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
We currently set the log ticket client ID when we reserve a
transaction. This client ID is only ever written to the log by
a CIL checkpoint or unmount records, and so anything using a high
level transaction allocated through xfs_trans_alloc() does not need
a log ticket client ID to be set.
For the CIL checkpoint, the client ID written to the journal is
always XFS_TRANSACTION, and for the unmount record it is always
XFS_LOG, and nothing else writes to the log. All of these operations
tell xlog_write() exactly what they need to write to the log (the
optype) and build their own opheaders for start, commit and unmount
records. Hence we no longer need to set the client id in either the
log ticket or the xfs_trans.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Remove the final case where xlog_write() has to prepend an opheader
to a log transaction. Similar to the start record, the commit record
is just an empty opheader with a XLOG_COMMIT_TRANS type, so we can
just make this the payload for the region being passed to
xlog_write() and remove the special handling in xlog_write() for
the commit record.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
So move the one-off start record writing in xlog_write() out into
the static header that the CIL push builds to write into the log
initially. This simplifes the xlog_write() logic a lot.
pahole on x86-64 confirms that the xlog_cil_trans_hdr is correctly
32 bit aligned and packed for copying the log op and transaction
headers directly into the log as a single log region copy.
struct xlog_cil_trans_hdr {
struct xlog_op_header oph[2]; /* 0 24 */
struct xfs_trans_header thdr; /* 24 16 */
struct xfs_log_iovec lhdr[2]; /* 40 32 */
/* size: 72, cachelines: 2, members: 3 */
/* last cacheline: 8 bytes */
};
A wart is needed to handle the fact that length of the region the
opheader points to doesn't include the opheader length. hence if
we embed the opheader, we have to substract the opheader length from
the length written into the opheader by the generic copying code.
This will eventually go away when everything is converted to
embedded opheaders.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
It is static code deep in the middle of the CIL push logic. Factor
it out into a helper so that it is clear and easy to modify
separately.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Secure erase is a very different operation from discard in that it is
a data integrity operation vs hint. Fully split the limits and helper
infrastructure to make the separation more clear.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Acked-by: Christoph Böhmwalder <christoph.boehmwalder@linbit.com> [drbd]
Acked-by: Ryusuke Konishi <konishi.ryusuke@gmail.com> [nifs2]
Acked-by: Jaegeuk Kim <jaegeuk@kernel.org> [f2fs]
Acked-by: Coly Li <colyli@suse.de> [bcache]
Acked-by: David Sterba <dsterba@suse.com> [btrfs]
Acked-by: Chao Yu <chao@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20220415045258.199825-27-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jan Kara reported a performance regression in dbench that he
bisected down to commit bad77c375e ("xfs: CIL checkpoint
flushes caches unconditionally").
Whilst developing the journal flush/fua optimisations this cache was
part of, it appeared to made a significant difference to
performance. However, now that this patchset has settled and all the
correctness issues fixed, there does not appear to be any
significant performance benefit to asynchronous cache flushes.
In fact, the opposite is true on some storage types and workloads,
where additional cache flushes that can occur from fsync heavy
workloads have measurable and significant impact on overall
throughput.
Local dbench testing shows little difference on dbench runs with
sync vs async cache flushes on either fast or slow SSD storage, and
no difference in streaming concurrent async transaction workloads
like fs-mark.
Fast NVME storage.
From `dbench -t 30`, CIL scale:
clients async sync
BW Latency BW Latency
1 935.18 0.855 915.64 0.903
8 2404.51 6.873 2341.77 6.511
16 3003.42 6.460 2931.57 6.529
32 3697.23 7.939 3596.28 7.894
128 7237.43 15.495 7217.74 11.588
512 5079.24 90.587 5167.08 95.822
fsmark, 32 threads, create w/ 64 byte xattr w/32k logbsize
create chown unlink
async 1m41s 1m16s 2m03s
sync 1m40s 1m19s 1m54s
Slower SATA SSD storage:
From `dbench -t 30`, CIL scale:
clients async sync
BW Latency BW Latency
1 78.59 15.792 83.78 10.729
8 367.88 92.067 404.63 59.943
16 564.51 72.524 602.71 76.089
32 831.66 105.984 870.26 110.482
128 1659.76 102.969 1624.73 91.356
512 2135.91 223.054 2603.07 161.160
fsmark, 16 threads, create w/32k logbsize
create unlink
async 5m06s 4m15s
sync 5m00s 4m22s
And on Jan's test machine:
5.18-rc8-vanilla 5.18-rc8-patched
Amean 1 71.22 ( 0.00%) 64.94 * 8.81%*
Amean 2 93.03 ( 0.00%) 84.80 * 8.85%*
Amean 4 150.54 ( 0.00%) 137.51 * 8.66%*
Amean 8 252.53 ( 0.00%) 242.24 * 4.08%*
Amean 16 454.13 ( 0.00%) 439.08 * 3.31%*
Amean 32 835.24 ( 0.00%) 829.74 * 0.66%*
Amean 64 1740.59 ( 0.00%) 1686.73 * 3.09%*
Performance and cache flush behaviour is restored to pre-regression
levels.
As such, we can now consider the async cache flush mechanism an
unnecessary exercise in premature optimisation and hence we can
now remove it and the infrastructure it requires completely.
Fixes: bad77c375e ("xfs: CIL checkpoint flushes caches unconditionally")
Reported-and-tested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
We've got a mess on our hands.
1. xfs_trans_commit() cannot cancel transactions because the mount is
shut down - that causes dirty, aborted, unlogged log items to sit
unpinned in memory and potentially get written to disk before the
log is shut down. Hence xfs_trans_commit() can only abort
transactions when xlog_is_shutdown() is true.
2. xfs_force_shutdown() is used in places to cause the current
modification to be aborted via xfs_trans_commit() because it may be
impractical or impossible to cancel the transaction directly, and
hence xfs_trans_commit() must cancel transactions when
xfs_is_shutdown() is true in this situation. But we can't do that
because of #1.
3. Log IO errors cause log shutdowns by calling xfs_force_shutdown()
to shut down the mount and then the log from log IO completion.
4. xfs_force_shutdown() can result in a log force being issued,
which has to wait for log IO completion before it will mark the log
as shut down. If #3 races with some other shutdown trigger that runs
a log force, we rely on xfs_force_shutdown() silently ignoring #3
and avoiding shutting down the log until the failed log force
completes.
5. To ensure #2 always works, we have to ensure that
xfs_force_shutdown() does not return until the the log is shut down.
But in the case of #4, this will result in a deadlock because the
log Io completion will block waiting for a log force to complete
which is blocked waiting for log IO to complete....
So the very first thing we have to do here to untangle this mess is
dissociate log shutdown triggers from mount shutdowns. We already
have xlog_forced_shutdown, which will atomically transistion to the
log a shutdown state. Due to internal asserts it cannot be called
multiple times, but was done simply because the only place that
could call it was xfs_do_force_shutdown() (i.e. the mount shutdown!)
and that could only call it once and once only. So the first thing
we do is remove the asserts.
We then convert all the internal log shutdown triggers to call
xlog_force_shutdown() directly instead of xfs_force_shutdown(). This
allows the log shutdown triggers to shut down the log without
needing to care about mount based shutdown constraints. This means
we shut down the log independently of the mount and the mount may
not notice this until it's next attempt to read or modify metadata.
At that point (e.g. xfs_trans_commit()) it will see that the log is
shutdown, error out and shutdown the mount.
To ensure that all the unmount behaviours and asserts track
correctly as a result of a log shutdown, propagate the shutdown up
to the mount if it is not already set. This keeps the mount and log
state in sync, and saves a huge amount of hassle where code fails
because of a log shutdown but only checks for mount shutdowns and
hence ends up doing the wrong thing. Cleaning up that mess is
an exercise for another day.
This enables us to address the other problems noted above in
followup patches.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Log items belong to the log, not the xfs_mount. Convert the mount
pointer in the log item to a xlog pointer in preparation for
upcoming log centric changes to the log items.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
When the AIL tries to flush the CIL, it relies on the CIL push
ending up on stable storage without having to wait for and
manipulate iclog state directly. However, if there is already a
pending CIL push when the AIL tries to flush the CIL, it won't set
the cil->xc_push_commit_stable flag and so the CIL push will not
actively flush the commit record iclog.
generic/530 when run on a single CPU test VM can trigger this fairly
reliably. This test exercises unlinked inode recovery, and can
result in inodes being pinned in memory by ongoing modifications to
the inode cluster buffer to record unlinked list modifications. As a
result, the first inode unlinked in a buffer can pin the tail of the
log whilst the inode cluster buffer is pinned by the current
checkpoint that has been pushed but isn't on stable storage because
because the cil->xc_push_commit_stable was not set. This results in
the log/AIL effectively deadlocking until something triggers the
commit record iclog to be pushed to stable storage (i.e. the
periodic log worker calling xfs_log_force()).
The fix is two-fold - first we should always set the
cil->xc_push_commit_stable when xlog_cil_flush() is called,
regardless of whether there is already a pending push or not.
Second, if the CIL is empty, we should trigger an iclog flush to
ensure that the iclogs of the last checkpoint have actually been
submitted to disk as that checkpoint may not have been run under
stable completion constraints.
Reported-and-tested-by: Matthew Wilcox <willy@infradead.org>
Fixes: 0020a190cf ("xfs: AIL needs asynchronous CIL forcing")
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Oh, let me count the ways that the kvmalloc API sucks dog eggs.
The problem is when we are logging lots of large objects, we hit
kvmalloc really damn hard with costly order allocations, and
behaviour utterly sucks:
- 49.73% xlog_cil_commit
- 31.62% kvmalloc_node
- 29.96% __kmalloc_node
- 29.38% kmalloc_large_node
- 29.33% __alloc_pages
- 24.33% __alloc_pages_slowpath.constprop.0
- 18.35% __alloc_pages_direct_compact
- 17.39% try_to_compact_pages
- compact_zone_order
- 15.26% compact_zone
5.29% __pageblock_pfn_to_page
3.71% PageHuge
- 1.44% isolate_migratepages_block
0.71% set_pfnblock_flags_mask
1.11% get_pfnblock_flags_mask
- 0.81% get_page_from_freelist
- 0.59% _raw_spin_lock_irqsave
- do_raw_spin_lock
__pv_queued_spin_lock_slowpath
- 3.24% try_to_free_pages
- 3.14% shrink_node
- 2.94% shrink_slab.constprop.0
- 0.89% super_cache_count
- 0.66% xfs_fs_nr_cached_objects
- 0.65% xfs_reclaim_inodes_count
0.55% xfs_perag_get_tag
0.58% kfree_rcu_shrink_count
- 2.09% get_page_from_freelist
- 1.03% _raw_spin_lock_irqsave
- do_raw_spin_lock
__pv_queued_spin_lock_slowpath
- 4.88% get_page_from_freelist
- 3.66% _raw_spin_lock_irqsave
- do_raw_spin_lock
__pv_queued_spin_lock_slowpath
- 1.63% __vmalloc_node
- __vmalloc_node_range
- 1.10% __alloc_pages_bulk
- 0.93% __alloc_pages
- 0.92% get_page_from_freelist
- 0.89% rmqueue_bulk
- 0.69% _raw_spin_lock
- do_raw_spin_lock
__pv_queued_spin_lock_slowpath
13.73% memcpy_erms
- 2.22% kvfree
On this workload, that's almost a dozen CPUs all trying to compact
and reclaim memory inside kvmalloc_node at the same time. Yet it is
regularly falling back to vmalloc despite all that compaction, page
and shrinker reclaim that direct reclaim is doing. Copying all the
metadata is taking far less CPU time than allocating the storage!
Direct reclaim should be considered extremely harmful.
This is a high frequency, high throughput, CPU usage and latency
sensitive allocation. We've got memory there, and we're using
kvmalloc to allow memory allocation to avoid doing lots of work to
try to do contiguous allocations.
Except it still does *lots of costly work* that is unnecessary.
Worse: the only way to avoid the slowpath page allocation trying to
do compaction on costly allocations is to turn off direct reclaim
(i.e. remove __GFP_RECLAIM_DIRECT from the gfp flags).
Unfortunately, the stupid kvmalloc API then says "oh, this isn't a
GFP_KERNEL allocation context, so you only get kmalloc!". This
cuts off the vmalloc fallback, and this leads to almost instant OOM
problems which ends up in filesystems deadlocks, shutdowns and/or
kernel crashes.
I want some basic kvmalloc behaviour:
- kmalloc for a contiguous range with fail fast semantics - no
compaction direct reclaim if the allocation enters the slow path.
- run normal vmalloc (i.e. GFP_KERNEL) if kmalloc fails
The really, really stupid part about this is these kvmalloc() calls
are run under memalloc_nofs task context, so all the allocations are
always reduced to GFP_NOFS regardless of the fact that kvmalloc
requires GFP_KERNEL to be passed in. IOWs, we're already telling
kvmalloc to behave differently to the gfp flags we pass in, but it
still won't allow vmalloc to be run with anything other than
GFP_KERNEL.
So, this patch open codes the kvmalloc() in the commit path to have
the above described behaviour. The result is we more than halve the
CPU time spend doing kvmalloc() in this path and transaction commits
with 64kB objects in them more than doubles. i.e. we get ~5x
reduction in CPU usage per costly-sized kvmalloc() invocation and
the profile looks like this:
- 37.60% xlog_cil_commit
16.01% memcpy_erms
- 8.45% __kmalloc
- 8.04% kmalloc_order_trace
- 8.03% kmalloc_order
- 7.93% alloc_pages
- 7.90% __alloc_pages
- 4.05% __alloc_pages_slowpath.constprop.0
- 2.18% get_page_from_freelist
- 1.77% wake_all_kswapds
....
- __wake_up_common_lock
- 0.94% _raw_spin_lock_irqsave
- 3.72% get_page_from_freelist
- 2.43% _raw_spin_lock_irqsave
- 5.72% vmalloc
- 5.72% __vmalloc_node_range
- 4.81% __get_vm_area_node.constprop.0
- 3.26% alloc_vmap_area
- 2.52% _raw_spin_lock
- 1.46% _raw_spin_lock
0.56% __alloc_pages_bulk
- 4.66% kvfree
- 3.25% vfree
- __vfree
- 3.23% __vunmap
- 1.95% remove_vm_area
- 1.06% free_vmap_area_noflush
- 0.82% _raw_spin_lock
- 0.68% _raw_spin_lock
- 0.92% _raw_spin_lock
- 1.40% kfree
- 1.36% __free_pages
- 1.35% __free_pages_ok
- 1.02% _raw_spin_lock_irqsave
It's worth noting that over 50% of the CPU time spent allocating
these shadow buffers is now spent on spinlocks. So the shadow buffer
allocation overhead is greatly reduced by getting rid of direct
reclaim from kmalloc, and could probably be made even less costly if
vmalloc() didn't use global spinlocks to protect it's structures.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
While I was running with KASAN and lockdep enabled, I stumbled upon an
KASAN report about a UAF to a freed CIL checkpoint. Looking at the
comment for xfs_log_item_in_current_chkpt, it seems pretty obvious to me
that the original patch to xfs_defer_finish_noroll should have done
something to lock the CIL to prevent it from switching the CIL contexts
while the predicate runs.
For upper level code that needs to know if a given log item is new
enough not to need relogging, add a new wrapper that takes the CIL
context lock long enough to sample the current CIL context. This is
kind of racy in that the CIL can switch the contexts immediately after
sampling, but that's ok because the consequence is that the defer ops
code is a little slow to relog items.
==================================================================
BUG: KASAN: use-after-free in xfs_log_item_in_current_chkpt+0x139/0x160 [xfs]
Read of size 8 at addr ffff88804ea5f608 by task fsstress/527999
CPU: 1 PID: 527999 Comm: fsstress Tainted: G D 5.16.0-rc4-xfsx #rc4
Call Trace:
<TASK>
dump_stack_lvl+0x45/0x59
print_address_description.constprop.0+0x1f/0x140
kasan_report.cold+0x83/0xdf
xfs_log_item_in_current_chkpt+0x139/0x160
xfs_defer_finish_noroll+0x3bb/0x1e30
__xfs_trans_commit+0x6c8/0xcf0
xfs_reflink_remap_extent+0x66f/0x10e0
xfs_reflink_remap_blocks+0x2dd/0xa90
xfs_file_remap_range+0x27b/0xc30
vfs_dedupe_file_range_one+0x368/0x420
vfs_dedupe_file_range+0x37c/0x5d0
do_vfs_ioctl+0x308/0x1260
__x64_sys_ioctl+0xa1/0x170
do_syscall_64+0x35/0x80
entry_SYSCALL_64_after_hwframe+0x44/0xae
RIP: 0033:0x7f2c71a2950b
Code: 0f 1e fa 48 8b 05 85 39 0d 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff
ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01
f0 ff ff 73 01 c3 48 8b 0d 55 39 0d 00 f7 d8 64 89 01 48
RSP: 002b:00007ffe8c0e03c8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 00005600862a8740 RCX: 00007f2c71a2950b
RDX: 00005600862a7be0 RSI: 00000000c0189436 RDI: 0000000000000004
RBP: 000000000000000b R08: 0000000000000027 R09: 0000000000000003
R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000005a
R13: 00005600862804a8 R14: 0000000000016000 R15: 00005600862a8a20
</TASK>
Allocated by task 464064:
kasan_save_stack+0x1e/0x50
__kasan_kmalloc+0x81/0xa0
kmem_alloc+0xcd/0x2c0 [xfs]
xlog_cil_ctx_alloc+0x17/0x1e0 [xfs]
xlog_cil_push_work+0x141/0x13d0 [xfs]
process_one_work+0x7f6/0x1380
worker_thread+0x59d/0x1040
kthread+0x3b0/0x490
ret_from_fork+0x1f/0x30
Freed by task 51:
kasan_save_stack+0x1e/0x50
kasan_set_track+0x21/0x30
kasan_set_free_info+0x20/0x30
__kasan_slab_free+0xed/0x130
slab_free_freelist_hook+0x7f/0x160
kfree+0xde/0x340
xlog_cil_committed+0xbfd/0xfe0 [xfs]
xlog_cil_process_committed+0x103/0x1c0 [xfs]
xlog_state_do_callback+0x45d/0xbd0 [xfs]
xlog_ioend_work+0x116/0x1c0 [xfs]
process_one_work+0x7f6/0x1380
worker_thread+0x59d/0x1040
kthread+0x3b0/0x490
ret_from_fork+0x1f/0x30
Last potentially related work creation:
kasan_save_stack+0x1e/0x50
__kasan_record_aux_stack+0xb7/0xc0
insert_work+0x48/0x2e0
__queue_work+0x4e7/0xda0
queue_work_on+0x69/0x80
xlog_cil_push_now.isra.0+0x16b/0x210 [xfs]
xlog_cil_force_seq+0x1b7/0x850 [xfs]
xfs_log_force_seq+0x1c7/0x670 [xfs]
xfs_file_fsync+0x7c1/0xa60 [xfs]
__x64_sys_fsync+0x52/0x80
do_syscall_64+0x35/0x80
entry_SYSCALL_64_after_hwframe+0x44/0xae
The buggy address belongs to the object at ffff88804ea5f600
which belongs to the cache kmalloc-256 of size 256
The buggy address is located 8 bytes inside of
256-byte region [ffff88804ea5f600, ffff88804ea5f700)
The buggy address belongs to the page:
page:ffffea00013a9780 refcount:1 mapcount:0 mapping:0000000000000000 index:0xffff88804ea5ea00 pfn:0x4ea5e
head:ffffea00013a9780 order:1 compound_mapcount:0
flags: 0x4fff80000010200(slab|head|node=1|zone=1|lastcpupid=0xfff)
raw: 04fff80000010200 ffffea0001245908 ffffea00011bd388 ffff888004c42b40
raw: ffff88804ea5ea00 0000000000100009 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
ffff88804ea5f500: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
ffff88804ea5f580: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>ffff88804ea5f600: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
^
ffff88804ea5f680: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
ffff88804ea5f700: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
==================================================================
Fixes: 4e919af782 ("xfs: periodically relog deferred intent items")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Replace m_flags feature checks with xfs_has_<feature>() calls and
rework the setup code to set flags in m_features.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
We only use the CIL workqueue in the CIL, so it makes no sense to
hang it off the xfs_mount and have to walk multiple pointers back up
to the mount when we have the CIL structures right there.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Because we use a single work structure attached to the CIL rather
than the CIL context, we can only queue a single work item at a
time. This results in the CIL being single threaded and limits
performance when it becomes CPU bound.
The design of the CIL is that it is pipelined and multiple commits
can be running concurrently, but the way the work is currently
implemented means that it is not pipelining as it was intended. The
critical work to switch the CIL context can take a few milliseconds
to run, but the rest of the CIL context flush can take hundreds of
milliseconds to complete. The context switching is the serialisation
point of the CIL, once the context has been switched the rest of the
context push can run asynchrnously with all other context pushes.
Hence we can move the work to the CIL context so that we can run
multiple CIL pushes at the same time and spread the majority of
the work out over multiple CPUs. We can keep the per-cpu CIL commit
state on the CIL rather than the context, because the context is
pinned to the CIL until the switch is done and we aggregate and
drain the per-cpu state held on the CIL during the context switch.
However, because we no longer serialise the CIL work, we can have
effectively unlimited CIL pushes in progress. We don't want to do
this - not only does it create contention on the iclogs and the
state machine locks, we can run the log right out of space with
outstanding pushes. Instead, limit the work concurrency to 4
concurrent works being processed at a time. This is enough
concurrency to remove the CIL from being a CPU bound bottleneck but
not enough to create new contention points or unbound concurrency
issues.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
The AIL pushing is stalling on log forces when it comes across
pinned items. This is happening on removal workloads where the AIL
is dominated by stale items that are removed from AIL when the
checkpoint that marks the items stale is committed to the journal.
This results is relatively few items in the AIL, but those that are
are often pinned as directories items are being removed from are
still being logged.
As a result, many push cycles through the CIL will first issue a
blocking log force to unpin the items. This can take some time to
complete, with tracing regularly showing push delays of half a
second and sometimes up into the range of several seconds. Sequences
like this aren't uncommon:
....
399.829437: xfsaild: last lsn 0x11002dd000 count 101 stuck 101 flushing 0 tout 20
<wanted 20ms, got 270ms delay>
400.099622: xfsaild: target 0x11002f3600, prev 0x11002f3600, last lsn 0x0
400.099623: xfsaild: first lsn 0x11002f3600
400.099679: xfsaild: last lsn 0x1100305000 count 16 stuck 11 flushing 0 tout 50
<wanted 50ms, got 500ms delay>
400.589348: xfsaild: target 0x110032e600, prev 0x11002f3600, last lsn 0x0
400.589349: xfsaild: first lsn 0x1100305000
400.589595: xfsaild: last lsn 0x110032e600 count 156 stuck 101 flushing 30 tout 50
<wanted 50ms, got 460ms delay>
400.950341: xfsaild: target 0x1100353000, prev 0x110032e600, last lsn 0x0
400.950343: xfsaild: first lsn 0x1100317c00
400.950436: xfsaild: last lsn 0x110033d200 count 105 stuck 101 flushing 0 tout 20
<wanted 20ms, got 200ms delay>
401.142333: xfsaild: target 0x1100361600, prev 0x1100353000, last lsn 0x0
401.142334: xfsaild: first lsn 0x110032e600
401.142535: xfsaild: last lsn 0x1100353000 count 122 stuck 101 flushing 8 tout 10
<wanted 10ms, got 10ms delay>
401.154323: xfsaild: target 0x1100361600, prev 0x1100361600, last lsn 0x1100353000
401.154328: xfsaild: first lsn 0x1100353000
401.154389: xfsaild: last lsn 0x1100353000 count 101 stuck 101 flushing 0 tout 20
<wanted 20ms, got 300ms delay>
401.451525: xfsaild: target 0x1100361600, prev 0x1100361600, last lsn 0x0
401.451526: xfsaild: first lsn 0x1100353000
401.451804: xfsaild: last lsn 0x1100377200 count 170 stuck 22 flushing 122 tout 50
<wanted 50ms, got 500ms delay>
401.933581: xfsaild: target 0x1100361600, prev 0x1100361600, last lsn 0x0
....
In each of these cases, every AIL pass saw 101 log items stuck on
the AIL (pinned) with very few other items being found. Each pass, a
log force was issued, and delay between last/first is the sleep time
+ the sync log force time.
Some of these 101 items pinned the tail of the log. The tail of the
log does slowly creep forward (first lsn), but the problem is that
the log is actually out of reservation space because it's been
running so many transactions that stale items that never reach the
AIL but consume log space. Hence we have a largely empty AIL, with
long term pins on items that pin the tail of the log that don't get
pushed frequently enough to keep log space available.
The problem is the hundreds of milliseconds that we block in the log
force pushing the CIL out to disk. The AIL should not be stalled
like this - it needs to run and flush items that are at the tail of
the log with minimal latency. What we really need to do is trigger a
log flush, but then not wait for it at all - we've already done our
waiting for stuff to complete when we backed off prior to the log
force being issued.
Even if we remove the XFS_LOG_SYNC from the xfs_log_force() call, we
still do a blocking flush of the CIL and that is what is causing the
issue. Hence we need a new interface for the CIL to trigger an
immediate background push of the CIL to get it moving faster but not
to wait on that to occur. While the CIL is pushing, the AIL can also
be pushing.
We already have an internal interface to do this -
xlog_cil_push_now() - but we need a wrapper for it to be used
externally. xlog_cil_force_seq() can easily be extended to do what
we need as it already implements the synchronous CIL push via
xlog_cil_push_now(). Add the necessary flags and "push current
sequence" semantics to xlog_cil_force_seq() and convert the AIL
pushing to use it.
One of the complexities here is that the CIL push does not guarantee
that the commit record for the CIL checkpoint is written to disk.
The current log force ensures this by submitting the current ACTIVE
iclog that the commit record was written to. We need the CIL to
actually write this commit record to disk for an async push to
ensure that the checkpoint actually makes it to disk and unpins the
pinned items in the checkpoint on completion. Hence we need to pass
down to the CIL push that we are doing an async flush so that it can
switch out the commit_iclog if necessary to get written to disk when
the commit iclog is finally released.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Because log recovery depends on strictly ordered start records as
well as strictly ordered commit records.
This is a zero day bug in the way XFS writes pipelined transactions
to the journal which is exposed by fixing the zero day bug that
prevents the CIL from pipelining checkpoints. This re-introduces
explicit concurrent commits back into the on-disk journal and hence
out of order start records.
The XFS journal commit code has never ordered start records and we
have relied on strict commit record ordering for correct recovery
ordering of concurrently written transactions. Unfortunately, root
cause analysis uncovered the fact that log recovery uses the LSN of
the start record for transaction commit processing. Hence, whilst
the commits are processed in strict order by recovery, the LSNs
associated with the commits can be out of order and so recovery may
stamp incorrect LSNs into objects and/or misorder intents in the AIL
for later processing. This can result in log recovery failures
and/or on disk corruption, sometimes silent.
Because this is a long standing log recovery issue, we can't just
fix log recovery and call it good. This still leaves older kernels
susceptible to recovery failures and corruption when replaying a log
from a kernel that pipelines checkpoints. There is also the issue
that in-memory ordering for AIL pushing and data integrity
operations are based on checkpoint start LSNs, and if the start LSN
is incorrect in the journal, it is also incorrect in memory.
Hence there's really only one choice for fixing this zero-day bug:
we need to strictly order checkpoint start records in ascending
sequence order in the log, the same way we already strictly order
commit records.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Now that we have a mechanism to guarantee that the callbacks
attached to an iclog are owned by the context that attaches them
until they drop their reference to the iclog via
xlog_state_release_iclog(), we can attach callbacks to the iclog at
any time we have an active reference to the iclog.
xlog_state_get_iclog_space() always guarantees that the commit
record will fit in the iclog it returns, so we can move this IO
callback setting to xlog_cil_set_ctx_write_state(), record the
commit iclog in the context and remove the need for the commit iclog
to be returned by xlog_write() altogether.
This, in turn, allows us to move the wakeup for ordered commit
record writes up into xlog_cil_set_ctx_write_state(), too, because
we have been guaranteed that this commit record will be physically
located in the iclog before any waiting commit record at a higher
sequence number will be granted iclog space.
This further cleans up the post commit record write processing in
the CIL push code, especially as xlog_state_release_iclog() will now
clean up the context when shutdown errors occur.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
So we can use it for start record ordering as well as commit record
ordering in future.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Pass the CIL context to xlog_write() rather than a pointer to a LSN
variable. Only the CIL checkpoint calls to xlog_write() need to know
about the start LSN of the writes, so rework xlog_write to directly
write the LSNs into the CIL context structure.
This removes the commit_lsn variable from xlog_cil_push_work(), so
now we only have to issue the commit record ordering wakeup from
there.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
It is only used by the CIL checkpoints, and is the counterpart to
start record formatting and writing that is already local to
xfs_log_cil.c.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
When the log is shutdown, it currently walks all the iclogs and runs
callbacks that are attached to the iclogs, regardless of whether the
iclog is queued for IO completion or not. This creates a problem for
contexts attaching callbacks to iclogs in that a racing shutdown can
run the callbacks even before the attaching context has finished
processing the iclog and releasing it for IO submission.
If the callback processing of the iclog frees the structure that is
attached to the iclog, then this leads to an UAF scenario that can
only be protected against by holding the icloglock from the point
callbacks are attached through to the release of the iclog. While we
currently do this, it is not practical or sustainable.
Hence we need to make shutdown processing the responsibility of the
context that holds active references to the iclog. We know that the
contexts attaching callbacks to the iclog must have active
references to the iclog, and that means they must be in either
ACTIVE or WANT_SYNC states. xlog_state_do_callback() will skip over
iclogs in these states -except- when the log is shut down.
xlog_state_do_callback() checks the state of the iclogs while
holding the icloglock, therefore the reference count/state change
that occurs in xlog_state_release_iclog() after the callbacks are
atomic w.r.t. shutdown processing.
We can't push the responsibility of callback cleanup onto the CIL
context because we can have ACTIVE iclogs that have callbacks
attached that have already been released. Hence we really need to
internalise the cleanup of callbacks into xlog_state_release_iclog()
processing.
Indeed, we already have that internalisation via:
xlog_state_release_iclog
drop last reference
->SYNCING
xlog_sync
xlog_write_iclog
if (log_is_shutdown)
xlog_state_done_syncing()
xlog_state_do_callback()
<process shutdown on iclog that is now in SYNCING state>
The problem is that xlog_state_release_iclog() aborts before doing
anything if the log is already shut down. It assumes that the
callbacks have already been cleaned up, and it doesn't need to do
any cleanup.
Hence the fix is to remove the xlog_is_shutdown() check from
xlog_state_release_iclog() so that reference counts are correctly
released from the iclogs, and when the reference count is zero we
always transition to SYNCING if the log is shut down. Hence we'll
always enter the xlog_sync() path in a shutdown and eventually end
up erroring out the iclog IO and running xlog_state_do_callback() to
process the callbacks attached to the iclog.
This allows us to stop processing referenced ACTIVE/WANT_SYNC iclogs
directly in the shutdown code, and in doing so gets rid of the UAF
vector that currently exists. This then decouples the adding of
callbacks to the iclogs from xlog_state_release_iclog() as we
guarantee that xlog_state_release_iclog() will process the callbacks
if the log has been shut down before xlog_state_release_iclog() has
been called.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
We don't need an iclog state field to tell us the log has been shut
down. We can just check the xlog_is_shutdown() instead. The avoids
the need to have shutdown overwrite the current iclog state while
being active used by the log code and so having to ensure that every
iclog state check handles XLOG_STATE_IOERROR appropriately.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Make it less shouty and a static inline before adding more calls
through the log code.
Also convert internal log code that uses XFS_FORCED_SHUTDOWN(mount)
to use xlog_is_shutdown(log) as well.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
There is no reason for this wrapper existing anymore. All the places
that use KM_NOFS allocation are within transaction contexts and
hence covered by memalloc_nofs_save/restore contexts. Hence we don't
need any special handling of vmalloc for large IOs anymore and
so special casing this code isn't necessary.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
There is a race between the new CIL async data device metadata IO
completion cache flush and the log tail in the iclog the flush
covers being updated. This can be seen by repeating generic/482 in a
loop and eventually log recovery fails with a failures such as this:
XFS (dm-3): Starting recovery (logdev: internal)
XFS (dm-3): bad inode magic/vsn daddr 228352 #0 (magic=0)
XFS (dm-3): Metadata corruption detected at xfs_inode_buf_verify+0x180/0x190, xfs_inode block 0x37c00 xfs_inode_buf_verify
XFS (dm-3): Unmount and run xfs_repair
XFS (dm-3): First 128 bytes of corrupted metadata buffer:
00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00000070: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
XFS (dm-3): metadata I/O error in "xlog_recover_items_pass2+0x55/0xc0" at daddr 0x37c00 len 32 error 117
Analysis of the logwrite replay shows that there were no writes to
the data device between the FUA @ write 124 and the FUA at write @
125, but log recovery @ 125 failed. The difference was the one log
write @ 125 moved the tail of the log forwards from (1,8) to (1,32)
and so the inode create intent in (1,8) was not replayed and so the
inode cluster was zero on disk when replay of the first inode item
in (1,32) was attempted.
What this meant was that the journal write that occurred at @ 125
did not ensure that metadata completed before the iclog was written
was correctly on stable storage. The tail of the log moved forward,
so IO must have been completed between the two iclog writes. This
means that there is a race condition between the unconditional async
cache flush in the CIL push work and the tail LSN that is written to
the iclog. This happens like so:
CIL push work AIL push work
------------- -------------
Add to committing list
start async data dev cache flush
.....
<flush completes>
<all writes to old tail lsn are stable>
xlog_write
.... push inode create buffer
<start IO>
.....
xlog_write(commit record)
.... <IO completes>
log tail moves
xlog_assign_tail_lsn()
start_lsn == commit_lsn
<no iclog preflush!>
xlog_state_release_iclog
__xlog_state_release_iclog()
<writes *new* tail_lsn into iclog>
xlog_sync()
....
submit_bio()
<tail in log moves forward without flushing written metadata>
Essentially, this can only occur if the commit iclog is issued
without a cache flush. If the iclog bio is submitted with
REQ_PREFLUSH, then it will guarantee that all the completed IO is
one stable storage before the iclog bio with the new tail LSN in it
is written to the log.
IOWs, the tail lsn that is written to the iclog needs to be sampled
*before* we issue the cache flush that guarantees all IO up to that
LSN has been completed.
To fix this without giving up the performance advantage of the
flush/FUA optimisations (e.g. g/482 runtime halves with 5.14-rc1
compared to 5.13), we need to ensure that we always issue a cache
flush if the tail LSN changes between the initial async flush and
the commit record being written. THis requires sampling the tail_lsn
before we start the flush, and then passing the sampled tail LSN to
xlog_state_release_iclog() so it can determine if the the tail LSN
has changed while writing the checkpoint. If the tail LSN has
changed, then it needs to set the NEED_FLUSH flag on the iclog and
we'll issue another cache flush before writing the iclog.
Fixes: eef983ffea ("xfs: journal IO cache flush reductions")
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
The iclogbuf ring attached to the struct xlog is circular, hence the
first and last iclogs in the ring can only be determined by
comparing them against the log->l_iclog pointer.
In xfs_cil_push_work(), we want to wait on previous iclogs that were
issued so that we can flush them to stable storage with the commit
record write, and it simply waits on the previous iclog in the ring.
This, however, leads to CIL push hangs in generic/019 like so:
task:kworker/u33:0 state:D stack:12680 pid: 7 ppid: 2 flags:0x00004000
Workqueue: xfs-cil/pmem1 xlog_cil_push_work
Call Trace:
__schedule+0x30b/0x9f0
schedule+0x68/0xe0
xlog_wait_on_iclog+0x121/0x190
? wake_up_q+0xa0/0xa0
xlog_cil_push_work+0x994/0xa10
? _raw_spin_lock+0x15/0x20
? xfs_swap_extents+0x920/0x920
process_one_work+0x1ab/0x390
worker_thread+0x56/0x3d0
? rescuer_thread+0x3c0/0x3c0
kthread+0x14d/0x170
? __kthread_bind_mask+0x70/0x70
ret_from_fork+0x1f/0x30
With other threads blocking in either xlog_state_get_iclog_space()
waiting for iclog space or xlog_grant_head_wait() waiting for log
reservation space.
The problem here is that the previous iclog on the ring might
actually be a future iclog. That is, if log->l_iclog points at
commit_iclog, commit_iclog is the first (oldest) iclog in the ring
and there are no previous iclogs pending as they have all completed
their IO and been activated again. IOWs, commit_iclog->ic_prev
points to an iclog that will be written in the future, not one that
has been written in the past.
Hence, in this case, waiting on the ->ic_prev iclog is incorrect
behaviour, and depending on the state of the future iclog, we can
end up with a circular ABA wait cycle and we hang.
The fix is made more complex by the fact that many iclogs states
cannot be used to determine if the iclog is a past or future iclog.
Hence we have to determine past iclogs by checking the LSN of the
iclog rather than their state. A past ACTIVE iclog will have a LSN
of zero, while a future ACTIVE iclog will have a LSN greater than
the current iclog. We don't wait on either of these cases.
Similarly, a future iclog that hasn't completed IO will have an LSN
greater than the current iclog and so we don't wait on them. A past
iclog that is still undergoing IO completion will have a LSN less
than the current iclog and those are the only iclogs that we need to
wait on.
Hence we can use the iclog LSN to determine what iclogs we need to
wait on here.
Fixes: 5fd9256ce156 ("xfs: separate CIL commit record IO")
Reported-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
The iclog callback chain has it's own lock. That was added way back
in 2008 by myself to alleviate severe lock contention on the
icloglock in commit 114d23aae5 ("[XFS] Per iclog callback chain
lock"). This was long before delayed logging took the icloglock out
of the hot transaction commit path and removed all contention on it.
Hence the separate ic_callback_lock doesn't serve any scalability
purpose anymore, and hasn't for close on a decade.
Further, we only attach callbacks to iclogs in one place where we
are already taking the icloglock soon after attaching the callbacks.
We also have to drop the icloglock to run callbacks and grab it
immediately afterwards again. So given that the icloglock is no
longer hot, making it cover callbacks again doesn't really change
the locking patterns very much at all.
We also need to extend the icloglock to cover callback addition to
fix a zero-day UAF in the CIL push code. This occurs when shutdown
races with xlog_cil_push_work() and the shutdown runs the callbacks
before the push releases the iclog. This results in the CIL context
structure attached to the iclog being freed by the callback before
the CIL push has finished referencing it, leading to UAF bugs.
Hence, to avoid this UAF, we need the callback attachment to be
atomic with post processing of the commit iclog and references to
the structures being attached to the iclog. This requires holding
the icloglock as that's the only way to serialise iclog state
against a shutdown in progress.
The result is we need to be using the icloglock to protect the
callback list addition and removal and serialise them with shutdown.
That makes the ic_callback_lock redundant and so it can be removed.
Fixes: 71e330b593 ("xfs: Introduce delayed logging core code")
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
In doing an investigation into AIL push stalls, I was looking at the
log force code to see if an async CIL push could be done instead.
This lead me to xfs_log_force_lsn() and looking at how it works.
xfs_log_force_lsn() is only called from inode synchronisation
contexts such as fsync(), and it takes the ip->i_itemp->ili_last_lsn
value as the LSN to sync the log to. This gets passed to
xlog_cil_force_lsn() via xfs_log_force_lsn() to flush the CIL to the
journal, and then used by xfs_log_force_lsn() to flush the iclogs to
the journal.
The problem is that ip->i_itemp->ili_last_lsn does not store a
log sequence number. What it stores is passed to it from the
->iop_committing method, which is called by xfs_log_commit_cil().
The value this passes to the iop_committing method is the CIL
context sequence number that the item was committed to.
As it turns out, xlog_cil_force_lsn() converts the sequence to an
actual commit LSN for the related context and returns that to
xfs_log_force_lsn(). xfs_log_force_lsn() overwrites it's "lsn"
variable that contained a sequence with an actual LSN and then uses
that to sync the iclogs.
This caused me some confusion for a while, even though I originally
wrote all this code a decade ago. ->iop_committing is only used by
a couple of log item types, and only inode items use the sequence
number it is passed.
Let's clean up the API, CIL structures and inode log item to call it
a sequence number, and make it clear that the high level code is
using CIL sequence numbers and not on-disk LSNs for integrity
synchronisation purposes.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
A hang with tasks stuck on the CIL hard throttle was reported and
largely diagnosed by Donald Buczek, who discovered that it was a
result of the CIL context space usage decrementing in committed
transactions once the hard throttle limit had been hit and processes
were already blocked. This resulted in the CIL push not waking up
those waiters because the CIL context was no longer over the hard
throttle limit.
The surprising aspect of this was the CIL space usage going
backwards regularly enough to trigger this situation. Assumptions
had been made in design that the relogging process would only
increase the size of the objects in the CIL, and so that space would
only increase.
This change and commit message fixes the issue and documents the
result of an audit of the triggers that can cause the CIL space to
go backwards, how large the backwards steps tend to be, the
frequency in which they occur, and what the impact on the CIL
accounting code is.
Even though the CIL ctx->space_used can go backwards, it will only
do so if the log item is already logged to the CIL and contains a
space reservation for it's entire logged state. This is tracked by
the shadow buffer state on the log item. If the item is not
previously logged in the CIL it has no shadow buffer nor log vector,
and hence the entire size of the logged item copied to the log
vector is accounted to the CIL space usage. i.e. it will always go
up in this case.
If the item has a log vector (i.e. already in the CIL) and the size
decreases, then the existing log vector will be overwritten and the
space usage will go down. This is the only condition where the space
usage reduces, and it can only occur when an item is already tracked
in the CIL. Hence we are safe from CIL space usage underruns as a
result of log items decreasing in size when they are relogged.
Typically this reduction in CIL usage occurs from metadata blocks
being free, such as when a btree block merge occurs or a directory
enter/xattr entry is removed and the da-tree is reduced in size.
This generally results in a reduction in size of around a single
block in the CIL, but also tends to increase the number of log
vectors because the parent and sibling nodes in the tree needs to be
updated when a btree block is removed. If a multi-level merge
occurs, then we see reduction in size of 2+ blocks, but again the
log vector count goes up.
The other vector is inode fork size changes, which only log the
current size of the fork and ignore the previously logged size when
the fork is relogged. Hence if we are removing items from the inode
fork (dir/xattr removal in shortform, extent record removal in
extent form, etc) the relogged size of the inode for can decrease.
No other log items can decrease in size either because they are a
fixed size (e.g. dquots) or they cannot be relogged (e.g. relogging
an intent actually creates a new intent log item and doesn't relog
the old item at all.) Hence the only two vectors for CIL context
size reduction are relogging inode forks and marking buffers active
in the CIL as stale.
Long story short: the majority of the code does the right thing and
handles the reduction in log item size correctly, and only the CIL
hard throttle implementation is problematic and needs fixing. This
patch makes that fix, as well as adds comments in the log item code
that result in items shrinking in size when they are relogged as a
clear reminder that this can and does happen frequently.
The throttle fix is based upon the change Donald proposed, though it
goes further to ensure that once the throttle is activated, it
captures all tasks until the CIL push issues a wakeup, regardless of
whether the CIL space used has gone back under the throttle
threshold.
This ensures that we prevent tasks reducing the CIL slightly under
the throttle threshold and then making more changes that push it
well over the throttle limit. This is acheived by checking if the
throttle wait queue is already active as a condition of throttling.
Hence once we start throttling, we continue to apply the throttle
until the CIL context push wakes everything on the wait queue.
We can use waitqueue_active() for the waitqueue manipulations and
checks as they are all done under the ctx->xc_push_lock. Hence the
waitqueue has external serialisation and we can safely peek inside
the wait queue without holding the internal waitqueue locks.
Many thanks to Donald for his diagnostic and analysis work to
isolate the cause of this hang.
Reported-and-tested-by: Donald Buczek <buczek@molgen.mpg.de>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Chandan Babu R <chandanrlinux@gmail.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Currently every journal IO is issued as REQ_PREFLUSH | REQ_FUA to
guarantee the ordering requirements the journal has w.r.t. metadata
writeback. THe two ordering constraints are:
1. we cannot overwrite metadata in the journal until we guarantee
that the dirty metadata has been written back in place and is
stable.
2. we cannot write back dirty metadata until it has been written to
the journal and guaranteed to be stable (and hence recoverable) in
the journal.
The ordering guarantees of #1 are provided by REQ_PREFLUSH. This
causes the journal IO to issue a cache flush and wait for it to
complete before issuing the write IO to the journal. Hence all
completed metadata IO is guaranteed to be stable before the journal
overwrites the old metadata.
The ordering guarantees of #2 are provided by the REQ_FUA, which
ensures the journal writes do not complete until they are on stable
storage. Hence by the time the last journal IO in a checkpoint
completes, we know that the entire checkpoint is on stable storage
and we can unpin the dirty metadata and allow it to be written back.
This is the mechanism by which ordering was first implemented in XFS
way back in 2002 by commit 95d97c36e5155075ba2eb22b17562cfcc53fcf96
("Add support for drive write cache flushing") in the xfs-archive
tree.
A lot has changed since then, most notably we now use delayed
logging to checkpoint the filesystem to the journal rather than
write each individual transaction to the journal. Cache flushes on
journal IO are necessary when individual transactions are wholly
contained within a single iclog. However, CIL checkpoints are single
transactions that typically span hundreds to thousands of individual
journal writes, and so the requirements for device cache flushing
have changed.
That is, the ordering rules I state above apply to ordering of
atomic transactions recorded in the journal, not to the journal IO
itself. Hence we need to ensure metadata is stable before we start
writing a new transaction to the journal (guarantee #1), and we need
to ensure the entire transaction is stable in the journal before we
start metadata writeback (guarantee #2).
Hence we only need a REQ_PREFLUSH on the journal IO that starts a
new journal transaction to provide #1, and it is not on any other
journal IO done within the context of that journal transaction.
The CIL checkpoint already issues a cache flush before it starts
writing to the log, so we no longer need the iclog IO to issue a
REQ_REFLUSH for us. Hence if XLOG_START_TRANS is passed
to xlog_write(), we no longer need to mark the first iclog in
the log write with REQ_PREFLUSH for this case. As an added bonus,
this ordering mechanism works for both internal and external logs,
meaning we can remove the explicit data device cache flushes from
the iclog write code when using external logs.
Given the new ordering semantics of commit records for the CIL, we
need iclogs containing commit records to issue a REQ_PREFLUSH. We
also require unmount records to do this. Hence for both
XLOG_COMMIT_TRANS and XLOG_UNMOUNT_TRANS xlog_write() calls we need
to mark the first iclog being written with REQ_PREFLUSH.
For both commit records and unmount records, we also want them
immediately on stable storage, so we want to also mark the iclogs
that contain these records to be marked REQ_FUA. That means if a
record is split across multiple iclogs, they are all marked REQ_FUA
and not just the last one so that when the transaction is completed
all the parts of the record are on stable storage.
And for external logs, unmount records need a pre-write data device
cache flush similar to the CIL checkpoint cache pre-flush as the
internal iclog write code does not do this implicitly anymore.
As an optimisation, when the commit record lands in the same iclog
as the journal transaction starts, we don't need to wait for
anything and can simply use REQ_FUA to provide guarantee #2. This
means that for fsync() heavy workloads, the cache flush behaviour is
completely unchanged and there is no degradation in performance as a
result of optimise the multi-IO transaction case.
The most notable sign that there is less IO latency on my test
machine (nvme SSDs) is that the "noiclogs" rate has dropped
substantially. This metric indicates that the CIL push is blocking
in xlog_get_iclog_space() waiting for iclog IO completion to occur.
With 8 iclogs of 256kB, the rate is appoximately 1 noiclog event to
every 4 iclog writes. IOWs, every 4th call to xlog_get_iclog_space()
is blocking waiting for log IO. With the changes in this patch, this
drops to 1 noiclog event for every 100 iclog writes. Hence it is
clear that log IO is completing much faster than it was previously,
but it is also clear that for large iclog sizes, this isn't the
performance limiting factor on this hardware.
With smaller iclogs (32kB), however, there is a substantial
difference. With the cache flush modifications, the journal is now
running at over 4000 write IOPS, and the journal throughput is
largely identical to the 256kB iclogs and the noiclog event rate
stays low at about 1:50 iclog writes. The existing code tops out at
about 2500 IOPS as the number of cache flushes dominate performance
and latency. The noiclog event rate is about 1:4, and the
performance variance is quite large as the journal throughput can
fall to less than half the peak sustained rate when the cache flush
rate prevents metadata writeback from keeping up and the log runs
out of space and throttles reservations.
As a result:
logbsize fsmark create rate rm -rf
before 32kb 152851+/-5.3e+04 5m28s
patched 32kb 221533+/-1.1e+04 5m24s
before 256kb 220239+/-6.2e+03 4m58s
patched 256kb 228286+/-9.2e+03 5m06s
The rm -rf times are included because I ran them, but the
differences are largely noise. This workload is largely metadata
read IO latency bound and the changes to the journal cache flushing
doesn't really make any noticable difference to behaviour apart from
a reduction in noiclog events from background CIL pushing.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Chandan Babu R <chandanrlinux@gmail.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
The CIL push is the only call to xlog_write that sets this variable
to true. The other callers don't need a start rec, and they tell
xlog_write what to do by passing the type of ophdr they need written
in the flags field. The need_start_rec parameter essentially tells
xlog_write to to write an extra ophdr with a XLOG_START_TRANS type,
so get rid of the variable to do this and pass XLOG_START_TRANS as
the flag value into xlog_write() from the CIL push.
$ size fs/xfs/xfs_log.o*
text data bss dec hex filename
27595 560 8 28163 6e03 fs/xfs/xfs_log.o.orig
27454 560 8 28022 6d76 fs/xfs/xfs_log.o.patched
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Chandan Babu R <chandanrlinux@gmail.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Currently every journal IO is issued as REQ_PREFLUSH | REQ_FUA to
guarantee the ordering requirements the journal has w.r.t. metadata
writeback. THe two ordering constraints are:
1. we cannot overwrite metadata in the journal until we guarantee
that the dirty metadata has been written back in place and is
stable.
2. we cannot write back dirty metadata until it has been written to
the journal and guaranteed to be stable (and hence recoverable) in
the journal.
These rules apply to the atomic transactions recorded in the
journal, not to the journal IO itself. Hence we need to ensure
metadata is stable before we start writing a new transaction to the
journal (guarantee #1), and we need to ensure the entire transaction
is stable in the journal before we start metadata writeback
(guarantee #2).
The ordering guarantees of #1 are currently provided by REQ_PREFLUSH
being added to every iclog IO. This causes the journal IO to issue a
cache flush and wait for it to complete before issuing the write IO
to the journal. Hence all completed metadata IO is guaranteed to be
stable before the journal overwrites the old metadata.
However, for long running CIL checkpoints that might do a thousand
journal IOs, we don't need every single one of these iclog IOs to
issue a cache flush - the cache flush done before the first iclog is
submitted is sufficient to cover the entire range in the log that
the checkpoint will overwrite because the CIL space reservation
guarantees the tail of the log (completed metadata) is already
beyond the range of the checkpoint write.
Hence we only need a full cache flush between closing off the CIL
checkpoint context (i.e. when the push switches it out) and issuing
the first journal IO. Rather than plumbing this through to the
journal IO, we can start this cache flush the moment the CIL context
is owned exclusively by the push worker. The cache flush can be in
progress while we process the CIL ready for writing, hence
reducing the latency of the initial iclog write. This is especially
true for large checkpoints, where we might have to process hundreds
of thousands of log vectors before we issue the first iclog write.
In these cases, it is likely the cache flush has already been
completed by the time we have built the CIL log vector chain.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Chandan Babu R <chandanrlinux@gmail.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
To allow for iclog IO device cache flush behaviour to be optimised,
we first need to separate out the commit record iclog IO from the
rest of the checkpoint so we can wait for the checkpoint IO to
complete before we issue the commit record.
This separation is only necessary if the commit record is being
written into a different iclog to the start of the checkpoint as the
upcoming cache flushing changes requires completion ordering against
the other iclogs submitted by the checkpoint.
If the entire checkpoint and commit is in the one iclog, then they
are both covered by the one set of cache flush primitives on the
iclog and hence there is no need to separate them for ordering.
Otherwise, we need to wait for all the previous iclogs to complete
so they are ordered correctly and made stable by the REQ_PREFLUSH
that the commit record iclog IO issues. This guarantees that if a
reader sees the commit record in the journal, they will also see the
entire checkpoint that commit record closes off.
This also provides the guarantee that when the commit record IO
completes, we can safely unpin all the log items in the checkpoint
so they can be written back because the entire checkpoint is stable
in the journal.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Chandan Babu R <chandanrlinux@gmail.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Delete repeated words in fs/xfs/.
{we, that, the, a, to, fork}
Change "it it" to "it is" in one location.
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
To: linux-fsdevel@vger.kernel.org
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: linux-xfs@vger.kernel.org
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Dave Chinner
Linus Torvalds
Dave Chinner
Christoph Hellwig
Darrick J. Wong
Randy Dunlap