crypto: qat - resolve race condition during AER recovery
During the PCI AER system's error recovery process, the kernel driver
may encounter a race condition with freeing the reset_data structure's
memory. If the device restart will take more than 10 seconds the function
scheduling that restart will exit due to a timeout, and the reset_data
structure will be freed. However, this data structure is used for
completion notification after the restart is completed, which leads
to a UAF bug.
This results in a KFENCE bug notice.
BUG: KFENCE: use-after-free read in adf_device_reset_worker+0x38/0xa0 [intel_qat]
Use-after-free read at 0x00000000bc56fddf (in kfence-#142):
adf_device_reset_worker+0x38/0xa0 [intel_qat]
process_one_work+0x173/0x340
To resolve this race condition, the memory associated to the container
of the work_struct is freed on the worker if the timeout expired,
otherwise on the function that schedules the worker.
The timeout detection can be done by checking if the caller is
still waiting for completion or not by using completion_done() function.
crypto: qat - change SLAs cleanup flow at shutdown
The implementation of the Rate Limiting (RL) feature includes the cleanup
of all SLAs during device shutdown. For each SLA, the firmware is notified
of the removal through an admin message, the data structures that take
into account the budgets are updated and the memory is freed.
However, this explicit cleanup is not necessary as (1) the device is
reset, and the firmware state is lost and (2) all RL data structures
are freed anyway.
In addition, if the device is unresponsive, for example after a PCI
AER error is detected, the admin interface might not be available.
This might slow down the shutdown sequence and cause a timeout in
the recovery flows which in turn makes the driver believe that the
device is not recoverable.
Fix by replacing the explicit SLAs removal with just a free of the
SLA data structures.
Lukas Bulwahn [Thu, 8 Feb 2024 09:33:27 +0000 (10:33 +0100)]
MAINTAINERS: adjust file entries after crypto vmx file movement
Commit 109303336a0c ("crypto: vmx - Move to arch/powerpc/crypto") moves the
crypto vmx files to arch/powerpc, but misses to adjust the file entries for
IBM Power VMX Cryptographic instructions and LINUX FOR POWERPC.
Hence, ./scripts/get_maintainer.pl --self-test=patterns complains about
broken references.
Adjust these file entries accordingly. To keep the matched files exact
after the movement, spell out each file name in the new directory.
Weili Qian [Wed, 7 Feb 2024 09:51:00 +0000 (17:51 +0800)]
crypto: hisilicon/qm - obtain stop queue status
The debugfs files 'dev_state' and 'dev_timeout' are added.
Users can query the current queue stop status through these two
files. And set the waiting timeout when the queue is released.
dev_state: if dev_timeout is set, dev_state indicates the status
of stopping the queue. 0 indicates that the queue is stopped
successfully. Other values indicate that the queue stops fail.
If dev_timeout is not set, the value of dev_state is 0;
dev_timeout: if the queue fails to stop, the queue is released
after waiting dev_timeout * 20ms.
Weili Qian [Wed, 7 Feb 2024 09:50:59 +0000 (17:50 +0800)]
crypto: hisilicon/qm - add stop function by hardware
Hardware V3 could be able to drain function by sending mailbox
to hardware which will trigger tasks in device to be flushed out.
When the function is reset, the function can be stopped by this way.
In the case when only TSME is enabled, it is useful to state that fact
too, so that users are aware that memory encryption is still enabled
even when the corresponding software variant of memory encryption is not
enabled.
crypto: rsa - restrict plaintext/ciphertext values more
SP 800-56Br2, Section 7.1.1 [1] specifies that:
1. If m does not satisfy 1 < m < (n – 1), output an indication that m is
out of range, and exit without further processing.
Similarly, Section 7.1.2 of the same standard specifies that:
1. If the ciphertext c does not satisfy 1 < c < (n – 1), output an
indication that the ciphertext is out of range, and exit without further
processing.
This range is slightly more conservative than RFC3447, as it also
excludes RSA fixed points 0, 1, and n - 1.
Mun Chun Yep [Fri, 2 Feb 2024 10:53:24 +0000 (18:53 +0800)]
crypto: qat - improve aer error reset handling
Rework the AER reset and recovery flow to take into account root port
integrated devices that gets reset between the error detected and the
slot reset callbacks.
In adf_error_detected() the devices is gracefully shut down. The worker
threads are disabled, the error conditions are notified to listeners and
through PFVF comms and finally the device is reset as part of
adf_dev_down().
In adf_slot_reset(), the device is brought up again. If SRIOV VFs were
enabled before reset, these are re-enabled and VFs are notified of
restarting through PFVF comms.
Furong Zhou [Fri, 2 Feb 2024 10:53:23 +0000 (18:53 +0800)]
crypto: qat - limit heartbeat notifications
When the driver detects an heartbeat failure, it starts the recovery
flow. Set a limit so that the number of events is limited in case the
heartbeat status is read too frequently.
Mun Chun Yep [Fri, 2 Feb 2024 10:53:21 +0000 (18:53 +0800)]
crypto: qat - add fatal error notification
Notify a fatal error condition and optionally reset the device in
the following cases:
* if the device reports an uncorrectable fatal error through an
interrupt
* if the heartbeat feature detects that the device is not
responding
This patch is based on earlier work done by Shashank Gupta.
Mun Chun Yep [Fri, 2 Feb 2024 10:53:20 +0000 (18:53 +0800)]
crypto: qat - re-enable sriov after pf reset
When a Physical Function (PF) is reset, SR-IOV gets disabled, making the
associated Virtual Functions (VFs) unavailable. Even after reset and
using pci_restore_state, VFs remain uncreated because the numvfs still
at 0. Therefore, it's necessary to reconfigure SR-IOV to re-enable VFs.
This commit introduces the ADF_SRIOV_ENABLED configuration flag to cache
the SR-IOV enablement state. SR-IOV is only re-enabled if it was
previously configured.
This commit also introduces a dedicated workqueue without
`WQ_MEM_RECLAIM` flag for enabling SR-IOV during Heartbeat and CPM error
resets, preventing workqueue flushing warning.
This patch is based on earlier work done by Shashank Gupta.
Mun Chun Yep [Fri, 2 Feb 2024 10:53:19 +0000 (18:53 +0800)]
crypto: qat - update PFVF protocol for recovery
Update the PFVF logic to handle restart and recovery. This adds the
following functions:
* adf_pf2vf_notify_fatal_error(): allows the PF to notify VFs that the
device detected a fatal error and requires a reset. This sends to
VF the event `ADF_PF2VF_MSGTYPE_FATAL_ERROR`.
* adf_pf2vf_wait_for_restarting_complete(): allows the PF to wait for
`ADF_VF2PF_MSGTYPE_RESTARTING_COMPLETE` events from active VFs
before proceeding with a reset.
* adf_pf2vf_notify_restarted(): enables the PF to notify VFs with
an `ADF_PF2VF_MSGTYPE_RESTARTED` event after recovery, indicating that
the device is back to normal. This prompts VF drivers switch back to
use the accelerator for workload processing.
These changes improve the communication and synchronization between PF
and VF drivers during system restart and recovery processes.
Furong Zhou [Fri, 2 Feb 2024 10:53:17 +0000 (18:53 +0800)]
crypto: qat - add fatal error notify method
Add error notify method to report a fatal error event to all the
subsystems registered. In addition expose an API,
adf_notify_fatal_error(), that allows to trigger a fatal error
notification asynchronously in the context of a workqueue.
This will be invoked when a fatal error is detected by the ISR or
through Heartbeat.
Add a mechanism that allows to inject a heartbeat error for testing
purposes.
A new attribute `inject_error` is added to debugfs for each QAT device.
Upon a write on this attribute, the driver will inject an error on the
device which can then be detected by the heartbeat feature.
Errors are breaking the device functionality thus they require a
device reset in order to be recovered.
This functionality is not compiled by default, to enable it
CRYPTO_DEV_QAT_ERROR_INJECTION must be set.
Qi Tao [Fri, 26 Jan 2024 09:38:27 +0000 (17:38 +0800)]
crypto: hisilicon/sec2 - fix some cleanup issues
This patch fixes following cleanup issues:
- The return value of the function is
inconsistent with the actual return type.
- After the pointer type is directly converted
to the `__le64` type, the program may crash
or produce unexpected results.
crypto: testmgr - remove unused xts4096 and xts512 algorithms from testmgr.c
Commit a93492cae30a ("crypto: ccree - remove data unit size support")
removed support for the xts512 and xts4096 algorithms, but left them
defined in testmgr.c. This patch removes those definitions.
Erick Archer [Sun, 21 Jan 2024 16:40:43 +0000 (17:40 +0100)]
crypto: qat - use kcalloc_node() instead of kzalloc_node()
As noted in the "Deprecated Interfaces, Language Features, Attributes,
and Conventions" documentation [1], size calculations (especially
multiplication) should not be performed in memory allocator (or similar)
function arguments due to the risk of them overflowing. This could lead
to values wrapping around and a smaller allocation being made than the
caller was expecting. Using those allocations could lead to linear
overflows of heap memory and other misbehaviors.
So, use the purpose specific kcalloc_node() function instead of the
argument count * size in the kzalloc_node() function.
Erick Archer [Sun, 21 Jan 2024 15:34:07 +0000 (16:34 +0100)]
crypto: sun8i-ce - Use kcalloc() instead of kzalloc()
As noted in the "Deprecated Interfaces, Language Features, Attributes,
and Conventions" documentation [1], size calculations (especially
multiplication) should not be performed in memory allocator (or similar)
function arguments due to the risk of them overflowing. This could lead
to values wrapping around and a smaller allocation being made than the
caller was expecting. Using those allocations could lead to linear
overflows of heap memory and other misbehaviors.
So, use the purpose specific kcalloc() function instead of the argument
size * count in the kzalloc() function.
Switch to raw_smp_processor_id() to prevent a number of
warnings from kernel debugging. We do not care about
preemption here, as the CPU number is only used as a
poor mans load balancing or device selection. If preemption
happens during an encrypt/decrypt operation a small performance
hit will occur but everything will continue to work, so just
ignore it. This commit is similar to e7a9b05ca4
("crypto: cavium - Fix smp_processor_id() warnings").
Ard Biesheuvel [Thu, 18 Jan 2024 17:06:37 +0000 (18:06 +0100)]
crypto: arm64/aes-ccm - Merge finalization into en/decrypt asm helpers
The C glue code already infers whether or not the current iteration is
the final one, by comparing walk.nbytes with walk.total. This means we
can easily inform the asm helpers of this as well, by conditionally
passing a pointer to the original IV, which is used in the finalization
of the MAC. This removes the need for a separate call into the asm code
to perform the finalization.
Ard Biesheuvel [Thu, 18 Jan 2024 17:06:36 +0000 (18:06 +0100)]
crypto: arm64/aes-ccm - Merge encrypt and decrypt tail handling
The encryption and decryption code paths are mostly identical, except
for a small difference where the plaintext input into the MAC is taken
from either the input or the output block.
We can factor this in quite easily using a vector bit select, and a few
additional XORs, without the need for branches. This way, we can use the
same tail handling logic on the encrypt and decrypt code paths, allowing
further consolidation of the asm helpers in a subsequent patch.
(In the main loop, adding just a handful of ALU instructions results in
a noticeable performance hit [around 5% on Apple M2], so those routines
are kept separate)
Ard Biesheuvel [Thu, 18 Jan 2024 17:06:35 +0000 (18:06 +0100)]
crypto: arm64/aes-ccm - Cache round keys and unroll AES loops
The CCM code as originally written attempted to use as few NEON
registers as possible, to avoid having to eagerly preserve/restore the
entire NEON register file at every call to kernel_neon_begin/end. At
that time, this API took a number of NEON registers as a parameter, and
only preserved that many registers.
Today, the NEON register file is restored lazily, and the old API is
long gone. This means we can use as many NEON registers as we can make
meaningful use of, which means in the AES case that we can keep all
round keys in registers rather than reloading each of them for each AES
block processed.
On Cortex-A53, this results in a speedup of more than 50%. (From 4
cycles per byte to 2.6 cycles per byte)
Ard Biesheuvel [Thu, 18 Jan 2024 17:06:34 +0000 (18:06 +0100)]
crypto: arm64/aes-ccm - Reuse existing MAC update for AAD input
CCM combines the counter (CTR) encryption mode with a MAC based on the
same block cipher. This MAC construction is a bit clunky: it invokes the
block cipher in a way that cannot be parallelized, resulting in poor CPU
pipeline efficiency.
The arm64 CCM code mitigates this by interleaving the encryption and MAC
at the AES round level, resulting in a substantial speedup. But this
approach does not apply to the additional authenticated data (AAD) which
is not encrypted.
This means the special asm routine dealing with the AAD is not any
better than the MAC update routine used by the arm64 AES block
encryption driver, so let's reuse that, and drop the special AES-CCM
version.
Ard Biesheuvel [Thu, 18 Jan 2024 17:06:33 +0000 (18:06 +0100)]
crypto: arm64/aes-ccm - Replace bytewise tail handling with NEON permute
Implement the CCM tail handling using a single sequence that uses
permute vectors and overlapping loads and stores, rather than going over
the tail byte by byte in a loop, and using scalar operations. This is
more efficient, even though the measured speedup is only around 1-2% on
the CPUs I have tried.
Ard Biesheuvel [Thu, 18 Jan 2024 17:06:32 +0000 (18:06 +0100)]
crypto: arm64/aes-ccm - Pass short inputs via stack buffer
In preparation for optimizing the CCM core asm code using permutation
vectors and overlapping loads and stores, ensure that inputs shorter
than the size of a AES block are passed via a buffer on the stack, in a
way that positions the data at the end of a 16 byte buffer. This removes
the need for the asm code to reason about a rare corner case where the
tail of the data cannot be read/written using a single NEON load/store
instruction.
While at it, tweak the copyright header and authorship to bring it up to
date.
Ard Biesheuvel [Thu, 18 Jan 2024 17:06:31 +0000 (18:06 +0100)]
crypto: arm64/aes-ccm - Keep NEON enabled during skcipher walk
Now that kernel mode NEON no longer disables preemption, we no longer
have to take care to disable and re-enable use of the NEON when calling
into the skcipher walk API. So just keep it enabled until done.
This reverts commit 57ead1bf1c54, which updated the CCM code to only
rely on walk.nbytes to check for failures returned from the skcipher
walk API, mostly for the common good rather than to fix a particular
problem in the code.
This change introduces a problem of its own: the skcipher walk is
started with the 'atomic' argument set to false, which means that the
skcipher walk API is permitted to sleep. Subsequently, it invokes
skcipher_walk_done() with preemption disabled on the final iteration of
the loop. This appears to work by accident, but it is arguably a bad
example, and providing a better example was the point of the original
patch.
Given that future changes to the CCM code will rely on the original
behavior of entering the loop even for zero sized inputs, let's just
revert this change entirely, and proceed from there.
The pointer secs is being assigned a value however secs is never
read afterwards. The pointer secs is redundant and can be removed.
Cleans up clang scan build warning:
warning: Although the value stored to 'secs' is used in the enclosing
expression, the value is never actually read from 'secs'
[deadcode.DeadStores]
Colin Ian King [Tue, 16 Jan 2024 10:43:02 +0000 (10:43 +0000)]
crypto: pcbc - remove redundant assignment to nbytes
The assignment to nbytes is redundant, the while loop needs
to just refer to the value in walk.nbytes and the value of
nbytes is being re-assigned inside the loop on both paths
of the following if-statement. Remove redundant assignment.
Cleans up clang scan build warning:
warning: Although the value stored to 'nbytes' is used in
the enclosing expression, the value is never actually read
from 'nbytes' [deadcode.DeadStores]
Weili Qian [Fri, 12 Jan 2024 10:25:45 +0000 (18:25 +0800)]
crypto: hisilicon/qm - support get device state
Support get device current state. The value 0 indicates that
the device is busy, and the value 1 indicates that the
device is idle. When the device is in suspended, 1 is returned.
Minjie Du [Tue, 9 Jan 2024 02:19:14 +0000 (10:19 +0800)]
crypto: iaa - Remove unnecessary debugfs_create_dir() error check in iaa_crypto_debugfs_init()
This patch removes the debugfs_create_dir() error checking in
iaa_crypto_debugfs_init(). Because the debugfs_create_dir() is developed
in a way that the caller can safely handle the errors that
occur during the creation of DebugFS nodes.
Tom Zanussi [Mon, 8 Jan 2024 22:53:48 +0000 (16:53 -0600)]
crypto: iaa - Remove header table code
The header table and related code is currently unused - it was
included and used for canned mode, but canned mode has been removed,
so this code can be safely removed as well.
This indirectly fixes a bug reported by Dan Carpenter.
Arnd Bergmann [Wed, 3 Jan 2024 16:26:02 +0000 (17:26 +0100)]
crypto: qat - avoid memcpy() overflow warning
The use of array_size() leads gcc to assume the memcpy() can have a larger
limit than actually possible, which triggers a string fortification warning:
In file included from include/linux/string.h:296,
from include/linux/bitmap.h:12,
from include/linux/cpumask.h:12,
from include/linux/sched.h:16,
from include/linux/delay.h:23,
from include/linux/iopoll.h:12,
from drivers/crypto/intel/qat/qat_common/adf_gen4_hw_data.c:3:
In function 'fortify_memcpy_chk',
inlined from 'adf_gen4_init_thd2arb_map' at drivers/crypto/intel/qat/qat_common/adf_gen4_hw_data.c:401:3:
include/linux/fortify-string.h:579:4: error: call to '__write_overflow_field' declared with attribute warning: detected write beyond size of field (1st parameter); maybe use struct_group()? [-Werror=attribute-warning]
579 | __write_overflow_field(p_size_field, size);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
include/linux/fortify-string.h:588:4: error: call to '__read_overflow2_field' declared with attribute warning: detected read beyond size of field (2nd parameter); maybe use struct_group()? [-Werror=attribute-warning]
588 | __read_overflow2_field(q_size_field, size);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Danny Tsen [Tue, 2 Jan 2024 20:58:56 +0000 (15:58 -0500)]
crypto: vmx - Move to arch/powerpc/crypto
Relocate all crypto files in vmx driver to arch/powerpc/crypto directory
and remove vmx directory.
drivers/crypto/vmx/aes.c rename to arch/powerpc/crypto/aes.c
drivers/crypto/vmx/aes_cbc.c rename to arch/powerpc/crypto/aes_cbc.c
drivers/crypto/vmx/aes_ctr.c rename to arch/powerpc/crypto/aes_ctr.c
drivers/crypto/vmx/aes_xts.c rename to arch/powerpc/crypto/aes_xts.c
drivers/crypto/vmx/aesp8-ppc.h rename to arch/powerpc/crypto/aesp8-ppc.h
drivers/crypto/vmx/aesp8-ppc.pl rename to arch/powerpc/crypto/aesp8-ppc.pl
drivers/crypto/vmx/ghash.c rename to arch/powerpc/crypto/ghash.c
drivers/crypto/vmx/ghashp8-ppc.pl rename to arch/powerpc/crypto/ghashp8-ppc.pl
drivers/crypto/vmx/vmx.c rename to arch/powerpc/crypto/vmx.c
Markus Elfring [Tue, 26 Dec 2023 10:00:20 +0000 (11:00 +0100)]
crypto: virtio - Less function calls in __virtio_crypto_akcipher_do_req() after error detection
The kfree() function was called in up to two cases by the
__virtio_crypto_akcipher_do_req() function during error handling
even if the passed variable contained a null pointer.
This issue was detected by using the Coccinelle software.
* Adjust jump targets.
* Delete two initialisations which became unnecessary
with this refactoring.
The commit "crypto: qat - generate dynamically arbiter mappings"
introduced a regression on qat_402xx devices.
This is reported when the driver probes the device, as indicated by
the following error messages:
4xxx 0000:0b:00.0: enabling device (0140 -> 0142)
4xxx 0000:0b:00.0: Generate of the thread to arbiter map failed
4xxx 0000:0b:00.0: Direct firmware load for qat_402xx_mmp.bin failed with error -2
The root cause of this issue was the omission of a necessary function
pointer required by the mapping algorithm during the implementation.
Fix it by adding the missing function pointer.
Linus Torvalds [Sun, 21 Jan 2024 22:01:12 +0000 (14:01 -0800)]
Merge tag 'bcachefs-2024-01-21' of https://evilpiepirate.org/git/bcachefs
Pull more bcachefs updates from Kent Overstreet:
"Some fixes, Some refactoring, some minor features:
- Assorted prep work for disk space accounting rewrite
- BTREE_TRIGGER_ATOMIC: after combining our trigger callbacks, this
makes our trigger context more explicit
- A few fixes to avoid excessive transaction restarts on
multithreaded workloads: fstests (in addition to ktest tests) are
now checking slowpath counters, and that's shaking out a few bugs
- Assorted tracepoint improvements
- Starting to break up bcachefs_format.h and move on disk types so
they're with the code they belong to; this will make room to start
documenting the on disk format better.
- A few minor fixes"
* tag 'bcachefs-2024-01-21' of https://evilpiepirate.org/git/bcachefs: (46 commits)
bcachefs: Improve inode_to_text()
bcachefs: logged_ops_format.h
bcachefs: reflink_format.h
bcachefs; extents_format.h
bcachefs: ec_format.h
bcachefs: subvolume_format.h
bcachefs: snapshot_format.h
bcachefs: alloc_background_format.h
bcachefs: xattr_format.h
bcachefs: dirent_format.h
bcachefs: inode_format.h
bcachefs; quota_format.h
bcachefs: sb-counters_format.h
bcachefs: counters.c -> sb-counters.c
bcachefs: comment bch_subvolume
bcachefs: bch_snapshot::btime
bcachefs: add missing __GFP_NOWARN
bcachefs: opts->compression can now also be applied in the background
bcachefs: Prep work for variable size btree node buffers
bcachefs: grab s_umount only if snapshotting
...
Linus Torvalds [Sun, 21 Jan 2024 19:14:40 +0000 (11:14 -0800)]
Merge tag 'timers-core-2024-01-21' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull timer updates from Thomas Gleixner:
"Updates for time and clocksources:
- A fix for the idle and iowait time accounting vs CPU hotplug.
The time is reset on CPU hotplug which makes the accumulated
systemwide time jump backwards.
- Assorted fixes and improvements for clocksource/event drivers"
* tag 'timers-core-2024-01-21' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
tick-sched: Fix idle and iowait sleeptime accounting vs CPU hotplug
clocksource/drivers/ep93xx: Fix error handling during probe
clocksource/drivers/cadence-ttc: Fix some kernel-doc warnings
clocksource/drivers/timer-ti-dm: Fix make W=n kerneldoc warnings
clocksource/timer-riscv: Add riscv_clock_shutdown callback
dt-bindings: timer: Add StarFive JH8100 clint
dt-bindings: timer: thead,c900-aclint-mtimer: separate mtime and mtimecmp regs
Kent Overstreet [Tue, 16 Jan 2024 21:20:21 +0000 (16:20 -0500)]
bcachefs: opts->compression can now also be applied in the background
The "apply this compression method in the background" paths now use the
compression option if background_compression is not set; this means that
setting or changing the compression option will cause existing data to
be compressed accordingly in the background.
Kent Overstreet [Tue, 16 Jan 2024 18:29:59 +0000 (13:29 -0500)]
bcachefs: Prep work for variable size btree node buffers
bcachefs btree nodes are big - typically 256k - and btree roots are
pinned in memory. As we're now up to 18 btrees, we now have significant
memory overhead in mostly empty btree roots.
And in the future we're going to start enforcing that certain btree node
boundaries exist, to solve lock contention issues - analagous to XFS's
AGIs.
Thus, we need to start allocating smaller btree node buffers when we
can. This patch changes code that refers to the filesystem constant
c->opts.btree_node_size to refer to the btree node buffer size -
btree_buf_bytes() - where appropriate.
In __bch2_ioctl_subvolume_create(), we grab s_umount unconditionally
and unlock it at the end of the function. There is a comment
"why do we need this lock?" about the lock coming from
commit 42d237320e98 ("bcachefs: Snapshot creation, deletion")
The reason is that __bch2_ioctl_subvolume_create() calls
sync_inodes_sb() which enforce locked s_umount to writeback all dirty
nodes before doing snapshot works.
Fix it by read locking s_umount for snapshotting only and unlocking
s_umount after sync_inodes_sb().
Colin Ian King [Tue, 16 Jan 2024 11:07:23 +0000 (11:07 +0000)]
bcachefs: remove redundant variable tmp
The variable tmp is being assigned a value but it isn't being
read afterwards. The assignment is redundant and so tmp can be
removed.
Cleans up clang scan build warning:
warning: Although the value stored to 'ret' is used in the enclosing
expression, the value is never actually read from 'ret'
[deadcode.DeadStores]
Kent Overstreet [Tue, 16 Jan 2024 01:37:23 +0000 (20:37 -0500)]
bcachefs: Fix excess transaction restarts in __bchfs_fallocate()
drop_locks_do() should not be used in a fastpath without first trying
the do in nonblocking mode - the unlock and relock will cause excessive
transaction restarts and potentially livelocking with other threads that
are contending for the same locks.
Kent Overstreet [Mon, 15 Jan 2024 22:59:51 +0000 (17:59 -0500)]
bcachefs: Better journal tracepoints
Factor out bch2_journal_bufs_to_text(), and use it in the
journal_entry_full() tracepoint; when we can't get a journal reservation
we need to know the outstanding journal entry sizes to know if the
problem is due to excessive flushing.
Kent Overstreet [Mon, 15 Jan 2024 22:56:22 +0000 (17:56 -0500)]
bcachefs: Avoid flushing the journal in the discard path
When issuing discards, we may need to flush the journal if there's too
many buckets that can't be discarded until a journal flush.
But the heuristic was bad; we should be comparing the number of buckets
that need to flushes against the number of free buckets, not the number
of buckets we saw.
Kent Overstreet [Mon, 15 Jan 2024 19:15:26 +0000 (14:15 -0500)]
bcachefs: bch2_kthread_io_clock_wait() no longer sleeps until full amount
Drop t he loop in bch2_kthread_io_clock_wait(): this allows the code
that uses it to be woken up for other reasons, and fixes a bug where
rebalance wouldn't wake up when a scan was requested.
This raises the possibility of spurious wakeups, but callers should
always be able to handle that reasonably well.
Kent Overstreet [Thu, 11 Jan 2024 04:47:04 +0000 (23:47 -0500)]
bcachefs: Reduce would_deadlock restarts
We don't have to take locks in any particular ordering - we'll make
forward progress just fine - but if we try to stick to an ordering, it
can help to avoid excessive would_deadlock transaction restarts.
This tweaks the reflink path to take extents btree locks in the right
order.
Kent Overstreet [Thu, 11 Jan 2024 04:08:30 +0000 (23:08 -0500)]
bcachefs: Don't log errors if BCH_WRITE_ALLOC_NOWAIT
Previously, we added logging in the write path to ensure that any
unexpected errors getting reported to userspace have a log message; but
BCH_WRITE_ALLOC_NOWAIT is a special case, it's used for promotes where
errors are expected and not reported out to userspace - so we need to
silence those.
projects / linux.git / log