Pass a queue_limits to blk_alloc_queue and apply it after validating and
capping the values using blk_validate_limits.  This will allow allocating
queues with valid queue limits instead of setting the values one at a
time later.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-9-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

Convert queue_discard_max_store to use queue_limits_commit_update to
check and update the max_discard_sectors limit and freeze the queue
before doing so to ensure we don't have requests in flight while
changing the limits.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-8-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

Add a new max_user_discard_sectors limit that mirrors max_user_sectors
and stores the value that the user manually set.  This now allows
updates of the max_hw_discard_sectors to not worry about the user
limit.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-7-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

Convert queue_max_sectors_store to use queue_limits_commit_update to
check and update the max_sectors limit and freeze the queue before
doing so to ensure we don't have requests in flight while changing
the limits.

Note that this removes the previously held queue_lock that doesn't
protect against any other reader or writer.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-6-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

Add a new queue_limits_{start,commit}_update pair of functions that
allows taking an atomic snapshot of queue limits, update it, and
commit it if it passes validity checking.  Also use the low-level
validation helper to implement blk_set_default_limits instead of
duplicating the initialization.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-5-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

blk_set_stacking_limits uses very little from blk_set_default_limits.
Open code these initializations in preparation for rewriting
blk_set_default_limits.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-4-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

Factor out a blk_apply_bdi_limits limits helper that can be used with
an explicit queue_limits argument, which will be useful later.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-3-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

The maximum number of open and active zones is a limit on the queue
and should be places there so that we can including it in the upcoming
queue limits batch update API.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240213073425.1621680-2-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>

There are four state machines in drbd that use a common infrastructure, with
a cast to an incompatible function type in REMEMBER_STATE_CHANGE that clang-16
now warns about:

drivers/block/drbd/drbd_state.c:1632:3: error: cast from 'int (*)(struct sk_buff *, unsigned int, struct drbd_resource_state_change *, enum drbd_notification_type)' to 'typeof (last_func)' (aka 'int (*)(struct sk_buff *, unsigned int, void *, enum drbd_notification_type)') converts to incompatible function type [-Werror,-Wcast-function-type-strict]
 1632 |                 REMEMBER_STATE_CHANGE(notify_resource_state_change,
      |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 1633 |                                       resource_state_change, NOTIFY_CHANGE);
      |                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/block/drbd/drbd_state.c:1619:17: note: expanded from macro 'REMEMBER_STATE_CHANGE'
 1619 |            last_func = (typeof(last_func))func; \
      |                        ^~~~~~~~~~~~~~~~~~~~~~~
drivers/block/drbd/drbd_state.c:1641:4: error: cast from 'int (*)(struct sk_buff *, unsigned int, struct drbd_connection_state_change *, enum drbd_notification_type)' to 'typeof (last_func)' (aka 'int (*)(struct sk_buff *, unsigned int, void *, enum drbd_notification_type)') converts to incompatible function type [-Werror,-Wcast-function-type-strict]
 1641 |                         REMEMBER_STATE_CHANGE(notify_connection_state_change,
      |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 1642 |                                               connection_state_change, NOTIFY_CHANGE);
      |                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Change these all to actually expect a void pointer to be passed, which
matches the caller.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20240213100354.457128-1-arnd@kernel.orgSigned-off-by: Jens Axboe <axboe@kernel.dk>

clang-16 complains about a control flow integrity (kcfi) violation
casting between incompatible pointers:

drivers/block/floppy.c:2001:11: error: cast from 'void (*)(void)' to 'done_f' (aka 'void (*)(int)') converts to incompatible function type [-Werror,-Wcast-function-type-strict]
 2001 |         .done           = (done_f)empty
      |                           ^~~~~~~~~~~~~

Just add another empty function with the correct prototype as a
workaround.

The warning is for code that was added before the start of the normal
git history, but I tracked it done to an early change in the reconstructed
linux-history.git.

Fixes: 598a477a ("Import 1.1.41")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/20240213095918.455478-1-arnd@kernel.orgSigned-off-by: Jens Axboe <axboe@kernel.dk>

NVM command set 1.0 (or later) mandates PI to be in the last bytes of
metadata. But this was not supported in the block-layer, and driver
registered a nop profile.

Since block-integrity can now handle flexible PI offset, change the
driver to support this configuration.
Signed-off-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240201130126.211402-4-joshi.k@samsung.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Block layer integrity processing assumes that protection information
(PI) is placed in the first bytes of each metadata block.

Remove this limitation and include the metadata before the PI in the
calculation of the guard tag.
Signed-off-by: Kanchan Joshi <joshi.k@samsung.com>
Signed-off-by: Chinmay Gameti <c.gameti@samsung.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240201130126.211402-3-joshi.k@samsung.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Allow computation using the existing guard value.
This is a prep patch.
Signed-off-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240201130126.211402-2-joshi.k@samsung.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Now that all callers pass in GFP_KERNEL to blkdev_zone_mgmt() and use
memalloc_no{io,fs}_{save,restore}() to define the allocation scope, we can
drop the gfp_mask parameter from blkdev_zone_mgmt() as well as
blkdev_zone_reset_all() and blkdev_zone_reset_all_emulated().
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Mike Snitzer <snitzer@kernel.org>
Link: https://lore.kernel.org/r/20240128-zonefs_nofs-v3-5-ae3b7c8def61@wdc.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Guard the calls to blkdev_zone_mgmt() with a memalloc_nofs scope.
This helps us getting rid of the GFP_NOFS argument to blkdev_zone_mgmt();
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Chao Yu <chao@kernel.org>
Link: https://lore.kernel.org/r/20240128-zonefs_nofs-v3-4-ae3b7c8def61@wdc.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Add a memalloc_nofs scope around all calls to blkdev_zone_mgmt(). This
allows us to further get rid of the GFP_NOFS argument for
blkdev_zone_mgmt().
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Link: https://lore.kernel.org/r/20240128-zonefs_nofs-v3-3-ae3b7c8def61@wdc.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Guard the calls to blkdev_zone_mgmt() with a memalloc_noio scope.
This helps us getting rid of the GFP_NOIO argument to blkdev_zone_mgmt();
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Mike Snitzer <snitzer@kernel.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20240128-zonefs_nofs-v3-2-ae3b7c8def61@wdc.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Pass GFP_KERNEL instead of GFP_NOFS to the blkdev_zone_mgmt() call in
zonefs_zone_mgmt().

As as zonefs_zone_mgmt() and zonefs_inode_zone_mgmt() are never called
from a place that can recurse back into the filesystem on memory reclaim,
it is save to call blkdev_zone_mgmt() with GFP_KERNEL.

Link: https://lore.kernel.org/all/ZZcgXI46AinlcBDP@casper.infradead.org/Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Acked-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20240128-zonefs_nofs-v3-1-ae3b7c8def61@wdc.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Once the discipline is associated with the device, deleting the device
takes care of decrementing the module's refcount.  Doing it manually on
this error path causes refcount to artificially decrease on each error
while it should just stay the same.

Fixes: c020d722 ("s390/dasd: fix panic during offline processing")
Signed-off-by: Miroslav Franc <mfranc@suse.cz>
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240209124522.3697827-3-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Some ERP errors still share the same message format and only add
different reason codes to it. These reason codes don't have any meaning
anymore.
Make the individual error messages more explicit and remove the reason
codes altogether. Comments around the error messages are also removed as
they provide no additional value anymore with more explicit messages.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240209124522.3697827-2-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Allow setting shared_tags through configfs, which could only be set as a
module parameter. For that purpose, delay tag_set initialization from
null_init() to null_add_dev(). Refer tag_set.ops as the flag to check if
tag_set is initialized or not.

The following parameters can not be set through configfs yet:

    timeout
    requeue
    init_hctx
Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240130042134.2463659-1-shinichiro.kawasaki@wdc.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Use the new KMEM_CACHE() macro instead of direct kmem_cache_create
to simplify the creation of SLAB caches.
Signed-off-by: Kunwu Chan <chentao@kylinos.cn>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20240131094323.146659-1-chentao@kylinos.cnSigned-off-by: Jens Axboe <axboe@kernel.dk>

When enlisting a bio into ->free_list_irq we protect the list by
disabling irqs. It's likely they're already disabled and performance of
local_irq_{save,restore}() is decent, but it's not zero cost.

Let's only use the irq cache when when we're serving a hard irq, which
allows to remove local_irq_{save,restore}(), and fall back to bio_free()
in all left cases.

Profiles indicate that the bio_put() cost is reduced by ~3.5 times
(1.76% -> 0.49%), and total throughput of a CPU bound benchmark improve
by around 1% (t/io_uring with high QD and several drives).
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/36d207540b7046c653cc16e5ff08fe7234b19f81.1707314970.git.asml.silence@gmail.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

bio_put_percpu_cache() puts all non-iopoll bios into the irq-safe list,
which entails disabling irqs. The overhead of that is not that bad when
interrupts are already off but getting worse otherwise. We can optimise
it when we're in the task context by using ->free_list directly just as
the IOPOLL path does.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/4774e1a0f905f96c63174b0f3e4f79f0d9b63246.1707314970.git.asml.silence@gmail.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

All log messages in dasd.c use the printk variants of pr_*(). They all
add the name of the affected device manually to the log message.
This can be simplified by using the dev_*() variants of printk, which
include the device information and make a separate call to dev_name()
unnecessary.

The KMSG_COMPONENT and the pr_fmt() definition can be dropped. Note that
this removes the "dasd: " prefix from the one pr_info() call in
dasd_init(). However, the log message already provides all relevant
information.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-10-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

PRINTK_HEADER was mainly used to prefix log messages with the module
name. Most components don't use this definition anymore. Either because
there are no log messages being generated anymore, or pr_*() were
replaced by dev_*(), which contains device and component information
already.

PRINTK_HEADER is also dropped in the function
dasd_3990_erp_handle_match_erp() in dasd_3990_erp.c from a panic() call
as panic() already provides all relevant information.

KMSG_COMPONENT was mainly used to identify a component in a long gone
kernel message catalog feature.

Remove both definition since they're either not used or alternatives
make the code slightly shorter and more readable.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-9-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Printing pointer in error messages doesn't add any value since the
addresses are hashed. Remove the %p format specifier and adapt the error
messages slightly.

Replace %p with %px in ERP to get the actual addresses since ERP is used
for debugging purposes only anyway.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-8-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

To reduce the information required for the string generation in the
sense dump functions, use the more concise dev_err() variant over
printk(KERN_ERR, ...) to improve code readability.

The dev_err() function provides the component and device name for free
and the separate dev_name() calls as well as the PRINTK_HEADER can be
dropped.

Dropping PRINTK_HEADER removes the "dasd(eckd):" for all lines. Only the
first line of a dev_err() call is prefixed with the component and device
(e.g. "dasd-eckd 0.0.95d0:").

The format specifier for printed pointers is also changed to unhashed
(%px) as this can help with debugging and servicing.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-7-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

The macros DEV_MESSAGE, MESSAGE, DEV_MESSAGE_LOG, and MESSAGE_LOG, are
not used and there is no history anymore of any usage. Remove them.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-6-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

All error messages for a failling dasd_smalloc_request() call are logged
via DBF, except one. There is no value in logging this particular
allocation failure via dev_err(). Move the message to DBF, too, to be
in line with the rest.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-5-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

In quite a few cases an errorstring is generated using snprintf() before
it's passed to dev_err(). This indirection is unnecessary and all
information can simply be passed directly to dev_err() instead.
The errrorstring and ERRORLENGTH definitions are removed entirely.

While at it, rephrase the error messages to provide more context where
possible. Also, fix a few incorrectly used format specifier (e.g. %x02
-> %02x) in those messages.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-4-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

sysfs_emit() should be used in show() functions. There are still a
couple of functions that use sprintf().
Replace outstanding occurrences of sprintf() in all show() functions
with sysfs_emit().
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-3-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

There are two variants of the device uid string. One containing the
virtual device unit information table (vduit) identifying the device as
a virtual device located on a real device in a z/VM environment. The
other variant does not contain those additional information.

Simplify the string generation with a shorter check of an existing vduit
embedded in the snprintf() calls.
Signed-off-by: Jan Höppner <hoeppner@linux.ibm.com>
Reviewed-by: Stefan Haberland <sth@linux.ibm.com>
Signed-off-by: Stefan Haberland <sth@linux.ibm.com>
Link: https://lore.kernel.org/r/20240208164248.540985-2-sth@linux.ibm.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Now that the driver core can properly handle constant struct bus_type,
move the rbd_bus_type variable to be a constant structure as well,
placing it into read-only memory which can not be modified at runtime.

Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Suggested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ricardo B. Marliere <ricardo@marliere.net>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240204-bus_cleanup-block-v1-1-fc77afd8d7cc@marliere.netSigned-off-by: Jens Axboe <axboe@kernel.dk>

After calling throtl_peek_queued(), the data direction can be determined so
there is no need to call bio_data_dir() to check the direction again.
Signed-off-by: Tang Yizhou <yizhou.tang@shopee.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240123081248.3752878-1-yizhou.tang@shopee.comSigned-off-by: Jens Axboe <axboe@kernel.dk>

Mark the task as having a cached timestamp when set assign it, so we
can efficiently check if it needs updating post being scheduled back in.
This covers both the actual schedule out case, which would've flushed
the plug, and the preemption case which doesn't touch the plugged
requests (for many reasons, one of them being then we'd need to have
preemption disabled around plug state manipulation).
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Querying the current time is the most costly thing we do in the block
layer per IO, and depending on kernel config settings, we may do it
many times per IO.

None of the callers actually need nsec granularity. Take advantage of
that by caching the current time in the plug, with the assumption here
being that any time checking will be temporally close enough that the
slight loss of precision doesn't matter.

If the block plug gets flushed, eg on preempt or schedule out, then
we invalidate the cached clock.

On a basic peak IOPS test case with iostats enabled, this changes
the performance from:

IOPS=108.41M, BW=52.93GiB/s, IOS/call=31/31
IOPS=108.43M, BW=52.94GiB/s, IOS/call=32/32
IOPS=108.29M, BW=52.88GiB/s, IOS/call=31/32
IOPS=108.35M, BW=52.91GiB/s, IOS/call=32/32
IOPS=108.42M, BW=52.94GiB/s, IOS/call=31/31
IOPS=108.40M, BW=52.93GiB/s, IOS/call=32/32
IOPS=108.31M, BW=52.89GiB/s, IOS/call=32/31

to

IOPS=118.79M, BW=58.00GiB/s, IOS/call=31/32
IOPS=118.62M, BW=57.92GiB/s, IOS/call=31/31
IOPS=118.80M, BW=58.01GiB/s, IOS/call=32/31
IOPS=118.78M, BW=58.00GiB/s, IOS/call=32/32
IOPS=118.69M, BW=57.95GiB/s, IOS/call=32/31
IOPS=118.62M, BW=57.92GiB/s, IOS/call=32/31
IOPS=118.63M, BW=57.92GiB/s, IOS/call=31/32

which is more than a 9% improvement in performance. Looking at perf diff,
we can see a huge reduction in time overhead:

    10.55%     -9.88%  [kernel.vmlinux]  [k] read_tsc
     1.31%     -1.22%  [kernel.vmlinux]  [k] ktime_get

Note that since this relies on blk_plug for the caching, it's only
applicable to the issue side. But this is where most of the time calls
happen anyway. On the completion side, cached time stamping is done with
struct io_comp patch, as long as the driver supports it.

It's also worth noting that the above testing doesn't enable any of the
higher cost CPU items on the block layer side, like wbt, cgroups,
iocost, etc, which all would add additional time querying and hence
overhead. IOW, results would likely look even better in comparison with
those enabled, as distros would do.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Convert any user of ktime_get_ns() to use blk_time_get_ns(), and
ktime_get() to blk_time_get(), so we have a unified API for querying the
current time in nanoseconds or as ktime.

No functional changes intended, this patch just wraps ktime_get_ns()
and ktime_get() with a block helper.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

In preparation for moving time keeping into blk.h, move the cgroup
related code for timestamps in here too. This will help avoid a circular
dependency, and also moves it into a more appropriate header as this one
is private to the block layer code.

Leave struct bio_issue in blk_types.h as it's a proper time definition.
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Share the main merge / split / integrity preparation code between the
cached request vs newly allocated request cases, and add comments
explaining the cached request handling.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Tested-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20240124092658.2258309-4-hch@lst.deSigned-off-by: Jens Axboe <axboe@kernel.dk>