We have updated limits after calling wbt_set_min_lat(). No need to
update again.
Reviewed-by: NBob Liu <bob.liu@oracle.com>
Signed-off-by: NYufen Yu <yuyufen@huawei.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

The default hard disk param sets latency targets at 50ms.  As the
default target percentiles are zero, these don't directly regulate
vrate; however, they're still used to calculate the period length -
100ms in this case.

This is excessively low.  A SATA drive with QD32 saturated with random
IOs can easily reach avg completion latency of several hundred msecs.
A period duration which is substantially lower than avg completion
latency can lead to wildly fluctuating vrate.

Let's bump up the default latency targets to 250ms so that the period
duration is sufficiently long.
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Some IOs may span multiple periods.  As latencies are collected on
completion, the inbetween periods won't register them and may
incorrectly decide to increase vrate.  nr_lagging tracks these IOs to
avoid those situations.  Currently, whenever there are IOs which are
spanning from the previous period, busy_level is reset to 0 if
negative thus suppressing vrate increase.

This has the following two problems.

* When latency target percentiles aren't set, vrate adjustment should
  only be governed by queue depth depletion; however, the current code
  keeps nr_lagging active which pulls in latency results and can keep
  down vrate unexpectedly.

* When lagging condition is detected, it resets the entire negative
  busy_level.  This turned out to be way too aggressive on some
  devices which sometimes experience extended latencies on a small
  subset of commands.  In addition, a lagging IO will be accounted as
  latency target miss on completion anyway and resetting busy_level
  amplifies its impact unnecessarily.

This patch fixes the above two problems by disabling nr_lagging
counting when latency target percentiles aren't set and blocking vrate
increases when there are lagging IOs while leaving busy_level as-is.
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

vrate_adj tracepoint traces vrate changes; however, it does so only
when busy_level is non-zero.  busy_level turning to zero can sometimes
be as interesting an event.  This patch also enables vrate_adj
tracepoint on other vrate related events - busy_level changes and
non-zero nr_lagging.
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

cecf5d87 ("block: split .sysfs_lock into two locks") starts to
release & acquire sysfs_lock before registering/un-registering elevator
queue during switching elevator for avoiding potential deadlock from
showing & storing 'queue/iosched' attributes and removing elevator's
kobject.

Turns out there isn't such deadlock because 'q->sysfs_lock' isn't
required in .show & .store of queue/iosched's attributes, and just
elevator's sysfs lock is acquired in elv_iosched_store() and
elv_iosched_show(). So it is safe to hold queue's sysfs lock when
registering/un-registering elevator queue.

The biggest issue is that commit cecf5d87 assumes that concurrent
write on 'queue/scheduler' can't happen. However, this assumption isn't
true, because kernfs_fop_write() only guarantees that concurrent write
aren't called on the same open file, but the write could be from
different open on the file. So we can't release & re-acquire queue's
sysfs lock during switching elevator, otherwise use-after-free on
elevator could be triggered.

Fixes the issue by not releasing queue's sysfs lock during switching
elevator.

Fixes: cecf5d87 ("block: split .sysfs_lock into two locks")
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: NBart Van Assche <bvanassche@acm.org>
Signed-off-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Commit c48dac13 ("block: don't hold q->sysfs_lock in elevator_init_mq")
removes q->sysfs_lock from elevator_init_mq(), but forgot to deal with
lockdep_assert_held() called in blk_mq_sched_free_requests() which is
run in failure path of elevator_init_mq().

blk_mq_sched_free_requests() is called in the following 3 functions:

	elevator_init_mq()
	elevator_exit()
	blk_cleanup_queue()

In blk_cleanup_queue(), blk_mq_sched_free_requests() is followed exactly
by 'mutex_lock(&q->sysfs_lock)'.

So moving the lockdep_assert_held() from blk_mq_sched_free_requests()
into elevator_exit() for fixing the report by syzbot.

Reported-by: syzbot+da3b7677bb913dc1b737@syzkaller.appspotmail.com
Fixed: c48dac13 ("block: don't hold q->sysfs_lock in elevator_init_mq")
Reviewed-by: NBart Van Assche <bvanassche@acm.org>
Reviewed-by: NDamien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Make sure that bsg_queue_rq() calls put_device() if an error is
encountered after get_device() was successful.

Fixes: cd2f076f ("bsg: convert to use blk-mq")
Signed-off-by: NMartin Wilck <mwilck@suse.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Changing the switch() statement to symbolic constants made the compiler
(at least clang-9, did not check gcc) notice that there is one enum value
that is not handled here:

block/t10-pi.c:62:11: error: enumeration value 'T10_PI_TYPE0_PROTECTION'
not handled in switch [-Werror,-Wswitch]

Add a BUG_ON statement if we ever get to t10_pi_verify function with
TYPE0 and replace the switch() statement with if/else clause for the
valid types.

Fixes: 9b2061b1a262 ("block: use symbolic constants for t10_pi type")
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: NMax Gurtovoy <maxg@mellanox.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

If equal to 0, the injection limit for a bfq_queue is pushed to 1
after a first sample of the total service time of the I/O requests of
the queue is computed (to allow injection to start). Yet, because of a
mistake in the branch that performs this action, the push may happen
also in some other case. This commit fixes this issue.
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: NPaolo Valente <paolo.valente@linaro.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

The update period of the injection limit has been tentatively set to
100 ms, to reduce fluctuations. This value however proved to cause,
occasionally, the limit to be decremented for some bfq_queue only
after the queue underwent excessive injection for a lot of time. This
commit reduces the period to 10 ms.
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: NPaolo Valente <paolo.valente@linaro.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Upon an increment attempt of the injection limit, the latter is
constrained not to become higher than twice the maximum number
max_rq_in_driver of I/O requests that have happened to be in service
in the drive. This high bound allows the injection limit to grow
beyond max_rq_in_driver, which may then cause max_rq_in_driver itself
to grow.

However, since the limit is incremented by only one unit at a time,
there is no need for such a high bound, and just max_rq_in_driver+1 is
enough.
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: NPaolo Valente <paolo.valente@linaro.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

BFQ updates the injection limit of each bfq_queue as a function of how
much the limit inflates the service times experienced by the I/O
requests of the queue. So only service times affected by injection
must be taken into account. Unfortunately, in the current
implementation of this update scheme, the service time of an I/O
request rq not affected by injection may happen to be considered in
the following case: there is no I/O request in service when rq
arrives.

This commit fixes this issue by making sure that only service times
affected by injection are considered for updating the injection
limit. In particular, the service time of an I/O request rq is now
considered only if at least one of the following two conditions holds:
- the destination bfq_queue for rq underwent injection before rq
arrival, and there is still I/O in service in the drive on rq arrival
(the service of such unfinished I/O may delay the service of rq);
- injection occurs between the arrival and the completion time of rq.
Tested-by: NOleksandr Natalenko <oleksandr@natalenko.name>
Signed-off-by: NPaolo Valente <paolo.valente@linaro.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Currently t10_pi_prepare/t10_pi_complete functions are called during the
NVMe and SCSi layers command preparetion/completion, but their actual
place should be the block layer since T10-PI is a general data integrity
feature that is used by block storage protocols. Introduce .prepare_fn
and .complete_fn callbacks within the integrity profile that each type
can implement according to its needs.
Suggested-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Suggested-by: NMartin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NMax Gurtovoy <maxg@mellanox.com>

Fixed to not call queue integrity functions if BLK_DEV_INTEGRITY
isn't defined in the config.
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Replace all hard-coded values with T10_PI_TYPES to make the code more
readable.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMartin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: NMax Gurtovoy <maxg@mellanox.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

In __blk_mq_end_request() if block stats needs update, we should
ensure now is valid instead of 0 even when iostat is disabled.
Signed-off-by: NHou Tao <houtao1@huawei.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Currently rq->data_len will be decreased by partial completion or
zeroed by completion, so when blk_stat_add() is invoked, data_len
will be zero and there will never be samples in poll_cb because
blk_mq_poll_stats_bkt() will return -1 if data_len is zero.

We could move blk_stat_add() back to __blk_mq_complete_request(),
but that would make the effort of trying to call ktime_get_ns()
once in vain. Instead we can reuse throtl_size field, and use
it for both block stats and block throttle, and adjust the
logic in blk_mq_poll_stats_bkt() accordingly.

Fixes: 4bc6339a ("block: move blk_stat_add() to __blk_mq_end_request()")
Tested-by: NPavel Begunkov <asml.silence@gmail.com>
Signed-off-by: NHou Tao <houtao1@huawei.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Since commit 795fe54c ("bfq: Add per-device weight"), bfq uses
blkg_conf_prep() and blkg_conf_finish(), which are not exported. So, it
causes linkage error if bfq compiled as a module.

Fixes: 795fe54c ("bfq: Add per-device weight")
Signed-off-by: NPavel Begunkov <asml.silence@gmail.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

cecf5d87 ("block: split .sysfs_lock into two locks") starts to
release & actuire sysfs_lock again during switching elevator. So it
isn't enough to prevent switching elevator from happening by simply
clearing QUEUE_FLAG_REGISTERED with holding sysfs_lock, because
in-progress switch still can move on after re-acquiring the lock,
meantime the flag of QUEUE_FLAG_REGISTERED won't get checked.

Fixes this issue by checking 'q->elevator' directly & locklessly after
q->kobj is removed in blk_unregister_queue(), this way is safe because
q->elevator can't be changed at that time.

Fixes: cecf5d87 ("block: split .sysfs_lock into two locks")
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Mike Snitzer <snitzer@redhat.com>
Cc: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Some devices may skip blk_pm_runtime_init() and have null pointer
in its request_queue->dev. For example, SCSI devices of UFS Well-Known
LUNs.

Currently the null pointer is checked by the user of
blk_set_runtime_active(), i.e., scsi_dev_type_resume(). It is better to
check it by blk_set_runtime_active() itself instead of by its users.
Signed-off-by: NStanley Chu <stanley.chu@mediatek.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Report debt and rename del_ms row to delay for consistency.
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Merges have the same problem that forced-bios had which is fixed by
the previous patch.  The cost of a merge is calculated at the time of
issue and force-advances vtime into the future.  Until global vtime
catches up, how the cgroup's hweight changes in the meantime doesn't
matter and it often leads to situations where the cost is calculated
at one hweight and paid at a very different one.  See the previous
patch for more details.

Fix it by never advancing vtime into the future for merges.  If budget
is available, vtime is advanced.  Otherwise, the cost is charged as
debt.

This brings merge cost handling in line with issue cost handling in
ioc_rqos_throttle().
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Currently, when a bio needs to be force-charged and there isn't enough
budget, vtime is simply pushed into the future.  This means that the
cost of the whole bio is scaled using the current hweight and then
charged immediately.  Until the global vtime advances beyond this
future vtime, the cgroup won't be allowed to issue normal IOs.

This is incorrect and can lead to, for example, exploding vrate or
extended stalls if vrate range is constrained.  Consider the following
scenario.

1. A cgroup with a very low hweight runs out of budget.

2. A storm of swap-out happens on it.  All of them are scaled
   according to the current low hweight and charged to vtime pushing
   it to a far future.

3. All other cgroups go idle and now the above cgroup has access to
   the whole device.  However, because vtime is already wound using
   the past low hweight, what its current hweight is doesn't matter
   until global vtime catches up to the local vtime.

4. As a result, either vrate gets ramped up extremely or the IOs stall
   while the underlying device is idle.

This is because the hweight the overage is calculated at is different
from the hweight that it's being paid at.

Fix it by remembering the overage in absoulte vtime and continuously
paying with the actual budget according to the current hweight at each
period.

Note that non-forced bios which wait already remembers the cost in
absolute vtime.  This brings forced-bio accounting in line.
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

ioc_pd_free() first cancels the hrtimers and then deactivates the
iocg.  However, the iocg timer can run inbetween and reschedule the
hrtimers which will end up running after the iocg is freed leading to
crashes like the following.

  general protection fault: 0000 [#1] SMP
  ...
  RIP: 0010:iocg_kick_delay+0xbe/0x1b0
  RSP: 0018:ffffc90003598ea0 EFLAGS: 00010046
  RAX: 1cee00fd69512b54 RBX: ffff8881bba48400 RCX: 00000000000003e8
  RDX: 0000000000000000 RSI: 0000000000000001 RDI: ffff8881bba48400
  RBP: 0000000000004e20 R08: 0000000000000002 R09: 00000000000003e8
  R10: 0000000000000000 R11: 0000000000000000 R12: ffffc90003598ef0
  R13: 00979f3810ad461f R14: ffff8881bba4b400 R15: 25439f950d26e1d1
  FS:  0000000000000000(0000) GS:ffff88885f800000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00007f64328c7e40 CR3: 0000000002409005 CR4: 00000000003606e0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  Call Trace:
   <IRQ>
   iocg_delay_timer_fn+0x3d/0x60
   __hrtimer_run_queues+0xfe/0x270
   hrtimer_interrupt+0xf4/0x210
   smp_apic_timer_interrupt+0x5e/0x120
   apic_timer_interrupt+0xf/0x20
   </IRQ>

Fix it by canceling hrtimers after deactivating the iocg.

Fixes: 7caa4715 ("blkcg: implement blk-iocost")
Reported-by: NDave Jones <davej@codemonkey.org.uk>
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This adds to BFQ the missing per-device weight interfaces:
blkio.bfq.weight_device on legacy and io.bfq.weight on unified. The
implementation pretty closely resembles what we had in CFQ and the parsing code
is basically reused.

Tests
=====

Using two cgroups and three block devices, having weights setup as:

Cgroup          test1           test2
============================================
default         100             500
sda             500             100
sdb             default         default
sdc             200             200

cgroup v1 runs
--------------

    sda.test1.out:   READ: bw=913MiB/s
    sda.test2.out:   READ: bw=183MiB/s

    sdb.test1.out:   READ: bw=213MiB/s
    sdb.test2.out:   READ: bw=1054MiB/s

    sdc.test1.out:   READ: bw=650MiB/s
    sdc.test2.out:   READ: bw=650MiB/s

cgroup v2 runs
--------------

    sda.test1.out:   READ: bw=915MiB/s
    sda.test2.out:   READ: bw=184MiB/s

    sdb.test1.out:   READ: bw=216MiB/s
    sdb.test2.out:   READ: bw=1069MiB/s

    sdc.test1.out:   READ: bw=621MiB/s
    sdc.test2.out:   READ: bw=622MiB/s
Signed-off-by: NFam Zheng <zhengfeiran@bytedance.com>
Acked-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NPaolo Valente <paolo.valente@linaro.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This function will be useful when we update weight from the soon-coming
per-device interface.
Signed-off-by: NFam Zheng <zhengfeiran@bytedance.com>
Reviewed-by: NPaolo Valente <paolo.valente@linaro.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

The comment of bfq_group_set_weight says the reading of prio_changed
should happen before the reading of weight, but a memory barrier is
missing here. Add it now, to match the smp_wmb() there.
Signed-off-by: NFam Zheng <zhengfeiran@bytedance.com>
Reviewed-by: NPaolo Valente <paolo.valente@linaro.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

The lookup logic is broken - 'e' will never be NULL, even if the
list is empty. Maintain lookup hit in a separate variable instead.

Fixes: a0958ba7 ("block: Improve default elevator selection")
Reported-by: NJulia Lawall <julia.lawall@lip6.fr>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

When elevator_init_mq() is called from blk_mq_init_allocated_queue(),
the only information known about the device is the number of hardware
queues as the block device scan by the device driver is not completed
yet for most drivers. The device type and elevator required features
are not set yet, preventing to correctly select the default elevator
most suitable for the device.

This currently affects all multi-queue zoned block devices which default
to the "none" elevator instead of the required "mq-deadline" elevator.
These drives currently include host-managed SMR disks connected to a
smartpqi HBA and null_blk block devices with zoned mode enabled.
Upcoming NVMe Zoned Namespace devices will also be affected.

Fix this by adding the boolean elevator_init argument to
blk_mq_init_allocated_queue() to control the execution of
elevator_init_mq(). Two cases exist:
1) elevator_init = false is used for calls to
   blk_mq_init_allocated_queue() within blk_mq_init_queue(). In this
   case, a call to elevator_init_mq() is added to __device_add_disk(),
   resulting in the delayed initialization of the queue elevator
   after the device driver finished probing the device information. This
   effectively allows elevator_init_mq() access to more information
   about the device.
2) elevator_init = true preserves the current behavior of initializing
   the elevator directly from blk_mq_init_allocated_queue(). This case
   is used for the special request based DM devices where the device
   gendisk is created before the queue initialization and device
   information (e.g. queue limits) is already known when the queue
   initialization is executed.

Additionally, to make sure that the elevator initialization is never
done while requests are in-flight (there should be none when the device
driver calls device_add_disk()), freeze and quiesce the device request
queue before calling blk_mq_init_sched() in elevator_init_mq().
Reviewed-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NDamien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

For block devices that do not specify required features, preserve the
current default elevator selection (mq-deadline for single queue
devices, none for multi-queue devices). However, for devices specifying
required features (e.g. zoned block devices ELEVATOR_F_ZBD_SEQ_WRITE
feature), select the first available elevator providing the required
features.

In all cases, default to "none" if no elevator is available or if the
initialization of the default elevator fails.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NDamien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Introduce the definition of elevator features through the
elevator_features flags in the elevator_type structure. Each flag can
represent a feature supported by an elevator. The first feature defined
by this patch is support for zoned block device sequential write
constraint with the flag ELEVATOR_F_ZBD_SEQ_WRITE, which is implemented
by the mq-deadline elevator using zone write locking.

Other possible features are IO priorities, write hints, latency targets
or single-LUN dual-actuator disks (for which the elevator could maintain
one LBA ordered list per actuator).

The required_elevator_features field is also added to the request_queue
structure to allow a device driver to specify elevator feature flags
that an elevator must support for the correct operation of the device
(e.g. device drivers for zoned block devices can have the
ELEVATOR_F_ZBD_SEQ_WRITE flag as a required feature).
The helper function blk_queue_required_elevator_features() is
defined for setting this new field.

With these two new fields in place, the elevator functions
elevator_match() and elevator_find() are modified to allow a user to set
only an elevator with a set of features that satisfies the device
required features. Elevators not matching the device requirements are
not shown in the device sysfs queue/scheduler file to prevent their use.

The "none" elevator can always be selected as before.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NDamien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

If the default elevator chosen is mq-deadline, elevator_init_mq() may
return an error if mq-deadline initialization fails, leading to
blk_mq_init_allocated_queue() returning an error, which in turn will
cause the block device initialization to fail and the device not being
exposed.

Instead of taking such extreme measure, handle mq-deadline
initialization failures in the same manner as when mq-deadline is not
available (no module to load), that is, default to the "none" scheduler.
With this change, elevator_init_mq() return type can be changed to void.
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NMing Lei <ming.lei@redhat.com>
Signed-off-by: NDamien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Instead of checking a queue tag_set BLK_MQ_F_NO_SCHED flag before
calling elevator_init_mq() to make sure that the queue supports IO
scheduling, use the elevator.c function elv_support_iosched() in
elevator_init_mq(). This does not introduce any functional change but
ensure that elevator_init_mq() does the right thing based on the queue
settings.
Reviewed-by: NMing Lei <ming.lei@redhat.com>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NDamien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Since the inclusion of blk-mq, elevator argument was not being
considered anymore, and it's utility died long with the legacy IO path,
now removed too.
Reviewed-by: NHannes Reinecke <hare@suse.com>
Reviewed-by: NBob Liu <bob.liu@oracle.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMarcos Paulo de Souza <marcos.souza.org@gmail.com>

Fold with doc removal patch.
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Commit 7211aef8 ("block: mq-deadline: Fix write completion
handling") added a call to blk_mq_sched_mark_restart_hctx() in
dd_dispatch_request() to make sure that write request dispatching does
not stall when all target zones are locked. This fix left a subtle race
when a write completion happens during a dispatch execution on another
CPU:

CPU 0: Dispatch			CPU1: write completion

dd_dispatch_request()
    lock(&dd->lock);
    ...
    lock(&dd->zone_lock);	dd_finish_request()
    rq = find request		lock(&dd->zone_lock);
    unlock(&dd->zone_lock);
    				zone write unlock
				unlock(&dd->zone_lock);
				...
				__blk_mq_free_request
                                      check restart flag (not set)
				      -> queue not run
    ...
    if (!rq && have writes)
        blk_mq_sched_mark_restart_hctx()
    unlock(&dd->lock)

Since the dispatch context finishes after the write request completion
handling, marking the queue as needing a restart is not seen from
__blk_mq_free_request() and blk_mq_sched_restart() not executed leading
to the dispatch stall under 100% write workloads.

Fix this by moving the call to blk_mq_sched_mark_restart_hctx() from
dd_dispatch_request() into dd_finish_request() under the zone lock to
ensure full mutual exclusion between write request dispatch selection
and zone unlock on write request completion.

Fixes: 7211aef8 ("block: mq-deadline: Fix write completion handling")
Cc: stable@vger.kernel.org
Reported-by: NHans Holmberg <Hans.Holmberg@wdc.com>
Reviewed-by: NHans Holmberg <hans.holmberg@wdc.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NDamien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This patch adds a helper function whether a queue can merge
the segments by the DMA MAP layer (e.g. via IOMMU).
Signed-off-by: NYoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au
Signed-off-by: NChristoph Hellwig <hch@lst.de>

ioc_cpd_alloc() forgot to check NULL return from kzalloc().  Add it.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reported-by: Nkbuild test robot <lkp@intel.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

blk_iocost_init() forgot to free its percpu stat on the error path.
Fix it.

Fixes: 7caa4715 ("blkcg: implement blk-iocost")
Reported-by: NHillf Danton <hdanton@sina.com>
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Add a script which can be used to generate device-specific iocost
linear model coefficients.
Signed-off-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

Instead of mucking with debugfs and ->pd_stat(), add drgn based
monitoring script.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Omar Sandoval <osandov@fb.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>

This patchset implements IO cost model based work-conserving
proportional controller.

While io.latency provides the capability to comprehensively prioritize
and protect IOs depending on the cgroups, its protection is binary -
the lowest latency target cgroup which is suffering is protected at
the cost of all others.  In many use cases including stacking multiple
workload containers in a single system, it's necessary to distribute
IO capacity with better granularity.

One challenge of controlling IO resources is the lack of trivially
observable cost metric.  The most common metrics - bandwidth and iops
- can be off by orders of magnitude depending on the device type and
IO pattern.  However, the cost isn't a complete mystery.  Given
several key attributes, we can make fairly reliable predictions on how
expensive a given stream of IOs would be, at least compared to other
IO patterns.

The function which determines the cost of a given IO is the IO cost
model for the device.  This controller distributes IO capacity based
on the costs estimated by such model.  The more accurate the cost
model the better but the controller adapts based on IO completion
latency and as long as the relative costs across differents IO
patterns are consistent and sensible, it'll adapt to the actual
performance of the device.

Currently, the only implemented cost model is a simple linear one with
a few sets of default parameters for different classes of device.
This covers most common devices reasonably well.  All the
infrastructure to tune and add different cost models is already in
place and a later patch will also allow using bpf progs for cost
models.

Please see the top comment in blk-iocost.c and documentation for
more details.

v2: Rebased on top of RQ_ALLOC_TIME changes and folded in Rik's fix
    for a divide-by-zero bug in current_hweight() triggered by zero
    inuse_sum.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Andy Newell <newella@fb.com>
Cc: Josef Bacik <jbacik@fb.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: NJens Axboe <axboe@kernel.dk>