When wbt_enable_default() is moved out of queue freezing in elevator_change(),
it can cause the wbt inflight counter to become negative (-1), leading to hung
tasks in the writeback path. Tasks get stuck in wbt_wait() because the counter
is in an inconsistent state.
The issue occurs because wbt_enable_default() could race with IO submission,
allowing the counter to be decremented before proper initialization. This manifests
as:
rq_wait[0]:
inflight: -1
has_waiters: True
rwb_enabled() checks the state, which can be updated exactly between wbt_wait()
(rq_qos_throttle()) and wbt_track()(rq_qos_track()), then the inflight counter
will become negative.
And results in hung task warnings like:
task:kworker/u24:39 state:D stack:0 pid:14767
Call Trace:
rq_qos_wait+0xb4/0x150
wbt_wait+0xa9/0x100
__rq_qos_throttle+0x24/0x40
blk_mq_submit_bio+0x672/0x7b0
...
Fix this by:
1. Splitting wbt_enable_default() into:
- __wbt_enable_default(): Returns true if wbt_init() should be called
- wbt_enable_default(): Wrapper for existing callers (no init)
- wbt_init_enable_default(): New function that checks and inits WBT
2. Using wbt_init_enable_default() in blk_register_queue() to ensure
proper initialization during queue registration
3. Move wbt_init() out of wbt_enable_default() which is only for enabling
disabled wbt from bfq and iocost, and wbt_init() isn't needed. Then the
original lock warning can be avoided.
4. Removing the ELEVATOR_FLAG_ENABLE_WBT_ON_EXIT flag and its handling
code since it's no longer needed
This ensures WBT is properly initialized before any IO can be submitted,
preventing the counter from going negative.
Cc: Nilay Shroff <nilay@linux.ibm.com>
Cc: Yu Kuai <yukuai@fnnas.com>
Cc: Guangwu Zhang <guazhang@redhat.com>
Fixes: 78c271344b6f ("block: move wbt_enable_default() out of queue freezing from sched ->exit()")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In commit 1e44bedbc921 ("block: unifying elevator change"), the
elevator_init_mq function was deleted, but its declaration in elevator.h
was overlooked. This patch fixes it.
Signed-off-by: Chengkaitao <chengkaitao@kylinos.cn>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The previous patch introduced ->alloc_sched_data and
->free_sched_data methods. This patch builds upon that
by now using these methods during elevator switch and
nr_hw_queue update.
It's also ensured that scheduler-specific data is
allocated and freed through the new callbacks outside
of the ->freeze_lock and ->elevator_lock locking contexts,
thereby preventing any dependency on pcpu_alloc_mutex.
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Yu Kuai <yukuai@fnnas.com>
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The recent lockdep splat [1] highlights a potential deadlock risk
involving ->elevator_lock and ->freeze_lock dependencies on -pcpu_alloc_
mutex. The trace shows that the issue occurs when the Kyber scheduler
allocates dynamic memory for its elevator data during initialization.
To address this, introduce two new elevator operation callbacks:
->alloc_sched_data and ->free_sched_data. The subsequent patch would
build upon these newly introduced methods to suppress lockdep splat[1].
[1] https://lore.kernel.org/all/CAGVVp+VNW4M-5DZMNoADp6o2VKFhi7KxWpTDkcnVyjO0=-D5+A@mail.gmail.com/
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch introduces a new structure, struct elevator_resources, to
group together all elevator-related resources that share the same
lifetime. As a first step, this change moves the elevator tag pointer
from struct elv_change_ctx into the new struct elevator_resources.
Additionally, rename blk_mq_alloc_sched_tags_batch() and
blk_mq_free_sched_tags_batch() to blk_mq_alloc_sched_res_batch() and
blk_mq_free_sched_res_batch(), respectively. Introduce two new wrapper
helpers, blk_mq_alloc_sched_res() and blk_mq_free_sched_res(), around
blk_mq_alloc_sched_tags() and blk_mq_free_sched_tags().
These changes pave the way for consolidating the allocation and freeing
of elevator-specific resources into common helper functions. This
refactoring improves encapsulation and prepares the code for future
extensions, allowing additional elevator-specific data to be added to
struct elevator_resources without cluttering struct elv_change_ctx.
Subsequent patches will extend struct elevator_resources to include
other elevator-related data.
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Yu Kuai <yukuai@fnnas.com>
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently, the nr_hw_queues update path manages two disjoint xarrays —
one for elevator tags and another for elevator type — both used during
elevator switching. Maintaining these two parallel structures for the
same purpose adds unnecessary complexity and potential for mismatched
state.
This patch unifies both xarrays into a single structure, struct
elv_change_ctx, which holds all per-queue elevator change context. A
single xarray, named elv_tbl, now maps each queue (q->id) in a tagset
to its corresponding elv_change_ctx entry, encapsulating the elevator
tags, type and name references.
This unification simplifies the code, improves maintainability, and
clarifies ownership of per-queue elevator state.
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Yu Kuai <yukuai@fnnas.com>
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Current depth_updated has some problems:
1) depth_updated() will be called for each hctx, while all elevators
will update async_depth for the disk level, this is not related to hctx;
2) In blk_mq_update_nr_requests(), if previous hctx update succeed and
this hctx update failed, q->nr_requests will not be updated, while
async_depth is already updated with new nr_reqeuests in previous
depth_updated();
3) All elevators are using q->nr_requests to calculate async_depth now,
however, q->nr_requests is still the old value when depth_updated() is
called from blk_mq_update_nr_requests();
Those problems are first from error path, then mq-deadline, and recently
for bfq and kyber, fix those problems by:
- pass in request_queue instead of hctx;
- move depth_updated() after q->nr_requests is updated in
blk_mq_update_nr_requests();
- add depth_updated() call inside init_sched() method to initialize
async_depth;
- remove init_hctx() method for mq-deadline and bfq that is useless now;
Fixes: 77f1e0a52d26 ("bfq: update internal depth state when queue depth changes")
Fixes: 39823b47bbd4 ("block/mq-deadline: Fix the tag reservation code")
Fixes: 42e6c6ce03fd ("lib/sbitmap: convert shallow_depth from one word to the whole sbitmap")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Li Nan <linan122@huawei.com>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Link: https://lore.kernel.org/r/20250821060612.1729939-2-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Recent lockdep reports [1] have revealed a potential deadlock caused by a
lock dependency between the percpu allocator lock and the elevator lock.
This issue can be avoided by ensuring that the allocation and release of
scheduler tags (sched_tags) are performed outside the elevator lock.
Furthermore, the queue does not need to be remain frozen during these
operations.
To address this, move all sched_tags allocations and deallocations outside
of both the ->elevator_lock and the ->freeze_lock. Since the lifetime of
the elevator queue and its associated sched_tags is closely tied, the
allocated sched_tags are now stored in the elevator queue structure. Then,
during the actual elevator switch (which runs under ->freeze_lock and
->elevator_lock), the pre-allocated sched_tags are assigned to the
appropriate q->hctx. Once the elevator switch is complete and the locks
are released, the old elevator queue and its associated sched_tags are
freed.
This commit specifically addresses the allocation/deallocation of sched_
tags during elevator switching. Note that sched_tags may also be allocated
in other contexts, such as during nr_hw_queues updates. Supporting that
use case will require batch allocation/deallocation, which will be handled
in a follow-up patch.
This restructuring ensures that sched_tags memory management occurs
entirely outside of the ->elevator_lock and ->freeze_lock context,
eliminating the lock dependency problem seen during scheduler updates.
[1] https://lore.kernel.org/all/0659ea8d-a463-47c8-9180-43c719e106eb@linux.ibm.com/
Reported-by: Stefan Haberland <sth@linux.ibm.com>
Closes: https://lore.kernel.org/all/0659ea8d-a463-47c8-9180-43c719e106eb@linux.ibm.com/
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Link: https://lore.kernel.org/r/20250730074614.2537382-3-nilay@linux.ibm.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In preparation for allocating sched_tags before freezing the request
queue and acquiring ->elevator_lock, move the elevator queue allocation
logic from the elevator ops ->init_sched callback into blk_mq_init_sched.
As elevator_alloc is now only invoked from block layer core, we don't
need to export it, so unexport elevator_alloc function.
This refactoring provides a centralized location for elevator queue
initialization, which makes it easier to store pre-allocated sched_tags
in the struct elevator_queue during later changes.
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Link: https://lore.kernel.org/r/20250730074614.2537382-2-nilay@linux.ibm.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
scheduler's ->exit() is called with queue frozen and elevator lock is held, and
wbt_enable_default() can't be called with queue frozen, otherwise the
following lockdep warning is triggered:
#6 (&q->rq_qos_mutex){+.+.}-{4:4}:
#5 (&eq->sysfs_lock){+.+.}-{4:4}:
#4 (&q->elevator_lock){+.+.}-{4:4}:
#3 (&q->q_usage_counter(io)#3){++++}-{0:0}:
#2 (fs_reclaim){+.+.}-{0:0}:
#1 (&sb->s_type->i_mutex_key#3){+.+.}-{4:4}:
#0 (&q->debugfs_mutex){+.+.}-{4:4}:
Fix the issue by moving wbt_enable_default() out of bfq's exit(), and
call it from elevator_change_done().
Meantime add disk->rqos_state_mutex for covering wbt state change, which
matches the purpose more than ->elevator_lock.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20250505141805.2751237-26-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Prepare for moving elv_register[unregister]_queue out of elevator_lock
& queue freezing, so we may have to call elv_unregister_queue() after
elevator ->exit() is called, then there is small window for user to
call into ->show()/store(), and user-after-free can be caused.
Fail to show/store elevator sysfs attribute if elevator is dying by
adding one new flag of ELEVATOR_FLAG_DYNG, which is protected by
elevator ->sysfs_lock.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20250505141805.2751237-20-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
sched debugfs shares same lifetime with scheduler's kobject, and same
lock(elevator lock), so move sched debugfs register/unregister into
elevator_register_queue() and elevator_unregister_queue().
Then we needn't blk_mq_debugfs_register() for us to register sched
debugfs any more.
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250505141805.2751237-7-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
ELEVATOR_FLAG_DISABLE_WBT is only used by BFQ to disallow wbt when BFQ is
in use. The flag is set in BFQ's init(), and cleared in BFQ's exit().
Making it as request queue flag, so that we can avoid to deal with elevator
switch race. Also it isn't graceful to checking one scheduler flag in
wbt_enable_default().
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Nilay Shroff <nilay@linux.ibm.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250505141805.2751237-3-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A queue's elevator can be updated either when modifying nr_hw_queues
or through the sysfs scheduler attribute. Currently, elevator switching/
updating is protected using q->sysfs_lock, but this has led to lockdep
splats[1] due to inconsistent lock ordering between q->sysfs_lock and
the freeze-lock in multiple block layer call sites.
As the scope of q->sysfs_lock is not well-defined, its (mis)use has
resulted in numerous lockdep warnings. To address this, introduce a new
q->elevator_lock, dedicated specifically for protecting elevator
switches/updates. And we'd now use this new q->elevator_lock instead of
q->sysfs_lock for protecting elevator switches/updates.
While at it, make elv_iosched_load_module() a static function, as it is
only called from elv_iosched_store(). Also, remove redundant parameters
from elv_iosched_load_module() function signature.
[1] https://lore.kernel.org/all/67637e70.050a0220.3157ee.000c.GAE@google.com/
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Link: https://lore.kernel.org/r/20250304102551.2533767-5-nilay@linux.ibm.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Requesting a module either succeeds or does nothing, return an error from
this method does not make sense.
Also move the load_module after the store method in the struct
declaration to keep the important show and store methods together.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Link: https://lore.kernel.org/r/20241008050841.104602-1-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit af2814149883 ("block: freeze the queue in queue_attr_store")
changed queue_attr_store() to always freeze a sysfs attribute queue
before calling the attribute store() method, to ensure that no IOs are
in-flight when an attribute value is being updated.
However, this change created a potential deadlock situation for the
scheduler queue attribute as changing the queue elevator with
elv_iosched_store() can result in a call to request_module() if the user
requested module is not already registered. If the file of the requested
module is stored on the block device of the frozen queue, a deadlock
will happen as the read operations triggered by request_module() will
wait for the queue freeze to end.
Solve this issue by introducing the load_module method in struct
queue_sysfs_entry, and to calling this method function in
queue_attr_store() before freezing the attribute queue.
The macro definition QUEUE_RW_LOAD_MODULE_ENTRY() is added to define a
queue sysfs attribute that needs loading a module.
The definition of the scheduler atrribute is changed to using
QUEUE_RW_LOAD_MODULE_ENTRY(), with the function
elv_iosched_load_module() defined as the load_module method.
elv_iosched_store() can then be simplified to remove the call to
request_module().
Reported-by: Richard W.M. Jones <rjones@redhat.com>
Reported-by: Jiri Jaburek <jjaburek@redhat.com>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219166
Fixes: af2814149883 ("block: freeze the queue in queue_attr_store")
Cc: stable@vger.kernel.org
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Tested-by: Richard W.M. Jones <rjones@redhat.com>
Link: https://lore.kernel.org/r/20240908000704.414538-1-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The kobject for the queue entries is embedded into a struct gendisk.
Pass it to the sysfs methods instead of the request_queue derived from
it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240627111407.476276-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The only elevator feature ever implemented is ELEVATOR_F_ZBD_SEQ_WRITE
for signaling that a scheduler implements zone write locking to tightly
control the dispatching order of write operations to zoned block
devices. With the removal of zone write locking support in mq-deadline
and the reliance of all block device drivers on the block layer zone
write plugging to control ordering of write operations to zones, the
elevator feature ELEVATOR_F_ZBD_SEQ_WRITE is completely unused.
Remove it, and also remove the now unused code for filtering the
possible schedulers for a block device based on required features.
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Tested-by: Hans Holmberg <hans.holmberg@wdc.com>
Tested-by: Dennis Maisenbacher <dennis.maisenbacher@wdc.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240408014128.205141-23-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Instead of passing a bool at_head, pass down the full flags from the
blk_mq_insert_request interface.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20230413064057.707578-20-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Make sure we have helpers for all relevant module refcount operations on
the elevator_type in elevator.h, and use them. Move the call to the get
helper in blk_mq_elv_switch_none a bit so that it is obvious with a less
verbose comment.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20221020064819.1469928-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit b5dc5d4d1f4f ("block,bfq: Disable writeback throttling") tries to
disable wbt for bfq, it's done by calling wbt_disable_default() in
bfq_init_queue(). However, wbt is still enabled if default elevator is
bfq:
device_add_disk
elevator_init_mq
bfq_init_queue
wbt_disable_default -> done nothing
blk_register_queue
wbt_enable_default -> wbt is enabled
Fix the problem by adding a new flag ELEVATOR_FLAG_DISBALE_WBT, bfq
will set the flag in bfq_init_queue, and following wbt_enable_default()
won't enable wbt while the flag is set.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20221019121518.3865235-7-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There are only one flag to indicate that elevator is registered currently,
prepare to add a flag to disable wbt if default elevator is bfq.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20221019121518.3865235-6-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Use the new blk_opf_t type for arguments and variables that represent
request flags or a bitwise combination of a request operation and
request flags. Rename the function arguments and also a structure member
that hold a request operation and flags from 'rw' into 'opf'.
This patch does not change any functionality.
Cc: Christoph Hellwig <hch@lst.de>
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Hannes Reinecke <hare@suse.de>
Cc: Damien Le Moal <damien.lemoal@wdc.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20220714180729.1065367-7-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Except for the features passed to blk_queue_required_elevator_features,
elevator.h is only needed internally to the block layer. Move the
ELEVATOR_F_* definitions to blkdev.h, and the move elevator.h to
block/, dropping all the spurious includes outside of that.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20210920123328.1399408-13-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
