Israel Rukshin [Sun, 18 Aug 2019 09:08:52 +0000 (12:08 +0300)]
nvme-rdma: Add TOS for rdma transport
For RDMA transports, TOS is an extension of IB QoS to provide clients
the ability to segregate traffic flows for different type of data.
RDMA CM abstract it for ULPs using rdma_set_service_type().
Internally, each traffic flow is represented by a connection with all of
its independent resources like that of a normal connection, and is
differentiated by service type. In other words, there can be multiple qp
connections between an IP pair and each supports a unique service type.
One of the TOS usage is bandwidth management which allows setting bandwidth
limits for QoS classes, e.g. 80% bandwidth to controllers at QoS class A
and 20% to controllers at QoS class B.
Note: In addition to the TOS configuration, QOS must be configured on the
relevant HCA on the target (send RDMA commands) and initiator to effect
the traffic.
Minwoo Im [Sun, 4 Aug 2019 07:50:48 +0000 (16:50 +0900)]
nvme: trace: support for Get LBA Status opcode parsed
This patch adds Get LBA Status command's opcode to the macro that is
used by the trace feature. Now we can see "get_lba_status" instead of
the opcode value itself.
Signed-off-by: Minwoo Im <minwoo.im.dev@gmail.com> Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Tom Wu [Thu, 8 Aug 2019 02:22:36 +0000 (02:22 +0000)]
nvmet: fix data units read and written counters in SMART log
In nvme spec 1.3 there is a definition for data write/read counters
from SMART log, (See section 5.14.1.2):
This value is reported in thousands (i.e., a value of 1
corresponds to 1000 units of 512 bytes read) and is rounded up.
However, in nvme target where value is reported with actual units,
but not thousands of units as the spec requires.
Signed-off-by: Tom Wu <tomwu@mellanox.com> Reviewed-by: Israel Rukshin <israelr@mellanox.com> Reviewed-by: Max Gurtovoy <maxg@mellanox.com> Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Simple polling support via socket busy_poll interface.
Although we do not shutdown interrupts but simply hammer
the socket poll, we can sometimes find completions faster
than the normal interrupt driven RX path.
We add per queue nr_cqe counter that resets every time
RX path is invoked such that .poll callback can return it
to stay consistent with the semantics.
Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
nvme-tcp: Use protocol specific operations while reading socket
Using socket specific read_sock() calls instead of directly calling
tcp_read_sock() helps lld module registered handlers if any, to be called
from nvme-tcp host.
This patch therefore replaces the tcp_read_sock() with socket specific
prot_ops.
Tejun Heo [Wed, 28 Aug 2019 22:05:58 +0000 (15:05 -0700)]
blkcg: implement blk-iocost
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: Tejun 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: Jens Axboe <axboe@kernel.dk>
Tejun Heo [Wed, 28 Aug 2019 22:05:57 +0000 (15:05 -0700)]
blk-mq: add optional request->alloc_time_ns
There are currently two start time timestamps - start_time_ns and
io_start_time_ns. The former marks the request allocation and and the
second issue-to-device time. The planned io.weight controller needs
to measure the total time bios take to execute after it leaves rq_qos
including the time spent waiting for request to become available,
which can easily dominate on saturated devices.
This patch adds request->alloc_time_ns which records when the request
allocation attempt started. As it isn't used for the usual stats,
make it optional behind CONFIG_BLK_RQ_ALLOC_TIME and
QUEUE_FLAG_RQ_ALLOC_TIME so that it can be compiled out when there are
no users and it's active only on queues which need it even when
compiled in.
v2: s/pre_start_time/alloc_time/ and add CONFIG_BLK_RQ_ALLOC_TIME
gating as suggested by Jens.
wbt already gets queue depth changed notification through
wbt_set_queue_depth(). Generalize it into
rq_qos_ops->queue_depth_changed() so that other rq_qos policies can
easily hook into the events too.
Tejun Heo [Wed, 28 Aug 2019 22:05:53 +0000 (15:05 -0700)]
blkcg: separate blkcg_conf_get_disk() out of blkg_conf_prep()
Separate out blkcg_conf_get_disk() so that it can be used by blkcg
policy interface file input parsers before the policy is actually
enabled. This doesn't introduce any functional changes.
Tejun Heo [Wed, 28 Aug 2019 22:05:51 +0000 (15:05 -0700)]
blkcg: pass @q and @blkcg into blkcg_pol_alloc_pd_fn()
Instead of @node, pass in @q and @blkcg so that the alloc function has
more context. This doesn't cause any behavior change and will be used
by io.weight implementation.
Jens Axboe [Tue, 27 Aug 2019 19:56:59 +0000 (13:56 -0600)]
Merge branch 'md-next' of git://git.kernel.org/pub/scm/linux/kernel/git/song/md into for-5.4/block
Pull MD fixes from Song.
* 'md-next' of git://git.kernel.org/pub/scm/linux/kernel/git/song/md:
raid5 improve too many read errors msg by adding limits
md: don't report active array_state until after revalidate_disk() completes.
md: only call set_in_sync() when it is expected to succeed.
Nigel Croxon [Wed, 21 Aug 2019 13:27:08 +0000 (09:27 -0400)]
raid5 improve too many read errors msg by adding limits
Often limits can be changed by admin. When discussing such things
it helps if you can provide "self-sustained" facts. Also
sometimes the admin thinks he changed a limit, but it did not
take effect for some reason or he changed the wrong thing.
V3: Only pr_warn when Faulty is 0.
V2: Add read_errors value to pr_warn.
Signed-off-by: Nigel Croxon <ncroxon@redhat.com> Signed-off-by: Song Liu <songliubraving@fb.com>
NeilBrown [Tue, 20 Aug 2019 00:21:09 +0000 (10:21 +1000)]
md: don't report active array_state until after revalidate_disk() completes.
Until revalidate_disk() has completed, the size of a new md array will
appear to be zero.
So we shouldn't report, through array_state, that the array is active
until that time.
udev rules check array_state to see if the array is ready. As soon as
it appear to be zero, fsck can be run. If it find the size to be
zero, it will fail.
So add a new flag to provide an interlock between do_md_run() and
array_state_show(). This flag is set while do_md_run() is active and
it prevents array_state_show() from reporting that the array is
active.
Before do_md_run() is called, ->pers will be NULL so array is
definitely not active.
After do_md_run() is called, revalidate_disk() will have run and the
array will be completely ready.
We also move various sysfs_notify*() calls out of md_run() into
do_md_run() after MD_NOT_READY is cleared. This ensure the
information is ready before the notification is sent.
Prior to v4.12, array_state_show() was called with the
mddev->reconfig_mutex held, which provided exclusion with do_md_run().
Note that MD_NOT_READY cleared twice. This is deliberate to cover
both success and error paths with minimal noise.
Fixes: b7b17c9b67e5 ("md: remove mddev_lock() from md_attr_show()") Cc: stable@vger.kernel.org (v4.12++) Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Song Liu <songliubraving@fb.com>
NeilBrown [Tue, 20 Aug 2019 00:21:09 +0000 (10:21 +1000)]
md: only call set_in_sync() when it is expected to succeed.
Since commit 4ad23a976413 ("MD: use per-cpu counter for
writes_pending"), set_in_sync() is substantially more expensive: it
can wait for a full RCU grace period which can be 10s of milliseconds.
So we should only call it when the cost is justified.
md_check_recovery() currently calls set_in_sync() every time it finds
anything to do (on non-external active arrays). For an array
performing resync or recovery, this will be quite often.
Each call will introduce a delay to the md thread, which can noticeable
affect IO submission latency.
In md_check_recovery() we only need to call set_in_sync() if
'safemode' was non-zero at entry, meaning that there has been not
recent IO. So we save this "safemode was nonzero" state, and only
call set_in_sync() if it was non-zero.
This measurably reduces mean and maximum IO submission latency during
resync/recovery.
Reported-and-tested-by: Jack Wang <jinpu.wang@cloud.ionos.com> Fixes: 4ad23a976413 ("MD: use per-cpu counter for writes_pending") Cc: stable@vger.kernel.org (v4.12+) Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Song Liu <songliubraving@fb.com>
Ming Lei [Tue, 27 Aug 2019 11:01:48 +0000 (19:01 +0800)]
block: split .sysfs_lock into two locks
The kernfs built-in lock of 'kn->count' is held in sysfs .show/.store
path. Meantime, inside block's .show/.store callback, q->sysfs_lock is
required.
However, when mq & iosched kobjects are removed via
blk_mq_unregister_dev() & elv_unregister_queue(), q->sysfs_lock is held
too. This way causes AB-BA lock because the kernfs built-in lock of
'kn-count' is required inside kobject_del() too, see the lockdep warning[1].
On the other hand, it isn't necessary to acquire q->sysfs_lock for
both blk_mq_unregister_dev() & elv_unregister_queue() because
clearing REGISTERED flag prevents storing to 'queue/scheduler'
from being happened. Also sysfs write(store) is exclusive, so no
necessary to hold the lock for elv_unregister_queue() when it is
called in switching elevator path.
So split .sysfs_lock into two: one is still named as .sysfs_lock for
covering sync .store, the other one is named as .sysfs_dir_lock
for covering kobjects and related status change.
sysfs itself can handle the race between add/remove kobjects and
showing/storing attributes under kobjects. For switching scheduler
via storing to 'queue/scheduler', we use the queue flag of
QUEUE_FLAG_REGISTERED with .sysfs_lock for avoiding the race, then
we can avoid to hold .sysfs_lock during removing/adding kobjects.
[1] lockdep warning
======================================================
WARNING: possible circular locking dependency detected 5.3.0-rc3-00044-g73277fc75ea0 #1380 Not tainted
------------------------------------------------------
rmmod/777 is trying to acquire lock: 00000000ac50e981 (kn->count#202){++++}, at: kernfs_remove_by_name_ns+0x59/0x72
but task is already holding lock: 00000000fb16ae21 (&q->sysfs_lock){+.+.}, at: blk_unregister_queue+0x78/0x10b
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
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: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Ming Lei [Tue, 27 Aug 2019 11:01:47 +0000 (19:01 +0800)]
block: add helper for checking if queue is registered
There are 4 users which check if queue is registered, so add one helper
to check it.
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> Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Ming Lei [Tue, 27 Aug 2019 11:01:46 +0000 (19:01 +0800)]
blk-mq: don't hold q->sysfs_lock in blk_mq_map_swqueue
blk_mq_map_swqueue() is called from blk_mq_init_allocated_queue()
and blk_mq_update_nr_hw_queues(). For the former caller, the kobject
isn't exposed to userspace yet. For the latter caller, hctx sysfs entries
and debugfs are un-registered before updating nr_hw_queues.
On the other hand, commit 2f8f1336a48b ("blk-mq: always free hctx after
request queue is freed") moves freeing hctx into queue's release
handler, so there won't be race with queue release path too.
So don't hold q->sysfs_lock in blk_mq_map_swqueue().
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> Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Ming Lei [Tue, 27 Aug 2019 11:01:45 +0000 (19:01 +0800)]
block: don't hold q->sysfs_lock in elevator_init_mq
The original comment says:
q->sysfs_lock must be held to provide mutual exclusion between
elevator_switch() and here.
Which is simply wrong. elevator_init_mq() is only called from
blk_mq_init_allocated_queue, which is always called before the request
queue is registered via blk_register_queue(), for dm-rq or normal rq
based driver. However, queue's kobject is only exposed and added to sysfs
in blk_register_queue(). So there isn't such race between elevator_switch()
and elevator_init_mq().
So avoid to hold q->sysfs_lock in elevator_init_mq().
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> Cc: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bart Van Assche [Tue, 27 Aug 2019 11:01:44 +0000 (19:01 +0800)]
block: Remove blk_mq_register_dev()
This function has no callers. Hence remove it.
Cc: Christoph Hellwig <hch@infradead.org> Cc: Ming Lei <ming.lei@redhat.com> Cc: Hannes Reinecke <hare@suse.com> Signed-off-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
There's an inherent mismatch between memcg and writeback. The former
trackes ownership per-page while the latter per-inode. This was a
deliberate design decision because honoring per-page ownership in the
writeback path is complicated, may lead to higher CPU and IO overheads
and deemed unnecessary given that write-sharing an inode across
different cgroups isn't a common use-case.
Combined with inode majority-writer ownership switching, this works
well enough in most cases but there are some pathological cases. For
example, let's say there are two cgroups A and B which keep writing to
different but confined parts of the same inode. B owns the inode and
A's memory is limited far below B's. A's dirty ratio can rise enough
to trigger balance_dirty_pages() sleeps but B's can be low enough to
avoid triggering background writeback. A will be slowed down without
a way to make writeback of the dirty pages happen.
This patch implements foreign dirty recording and foreign mechanism so
that when a memcg encounters a condition as above it can trigger
flushes on bdi_writebacks which can clean its pages. Please see the
comment on top of mem_cgroup_track_foreign_dirty_slowpath() for
details.
Tejun Heo [Mon, 26 Aug 2019 16:06:53 +0000 (09:06 -0700)]
bdi: Add bdi->id
There currently is no way to universally identify and lookup a bdi
without holding a reference and pointer to it. This patch adds an
non-recycling bdi->id and implements bdi_get_by_id() which looks up
bdis by their ids. This will be used by memcg foreign inode flushing.
I left bdi_list alone for simplicity and because while rb_tree does
support rcu assignment it doesn't seem to guarantee lossless walk when
walk is racing aginst tree rebalance operations.
Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Tejun Heo [Mon, 26 Aug 2019 16:06:52 +0000 (09:06 -0700)]
writeback: Generalize and expose wb_completion
wb_completion is used to track writeback completions. We want to use
it from memcg side for foreign inode flushes. This patch updates it
to remember the target waitq instead of assuming bdi->wb_waitq and
expose it outside of fs-writeback.c.
Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch creates a helper function for handling zoned block device
operations.
This patch also restructured the code for null_blk_zoned.c and uses the
pattern to return blk_status_t and catch the error in the function
null_handle_cmd() into cmd->error variable instead of setting it up in
the deeper layer just like the way it is done for flush, badblocks and
memory backed case in the null_handle_cmd(). We also move
null_handle_zoned() to the null_blk_zoned.c to keep the zoned code
separate.
This patch creates a helper for handling requests when null_blk is
memory backed in the null_handle_cmd(). Although the helper is very
simple right now, it makes the code flow consistent with the rest of
code in the null_handle_cmd() and provides a uniform code structure
for future code.
This is a preparation patch which moves the duplicate code for sectors
and nr_sectors calculations for bio vs request mode into their
respective callers (null_queue_bio(), null_qeueue_req()). Now the core
function only deals with the respective actions and commands instead of
having to calculte the bio vs req operations and different sector
related variables. We also move the flush command handling at the top
which significantly simplifies the rest of the code.
block: move same page handling from __bio_add_pc_page to the callers
Hiding page refcount manipulation inside a low-level bio helper is
somewhat awkward. Instead return the same page information to the
callers, where it fits in much better.
Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Passsthrough bio handling should be the same as normal bio handling,
except that we need to take hardware limitations into account. Thus
use the common try_merge implementation after checking the hardware
limits. This changes behavior in that we now also check segment
and dma boundary settings for same page merges, which is a little
more work but has no effect as those need to be larger than the
page size.
Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we can add more data into an existing segment we do not create a gap
per definition, so move the check for a gap after the attempt to merge
into the segment.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
that limited the number of devices to 256. Before the patch we could
create 1000s of devices, but the patch switched us from using our
own thread to using a work queue which has a default limit of 256
active works.
The problem is that our recv_work function sits in a loop until
disconnection but only handles IO for one connection. The work is
started when the connection is started/restarted, but if we end up
creating 257 or more connections, the queue_work call just queues
connection257+'s recv_work and that waits for connection 1 - 256's
recv_work to be disconnected and that work instance completing.
Instead of reverting back to kthreads, this has us allocate a
workqueue_struct per device, so we can block in the work.
Cc: stable@vger.kernel.org Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Mike Christie <mchristi@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
where before the patch userspace would set the timeout to 0 to disable
it. With the above patch, a zero timeout tells the block layer to use
the default value of 30 seconds. For setups where commands can take a
long time or experience transient issues like network disruptions this
then results in IO errors being sent to the application.
To fix this, the patch still uses the common block layer timeout
framework, but if zero is set, nbd just logs a message and then resets
the timer when it expires.
Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Mike Christie <mchristi@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
where if the timeout handler runs when the completion path is and we fail
to grab the mutex in the timeout handler we will leave a config reference
and cannot free the config later.
Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Mike Christie <mchristi@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
In the function 'response_parse', num_entries will never be 0 as
slen is checked for 0. Hence, the condition 'if (num_entries == 0)'
can never be true.
Signed-off-by: Revanth Rajashekar <revanth.rajashekar@intel.com> Reviewed-by: Scott Bauer <sbauer@plzdonthack.me> Reviewed-by: Jon Derrick <jonathan.derrick@intel.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Junxiao Bi [Fri, 16 Aug 2019 21:12:33 +0000 (14:12 -0700)]
block: remove struct request_queue queue_head
The dispatch list is not used any more, as the legacy block IO stack
has been removed.
Reviewed-by: Bart Van Assche <bvanassche@acm.org> Reviewed-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Tejun Heo [Fri, 2 Aug 2019 19:08:13 +0000 (12:08 -0700)]
writeback, cgroup: inode_switch_wbs() shouldn't give up on wb_switch_rwsem trylock fail
As inode wb switching may make sync(2) miss some inodes, they're
synchronized using wb_switch_rwsem so that no wb switching happens
while sync(2) is in progress. In addition to synchronizing the actual
switching, the rwsem is also used to prevent queueing new switch
attempts while sync(2) is in progress. This is to avoid queueing too
many instances while the rwsem is held by sync(2). Unfortunately,
this is too agressive and can block wb switching for a long time if
sync(2) is frequent.
The goal is avoiding expolding the number of scheduled switches, not
avoiding scheduling anything. Let's use wb_switch_rwsem only for
synchronizing the actual switching and sync(2) and use
isw_nr_in_flight instead for limiting the maximum number of scheduled
switches. The limit is set to 1024 which should be more than enough
while still avoiding extreme situations.
Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Tejun Heo [Thu, 15 Aug 2019 19:25:28 +0000 (12:25 -0700)]
writeback, cgroup: Adjust WB_FRN_TIME_CUT_DIV to accelerate foreign inode switching
WB_FRN_TIME_CUT_DIV is used to tell the foreign inode detection logic
to ignore short writeback rounds to prevent getting confused by a
burst of short writebacks. The parameter is currently 2 meaning that
anything smaller than half of the running average writback duration
will be ignored.
This is unnecessarily aggressive. The detection logic uses 16 history
slots and is already reasonably protected against some short bursts
confusing it and the current parameter can lead to tens of seconds of
missed detection depending on the writeback pattern.
Let's change the parameter to 8, so that it only ignores writeback
with are smaller than 12.5% of the current running average.
v2: Add comment explaining what's going on with the foreign detection
parameters.
Reviewed-by: Jan Kara <jack@suse.cz> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Johannes Weiner [Thu, 8 Aug 2019 19:03:00 +0000 (15:03 -0400)]
block: annotate refault stalls from IO submission
psi tracks the time tasks wait for refaulting pages to become
uptodate, but it does not track the time spent submitting the IO. The
submission part can be significant if backing storage is contended or
when cgroup throttling (io.latency) is in effect - a lot of time is
spent in submit_bio(). In that case, we underreport memory pressure.
Annotate submit_bio() to account submission time as memory stall when
the bio is reading userspace workingset pages.
Jens Axboe [Fri, 9 Aug 2019 04:11:28 +0000 (22:11 -0600)]
lightnvm: remove unused 'geo' variable
A previous commit correctly removed set-but-not-read variables, but
this left two new variables now unused. Kill them.
Fixes: ba6f7da99aaf ("lightnvm: remove set but not used variables 'data_len' and 'rq_len'") Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Alessio Balsini [Wed, 7 Aug 2019 00:48:28 +0000 (01:48 +0100)]
loop: Add LOOP_SET_DIRECT_IO to compat ioctl
Enabling Direct I/O with loop devices helps reducing memory usage by
avoiding double caching. 32 bit applications running on 64 bits systems
are currently not able to request direct I/O because is missing from the
lo_compat_ioctl.
This patch fixes the compatibility issue mentioned above by exporting
LOOP_SET_DIRECT_IO as additional lo_compat_ioctl() entry.
The input argument for this ioctl is a single long converted to a 1-bit
boolean, so compatibility is preserved.
Zhou Wang [Wed, 24 Jul 2019 03:54:23 +0000 (11:54 +0800)]
lib: scatterlist: Fix to support no mapped sg
In function sg_split, the second sg_calculate_split will return -EINVAL
when in_mapped_nents is 0.
Indeed there is no need to do second sg_calculate_split and sg_split_mapped
when in_mapped_nents is 0, as in_mapped_nents indicates no mapped entry in
original sgl.
Signed-off-by: Zhou Wang <wangzhou1@hisilicon.com> Acked-by: Robert Jarzmik <robert.jarzmik@free.fr> Signed-off-by: Jens Axboe <axboe@kernel.dk>
YueHaibing [Wed, 7 Aug 2019 13:18:47 +0000 (21:18 +0800)]
lightnvm: remove set but not used variables 'data_len' and 'rq_len'
drivers/lightnvm/pblk-read.c: In function pblk_submit_read_gc:
drivers/lightnvm/pblk-read.c:423:6: warning: variable data_len set but not used [-Wunused-but-set-variable]
drivers/lightnvm/pblk-recovery.c: In function pblk_recov_scan_oob:
drivers/lightnvm/pblk-recovery.c:368:15: warning: variable rq_len set but not used [-Wunused-but-set-variable]
They are not used since commit 48e5da725581 ("lightnvm:
move metadata mapping to lower level driver")
Jens Axboe [Wed, 7 Aug 2019 18:26:53 +0000 (12:26 -0600)]
Merge branch 'md-next' of https://github.com/liu-song-6/linux into for-5.4/block
Pull MD changes from Song.
* 'md-next' of https://github.com/liu-song-6/linux:
raid1: factor out a common routine to handle the completion of sync write
md: don't call spare_active in md_reap_sync_thread if all member devices can't work
md: don't set In_sync if array is frozen
md: allow last device to be forcibly removed from RAID1/RAID10.
md: Convert to use int_pow()
md/raid10: end bio when the device faulty
md/raid1: end bio when the device faulty
md/raid6: Set R5_ReadError when there is read failure on parity disk
raid1: use an int as the return value of raise_barrier()
md: don't call spare_active in md_reap_sync_thread if all member devices can't work
When add one disk to array, the md_reap_sync_thread is responsible
to activate the spare and set In_sync flag for the new member in
spare_active().
But if raid1 has one member disk A, and disk B is added to the array.
Then we offline A before all the datas are synchronized from A to B,
obviously B doesn't have the latest data as A, but B is still marked
with In_sync flag.
So let's not call spare_active under the condition, otherwise B is
still showed with 'U' state which is not correct.
Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com> Signed-off-by: Song Liu <songliubraving@fb.com>
Then from kernel side, Manage_add invokes the path (add_new_disk ->
validate_super = super_1_validate) to set In_sync flag.
Since In_sync means "device is in_sync with rest of array", and the new
added disk need to resync thread to help the synchronization of data.
And md_reap_sync_thread would call spare_active to set In_sync for the
new added disk finally. So don't set In_sync if array is in frozen.
Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com> Signed-off-by: Song Liu <songliubraving@fb.com>
md: allow last device to be forcibly removed from RAID1/RAID10.
When the 'last' device in a RAID1 or RAID10 reports an error,
we do not mark it as failed. This would serve little purpose
as there is no risk of losing data beyond that which is obviously
lost (as there is with RAID5), and there could be other sectors
on the device which are readable, and only readable from this device.
This in general this maximises access to data.
However the current implementation also stops an admin from removing
the last device by direct action. This is rarely useful, but in many
case is not harmful and can make automation easier by removing special
cases.
Also, if an attempt to write metadata fails the device must be marked
as faulty, else an infinite loop will result, attempting to update
the metadata on all non-faulty devices.
So add 'fail_last_dev' member to 'struct mddev', then we can bypasses
the 'last disk' checks for RAID1 and RAID10, and control the behavior
per array by change sysfs node.
Signed-off-by: NeilBrown <neilb@suse.de>
[add sysfs node for fail_last_dev by Guoqing] Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com> Signed-off-by: Song Liu <songliubraving@fb.com>
When write bio return error, it would be added to conf->retry_list
and wait for raid1d thread to retry write and acknowledge badblocks.
In narrow_write_error(), the error bio will be split in the unit of
badblock shift (such as one sector) and raid1d thread issues them
one by one. Until all of the splited bio has finished, raid1d thread
can go on processing other things, which is time consuming.
But, there is a scene for error handling that is not necessary.
When the device has been set faulty, flush_bio_list() may end
bios in pending_bio_list with error status. Since these bios
has not been issued to the device actually, error handlding to
retry write and acknowledge badblocks make no sense.
Even without that scene, when the device is faulty, badblocks info
can not be written out to the device. Thus, we also no need to
handle the error IO.
Signed-off-by: Yufen Yu <yuyufen@huawei.com> Signed-off-by: Song Liu <songliubraving@fb.com>
Xiao Ni [Mon, 8 Jul 2019 02:14:32 +0000 (10:14 +0800)]
md/raid6: Set R5_ReadError when there is read failure on parity disk
7471fb77ce4d ("md/raid6: Fix anomily when recovering a single device in
RAID6.") avoids rereading P when it can be computed from other members.
However, this misses the chance to re-write the right data to P. This
patch sets R5_ReadError if the re-read fails.
Also, when re-read is skipped, we also missed the chance to reset
rdev->read_errors to 0. It can fail the disk when there are many read
errors on P member disk (other disks don't have read error)
V2: upper layer read request don't read parity/Q data. So there is no
need to consider such situation.
This is Reported-by: kbuild test robot <lkp@intel.com>
Fixes: 7471fb77ce4d ("md/raid6: Fix anomily when recovering a single device in RAID6.") Cc: <stable@vger.kernel.org> #4.4+ Signed-off-by: Xiao Ni <xni@redhat.com> Signed-off-by: Song Liu <songliubraving@fb.com>
Ming Lei [Thu, 25 Jul 2019 09:41:46 +0000 (17:41 +0800)]
blk-mq: balance mapping between present CPUs and queues
Spread queues among present CPUs first, then building mapping on other
non-present CPUs.
So we can minimize count of dead queues which are mapped by un-present
CPUs only. Then bad IO performance can be avoided by unbalanced mapping
between present CPUs and queues.
The similar policy has been applied on Managed IRQ affinity.
Cc: Yi Zhang <yi.zhang@redhat.com> Reported-by: Yi Zhang <yi.zhang@redhat.com> Reviewed-by: Bob Liu <bob.liu@oracle.com> Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Ming Lei [Thu, 25 Jul 2019 02:05:00 +0000 (10:05 +0800)]
scsi: implement .cleanup_rq callback
Implement .cleanup_rq() callback for freeing driver private part
of the request. Then we can avoid to leak this part if the request isn't
completed by SCSI, and freed by blk-mq or upper layer(such as dm-rq) finally.
Cc: Ewan D. Milne <emilne@redhat.com> Cc: Bart Van Assche <bvanassche@acm.org> Cc: Hannes Reinecke <hare@suse.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Mike Snitzer <snitzer@redhat.com> Cc: dm-devel@redhat.com Cc: <stable@vger.kernel.org> Fixes: 396eaf21ee17 ("blk-mq: improve DM's blk-mq IO merging via blk_insert_cloned_request feedback") Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Ming Lei [Thu, 25 Jul 2019 02:04:59 +0000 (10:04 +0800)]
blk-mq: add callback of .cleanup_rq
SCSI maintains its own driver private data hooked off of each SCSI
request, and the pridate data won't be freed after scsi_queue_rq()
returns BLK_STS_RESOURCE or BLK_STS_DEV_RESOURCE. An upper layer driver
(e.g. dm-rq) may need to retry these SCSI requests, before SCSI has
fully dispatched them, due to a lower level SCSI driver's resource
limitation identified in scsi_queue_rq(). Currently SCSI's per-request
private data is leaked when the upper layer driver (dm-rq) frees and
then retries these requests in response to BLK_STS_RESOURCE or
BLK_STS_DEV_RESOURCE returns from scsi_queue_rq().
This usecase is so specialized that it doesn't warrant training an
existing blk-mq interface (e.g. blk_mq_free_request) to allow SCSI to
account for freeing its driver private data -- doing so would add an
extra branch for handling a special case that all other consumers of
SCSI (and blk-mq) won't ever need to worry about.
So the most pragmatic way forward is to delegate freeing SCSI driver
private data to the upper layer driver (dm-rq). Do so by adding
new .cleanup_rq callback and calling a new blk_mq_cleanup_rq() method
from dm-rq. A following commit will implement the .cleanup_rq() hook
in scsi_mq_ops.
Cc: Ewan D. Milne <emilne@redhat.com> Cc: Bart Van Assche <bvanassche@acm.org> Cc: Hannes Reinecke <hare@suse.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Mike Snitzer <snitzer@redhat.com> Cc: dm-devel@redhat.com Cc: <stable@vger.kernel.org> Fixes: 396eaf21ee17 ("blk-mq: improve DM's blk-mq IO merging via blk_insert_cloned_request feedback") Signed-off-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch implements the zone reset all operation for sd_zbc.c. We add
a new boolean parameter for the sd_zbc_setup_reset_cmd() to indicate
REQ_OP_ZONE_RESET_ALL command setup. Along with that we add support in
the completion path for the zone reset all.
Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Acked-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
This implements REQ_OP_ZONE_RESET_ALL as a special case of the block
device zone reset operations where we just simply issue bio with the
newly introduced req op.
We issue this req op when the number of sectors is equal to the device's
partition's number of sectors and device has no partitions.
We also add support so that blk_op_str() can print the new reset-all
zone operation.
This patch also adds a generic make request check for newly
introduced REQ_OP_ZONE_RESET_ALL req_opf. We simply return error
when queue is zoned and reset-all flag is not set for
REQ_OP_ZONE_RESET_ALL.
Reviewed-by: Hannes Reinecke <hare@suse.com> Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
This patch introduces a new request operation REQ_OP_ZONE_RESET_ALL.
This is useful for the applications like mkfs where it needs to reset
all the zones present on the underlying block device. As part for this
patch we also introduce new QUEUE_FLAG_ZONE_RESETALL which indicates the
queue zone reset all capability and corresponding helper macro.
Reviewed-by: Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bart Van Assche [Thu, 1 Aug 2019 22:50:44 +0000 (15:50 -0700)]
block: Improve physical block alignment of split bios
Consider the following example:
* The logical block size is 4 KB.
* The physical block size is 8 KB.
* max_sectors equals (16 KB >> 9) sectors.
* A non-aligned 4 KB and an aligned 64 KB bio are merged into a single
non-aligned 68 KB bio.
The current behavior is to split such a bio into (16 KB + 16 KB + 16 KB
+ 16 KB + 4 KB). The start of none of these five bio's is aligned to a
physical block boundary.
This patch ensures that such a bio is split into four aligned and
one non-aligned bio instead of being split into five non-aligned bios.
This improves performance because most block devices can handle aligned
requests faster than non-aligned requests.
Since the physical block size is larger than or equal to the logical
block size, this patch preserves the guarantee that the returned
value is a multiple of the logical block size.
Cc: Christoph Hellwig <hch@infradead.org> Cc: Ming Lei <ming.lei@redhat.com> Cc: Hannes Reinecke <hare@suse.com> Signed-off-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bart Van Assche [Thu, 1 Aug 2019 22:50:43 +0000 (15:50 -0700)]
block: Simplify blk_bio_segment_split()
Move the max_sectors check into bvec_split_segs() such that a single
call to that function can do all the necessary checks. This patch
optimizes the fast path further, namely if a bvec fits in a page.
Cc: Christoph Hellwig <hch@lst.de> Cc: Ming Lei <ming.lei@redhat.com> Cc: Hannes Reinecke <hare@suse.com> Signed-off-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bart Van Assche [Thu, 1 Aug 2019 22:50:42 +0000 (15:50 -0700)]
block: Simplify bvec_split_segs()
Simplify this function by by removing two if-tests. Other than requiring
that the @sectors pointer is not NULL, this patch does not change the
behavior of bvec_split_segs().
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Cc: Christoph Hellwig <hch@lst.de> Cc: Ming Lei <ming.lei@redhat.com> Cc: Hannes Reinecke <hare@suse.com> Signed-off-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bart Van Assche [Thu, 1 Aug 2019 22:50:41 +0000 (15:50 -0700)]
block: Document the bio splitting functions
Since what the bio splitting functions do is nontrivial, document these
functions.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Cc: Christoph Hellwig <hch@infradead.org> Cc: Ming Lei <ming.lei@redhat.com> Cc: Hannes Reinecke <hare@suse.com> Signed-off-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bart Van Assche [Thu, 1 Aug 2019 22:50:40 +0000 (15:50 -0700)]
block: Declare several function pointer arguments 'const'
Make it clear to the compiler and also to humans that the functions
that query request queue properties do not modify any member of the
request_queue data structure.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Cc: Christoph Hellwig <hch@infradead.org> Cc: Ming Lei <ming.lei@redhat.com> Cc: Hannes Reinecke <hare@suse.com> Signed-off-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
projects / mirror_ubuntu-hirsute-kernel.git / log