The enum values for QPTYPE, PRTYPE and CMS are off by 1 from the
values defined in figure 42 of the NVM Express over Fabrics 1.0:
http://www.nvmexpress.org/wp-content/uploads/NVMe_over_Fabrics_1_0_Gold_20160605-1.pdf
Fix our enums to match the final spec.
Signed-off-by: Roland Dreier <roland@purestorage.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Constify all instances of blk_mq_ops, as they are never modified.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This adds a new module parameter to null_blk, blocking. If set, null_blk
will set the BLK_MQ_F_BLOCKING flag, indicating that it sometimes/always
needs to block in its ->queue_rq() function. The intent is to help find
regressions in blocking drivers, since not many of them exist.
If null_blk is loaded with submit_queues > 1 and blocking=1, this
shows the regression recently fixed by bf4907c05e.
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Commit a4d907b6a3 unified the single and multi queue request handlers,
but in the process, it also screwed up the locking balance and calls
blk_mq_try_issue_directly() with the ctx preempt lock held. This is a
problem for drivers that have set BLK_MQ_F_BLOCKING, since now they
can't reliably sleep.
While in there, protect against similar issues in the future, by adding
a might_sleep() trigger in the BLOCKING path for direct issue or queue
run.
Reported-by: Josef Bacik <josef@toxicpanda.com>
Tested-by: Josef Bacik <josef@toxicpanda.com>
Fixes: a4d907b6a3 ("blk-mq: streamline blk_mq_make_request")
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
trivial fix to spelling mistake in pr_err error message
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
In blk_mq_alloc_request_hctx, blk_mq_sched_get_request doesn't
get sw context so we don't need to put the context with
blk_mq_put_ctx. Unless, we will see preempt counter underflow.
Cc: Omar Sandoval <osandov@fb.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@fb.com>
In blk_mq_alloc_request_hctx, blk_mq_sched_get_request doesn't
get sw context so we don't need to put the context with
blk_mq_put_ctx. Unless, we will see preempt counter underflow.
Cc: Omar Sandoval <osandov@fb.com>
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@fb.com>
Pull thermal management fixes from Zhang Rui:
- Fix a potential deadlock in cpu_cooling driver, which was introduced
in 4.11-rc1. (Matthew Wilcox)
- Fix the cpu_cooling and devfreq_cooling code to handle possible error
return value from OPP calls, together with three minor fixes in the
same patch series. (Viresh Kumar)
* 'for-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux:
thermal: cpu_cooling: Check OPP for errors
thermal: cpu_cooling: Replace dev_warn with dev_err
thermal: devfreq: Check OPP for errors
thermal: devfreq_cooling: Replace dev_warn with dev_err
thermal: devfreq: Simplify expression
thermal: Fix potential deadlock in cpu_cooling
Pull block fixes from Jens Axboe:
"Five fixes for this series:
- a fix from me to ensure that blk-mq drivers that terminate IO in
their ->queue_rq() handler by returning QUEUE_ERROR don't stall
with a scheduler enabled.
- four nbd fixes from Josef and Ratna, fixing various problems that
are critical enough to go in for this cycle. They have been well
tested"
* 'for-linus' of git://git.kernel.dk/linux-block:
nbd: replace kill_bdev() with __invalidate_device()
nbd: set queue timeout properly
nbd: set rq->errors to actual error code
nbd: handle ERESTARTSYS properly
blk-mq: include errors in did_work calculation
Merge xfrm_user validation fixes from Andy Whitcroft:
"Two patches we are applying to Ubuntu for XFRM_MSG_NEWAE validation
issue reported by ZDI.
The first of these is the primary fix, and the second is for a more
theoretical issue that Kees pointed out when reviewing the first"
* emailed patches from Andy Whitcroft <apw@canonical.com>:
xfrm_user: validate XFRM_MSG_NEWAE incoming ESN size harder
xfrm_user: validate XFRM_MSG_NEWAE XFRMA_REPLAY_ESN_VAL replay_window
Currently we return true in blk_mq_dispatch_rq_list() if we queued IO
successfully, but we really want to return whether or not the we made
progress. Progress includes if we got an error return. If we don't,
this can lead to a hang in blk_mq_sched_dispatch_requests() when a
driver is draining IO by returning BLK_MQ_QUEUE_ERROR instead of
manually ending the IO in error and return BLK_MQ_QUEUE_OK.
Tested-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
When try to issue a request directly and we fail we will requeue the
request, but call blk_mq_end_request() as well. This leads to the
completed request being on a queuelist and getting ended twice, which
causes list corruption in schedulers and other shenanigans.
Signed-off-by: Josef Bacik <jbacik@fb.com>
Reviewed-by: Ming Lei <tom.leiming@gmail.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@fb.com>
blkg_conf_prep() currently calls blkg_lookup_create() while holding
request queue spinlock. This means allocating memory for struct
blkcg_gq has to be made non-blocking. This causes occasional -ENOMEM
failures in call paths like below:
pcpu_alloc+0x68f/0x710
__alloc_percpu_gfp+0xd/0x10
__percpu_counter_init+0x55/0xc0
cfq_pd_alloc+0x3b2/0x4e0
blkg_alloc+0x187/0x230
blkg_create+0x489/0x670
blkg_lookup_create+0x9a/0x230
blkg_conf_prep+0x1fb/0x240
__cfqg_set_weight_device.isra.105+0x5c/0x180
cfq_set_weight_on_dfl+0x69/0xc0
cgroup_file_write+0x39/0x1c0
kernfs_fop_write+0x13f/0x1d0
__vfs_write+0x23/0x120
vfs_write+0xc2/0x1f0
SyS_write+0x44/0xb0
entry_SYSCALL_64_fastpath+0x18/0xad
In the code path above, percpu allocator cannot call vmalloc() due to
queue spinlock.
A failure in this call path gives grief to tools which are trying to
configure io weights. We see occasional failures happen shortly after
reboots even when system is not under any memory pressure. Machines
with a lot of cpus are more vulnerable to this condition.
Do struct blkcg_gq allocations outside the queue spinlock to allow
blocking during memory allocations.
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tahsin Erdogan <tahsin@google.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@fb.com>
I inadvertently applied the v5 version of this patch, whereas
the agreed upon version was v5. Revert this one so we can apply
the right one.
This reverts commit 7fc6b87a9f.
Merge PTRACE_SETREGSET leakage fixes from Dave Martin:
"This series is the collection of fixes I proposed on this topic, that
have not yet appeared upstream or in the stable branches,
The issue can leak kernel stack, but doesn't appear to allow userspace
to attack the kernel directly. The affected architectures are c6x,
h8300, metag, mips and sparc.
[ Mark Salter points out that c6x has no MMU or other mechanism to
prevent userspace access to kernel code or data on c6x, but it
doesn't hurt to clean that case up too. ]
The bugs arise from use of user_regset_copyin(). Users of
user_regset_copyin() can work in one of two ways:
1) Copy directly to thread_struct or equivalent. (This seems to be
the design assumption of the regset API, and is the most common
approach.)
2) Copy to a local variable and then transfer to thread_struct. (A
significant minority of cases.)
Buggy code typically involves approach 2"
* emailed patches from Dave Martin <Dave.Martin@arm.com>:
sparc/ptrace: Preserve previous registers for short regset write
mips/ptrace: Preserve previous registers for short regset write
metag/ptrace: Reject partial NT_METAG_RPIPE writes
metag/ptrace: Provide default TXSTATUS for short NT_PRSTATUS
metag/ptrace: Preserve previous registers for short regset write
h8300/ptrace: Fix incorrect register transfer count
c6x/ptrace: Remove useless PTRACE_SETREGSET implementation
Ensure that if userspace supplies insufficient data to PTRACE_SETREGSET
to fill all the registers, the thread's old registers are preserved.
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ensure that if userspace supplies insufficient data to PTRACE_SETREGSET
to fill all the registers, the thread's old registers are preserved.
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's not clear what behaviour is sensible when doing partial write of
NT_METAG_RPIPE, so just don't bother.
This patch assumes that userspace will never rely on a partial SETREGSET
in this case, since it's not clear what should happen anyway.
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Acked-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ensure that if userspace supplies insufficient data to PTRACE_SETREGSET
to fill TXSTATUS, a well-defined default value is used, based on the
task's current value.
Suggested-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ensure that if userspace supplies insufficient data to PTRACE_SETREGSET
to fill all the registers, the thread's old registers are preserved.
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Acked-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
regs_set() and regs_get() are vulnerable to an off-by-1 buffer overrun
if CONFIG_CPU_H8S is set, since this adds an extra entry to
register_offset[] but not to user_regs_struct.
So, iterate over user_regs_struct based on its actual size, not based on
the length of register_offset[].
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
gpr_set won't work correctly and can never have been tested, and the
correct behaviour is not clear due to the endianness-dependent task
layout.
So, just remove it. The core code will now return -EOPNOTSUPPORT when
trying to set NT_PRSTATUS on this architecture until/unless a correct
implementation is supplied.
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Kees Cook has pointed out that xfrm_replay_state_esn_len() is subject to
wrapping issues. To ensure we are correctly ensuring that the two ESN
structures are the same size compare both the overall size as reported
by xfrm_replay_state_esn_len() and the internal length are the same.
CVE-2017-7184
Signed-off-by: Andy Whitcroft <apw@canonical.com>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When a new xfrm state is created during an XFRM_MSG_NEWSA call we
validate the user supplied replay_esn to ensure that the size is valid
and to ensure that the replay_window size is within the allocated
buffer. However later it is possible to update this replay_esn via a
XFRM_MSG_NEWAE call. There we again validate the size of the supplied
buffer matches the existing state and if so inject the contents. We do
not at this point check that the replay_window is within the allocated
memory. This leads to out-of-bounds reads and writes triggered by
netlink packets. This leads to memory corruption and the potential for
priviledge escalation.
We already attempt to validate the incoming replay information in
xfrm_new_ae() via xfrm_replay_verify_len(). This confirms that the user
is not trying to change the size of the replay state buffer which
includes the replay_esn. It however does not check the replay_window
remains within that buffer. Add validation of the contained
replay_window.
CVE-2017-7184
Signed-off-by: Andy Whitcroft <apw@canonical.com>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
CONFIG_DEBUG_TEST_DRIVER_REMOVE found a possible leak of q->rq_wb when a
request queue is reregistered. This has been a problem since wbt was
introduced, but the WARN_ON(!list_empty(&stats->callbacks)) in the
blk-stat rework exposed it. Fix it by cleaning up wbt when we unregister
the queue.
Fixes: 87760e5eef ("block: hook up writeback throttling")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
blk_alloc_queue_node() already allocates q->stats, so
blk_mq_init_allocated_queue() is overwriting it with a new allocation.
Fixes: a83b576c9c ("block: fix stacked driver stats init and free")
Reviewed-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Now that the remaining drivers have been converted to one request queue
per gendisk, let's warn if a request queue gets registered more than
once. This will catch future drivers which might do it inadvertently or
any old drivers that I may have missed.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Before commit 780db2071a(blk-mq: decouble blk-mq freezing
from generic bypassing), the dying flag is checked before
entering queue, and Tejun converts the checking into .mq_freeze_depth,
and assumes the counter is increased just after dying flag
is set. Unfortunately we doesn't do that in blk_set_queue_dying().
This patch calls blk_freeze_queue_start() in blk_set_queue_dying(),
so that we can block new I/O coming once the queue is set as dying.
Given blk_set_queue_dying() is always called in remove path
of block device, and queue will be cleaned up later, we don't
need to worry about undoing the counter.
Cc: Tejun Heo <tj@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
As the .q_usage_counter is used by both legacy and
mq path, we need to block new I/O if queue becomes
dead in blk_queue_enter().
So rename it and we can use this function in both
paths.
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
Without the barrier, reading DEAD flag of .q_usage_counter
and reading .mq_freeze_depth may be reordered, then the
following wait_event_interruptible() may never return.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
This patch adds comment on two races related with
timeout handler:
- requeue from queue busy vs. timeout
- rq free & reallocation vs. timeout
Both the races themselves and current solution aren't
explicit enough, so add comments on them.
Cc: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
blkg_conf_prep() currently calls blkg_lookup_create() while holding
request queue spinlock. This means allocating memory for struct
blkcg_gq has to be made non-blocking. This causes occasional -ENOMEM
failures in call paths like below:
pcpu_alloc+0x68f/0x710
__alloc_percpu_gfp+0xd/0x10
__percpu_counter_init+0x55/0xc0
cfq_pd_alloc+0x3b2/0x4e0
blkg_alloc+0x187/0x230
blkg_create+0x489/0x670
blkg_lookup_create+0x9a/0x230
blkg_conf_prep+0x1fb/0x240
__cfqg_set_weight_device.isra.105+0x5c/0x180
cfq_set_weight_on_dfl+0x69/0xc0
cgroup_file_write+0x39/0x1c0
kernfs_fop_write+0x13f/0x1d0
__vfs_write+0x23/0x120
vfs_write+0xc2/0x1f0
SyS_write+0x44/0xb0
entry_SYSCALL_64_fastpath+0x18/0xad
In the code path above, percpu allocator cannot call vmalloc() due to
queue spinlock.
A failure in this call path gives grief to tools which are trying to
configure io weights. We see occasional failures happen shortly after
reboots even when system is not under any memory pressure. Machines
with a lot of cpus are more vulnerable to this condition.
Update blkg_create() function to temporarily drop the rcu and queue
locks when it is allowed by gfp mask.
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Tahsin Erdogan <tahsin@google.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@fb.com>
Compile-tested only (by hacking it to compile on x86).
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Fixes to multiple issues in virtio. Most notably
a regression fix for crashes reported by Fedora users.
Hybernate is still reportedly broken, working on it.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJY2qEcAAoJECgfDbjSjVRpM/oH/3GPZOh9/tMzDFDaDljqtWQy
PGVb74/3+O55xOOq9nyyS3+6BlCXmiUcynxg61QUOUqUuHPPdH/OntyyPgG0pYkx
271W81C1yc2xFp/qkOiMWKiPmsbJ7ykVg37NWtxm7Phf4RgX3wgymq87hWr4Td1G
q9k6oyMCmvJUECJVxOVHjPt+oYQ7zQkFBNB8kSNlj67gbe533jkPt46MMlXbX7fQ
lPdJTnLXN/GQxnVtw5AAiWF87z0wNVUefrLe9sHW3KOeGBdne4NXblvz3WF/iPq4
N96thgm7QOP3NgAqbaUa7Fb0+jxyi2DNYFrVPxnf+nOOQy/AVUX6GRZJ2Tu6gF0=
=oSO5
-----END PGP SIGNATURE-----
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
Pull virtio fixes from Michael Tsirkin:
"Fixes to multiple issues in virtio.
Most notably a regression fix for crashes reported by Fedora users.
Hibernate is still reportedly broken, working on it"
* tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost:
virtio_balloon: prevent uninitialized variable use
virtio-balloon: use actual number of stats for stats queue buffers
virtio_balloon: init 1st buffer in stats vq
virtio_pci: fix out of bound access for msix_names
The latest gcc-7.0.1 snapshot reports a new warning:
virtio/virtio_balloon.c: In function 'update_balloon_stats':
virtio/virtio_balloon.c:258:26: error: 'events[2]' is used uninitialized in this function [-Werror=uninitialized]
virtio/virtio_balloon.c:260:26: error: 'events[3]' is used uninitialized in this function [-Werror=uninitialized]
virtio/virtio_balloon.c:261:56: error: 'events[18]' is used uninitialized in this function [-Werror=uninitialized]
virtio/virtio_balloon.c:262:56: error: 'events[17]' is used uninitialized in this function [-Werror=uninitialized]
This seems absolutely right, so we should add an extra check to
prevent copying uninitialized stack data into the statistics.
>From all I can tell, this has been broken since the statistics code
was originally added in 2.6.34.
Fixes: 9564e138b1 ("virtio: Add memory statistics reporting to the balloon driver (V4)")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The virtio balloon driver contained a not-so-obvious invariant that
update_balloon_stats has to update exactly VIRTIO_BALLOON_S_NR counters
in order to send valid stats to the host. This commit fixes it by having
update_balloon_stats return the actual number of counters, and its
callers use it when pushing buffers to the stats virtqueue.
Note that it is still out of spec to change the number of counters
at run-time. "Driver MUST supply the same subset of statistics in all
buffers submitted to the statsq."
Suggested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When init_vqs runs, virtio_balloon.stats is either uninitialized or
contains stale values. The host updates its state with garbage data
because it has no way of knowing that this is just a marker buffer
used for signaling.
This patch updates the stats before pushing the initial buffer.
Alternative fixes:
* Push an empty buffer in init_vqs. Not easily done with the current
virtio implementation and violates the spec "Driver MUST supply the
same subset of statistics in all buffers submitted to the statsq".
* Push a buffer with invalid tags in init_vqs. Violates the same
spec clause, plus "invalid tag" is not really defined.
Note: the spec says:
When using the legacy interface, the device SHOULD ignore all values in
the first buffer in the statsq supplied by the driver after device
initialization. Note: Historically, drivers supplied an uninitialized
buffer in the first buffer.
Unfortunately QEMU does not seem to implement the recommendation
even for the legacy interface.
Cc: stable@vger.kernel.org
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Fedora has received multiple reports of crashes when running
4.11 as a guest
https://bugzilla.redhat.com/show_bug.cgi?id=1430297https://bugzilla.redhat.com/show_bug.cgi?id=1434462https://bugzilla.kernel.org/show_bug.cgi?id=194911https://bugzilla.redhat.com/show_bug.cgi?id=1433899
The crashes are not always consistent but they are generally
some flavor of oops or GPF in virtio related code. Multiple people
have done bisections (Thank you Thorsten Leemhuis and
Richard W.M. Jones) and found this commit to be at fault
07ec51480b is the first bad commit
commit 07ec51480b
Author: Christoph Hellwig <hch@lst.de>
Date: Sun Feb 5 18:15:19 2017 +0100
virtio_pci: use shared interrupts for virtqueues
The issue seems to be an out of bounds access to the msix_names
array corrupting kernel memory.
Fixes: 07ec51480b ("virtio_pci: use shared interrupts for virtqueues")
Reported-by: Laura Abbott <labbott@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Richard W.M. Jones <rjones@redhat.com>
Tested-by: Thorsten Leemhuis <linux@leemhuis.info>
One hard problem adding .low limit is to detect idle cgroup. If one
cgroup doesn't dispatch enough IO against its low limit, we must have a
mechanism to determine if other cgroups dispatch more IO. We added the
think time detection mechanism before, but it doesn't work for all
workloads. Here we add a latency based approach.
We already have mechanism to calculate latency threshold for each IO
size. For every IO dispatched from a cgorup, we compare its latency
against its threshold and record the info. If most IO latency is below
threshold (in the code I use 75%), the cgroup could be treated idle and
other cgroups can dispatch more IO.
Currently this latency target check is only for SSD as we can't
calcualte the latency target for hard disk. And this is only for cgroup
leaf node so far.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
User configures latency target, but the latency threshold for each
request size isn't fixed. For a SSD, the IO latency highly depends on
request size. To calculate latency threshold, we sample some data, eg,
average latency for request size 4k, 8k, 16k, 32k .. 1M. The latency
threshold of each request size will be the sample latency (I'll call it
base latency) plus latency target. For example, the base latency for
request size 4k is 80us and user configures latency target 60us. The 4k
latency threshold will be 80 + 60 = 140us.
To sample data, we calculate the order base 2 of rounded up IO sectors.
If the IO size is bigger than 1M, it will be accounted as 1M. Since the
calculation does round up, the base latency will be slightly smaller
than actual value. Also if there isn't any IO dispatched for a specific
IO size, we will use the base latency of smaller IO size for this IO
size.
But we shouldn't sample data at any time. The base latency is supposed
to be latency where disk isn't congested, because we use latency
threshold to schedule IOs between cgroups. If disk is congested, the
latency is higher, using it for scheduling is meaningless. Hence we only
do the sampling when block throttling is in the LOW limit, with
assumption disk isn't congested in such state. If the assumption isn't
true, eg, low limit is too high, calculated latency threshold will be
higher.
Hard disk is completely different. Latency depends on spindle seek
instead of request size. Currently this feature is SSD only, we probably
can use a fixed threshold like 4ms for hard disk though.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
