Queue deletion is done asynchronous when the last reference on the queue
is dropped. Thus, in order to make sure we don't over allocate under a
connect/disconnect storm, we let queue deletion complete before making
forward progress.
However, given that we flush the system_wq from rdma_cm context which
runs from a workqueue context, we can have a circular locking complaint
[1]. Fix that by using a private workqueue for queue deletion.
[1]:
======================================================
WARNING: possible circular locking dependency detected
4.19.0-rc4-dbg+ #3 Not tainted
------------------------------------------------------
kworker/5:0/39 is trying to acquire lock:
00000000a10b6db9 (&id_priv->handler_mutex){+.+.}, at: rdma_destroy_id+0x6f/0x440 [rdma_cm]
but task is already holding lock:
00000000331b4e2c ((work_completion)(&queue->release_work)){+.+.}, at: process_one_work+0x3ed/0xa20
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #3 ((work_completion)(&queue->release_work)){+.+.}:
process_one_work+0x474/0xa20
worker_thread+0x63/0x5a0
kthread+0x1cf/0x1f0
ret_from_fork+0x24/0x30
-> #2 ((wq_completion)"events"){+.+.}:
flush_workqueue+0xf3/0x970
nvmet_rdma_cm_handler+0x133d/0x1734 [nvmet_rdma]
cma_ib_req_handler+0x72f/0xf90 [rdma_cm]
cm_process_work+0x2e/0x110 [ib_cm]
cm_req_handler+0x135b/0x1c30 [ib_cm]
cm_work_handler+0x2b7/0x38cd [ib_cm]
process_one_work+0x4ae/0xa20
nvmet_rdma:nvmet_rdma_cm_handler: nvmet_rdma: disconnected (10): status 0 id 0000000040357082
worker_thread+0x63/0x5a0
kthread+0x1cf/0x1f0
ret_from_fork+0x24/0x30
nvme nvme0: Reconnecting in 10 seconds...
-> #1 (&id_priv->handler_mutex/1){+.+.}:
__mutex_lock+0xfe/0xbe0
mutex_lock_nested+0x1b/0x20
cma_ib_req_handler+0x6aa/0xf90 [rdma_cm]
cm_process_work+0x2e/0x110 [ib_cm]
cm_req_handler+0x135b/0x1c30 [ib_cm]
cm_work_handler+0x2b7/0x38cd [ib_cm]
process_one_work+0x4ae/0xa20
worker_thread+0x63/0x5a0
kthread+0x1cf/0x1f0
ret_from_fork+0x24/0x30
-> #0 (&id_priv->handler_mutex){+.+.}:
lock_acquire+0xc5/0x200
__mutex_lock+0xfe/0xbe0
mutex_lock_nested+0x1b/0x20
rdma_destroy_id+0x6f/0x440 [rdma_cm]
nvmet_rdma_release_queue_work+0x8e/0x1b0 [nvmet_rdma]
process_one_work+0x4ae/0xa20
worker_thread+0x63/0x5a0
kthread+0x1cf/0x1f0
ret_from_fork+0x24/0x30
Fixes: 777dc82395 ("nvmet-rdma: occasionally flush ongoing controller teardown")
Reported-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Tested-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Make current_path an array with an entry for every possible node, and
cache the best path on a per-node basis. Take the node distance into
account when selecting it. This is primarily useful for dual-ported PCIe
devices which are connected to PCIe root ports on different sockets.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Reviewed-by: Hannes Reinecke <hare@suse.com>
If we know that the I/O size exceeds our inline bio vec, no
point using it and split the rest to begin with. We could
in theory reuse the inline bio and only allocate the bio_vec,
but its really not worth optimizing for.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When an io is rejected by nvmf_check_ready() due to validation of the
controller state, the nvmf_fail_nonready_command() will normally return
BLK_STS_RESOURCE to requeue and retry. However, if the controller is
dying or the I/O is marked for NVMe multipath, the I/O is failed so that
the controller can terminate or so that the io can be issued on a
different path. Unfortunately, as this reject point is before the
transport has accepted the command, blk-mq ends up completing the I/O
and never calls nvme_complete_rq(), which is where multipath may preserve
or re-route the I/O. The end result is, the device user ends up seeing an
EIO error.
Example: single path connectivity, controller is under load, and a reset
is induced. An I/O is received:
a) while the reset state has been set but the queues have yet to be
stopped; or
b) after queues are started (at end of reset) but before the reconnect
has completed.
The I/O finishes with an EIO status.
This patch makes the following changes:
- Adds the HOST_PATH_ERROR pathing status from TP4028
- Modifies the reject point such that it appears to queue successfully,
but actually completes the io with the new pathing status and calls
nvme_complete_rq().
- nvme_complete_rq() recognizes the new status, avoids resetting the
controller (likely was already done in order to get this new status),
and calls the multipather to clear the current path that errored.
This allows the next command (retry or new command) to select a new
path if there is one.
Signed-off-by: James Smart <jsmart2021@gmail.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
This patch adds a new event for nvme async event notification.
We print the async event in the decoded format when we recognize
the event otherwise we just dump the result.
Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
The fc transport device should allow for a rediscovery, as userspace
might have lost the events. Example is udev events not handled during
system startup.
This patch add a sysfs entry 'nvme_discovery' on the fc class to
have it replay all udev discovery events for all local port/remote
port address pairs.
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Currently, if a targetport has been connected to via the nvmet config
(in other words, the add_port() transport routine called, and the nvmet
port pointer stored for using in upcalls on new io), and if the
targetport is then removed (say the lldd driver decides to unload or
fully reset its hardware) and then re-added (the lldd driver reloads or
reinits its hardware), the port pointer has been lost so there's no way
to continue to post commands up to nvmet via the transport port.
Correct by allocating a small "port context" structure that will be
linked to by the targetport. The context will save the targetport WWN's
and the nvmet port pointer to use for it. Initial allocation will occur
when the targetport is bound to via add_port. The context will be
deallocated when remove_port() is called. If a targetport is removed
while nvmet has the active port context, the targetport will be unlinked
from the port context before removal. If a new targetport is registered,
the port contexts without a binding are looked through and if the WWN's
match (so it's the same as nvmet's port context) the port context is
linked to the new target port. Thus new io can be received on the new
targetport and operation resumes with nvmet.
Additionally, this also resolves nvmet configuration changing out from
underneath of the nvme-fc target port (for example: a nvmetcli clear).
Signed-off-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
This patch removes the redundant module prefix used in the pr_err() when
nvmet_get_smart_log_nsid() failed to find the namespace provided as a part
of smart-log command.
Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Merge -rc6 in, for two reasons:
1) Resolve a trivial conflict in the blk-mq-tag.c documentation
2) A few important regression fixes went into upstream directly, so
they aren't in the 4.20 branch.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
* tag 'v4.19-rc6': (780 commits)
Linux 4.19-rc6
MAINTAINERS: fix reference to moved drivers/{misc => auxdisplay}/panel.c
cpufreq: qcom-kryo: Fix section annotations
perf/core: Add sanity check to deal with pinned event failure
xen/blkfront: correct purging of persistent grants
Revert "xen/blkfront: When purging persistent grants, keep them in the buffer"
selftests/powerpc: Fix Makefiles for headers_install change
blk-mq: I/O and timer unplugs are inverted in blktrace
dax: Fix deadlock in dax_lock_mapping_entry()
x86/boot: Fix kexec booting failure in the SEV bit detection code
bcache: add separate workqueue for journal_write to avoid deadlock
drm/amd/display: Fix Edid emulation for linux
drm/amd/display: Fix Vega10 lightup on S3 resume
drm/amdgpu: Fix vce work queue was not cancelled when suspend
Revert "drm/panel: Add device_link from panel device to DRM device"
xen/blkfront: When purging persistent grants, keep them in the buffer
clocksource/drivers/timer-atmel-pit: Properly handle error cases
block: fix deadline elevator drain for zoned block devices
ACPI / hotplug / PCI: Don't scan for non-hotplug bridges if slot is not bridge
drm/syncobj: Don't leak fences when WAIT_FOR_SUBMIT is set
...
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Miguel writes:
"A trivial fix for auxdisplay
- MAINTAINERS reference fix for moved file
Reported by Joe Perches"
* tag 'auxdisplay-for-greg-v4.19-rc6' of https://github.com/ojeda/linux:
MAINTAINERS: fix reference to moved drivers/{misc => auxdisplay}/panel.c
Dan writes:
"filesystem-dax for 4.19-rc6
Fix a deadlock in the new for 4.19 dax_lock_mapping_entry() routine."
* tag 'libnvdimm-fixes2-4.19-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
dax: Fix deadlock in dax_lock_mapping_entry()
Jens writes:
"Block fixes for 4.19-rc6
A set of fixes that should go into this release. This pull request
contains:
- A fix (hopefully) for the persistent grants for xen-blkfront. A
previous fix from this series wasn't complete, hence reverted, and
this one should hopefully be it. (Boris Ostrovsky)
- Fix for an elevator drain warning with SMR devices, which is
triggered when you switch schedulers (Damien)
- bcache deadlock fix (Guoju Fang)
- Fix for the block unplug tracepoint, which has had the
timer/explicit flag reverted since 4.11 (Ilya)
- Fix a regression in this series where the blk-mq timeout hook is
invoked with the RCU read lock held, hence preventing it from
blocking (Keith)
- NVMe pull from Christoph, with a single multipath fix (Susobhan Dey)"
* tag 'for-linus-20180929' of git://git.kernel.dk/linux-block:
xen/blkfront: correct purging of persistent grants
Revert "xen/blkfront: When purging persistent grants, keep them in the buffer"
blk-mq: I/O and timer unplugs are inverted in blktrace
bcache: add separate workqueue for journal_write to avoid deadlock
xen/blkfront: When purging persistent grants, keep them in the buffer
block: fix deadline elevator drain for zoned block devices
blk-mq: Allow blocking queue tag iter callbacks
nvme: properly propagate errors in nvme_mpath_init
Thomas writes:
"A single fix for the AMD memory encryption boot code so it does not
read random garbage instead of the cached encryption bit when a kexec
kernel is allocated above the 32bit address limit."
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/boot: Fix kexec booting failure in the SEV bit detection code
Thomas writes:
"Three small fixes for clocksource drivers:
- Proper error handling in the Atmel PIT driver
- Add CLOCK_SOURCE_SUSPEND_NONSTOP for TI SoCs so suspend works again
- Fix the next event function for Facebook Backpack-CMM BMC chips so
usleep(100) doesnt sleep several milliseconds"
* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
clocksource/drivers/timer-atmel-pit: Properly handle error cases
clocksource/drivers/fttmr010: Fix set_next_event handler
clocksource/drivers/ti-32k: Add CLOCK_SOURCE_SUSPEND_NONSTOP flag for non-am43 SoCs
Thomas writes:
"A single fix for a missing sanity check when a pinned event is tried
to be read on the wrong CPU due to a legit event scheduling failure."
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf/core: Add sanity check to deal with pinned event failure
Rafael writes:
"Power management fix for 4.19-rc6
Fix incorrect __init and __exit annotations in the Qualcomm
Kryo cpufreq driver (Nathan Chancellor)."
* tag 'pm-4.19-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
cpufreq: qcom-kryo: Fix section annotations
There is currently a warning when building the Kryo cpufreq driver into
the kernel image:
WARNING: vmlinux.o(.text+0x8aa424): Section mismatch in reference from
the function qcom_cpufreq_kryo_probe() to the function
.init.text:qcom_cpufreq_kryo_get_msm_id()
The function qcom_cpufreq_kryo_probe() references
the function __init qcom_cpufreq_kryo_get_msm_id().
This is often because qcom_cpufreq_kryo_probe lacks a __init
annotation or the annotation of qcom_cpufreq_kryo_get_msm_id is wrong.
Remove the '__init' annotation from qcom_cpufreq_kryo_get_msm_id
so that there is no more mismatch warning.
Additionally, Nick noticed that the remove function was marked as
'__init' when it should really be marked as '__exit'.
Fixes: 46e2856b8e (cpufreq: Add Kryo CPU scaling driver)
Fixes: 5ad7346b4a (cpufreq: kryo: Add module remove and exit)
Reported-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: 4.18+ <stable@vger.kernel.org> # 4.18+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Christoph writes:
"dma mapping fix for 4.19-rc6
fix a missing Kconfig symbol for commits introduced in 4.19-rc"
* tag 'dma-mapping-4.19-3' of git://git.infradead.org/users/hch/dma-mapping:
dma-mapping: add the missing ARCH_HAS_SYNC_DMA_FOR_CPU_ALL declaration
Dmitry writes:
"Input updates for v4.19-rc5
Just a few driver fixes"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
Input: uinput - allow for max == min during input_absinfo validation
Input: elantech - enable middle button of touchpad on ThinkPad P72
Input: atakbd - fix Atari CapsLock behaviour
Input: atakbd - fix Atari keymap
Input: egalax_ts - add system wakeup support
Input: gpio-keys - fix a documentation index issue
Mark writes:
"spi: Fixes for v4.19
Quite a few fixes for the Renesas drivers in here, plus a fix for the
Tegra driver and some documentation fixes for the recently added
spi-mem code. The Tegra fix is relatively large but fairly
straightforward and mechanical, it runs on probe so it's been
reasonably well covered in -next testing."
* tag 'spi-fix-v4.19-rc5' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
spi: spi-mem: Move the DMA-able constraint doc to the kerneldoc header
spi: spi-mem: Add missing description for data.nbytes field
spi: rspi: Fix interrupted DMA transfers
spi: rspi: Fix invalid SPI use during system suspend
spi: sh-msiof: Fix handling of write value for SISTR register
spi: sh-msiof: Fix invalid SPI use during system suspend
spi: gpio: Fix copy-and-paste error
spi: tegra20-slink: explicitly enable/disable clock
Mark writes:
"regulator: Fixes for 4.19
A collection of fairly minor bug fixes here, a couple of driver
specific ones plus two core fixes. There's one fix for the new
suspend state code which fixes some confusion with constant values
that are supposed to indicate noop operation and another fixing a
race condition with the creation of sysfs files on new regulators."
* tag 'regulator-v4.19-rc5' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
regulator: fix crash caused by null driver data
regulator: Fix 'do-nothing' value for regulators without suspend state
regulator: da9063: fix DT probing with constraints
regulator: bd71837: Disable voltage monitoring for LDO3/4
Michael writes:
"powerpc fixes for 4.19 #3
A reasonably big batch of fixes due to me being away for a few weeks.
A fix for the TM emulation support on Power9, which could result in
corrupting the guest r11 when running under KVM.
Two fixes to the TM code which could lead to userspace GPR corruption
if we take an SLB miss at exactly the wrong time.
Our dynamic patching code had a bug that meant we could patch freed
__init text, which could lead to corrupting userspace memory.
csum_ipv6_magic() didn't work on little endian platforms since we
optimised it recently.
A fix for an endian bug when reading a device tree property telling
us how many storage keys the machine has available.
Fix a crash seen on some configurations of PowerVM when migrating the
partition from one machine to another.
A fix for a regression in the setup of our CPU to NUMA node mapping
in KVM guests.
A fix to our selftest Makefiles to make them work since a recent
change to the shared Makefile logic."
* tag 'powerpc-4.19-3' of https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
selftests/powerpc: Fix Makefiles for headers_install change
powerpc/numa: Use associativity if VPHN hcall is successful
powerpc/tm: Avoid possible userspace r1 corruption on reclaim
powerpc/tm: Fix userspace r13 corruption
powerpc/pseries: Fix unitialized timer reset on migration
powerpc/pkeys: Fix reading of ibm, processor-storage-keys property
powerpc: fix csum_ipv6_magic() on little endian platforms
powerpc/powernv/ioda2: Reduce upper limit for DMA window size (again)
powerpc: Avoid code patching freed init sections
KVM: PPC: Book3S HV: Fix guest r11 corruption with POWER9 TM workarounds
We apply a smoothing to the scale changes in order to keep sawtoothy
behavior from occurring. However our window for checking if we've
missed our target can sometimes be lower than the smoothing interval
(500ms), especially on faster drives like ssd's. In order to deal with
this keep track of the running tally of the previous intervals that we
threw away because we had already done a scale event recently.
This is needed for the ssd case as these low latency drives will have
bursts of latency, and if it happens to be ok for the window that
directly follows the opening of the scale window we could unthrottle
when previous windows we were missing our target.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We use an average latency approach for determining if we're missing our
latency target. This works well for rotational storage where we have
generally consistent latencies, but for ssd's and other low latency
devices you have more of a spikey behavior, which means we often won't
throttle misbehaving groups because a lot of IO completes at drastically
faster times than our latency target. Instead keep track of how many
IO's miss our target and how many IO's are done in our time window. If
the p(90) latency is above our target then we know we need to throttle.
With this change in place we are seeing the same throttling behavior
with our testcase on ssd's as we see with rotational drives.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is logic to keep cgroups that haven't done a lot of IO in the most
recent scale window from being punished for over-active higher priority
groups. However for things like ssd's where the windows are pretty
short we'll end up with small numbers of samples, so 5% of samples will
come out to 0 if there aren't enough. Make the floor 1 sample to keep
us from improperly bailing out of scaling down.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Hitting the case where blk_queue_depth() returned 1 uncovered the fact
that iolatency doesn't actually handle this case properly, it simply
doesn't scale down anybody. For this case we should go straight into
applying the time delay, which we weren't doing. Since we already limit
the floor at 1 request this if statement is not needed, and this allows
us to set our depth to 1 which allows us to apply the delay if needed.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We were using blk_queue_depth() assuming that it would return
nr_requests, but we hit a case in production on drives that had to have
NCQ turned off in order for them to not shit the bed which resulted in a
qd of 1, even though the nr_requests was much larger. iolatency really
only cares about requests we are allowed to queue up, as any io that
get's onto the request list is going to be serviced soonish, so we want
to be throttling before the bio gets onto the request list. To make
iolatency work as expected, simply use q->nr_requests instead of
blk_queue_depth() as that is what we actually care about.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Dave writes:
"drm fixes for 4.19-rc6
Looks like a pretty normal week for graphics,
core: syncobj fix, panel link regression revert
amd: suspend/resume fixes, EDID emulation fix
mali-dp: NV12 writeback and vblank reset fixes
etnaviv: DMA setup fix"
* tag 'drm-fixes-2018-09-28' of git://anongit.freedesktop.org/drm/drm:
drm/amd/display: Fix Edid emulation for linux
drm/amd/display: Fix Vega10 lightup on S3 resume
drm/amdgpu: Fix vce work queue was not cancelled when suspend
Revert "drm/panel: Add device_link from panel device to DRM device"
drm/syncobj: Don't leak fences when WAIT_FOR_SUBMIT is set
drm/malidp: Fix writeback in NV12
drm: mali-dp: Call drm_crtc_vblank_reset on device init
drm/etnaviv: add DMA configuration for etnaviv platform device
Palmer writes:
"A Single RISC-V Update for 4.19-rc6
The Debian guys have been pushing on our port and found some
unversioned symbols leaking into modules. This PR contains a single
fix for that issue."
* tag 'riscv-for-linus-4.19-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/palmer/riscv-linux:
RISC-V: include linux/ftrace.h in asm-prototypes.h
NSEC_PER_SEC has type long, so 5 * NSEC_PER_SEC is calculated as a long.
However, 5 seconds is 5,000,000,000 nanoseconds, which overflows a
32-bit long. Make sure all of the targets are calculated as 64-bit
values.
Fixes: 6e25cb01ea ("kyber: implement improved heuristics")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit a46b53672b ("xen/blkfront: cleanup
stale persistent grants") introduced a regression as purged persistent
grants were not pu into the list of free grants again. Correct that.
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Use new-style DEVICE_ATTR_RO/DEVICE_ATTR_RW to create the sysfs attributes
and register the disk with default sysfs attribute groups.
Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Register default sysfs groups during device_add_disk() to avoid a
race condition with udev during startup.
Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Ed L. Cachin <ed.cashin@acm.org>
Reviewed-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We should be registering the ns_id attribute as default sysfs
attribute groups, otherwise we have a race condition between
the uevent and the attributes appearing in sysfs.
Suggested-by: Bart van Assche <bvanassche@acm.org>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Update device_add_disk() to take an 'groups' argument so that
individual drivers can register a device with additional sysfs
attributes.
This avoids race condition the driver would otherwise have if these
groups were to be created with sysfs_add_groups().
Signed-off-by: Martin Wilck <martin.wilck@suse.com>
Signed-off-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit b2d35fa5fc ("selftests: add headers_install to lib.mk")
introduced a requirement that Makefiles more than one level below the
selftests directory need to define top_srcdir, but it didn't update
any of the powerpc Makefiles.
This broke building all the powerpc selftests with eg:
make[1]: Entering directory '/src/linux/tools/testing/selftests/powerpc'
BUILD_TARGET=/src/linux/tools/testing/selftests/powerpc/alignment; mkdir -p $BUILD_TARGET; make OUTPUT=$BUILD_TARGET -k -C alignment all
make[2]: Entering directory '/src/linux/tools/testing/selftests/powerpc/alignment'
../../lib.mk:20: ../../../../scripts/subarch.include: No such file or directory
make[2]: *** No rule to make target '../../../../scripts/subarch.include'.
make[2]: Failed to remake makefile '../../../../scripts/subarch.include'.
Makefile:38: recipe for target 'alignment' failed
Fix it by setting top_srcdir in the affected Makefiles.
Fixes: b2d35fa5fc ("selftests: add headers_install to lib.mk")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
When debugging Kyber, it's really useful to know what latencies we've
been having, how the domain depths have been adjusted, and if we've
actually been throttling. Add three tracepoints, kyber_latency,
kyber_adjust, and kyber_throttled, to record that.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Kyber's current heuristics have a few flaws:
- It's based on the mean latency, but p99 latency tends to be more
meaningful to anyone who cares about latency. The mean can also be
skewed by rare outliers that the scheduler can't do anything about.
- The statistics calculations are purely time-based with a short window.
This works for steady, high load, but is more sensitive to outliers
with bursty workloads.
- It only considers the latency once an I/O has been submitted to the
device, but the user cares about the time spent in the kernel, as
well.
These are shortcomings of the generic blk-stat code which doesn't quite
fit the ideal use case for Kyber. So, this replaces the statistics with
a histogram used to calculate percentiles of total latency and I/O
latency, which we then use to adjust depths in a slightly more
intelligent manner:
- Sync and async writes are now the same domain.
- Discards are a separate domain.
- Domain queue depths are scaled by the ratio of the p99 total latency
to the target latency (e.g., if the p99 latency is double the target
latency, we will double the queue depth; if the p99 latency is half of
the target latency, we can halve the queue depth).
- We use the I/O latency to determine whether we should scale queue
depths down: we will only scale down if any domain's I/O latency
exceeds the target latency, which is an indicator of congestion in the
device.
These new heuristics are just as scalable as the heuristics they
replace.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The domain token sbitmaps are currently initialized to the device queue
depth or 256, whichever is larger, and immediately resized to the
maximum depth for that domain (256, 128, or 64 for read, write, and
other, respectively). The sbitmap is never resized larger than that, so
it's unnecessary to allocate a bitmap larger than the maximum depth.
Let's just allocate it to the maximum depth to begin with. This will use
marginally less memory, and more importantly, give us a more appropriate
number of bits per sbitmap word.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Kyber will need this in a future change if it is built as a module.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
