04 Nov, 2009
2 commits
-
CFQ has an optimization for cooperated applications. if several
io-context have close requests, they will get boost. But the
optimization get abused. Considering thread a, b, which work on one
file. a reads sectors s, s+2, s+4, ...; b reads sectors s+1, s+3, s
+5, ... Both a and b are sequential read, so they can open idle window.
a reads a sector s and goes to idle window and wakeup b. b reads sector
s+1, since in current implementation, cfq_should_preempt() thinks a and
b are cooperators, b will preempt a. b then reads sector s+1 and goes to
idle window and wakeup a. for the same reason, a will preempt b and
reads s+2. a and b will continue the circle. The circle will be very
long, and a and b will occupy whole disk queue. Other applications will
nearly have no chance to run.Fix this limiting coop preempt until a queue is scheduled normally
again.Signed-off-by: Shaohua Li
Acked-by: Jeff Moyer
Signed-off-by: Jens Axboe -
Commit a6151c3a5c8e1ff5a28450bc8d6a99a2a0add0a7 inadvertently reversed
a preempt condition check, potentially causing a performance regression.
Make the meta check correct again.Signed-off-by: Jens Axboe
24 Oct, 2009
1 commit
-
With 2.6.32-rc5 in a KVM guest using dm and virtio_blk, we see the
following errors:end_request: I/O error, dev vda, sector 0
end_request: I/O error, dev vda, sector 0The errors go away if dm stops submitting empty barriers, by reverting:
commit 52b1fd5a27c625c78373e024bf570af3c9d44a79
Author: Mikulas Patocka
dm: send empty barriers to targets in dm_flushWe should silently error all barriers, even empty barriers, on devices
like virtio_blk which don't support them.See also:
https://bugzilla.redhat.com/514901
Signed-off-by: Mark McLoughlin
Signed-off-by: Mike Snitzer
Acked-by: Alasdair G Kergon
Acked-by: Mikulas Patocka
Cc: Rusty Russell
Cc: Neil Brown
Cc: Christoph Hellwig
Signed-off-by: Jens Axboe
12 Oct, 2009
1 commit
-
Fix kernel-doc notation in blk-settings.c::blk_queue_max_discard_sectors().
Signed-off-by: Randy Dunlap
Signed-off-by: Jens Axboe
09 Oct, 2009
1 commit
-
elv_iosched_store() ignore the return value of strstrip(). It makes small
inconsistent behavior.This patch fixes it.
====================================
# cd /sys/block/{blockdev}/queuecase1:
# echo "anticipatory" > scheduler
# cat scheduler
noop [anticipatory] deadline cfqcase2:
# echo "anticipatory " > scheduler
# cat scheduler
noop [anticipatory] deadline cfqcase3:
# echo " anticipatory" > scheduler
bash: echo: write error: Invalid argument
====================================
# cd /sys/block/{blockdev}/queuecase1:
# echo "anticipatory" > scheduler
# cat scheduler
noop [anticipatory] deadline cfqcase2:
# echo "anticipatory " > scheduler
# cat scheduler
noop [anticipatory] deadline cfqcase3:
# echo " anticipatory" > scheduler
noop [anticipatory] deadline cfqCc: Li Zefan
Cc: Jens Axboe
Signed-off-by: KOSAKI Motohiro
Signed-off-by: Jens Axboe
08 Oct, 2009
3 commits
-
If the average think time is larger than the remaining time slice
for any given queue, don't allow it to idle. A succesful idle also
means that we need to dispatch and complete a request, so if we don't
even have time left for the idle process, we would overrun the slice
in any case.Signed-off-by: Corrado Zoccolo
Signed-off-by: Jens Axboe -
Saves 16 bytes of text, woohoo. But the more important point is
that it makes the code more readable when returning bool for 0/1
cases.Signed-off-by: Jens Axboe
-
CFQ enables idle only for processes that think less than the allowed
idle time. Since idle time is lower for seeky queues, we should use the
correct value in the comparison.Signed-off-by: Corrado Zoccolo
Signed-off-by: Jens Axboe
07 Oct, 2009
4 commits
-
We should subtract the slice residual from the rb tree key, since
a negative residual count indicates that the cfqq overran its slice
the last time. Hence we want to add the overrun time, to position
it a bit further away in the service tree.Reported-by: Corrado Zoccolo
Signed-off-by: Jens Axboe -
Makes the whole thing easier to read, cfq_dispatch_requests() was
a bit messy before.Signed-off-by: Jens Axboe
-
Makes it easier to read than the 0.
Signed-off-by: Jens Axboe
-
Commit a9327cac440be4d8333bba975cbbf76045096275 added seperate read
and write statistics of in_flight requests. And exported the number
of read and write requests in progress seperately through sysfs.But Corrado Zoccolo reported getting strange
output from "iostat -kx 2". Global values for service time and
utilization were garbage. For interval values, utilization was always
100%, and service time is higher than normal.So this was reverted by commit 0f78ab9899e9d6acb09d5465def618704255963b
The problem was in part_round_stats_single(), I missed the following:
if (now == part->stamp)
return;- if (part->in_flight) {
+ if (part_in_flight(part)) {
__part_stat_add(cpu, part, time_in_queue,
part_in_flight(part) * (now - part->stamp));
__part_stat_add(cpu, part, io_ticks, (now - part->stamp));With this chunk included, the reported regression gets fixed.
Signed-off-by: Nikanth Karthikesan
--
Signed-off-by: Jens Axboe
05 Oct, 2009
5 commits
-
It was briefly introduced to allow CFQ to to delayed scheduling,
but we ended up removing that feature again. So lets kill the
function and export, and just switch CFQ back to the normal work
schedule since it is now passing in a '0' delay from all call
sites.Signed-off-by: Jens Axboe
-
The RR service tree is indexed by a key that is relative to current jiffies.
This can cause problems on jiffies wraparound.The patch fixes it using time_before comparison, and changing
the add_front path to use a relative number, too.Signed-off-by: Corrado Zoccolo
Signed-off-by: Jens Axboe -
cfq uses rq->start_time as the fifo indicator, but that field may
get modified prior to cfq doing it's fifo list adjustment when
a request gets merged with another request. This can cause the
fifo list to become unordered.Reported-by: Corrado Zoccolo
Signed-off-by: Jens Axboe -
This reverts commit a9327cac440be4d8333bba975cbbf76045096275.
Corrado Zoccolo reports:
"with 2.6.32-rc1 I started getting the following strange output from
"iostat -kx 2":
Linux 2.6.31bisect (et2) 04/10/2009 _i686_ (2 CPU)avg-cpu: %user %nice %system %iowait %steal %idle
10,70 0,00 3,16 15,75 0,00 70,38Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s
avgrq-sz avgqu-sz await svctm %util
sda 18,22 0,00 0,67 0,01 14,77 0,02
43,94 0,01 10,53 39043915,03 2629219,87
sdb 60,89 9,68 50,79 3,04 1724,43 50,52
65,95 0,70 13,06 488437,47 2629219,87avg-cpu: %user %nice %system %iowait %steal %idle
2,72 0,00 0,74 0,00 0,00 96,53Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s
avgrq-sz avgqu-sz await svctm %util
sda 0,00 0,00 0,00 0,00 0,00 0,00
0,00 0,00 0,00 0,00 100,00
sdb 0,00 0,00 0,00 0,00 0,00 0,00
0,00 0,00 0,00 0,00 100,00avg-cpu: %user %nice %system %iowait %steal %idle
6,68 0,00 0,99 0,00 0,00 92,33Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s
avgrq-sz avgqu-sz await svctm %util
sda 0,00 0,00 0,00 0,00 0,00 0,00
0,00 0,00 0,00 0,00 100,00
sdb 0,00 0,00 0,00 0,00 0,00 0,00
0,00 0,00 0,00 0,00 100,00avg-cpu: %user %nice %system %iowait %steal %idle
4,40 0,00 0,73 1,47 0,00 93,40Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s
avgrq-sz avgqu-sz await svctm %util
sda 0,00 0,00 0,00 0,00 0,00 0,00
0,00 0,00 0,00 0,00 100,00
sdb 0,00 4,00 0,00 3,00 0,00 28,00
18,67 0,06 19,50 333,33 100,00Global values for service time and utilization are garbage. For
interval values, utilization is always 100%, and service time is
higher than normal.I bisected it down to:
[a9327cac440be4d8333bba975cbbf76045096275] Seperate read and write
statistics of in_flight requests
and verified that reverting just that commit indeed solves the issue
on 2.6.32-rc1."So until this is debugged, revert the bad commit.
Signed-off-by: Jens Axboe
-
We cannot delay for the first dispatch of the async queue if it
hasn't dispatched at all, since that could present a local user
DoS attack vector using an app that just did slow timed sync reads
while filling memory.Signed-off-by: Jens Axboe
04 Oct, 2009
3 commits
-
Not all users of the topology information want to use libblkid. Provide
the topology information through bdev ioctls.Also clarify sector size comments for existing BLK ioctls.
Signed-off-by: Martin K. Petersen
Signed-off-by: Jens Axboe -
We should use the sysfs modified slice sync value, in case it differs
from the default.Signed-off-by: Jens Axboe
-
Don't think that's necessarily a perfect description of what this
option fiddles with, but it's probably better than 'desktop'.Signed-off-by: Jens Axboe
03 Oct, 2009
3 commits
-
This slowly ramps up the async queue depth based on the time
passed since the sync IO, and doesn't allow async at all until
a sync slice period has passed.Signed-off-by: Jens Axboe
-
o Do not allow more than max_dispatch requests from an async queue, if some
sync request has finished recently. This is in the hope that sync activity
is still going on in the system and we might receive a sync request soon.
Most likely from a sync queue which finished a request and we did not enable
idling on it.Signed-off-by: Vivek Goyal
Signed-off-by: Jens Axboe -
This is basically identical to what Vivek Goyal posted, but combined
into one and labelled 'desktop' instead of 'fairness'. The goal
is to continue to improve on the latency side of things as it relates
to interactiveness, keeping the questionable bits under this sysfs
tunable so it would be easy for throughput-only people to turn off.Apart from adding the interactive sysfs knob, it also adds the
behavioural change of allowing slice idling even if the hardware
does tagged command queuing.Signed-off-by: Jens Axboe
02 Oct, 2009
6 commits
-
Since 2.6.31 now has request-based device-mapper, it's useful to have
a tracepoint for request-remapping as well as bio-remapping.
This patch adds a tracepoint for request-remapping, trace_block_rq_remap().Signed-off-by: Kiyoshi Ueda
Signed-off-by: Jun'ichi Nomura
Cc: Alasdair G Kergon
Cc: Li Zefan
Signed-off-by: Jens Axboe -
Currently we set the bio size to the byte equivalent of the blocks to
be trimmed when submitting the initial DISCARD ioctl. That means it
is subject to the max_hw_sectors limitation of the HBA which is
much lower than the size of a DISCARD request we can support.
Add a separate max_discard_sectors tunable to limit the size for discard
requests.We limit the max discard request size in bytes to 32bit as that is the
limit for bio->bi_size. This could be much larger if we had a way to pass
that information through the block layer.Signed-off-by: Christoph Hellwig
Signed-off-by: Jens Axboe -
prepare_discard_fn() was being called in a place where memory allocation
was effectively impossible. This makes it inappropriate for all but
the most trivial translations of Linux's DISCARD operation to the block
command set. Additionally adding a payload there makes the ownership
of the bio backing unclear as it's now allocated by the device driver
and not the submitter as usual.It is replaced with QUEUE_FLAG_DISCARD which is used to indicate whether
the queue supports discard operations or not. blkdev_issue_discard now
allocates a one-page, sector-length payload which is the right thing
for the common ATA and SCSI implementations.The mtd implementation of prepare_discard_fn() is replaced with simply
checking for the request being a discard.Largely based on a previous patch from Matthew Wilcox
which did the prepare_discard_fn but not the different payload allocation
yet.Signed-off-by: Christoph Hellwig
Signed-off-by: Jens Axboe -
Add missing blk_trace_remove_sysfs to be in pair with blk_trace_init_sysfs
introduced in commit 1d54ad6da9192fed5dd3b60224d9f2dfea0dcd82.
Release kobject also in case the request_fn is NULL.Problem was noticed via kmemleak backtrace when some sysfs entries were
note properly destroyed during device removal:unreferenced object 0xffff88001aa76640 (size 80):
comm "lvcreate", pid 2120, jiffies 4294885144
hex dump (first 32 bytes):
01 00 00 00 00 00 00 00 f0 65 a7 1a 00 88 ff ff .........e......
90 66 a7 1a 00 88 ff ff 86 1d 53 81 ff ff ff ff .f........S.....
backtrace:
[] kmemleak_alloc+0x26/0x60
[] kmem_cache_alloc+0x133/0x1c0
[] sysfs_new_dirent+0x41/0x120
[] sysfs_add_file_mode+0x3c/0xb0
[] internal_create_group+0xc1/0x1a0
[] sysfs_create_group+0x13/0x20
[] blk_trace_init_sysfs+0x14/0x20
[] blk_register_queue+0x3c/0xf0
[] add_disk+0x94/0x160
[] dm_create+0x598/0x6e0 [dm_mod]
[] dev_create+0x51/0x350 [dm_mod]
[] ctl_ioctl+0x1a3/0x240 [dm_mod]
[] dm_compat_ctl_ioctl+0x12/0x20 [dm_mod]
[] compat_sys_ioctl+0xcd/0x4f0
[] sysenter_dispatch+0x7/0x2c
[] 0xffffffffffffffffSigned-off-by: Zdenek Kabelac
Reviewed-by: Li Zefan
Signed-off-by: Jens Axboe -
Stacking devices do not have an inherent max_hw_sector limit. Set the
default to INT_MAX so we are bounded only by capabilities of the
underlying storage.Signed-off-by: Martin K. Petersen
Signed-off-by: Jens Axboe -
The topology changes unintentionally caused SAFE_MAX_SECTORS to be set
for stacking devices. Set the default limit to BLK_DEF_MAX_SECTORS and
provide SAFE_MAX_SECTORS in blk_queue_make_request() for legacy hw
drivers that depend on the old behavior.Acked-by: Mike Snitzer
Signed-off-by: Martin K. Petersen
Signed-off-by: Jens Axboe
20 Sep, 2009
1 commit
-
This allows subsytems to provide devtmpfs with non-default permissions
for the device node. Instead of the default mode of 0600, null, zero,
random, urandom, full, tty, ptmx now have a mode of 0666, which allows
non-privileged processes to access standard device nodes in case no
other userspace process applies the expected permissions.This also fixes a wrong assignment in pktcdvd and a checkpatch.pl complain.
Signed-off-by: Kay Sievers
Signed-off-by: Greg Kroah-Hartman
16 Sep, 2009
3 commits
-
* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core-2.6:
Driver Core: devtmpfs - kernel-maintained tmpfs-based /dev
debugfs: Modify default debugfs directory for debugging pktcdvd.
debugfs: Modified default dir of debugfs for debugging UHCI.
debugfs: Change debugfs directory of IWMC3200
debugfs: Change debuhgfs directory of trace-events-sample.h
debugfs: Fix mount directory of debugfs by default in events.txt
hpilo: add poll f_op
hpilo: add interrupt handler
hpilo: staging for interrupt handling
driver core: platform_device_add_data(): use kmemdup()
Driver core: Add support for compatibility classes
uio: add generic driver for PCI 2.3 devices
driver-core: move dma-coherent.c from kernel to driver/base
mem_class: fix bug
mem_class: use minor as index instead of searching the array
driver model: constify attribute groups
UIO: remove 'default n' from Kconfig
Driver core: Add accessor for device platform data
Driver core: move dev_get/set_drvdata to drivers/base/dd.c
Driver core: add new device to bus's list before probing -
Let attribute group vectors be declared "const". We'd
like to let most attribute metadata live in read-only
sections... this is a start.Signed-off-by: David Brownell
Signed-off-by: Greg Kroah-Hartman -
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu: (46 commits)
powerpc64: convert to dynamic percpu allocator
sparc64: use embedding percpu first chunk allocator
percpu: kill lpage first chunk allocator
x86,percpu: use embedding for 64bit NUMA and page for 32bit NUMA
percpu: update embedding first chunk allocator to handle sparse units
percpu: use group information to allocate vmap areas sparsely
vmalloc: implement pcpu_get_vm_areas()
vmalloc: separate out insert_vmalloc_vm()
percpu: add chunk->base_addr
percpu: add pcpu_unit_offsets[]
percpu: introduce pcpu_alloc_info and pcpu_group_info
percpu: move pcpu_lpage_build_unit_map() and pcpul_lpage_dump_cfg() upward
percpu: add @align to pcpu_fc_alloc_fn_t
percpu: make @dyn_size mandatory for pcpu_setup_first_chunk()
percpu: drop @static_size from first chunk allocators
percpu: generalize first chunk allocator selection
percpu: build first chunk allocators selectively
percpu: rename 4k first chunk allocator to page
percpu: improve boot messages
percpu: fix pcpu_reclaim() locking
...Fix trivial conflict as by Tejun Heo in kernel/sched.c
15 Sep, 2009
1 commit
-
* 'for-2.6.32' of git://git.kernel.dk/linux-2.6-block: (29 commits)
block: use blkdev_issue_discard in blk_ioctl_discard
Make DISCARD_BARRIER and DISCARD_NOBARRIER writes instead of reads
block: don't assume device has a request list backing in nr_requests store
block: Optimal I/O limit wrapper
cfq: choose a new next_req when a request is dispatched
Seperate read and write statistics of in_flight requests
aoe: end barrier bios with EOPNOTSUPP
block: trace bio queueing trial only when it occurs
block: enable rq CPU completion affinity by default
cfq: fix the log message after dispatched a request
block: use printk_once
cciss: memory leak in cciss_init_one()
splice: update mtime and atime on files
block: make blk_iopoll_prep_sched() follow normal 0/1 return convention
cfq-iosched: get rid of must_alloc flag
block: use interrupts disabled version of raise_softirq_irqoff()
block: fix comment in blk-iopoll.c
block: adjust default budget for blk-iopoll
block: fix long lines in block/blk-iopoll.c
block: add blk-iopoll, a NAPI like approach for block devices
...
14 Sep, 2009
5 commits
-
blk_ioctl_discard duplicates large amounts of code from blkdev_issue_discard,
the only difference between the two is that blkdev_issue_discard needs to
send a barrier discard request and blk_ioctl_discard a non-barrier one,
and blk_ioctl_discard needs to wait on the request. To facilitates this
add a flags argument to blkdev_issue_discard to control both aspects of the
behaviour. This will be very useful later on for using the waiting
funcitonality for other callers.Based on an earlier patch from Matthew Wilcox .
Signed-off-by: Christoph Hellwig
Signed-off-by: Jens Axboe -
Stacked devices do not. For now, just error out with -EINVAL. Later
we could make the limit apply on stacked devices too, for throttling
reasons.This fixes
5a54cd13353bb3b88887604e2c980aa01e314309
and should go into 2.6.31 stable as well.
Cc: stable@kernel.org
Signed-off-by: Jens Axboe -
Implement blk_limits_io_opt() and make blk_queue_io_opt() a wrapper
around it. DM needs this to avoid poking at the queue_limits directly.Signed-off-by: Martin K. Petersen
Signed-off-by: Mike Snitzer
Signed-off-by: Jens Axboe -
This patch addresses http://bugzilla.kernel.org/show_bug.cgi?id=13401, a
regression introduced in 2.6.30.From the bug report:
Signed-off-by: Jens Axboe
-
Currently, there is a single in_flight counter measuring the number of
requests in the request_queue. But some monitoring tools would like to
know how many read requests and write requests are in progress. Split the
current in_flight counter into two seperate counters for read and write.This information is exported as a sysfs attribute, as changing the
currently available stat files would break the existing tools.Signed-off-by: Nikanth Karthikesan
Signed-off-by: Jens Axboe
11 Sep, 2009
1 commit
-
If BIO is discarded or cross over end of device,
BIO queueing trial doesn't occur.Actually the trace was called just before make_request at first:
[PATCH] Block queue IO tracing support (blktrace) as of 2006-03-23
2056a782f8e7e65fd4bfd027506b4ce1c5e9ccd4And then 2 patches added some checks between them:
[PATCH] md: check bio address after mapping through partitions
5ddfe9691c91a244e8d1be597b6428fcefd58103,
[BLOCK] Don't allow empty barriers to be passed down to
queues that don't grok them
51fd77bd9f512ab6cc9df0733ba1caaab89eb957It breaks original goal.
Let's trace it only when it happens.Signed-off-by: Minchan Kim
Acked-by: Wu Fengguang
Cc: Li Zefan
Signed-off-by: Jens Axboe
