If wake_affine() pulls a task to another node for any reason and the node is
no longer preferred then temporarily stop automatic NUMA balancing pulling
the task back. Otherwise, tasks with a strong waker/wakee relationship
may constantly fight automatic NUMA balancing over where a task should
be placed.
Once again netperf is interesting here. The performance barely changes
but automatic NUMA balancing is interesting:
Hmean send-64 354.67 ( 0.00%) 352.15 ( -0.71%)
Hmean send-128 702.91 ( 0.00%) 693.84 ( -1.29%)
Hmean send-256 1350.07 ( 0.00%) 1344.19 ( -0.44%)
Hmean send-1024 5124.38 ( 0.00%) 4941.24 ( -3.57%)
Hmean send-2048 9687.44 ( 0.00%) 9624.45 ( -0.65%)
Hmean send-3312 14577.64 ( 0.00%) 14514.35 ( -0.43%)
Hmean send-4096 16393.62 ( 0.00%) 16488.30 ( 0.58%)
Hmean send-8192 26877.26 ( 0.00%) 26431.63 ( -1.66%)
Hmean send-16384 38683.43 ( 0.00%) 38264.91 ( -1.08%)
Hmean recv-64 354.67 ( 0.00%) 352.15 ( -0.71%)
Hmean recv-128 702.91 ( 0.00%) 693.84 ( -1.29%)
Hmean recv-256 1350.07 ( 0.00%) 1344.19 ( -0.44%)
Hmean recv-1024 5124.38 ( 0.00%) 4941.24 ( -3.57%)
Hmean recv-2048 9687.43 ( 0.00%) 9624.45 ( -0.65%)
Hmean recv-3312 14577.59 ( 0.00%) 14514.35 ( -0.43%)
Hmean recv-4096 16393.55 ( 0.00%) 16488.20 ( 0.58%)
Hmean recv-8192 26876.96 ( 0.00%) 26431.29 ( -1.66%)
Hmean recv-16384 38682.41 ( 0.00%) 38263.94 ( -1.08%)
NUMA alloc hit 1465986 1423090
NUMA alloc miss 0 0
NUMA interleave hit 0 0
NUMA alloc local 1465897 1423003
NUMA base PTE updates 1473 1420
NUMA huge PMD updates 0 0
NUMA page range updates 1473 1420
NUMA hint faults 1383 1312
NUMA hint local faults 451 124
NUMA hint local percent 32 9
There is a slight degrading in performance but there are slightly fewer
NUMA faults. There is a large drop in the percentage of local faults but
the bulk of migrations for netperf are in small shared libraries so it's
reflecting the fact that automatic NUMA balancing has backed off. This is
a case where despite wake_affine() and automatic NUMA balancing fighting
for placement that there is a marginal benefit to rescheduling to local
data quickly. However, it should be noted that wake_affine() and automatic
NUMA balancing fighting each other constantly is undesirable.
However, the benefit in other cases is large. This is the result for NAS
with the D class sizing on a 4-socket machine:
nas-mpi
4.15.0 4.15.0
sdnuma-v1r23 delayretry-v1r23
Time cg.D 557.00 ( 0.00%) 431.82 ( 22.47%)
Time ep.D 77.83 ( 0.00%) 79.01 ( -1.52%)
Time is.D 26.46 ( 0.00%) 26.64 ( -0.68%)
Time lu.D 727.14 ( 0.00%) 597.94 ( 17.77%)
Time mg.D 191.35 ( 0.00%) 146.85 ( 23.26%)
4.15.0 4.15.0
sdnuma-v1r23delayretry-v1r23
User 75665.20 70413.30
System 20321.59 8861.67
Elapsed 766.13 634.92
Minor Faults 16528502 7127941
Major Faults 4553 5068
NUMA alloc local 6963197 6749135
NUMA base PTE updates 366409093 107491434
NUMA huge PMD updates 687556 198880
NUMA page range updates 718437765 209317994
NUMA hint faults 13643410 4601187
NUMA hint local faults 9212593 3063996
NUMA hint local percent 67 66
Note the massive reduction in system CPU usage even though the percentage
of local faults is barely affected. There is a massive reduction in the
number of PTE updates showing that automatic NUMA balancing has backed off.
A critical observation is also that there is a massive reduction in minor
faults which is due to far fewer NUMA hinting faults being trapped.
There were questions on NAS OMP and how it behaved related to threads
being bound to CPUs. First, there are more gains than losses with this
patch applied and a reduction in system CPU usage:
nas-omp
4.16.0-rc1 4.16.0-rc1
sdnuma-v2r1 delayretry-v2r1
Time bt.D 436.71 ( 0.00%) 430.05 ( 1.53%)
Time cg.D 201.02 ( 0.00%) 180.87 ( 10.02%)
Time ep.D 32.84 ( 0.00%) 32.68 ( 0.49%)
Time is.D 9.63 ( 0.00%) 9.64 ( -0.10%)
Time lu.D 331.20 ( 0.00%) 304.80 ( 7.97%)
Time mg.D 54.87 ( 0.00%) 52.72 ( 3.92%)
Time sp.D 1108.78 ( 0.00%) 917.10 ( 17.29%)
Time ua.D 378.81 ( 0.00%) 398.83 ( -5.28%)
4.16.0-rc1 4.16.0-rc1
sdnuma-v2r1delayretry-v2r1
User 305633.08 296751.91
System 451.75 357.80
Elapsed 2595.73 2368.13
However, it does not close the gap between binding and being unbound. There
is negligible difference between the performance of the baseline and a
patched kernel when threads are bound so it is not presented here:
4.16.0-rc1 4.16.0-rc1
delayretry-bind delayretry-unbound
Time bt.D 385.02 ( 0.00%) 430.05 ( -11.70%)
Time cg.D 144.02 ( 0.00%) 180.87 ( -25.59%)
Time ep.D 32.85 ( 0.00%) 32.68 ( 0.52%)
Time is.D 10.52 ( 0.00%) 9.64 ( 8.37%)
Time lu.D 285.31 ( 0.00%) 304.80 ( -6.83%)
Time mg.D 43.21 ( 0.00%) 52.72 ( -22.01%)
Time sp.D 820.24 ( 0.00%) 917.10 ( -11.81%)
Time ua.D 337.09 ( 0.00%) 398.83 ( -18.32%)
4.16.0-rc1 4.16.0-rc1
delayretry-binddelayretry-unbound
User 277731.25 296751.91
System 261.29 357.80
Elapsed 2100.55 2368.13
Unfortunately, while performance is improved by the patch, there is still
quite a long way to go before it's equivalent to hard binding.
Other workloads like hackbench, tbench, dbench and schbench are barely
affected. dbench shows a mix of gains and losses depending on the machine
although in general, the results are more stable.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-7-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
find_idlest_group() compares a local group with each other group to select
the one that is most idle. When comparing groups in different NUMA domains,
a very slight imbalance is enough to select a remote NUMA node even if the
runnable load on both groups is 0 or close to 0. This ignores the cost of
remote accesses entirely and is a problem when selecting the CPU for a
newly forked task to run on. This is problematic when a forking server
is almost guaranteed to run on a remote node incurring numerous remote
accesses and potentially causing automatic NUMA balancing to try migrate
the task back or migrate the data to another node. Similar weirdness is
observed if a basic shell command pipes output to another as each process
in the pipeline is likely to start on different nodes and then get adjusted
later by wake_affine().
This patch adds imbalance to remote domains when considering whether to
select CPUs from remote domains. If the local domain is selected, imbalance
will still be used to try select a CPU from a lower scheduler domain's group
instead of stacking tasks on the same CPU.
A variety of workloads and machines were tested and as expected, there is no
difference on UMA. The difference on NUMA can be dramatic. This is a comparison
of elapsed times running the git regression test suite. It's fork-intensive with
short-lived processes:
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
Elapsed min 1706.06 ( 0.00%) 1435.94 ( 15.83%)
Elapsed mean 1709.53 ( 0.00%) 1436.98 ( 15.94%)
Elapsed stddev 2.16 ( 0.00%) 1.01 ( 53.38%)
Elapsed coeffvar 0.13 ( 0.00%) 0.07 ( 44.54%)
Elapsed max 1711.59 ( 0.00%) 1438.01 ( 15.98%)
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
User 5434.12 5188.41
System 4878.77 3467.09
Elapsed 10259.06 8624.21
That shows a considerable reduction in elapsed times. It's important to
note that automatic NUMA balancing does not affect this load as processes
are too short-lived.
There is also a noticable impact on hackbench such as this example using
processes and pipes:
hackbench-process-pipes
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
Amean 1 1.0973 ( 0.00%) 0.9393 ( 14.40%)
Amean 4 1.3427 ( 0.00%) 1.3730 ( -2.26%)
Amean 7 1.4233 ( 0.00%) 1.6670 ( -17.12%)
Amean 12 3.0250 ( 0.00%) 3.3013 ( -9.13%)
Amean 21 9.0860 ( 0.00%) 9.5343 ( -4.93%)
Amean 30 14.6547 ( 0.00%) 13.2433 ( 9.63%)
Amean 48 22.5447 ( 0.00%) 20.4303 ( 9.38%)
Amean 79 29.2010 ( 0.00%) 26.7853 ( 8.27%)
Amean 110 36.7443 ( 0.00%) 35.8453 ( 2.45%)
Amean 141 45.8533 ( 0.00%) 42.6223 ( 7.05%)
Amean 172 55.1317 ( 0.00%) 50.6473 ( 8.13%)
Amean 203 64.4420 ( 0.00%) 58.3957 ( 9.38%)
Amean 234 73.2293 ( 0.00%) 67.1047 ( 8.36%)
Amean 265 80.5220 ( 0.00%) 75.7330 ( 5.95%)
Amean 296 88.7567 ( 0.00%) 82.1533 ( 7.44%)
It's not a universal win as there are occasions when spreading wide and
quickly is a benefit but it's more of a win than it is a loss. For other
workloads, there is little difference but netperf is interesting. Without
the patch, the server and client starts on different nodes but quickly get
migrated due to wake_affine. Hence, the difference is overall performance
is marginal but detectable:
4.15.0 4.15.0
noexit-v1r23 sdnuma-v1r23
Hmean send-64 349.09 ( 0.00%) 354.67 ( 1.60%)
Hmean send-128 699.16 ( 0.00%) 702.91 ( 0.54%)
Hmean send-256 1316.34 ( 0.00%) 1350.07 ( 2.56%)
Hmean send-1024 5063.99 ( 0.00%) 5124.38 ( 1.19%)
Hmean send-2048 9705.19 ( 0.00%) 9687.44 ( -0.18%)
Hmean send-3312 14359.48 ( 0.00%) 14577.64 ( 1.52%)
Hmean send-4096 16324.20 ( 0.00%) 16393.62 ( 0.43%)
Hmean send-8192 26112.61 ( 0.00%) 26877.26 ( 2.93%)
Hmean send-16384 37208.44 ( 0.00%) 38683.43 ( 3.96%)
Hmean recv-64 349.09 ( 0.00%) 354.67 ( 1.60%)
Hmean recv-128 699.16 ( 0.00%) 702.91 ( 0.54%)
Hmean recv-256 1316.34 ( 0.00%) 1350.07 ( 2.56%)
Hmean recv-1024 5063.99 ( 0.00%) 5124.38 ( 1.19%)
Hmean recv-2048 9705.16 ( 0.00%) 9687.43 ( -0.18%)
Hmean recv-3312 14359.42 ( 0.00%) 14577.59 ( 1.52%)
Hmean recv-4096 16323.98 ( 0.00%) 16393.55 ( 0.43%)
Hmean recv-8192 26111.85 ( 0.00%) 26876.96 ( 2.93%)
Hmean recv-16384 37206.99 ( 0.00%) 38682.41 ( 3.97%)
However, what is very interesting is how automatic NUMA balancing behaves.
Each netperf instance runs long enough for balancing to activate:
NUMA base PTE updates 4620 1473
NUMA huge PMD updates 0 0
NUMA page range updates 4620 1473
NUMA hint faults 4301 1383
NUMA hint local faults 1309 451
NUMA hint local percent 30 32
NUMA pages migrated 1335 491
AutoNUMA cost 21% 6%
There is an unfortunate number of remote faults although tracing indicated
that the vast majority are in shared libraries. However, the tendency to
start tasks on the same node if there is capacity means that there were
far fewer PTE updates and faults incurred overall.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-6-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When a task exits, it notifies the parent that it has exited. This is a
sync wakeup and the exiting task may pull the parent towards the wakers
CPU. For simple workloads like using a shell, it was observed that the
shell is pulled across nodes by exiting processes. This is daft as the
parent may be long-lived and properly placed. This patch special cases a
sync wakeup on exit to avoid pulling tasks across nodes. Testing on a range
of workloads and machines showed very little differences in performance
although there was a small 3% boost on some machines running a shellscript
intensive workload (git regression test suite).
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-5-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
wake_affine_weight() will consider migrating a task to, or near, the current
CPU if there is a load imbalance. If the CPUs share LLC then either CPU
is valid as a search-for-idle-sibling target and equally appropriate for
stacking two tasks on one CPU if an idle sibling is unavailable. If they do
not share cache then a cross-node migration potentially impacts locality
so while they are equal from a CPU capacity point of view, they are not
equal in terms of memory locality. In either case, it's more appropriate
to migrate only if there is a difference in their effective load.
This patch modifies wake_affine_weight() to only consider migrating a task
if there is a load imbalance for normal wakeups but will allow potential
stacking if the loads are equal and it's a sync wakeup.
For the most part, the different in performance is marginal. For example,
on a 4-socket server running netperf UDP_STREAM on localhost the differences
are as follows:
4.15.0 4.15.0
16rc0 noequal-v1r23
Hmean send-64 355.47 ( 0.00%) 349.50 ( -1.68%)
Hmean send-128 697.98 ( 0.00%) 693.35 ( -0.66%)
Hmean send-256 1328.02 ( 0.00%) 1318.77 ( -0.70%)
Hmean send-1024 5051.83 ( 0.00%) 5051.11 ( -0.01%)
Hmean send-2048 9637.02 ( 0.00%) 9601.34 ( -0.37%)
Hmean send-3312 14355.37 ( 0.00%) 14414.51 ( 0.41%)
Hmean send-4096 16464.97 ( 0.00%) 16301.37 ( -0.99%)
Hmean send-8192 26722.42 ( 0.00%) 26428.95 ( -1.10%)
Hmean send-16384 38137.81 ( 0.00%) 38046.11 ( -0.24%)
Hmean recv-64 355.47 ( 0.00%) 349.50 ( -1.68%)
Hmean recv-128 697.98 ( 0.00%) 693.35 ( -0.66%)
Hmean recv-256 1328.02 ( 0.00%) 1318.77 ( -0.70%)
Hmean recv-1024 5051.83 ( 0.00%) 5051.11 ( -0.01%)
Hmean recv-2048 9636.95 ( 0.00%) 9601.30 ( -0.37%)
Hmean recv-3312 14355.32 ( 0.00%) 14414.48 ( 0.41%)
Hmean recv-4096 16464.74 ( 0.00%) 16301.16 ( -0.99%)
Hmean recv-8192 26721.63 ( 0.00%) 26428.17 ( -1.10%)
Hmean recv-16384 38136.00 ( 0.00%) 38044.88 ( -0.24%)
Stddev send-64 7.30 ( 0.00%) 4.75 ( 34.96%)
Stddev send-128 15.15 ( 0.00%) 22.38 ( -47.66%)
Stddev send-256 13.99 ( 0.00%) 19.14 ( -36.81%)
Stddev send-1024 105.73 ( 0.00%) 67.38 ( 36.27%)
Stddev send-2048 294.57 ( 0.00%) 223.88 ( 24.00%)
Stddev send-3312 302.28 ( 0.00%) 271.74 ( 10.10%)
Stddev send-4096 195.92 ( 0.00%) 121.10 ( 38.19%)
Stddev send-8192 399.71 ( 0.00%) 563.77 ( -41.04%)
Stddev send-16384 1163.47 ( 0.00%) 1103.68 ( 5.14%)
Stddev recv-64 7.30 ( 0.00%) 4.75 ( 34.96%)
Stddev recv-128 15.15 ( 0.00%) 22.38 ( -47.66%)
Stddev recv-256 13.99 ( 0.00%) 19.14 ( -36.81%)
Stddev recv-1024 105.73 ( 0.00%) 67.38 ( 36.27%)
Stddev recv-2048 294.59 ( 0.00%) 223.89 ( 24.00%)
Stddev recv-3312 302.24 ( 0.00%) 271.75 ( 10.09%)
Stddev recv-4096 196.03 ( 0.00%) 121.14 ( 38.20%)
Stddev recv-8192 399.86 ( 0.00%) 563.65 ( -40.96%)
Stddev recv-16384 1163.79 ( 0.00%) 1103.86 ( 5.15%)
The difference in overall performance is marginal but note that most
measurements are less variable. There were similar observations for other
netperf comparisons. hackbench with sockets or threads with processes or
threads showed minor difference with some reduction of migration. tbench
showed only marginal differences that were within the noise. dbench,
regardless of filesystem, showed minor differences all of which are
within noise. Multiple machines, both UMA and NUMA were tested without
any regressions showing up.
The biggest risk with a patch like this is affecting wakeup latencies.
However, the schbench load from Facebook which is very sensitive to wakeup
latency showed a mixed result with mostly improvements in wakeup latency:
4.15.0 4.15.0
16rc0 noequal-v1r23
Lat 50.00th-qrtle-1 38.00 ( 0.00%) 38.00 ( 0.00%)
Lat 75.00th-qrtle-1 49.00 ( 0.00%) 41.00 ( 16.33%)
Lat 90.00th-qrtle-1 52.00 ( 0.00%) 50.00 ( 3.85%)
Lat 95.00th-qrtle-1 54.00 ( 0.00%) 51.00 ( 5.56%)
Lat 99.00th-qrtle-1 63.00 ( 0.00%) 60.00 ( 4.76%)
Lat 99.50th-qrtle-1 66.00 ( 0.00%) 61.00 ( 7.58%)
Lat 99.90th-qrtle-1 78.00 ( 0.00%) 65.00 ( 16.67%)
Lat 50.00th-qrtle-2 38.00 ( 0.00%) 38.00 ( 0.00%)
Lat 75.00th-qrtle-2 42.00 ( 0.00%) 43.00 ( -2.38%)
Lat 90.00th-qrtle-2 46.00 ( 0.00%) 48.00 ( -4.35%)
Lat 95.00th-qrtle-2 49.00 ( 0.00%) 50.00 ( -2.04%)
Lat 99.00th-qrtle-2 55.00 ( 0.00%) 57.00 ( -3.64%)
Lat 99.50th-qrtle-2 58.00 ( 0.00%) 60.00 ( -3.45%)
Lat 99.90th-qrtle-2 65.00 ( 0.00%) 68.00 ( -4.62%)
Lat 50.00th-qrtle-4 41.00 ( 0.00%) 41.00 ( 0.00%)
Lat 75.00th-qrtle-4 45.00 ( 0.00%) 46.00 ( -2.22%)
Lat 90.00th-qrtle-4 50.00 ( 0.00%) 50.00 ( 0.00%)
Lat 95.00th-qrtle-4 54.00 ( 0.00%) 53.00 ( 1.85%)
Lat 99.00th-qrtle-4 61.00 ( 0.00%) 61.00 ( 0.00%)
Lat 99.50th-qrtle-4 65.00 ( 0.00%) 64.00 ( 1.54%)
Lat 99.90th-qrtle-4 76.00 ( 0.00%) 82.00 ( -7.89%)
Lat 50.00th-qrtle-8 48.00 ( 0.00%) 46.00 ( 4.17%)
Lat 75.00th-qrtle-8 55.00 ( 0.00%) 54.00 ( 1.82%)
Lat 90.00th-qrtle-8 60.00 ( 0.00%) 59.00 ( 1.67%)
Lat 95.00th-qrtle-8 63.00 ( 0.00%) 63.00 ( 0.00%)
Lat 99.00th-qrtle-8 71.00 ( 0.00%) 69.00 ( 2.82%)
Lat 99.50th-qrtle-8 74.00 ( 0.00%) 73.00 ( 1.35%)
Lat 99.90th-qrtle-8 98.00 ( 0.00%) 90.00 ( 8.16%)
Lat 50.00th-qrtle-16 56.00 ( 0.00%) 55.00 ( 1.79%)
Lat 75.00th-qrtle-16 68.00 ( 0.00%) 67.00 ( 1.47%)
Lat 90.00th-qrtle-16 77.00 ( 0.00%) 78.00 ( -1.30%)
Lat 95.00th-qrtle-16 82.00 ( 0.00%) 84.00 ( -2.44%)
Lat 99.00th-qrtle-16 90.00 ( 0.00%) 93.00 ( -3.33%)
Lat 99.50th-qrtle-16 93.00 ( 0.00%) 97.00 ( -4.30%)
Lat 99.90th-qrtle-16 110.00 ( 0.00%) 110.00 ( 0.00%)
Lat 50.00th-qrtle-32 68.00 ( 0.00%) 62.00 ( 8.82%)
Lat 75.00th-qrtle-32 90.00 ( 0.00%) 83.00 ( 7.78%)
Lat 90.00th-qrtle-32 110.00 ( 0.00%) 100.00 ( 9.09%)
Lat 95.00th-qrtle-32 122.00 ( 0.00%) 111.00 ( 9.02%)
Lat 99.00th-qrtle-32 145.00 ( 0.00%) 133.00 ( 8.28%)
Lat 99.50th-qrtle-32 154.00 ( 0.00%) 143.00 ( 7.14%)
Lat 99.90th-qrtle-32 2316.00 ( 0.00%) 515.00 ( 77.76%)
Lat 50.00th-qrtle-35 69.00 ( 0.00%) 72.00 ( -4.35%)
Lat 75.00th-qrtle-35 92.00 ( 0.00%) 95.00 ( -3.26%)
Lat 90.00th-qrtle-35 111.00 ( 0.00%) 114.00 ( -2.70%)
Lat 95.00th-qrtle-35 122.00 ( 0.00%) 124.00 ( -1.64%)
Lat 99.00th-qrtle-35 142.00 ( 0.00%) 144.00 ( -1.41%)
Lat 99.50th-qrtle-35 150.00 ( 0.00%) 154.00 ( -2.67%)
Lat 99.90th-qrtle-35 6104.00 ( 0.00%) 5640.00 ( 7.60%)
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Giovanni Gherdovich <ggherdovich@suse.cz>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180213133730.24064-4-mgorman@techsingularity.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The default values for nreader and nwriter are apparently not all that
user-friendly, resulting in people doing scalability tests that ran all
runs at large scale. This commit therefore makes both the nreaders and
nwriters module default to the number of CPUs, and adds a comment to
rcuperf.c stating that the number of CPUs should be specified using the
nr_cpus kernel boot parameter. This commit also eliminates the redundant
rcuperf scripting specification of default values for these parameters.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The rcu_torture_writer() function adapts to requested testing from module
parameters as well as the function pointers in the structure referenced
by cur_ops. However, as long as the module parameters do not conflict
with the function pointers, this adaptation is silent. This silence can
result in confusion as to exactly what was tested, which could in turn
result in untested RCU code making its way into mainline.
This commit therefore makes rcu_torture_writer() announce exactly which
portions of RCU's API it ends up testing.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
During boot, normal grace periods are processed as expedited. When
rcutorture is built into the kernel, it starts during boot and thus
detects that normal grace periods are unconditionally expedited.
Therefore, rcutorture concludes that there is no point in trying
to dynamically enable expediting, do it disables this aspect of testing,
which is a bit of an overreaction to the temporary boot-time expediting.
This commit therefore rechecks forced expediting throughout the test,
enabling dynamic expediting if normal grace periods are processed
normally at any point.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Currently the rcu_torture_fakewriter() function invokes cur_ops->sync()
and cur_ops->exp_sync() without first checking to see if they are in
fact non-NULL. This results in kernel NULL pointer dereferences when
testing RCU implementations that choose not to provide the full set of
primitives. Given that it is perfectly reasonable to have specialized
RCU implementations that provide only a subset of the RCU API, this is
a bug in rcutorture.
This commit therefore makes rcu_torture_fakewriter() check function
pointers before invoking them, thus allowing it to test subsetted
RCU implementations.
Reported-by: Lihao Liang <lianglihao@huawei.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
This commit moves to __func__ for function names and for KBUILD_MODNAME
for module names, all in the name of better resilience to change.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
This commit replaces array-allocation calls to kzalloc() with
equivalent calls to kcalloc().
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The code in srcu_gp_end() inserts a delay every 0x3ff grace periods in
order to prevent SRCU grace-period work from consuming an entire CPU
when there is a long sequence of expedited SRCU grace-period requests.
However, all of SRCU's grace-period work is carried out in workqueues,
which are in turn within kthreads, which are automatically throttled as
needed by the scheduler. In particular, if there is plenty of idle time,
there is no point in throttling.
This commit therefore removes the expedited SRCU grace-period throttling.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Of course, compilers will optimize out a dead code. Anyway, remove
any dead code for better readibility.
Signed-off-by: Byungchul Park <byungchul.park@lge.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Currently, given a multi-level srcu_node tree, SRCU can scan the full
set of srcu_data structures at each level when cleaning up after a grace
period. This, though harmless otherwise, represents pointless overhead.
This commit therefore eliminates this overhead by scanning the srcu_data
structures only when traversing the leaf srcu_node structures.
Signed-off-by: Ildar Ismagilov <devix84@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
SRCU checks each srcu_data structure's grace-period number for counter
wrap four times per cycle by default. This frequency guarantees that
normal comparisons will detect potential wrap. However, the expedited
grace-period number is not checked. The consquences are not too horrible
(a failure to expedite a grace period when requested), but it would be
good to avoid such things. This commit therefore adds this check to
the expedited grace-period number.
Signed-off-by: Ildar Ismagilov <devix84@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
This commit moves to __func__ for function names in the name of better
resilience to change.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
This commit reworks the first loop in sync_rcu_exp_select_cpus()
to avoid doing unnecssary stores to other CPUs' rcu_data
structures. This speeds up that first loop by roughly a factor of
two on an old x86 system. In the case where the system is mostly
idle, this loop incurs a large fraction of the overhead of the
synchronize_rcu_expedited(). There is less benefit on busy systems
because the overhead of the smp_call_function_single() in the second
loop dominates in that case.
However, it is not unusual to do configuration chances involving
RCU grace periods (both expedited and normal) while the system is
mostly idle, so this optimization is worth doing.
While we are in the area, this commit also adds parentheses to arguments
used by the for_each_leaf_node_possible_cpu() macro.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
If a CPU is transitioning to or from offline state, an expedited
grace period may undergo a timed wait. This timed wait can unduly
delay grace periods, so this commit adds a trace statement to make
it visible.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
This commit adds more tracing of expedited grace periods to enable
improved debugging of slowdowns.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The srcu_funnel_exp_start() function checks to see if the srcu_struct
structure's expedited grace period counter needs updating to reflect a
newly arrived request for an expedited SRCU grace period. Unfortunately,
the check is backwards, so this commit reverses the sense of the test.
Signed-off-by: Ildar Ismagilov <devix84@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Commits c0b334c5bf and ea9b0c8a26 introduced new sparse warnings
by accessing rcu_node->lock directly and ignoring the __private
marker. Introduce a new wrapper and use it. Also fix a similar problem
in srcutree.c introduced by a3883df393.
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
RCU's nxttail has been optimized to be a rcu_segcblist, which is
a multi-tailed linked list with macros defined for the indexes for
each tail. The indexes have been defined in linux/rcu_segcblist.h,
so this commit removes the redundant definitions in kernel/rcu/tree.h.
Signed-off-by: Liu Changcheng <changcheng.liu@intel.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The kernel/rcu/rcu.h file has a pair of consecutive #ifdefs on
CONFIG_TINY_RCU, so this commit consolidates them, thus saving a few
lines of code.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
It is not always obvious that the stack dump from a starved grace-period
kthread isn't instead that of a CPU stalling the current grace period.
This commit therefore adds a pr_err() flagging these dumps.
Reported-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The debugfs interface displayed statistics on RCU-pending checks but
this interface has since been removed. This commit therefore removes the
no-longer-used rcu_state structure's ->n_force_qs_lh and ->n_force_qs_ngp
fields along with their updates. (Though the ->n_force_qs_ngp field
was actually not used at all, embarrassingly enough.)
If this information proves necessary in the future, the corresponding
event traces will be added.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The debugfs interface displayed statistics on RCU-pending checks
but this interface has since been removed. This commit therefore
removes the no-longer-used rcu_data structure's ->n_rcu_pending,
->n_rp_core_needs_qs, ->n_rp_report_qs, ->n_rp_cb_ready,
->n_rp_cpu_needs_gp, ->n_rp_gp_completed, ->n_rp_gp_started,
->n_rp_nocb_defer_wakeup, and ->n_rp_need_nothing fields along with
their updates.
If this information proves necessary in the future, the corresponding
event traces will be added.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The debugfs interface displayed statistics on RCU callback invocation but
this interface has since been removed. This commit therefore removes the
no-longer-used rcu_data structure's ->n_cbs_invoked and ->n_nocbs_invoked
fields along with their updates.
If this information proves necessary in the future, the corresponding
event traces will be added.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The debugfs interface displayed statistics on RCU priority boosting,
but this interface has since been removed. This commit therefore
removes the no-longer-used rcu_data structure's ->n_tasks_boosted,
->n_exp_boosts, and ->n_exp_boosts and their updates.
If this information proves necessary in the future, the corresponding
event traces will be added.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
In CONFIG_RCU_NOCB_CPU=y kernels, if the boot parameters indicate that
none of the CPUs should in fact be offloaded, the following somewhat
obtuse message appears:
Offload RCU callbacks from CPUs: .
This commit therefore makes the message at least grammatically correct
in this case:
Offload RCU callbacks from CPUs: (none)
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Pull perf updates from Thomas Gleixner:
"Perf tool updates and kprobe fixes:
- perf_mmap overwrite mode fixes/overhaul, prep work to get 'perf
top' using it, making it bearable to use it in large core count
systems such as Knights Landing/Mill Intel systems (Kan Liang)
- s/390 now uses syscall.tbl, just like x86-64 to generate the
syscall table id -> string tables used by 'perf trace' (Hendrik
Brueckner)
- Use strtoull() instead of home grown function (Andy Shevchenko)
- Synchronize kernel ABI headers, v4.16-rc1 (Ingo Molnar)
- Document missing 'perf data --force' option (Sangwon Hong)
- Add perf vendor JSON metrics for ARM Cortex-A53 Processor (William
Cohen)
- Improve error handling and error propagation of ftrace based
kprobes so failures when installing kprobes are not silently
ignored and create disfunctional tracepoints"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (27 commits)
kprobes: Propagate error from disarm_kprobe_ftrace()
kprobes: Propagate error from arm_kprobe_ftrace()
Revert "tools include s390: Grab a copy of arch/s390/include/uapi/asm/unistd.h"
perf s390: Rework system call table creation by using syscall.tbl
perf s390: Grab a copy of arch/s390/kernel/syscall/syscall.tbl
tools/headers: Synchronize kernel ABI headers, v4.16-rc1
perf test: Fix test trace+probe_libc_inet_pton.sh for s390x
perf data: Document missing --force option
perf tools: Substitute yet another strtoull()
perf top: Check the latency of perf_top__mmap_read()
perf top: Switch default mode to overwrite mode
perf top: Remove lost events checking
perf hists browser: Add parameter to disable lost event warning
perf top: Add overwrite fall back
perf evsel: Expose the perf_missing_features struct
perf top: Check per-event overwrite term
perf mmap: Discard legacy interface for mmap read
perf test: Update mmap read functions for backward-ring-buffer test
perf mmap: Introduce perf_mmap__read_event()
perf mmap: Introduce perf_mmap__read_done()
...
Pull irq updates from Thomas Gleixner:
"A small set of updates mostly for irq chip drivers:
- MIPS GIC fix for spurious, masked interrupts
- fix for a subtle IPI bug in GICv3
- do not probe GICv3 ITSs that are marked as disabled
- multi-MSI support for GICv2m
- various small cleanups"
* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
irqdomain: Re-use DEFINE_SHOW_ATTRIBUTE() macro
irqchip/bcm: Remove hashed address printing
irqchip/gic-v2m: Add PCI Multi-MSI support
irqchip/gic-v3: Ignore disabled ITS nodes
irqchip/gic-v3: Use wmb() instead of smb_wmb() in gic_raise_softirq()
irqchip/gic-v3: Change pr_debug message to pr_devel
irqchip/mips-gic: Avoid spuriously handling masked interrupts
In case of threaded interrupts the thread follows the affinity setting of
the hard interrupt. The related function uses the affinity mask which was
set by either from user space or via one of the kernel mechanisms. This
mask can be wider than the resulting effective affinity of the hard
interrupt. As a consequence the thread might become affine to a completely
different CPU.
Use the effective interrupt affinity if the architecture supports it, so
the hard interrupt and the thread stay on the same CPU.
Reported-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
...instead of open coding file operations followed by custom ->open()
callbacks per each attribute.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Improve error handling when arming ftrace-based kprobes. Specifically, if
we fail to arm a ftrace-based kprobe, register_kprobe()/enable_kprobe()
should report an error instead of success. Previously, this has lead to
confusing situations where register_kprobe() would return 0 indicating
success, but the kprobe would not be functional if ftrace registration
during the kprobe arming process had failed. We should therefore take any
errors returned by ftrace into account and propagate this error so that we
do not register/enable kprobes that cannot be armed. This can happen if,
for example, register_ftrace_function() finds an IPMODIFY conflict (since
kprobe_ftrace_ops has this flag set) and returns an error. Such a conflict
is possible since livepatches also set the IPMODIFY flag for their ftrace_ops.
arm_all_kprobes() keeps its current behavior and attempts to arm all
kprobes. It returns the last encountered error and gives a warning if
not all probes could be armed.
This patch is based on Petr Mladek's original patchset (patches 2 and 3)
back in 2015, which improved kprobes error handling, found here:
https://lkml.org/lkml/2015/2/26/452
However, further work on this had been paused since then and the patches
were not upstreamed.
Based-on-patches-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Jessica Yu <jeyu@kernel.org>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Cc: David S . Miller <davem@davemloft.net>
Cc: Jiri Kosina <jikos@kernel.org>
Cc: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Miroslav Benes <mbenes@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: live-patching@vger.kernel.org
Link: http://lkml.kernel.org/r/20180109235124.30886-2-jeyu@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
syzkaller recently triggered OOM during percpu map allocation;
while there is work in progress by Dennis Zhou to add __GFP_NORETRY
semantics for percpu allocator under pressure, there seems also a
missing bpf_map_precharge_memlock() check in array map allocation.
Given today the actual bpf_map_charge_memlock() happens after the
find_and_alloc_map() in syscall path, the bpf_map_precharge_memlock()
is there to bail out early before we go and do the map setup work
when we find that we hit the limits anyway. Therefore add this for
array map as well.
Fixes: 6c90598174 ("bpf: pre-allocate hash map elements")
Fixes: a10423b87a ("bpf: introduce BPF_MAP_TYPE_PERCPU_ARRAY map")
Reported-by: syzbot+adb03f3f0bb57ce3acda@syzkaller.appspotmail.com
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Dennis Zhou <dennisszhou@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Commit 6f1982fedd ("sched/isolation: Handle the nohz_full= parameter")
broke CONFIG_NO_HZ_FULL_ALL=y kernels. This breakage is due to the code
under CONFIG_NO_HZ_FULL_ALL failing to invoke the shiny new housekeeping
functions. This means that rcutorture scenario TREE04 now emits RCU CPU
stall warnings due to the RCU grace-period kthreads not being awakened
at a time of their choosing, or perhaps even not at all:
[ 27.731422] rcu_bh kthread starved for 21001 jiffies! g18446744073709551369 c18446744073709551368 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x402 ->cpu=3
[ 27.731423] rcu_bh I14936 9 2 0x80080000
[ 27.731435] Call Trace:
[ 27.731440] __schedule+0x31a/0x6d0
[ 27.731442] schedule+0x31/0x80
[ 27.731446] schedule_timeout+0x15a/0x320
[ 27.731453] ? call_timer_fn+0x130/0x130
[ 27.731457] rcu_gp_kthread+0x66c/0xea0
[ 27.731458] ? rcu_gp_kthread+0x66c/0xea0
Because no one has complained about CONFIG_NO_HZ_FULL_ALL=y being broken,
I hypothesize that no one is in fact using it, other than rcutorture.
This commit therefore eliminates CONFIG_NO_HZ_FULL_ALL and updates
rcutorture's config files to instead use the nohz_full= kernel parameter
to put the desired CPUs into nohz_full mode.
Fixes: 6f1982fedd ("sched/isolation: Handle the nohz_full= parameter")
Reported-by: kernel test robot <xiaolong.ye@intel.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Chris Metcalf <cmetcalf@mellanox.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Wanpeng Li <kernellwp@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Since rcu_boot_init_percpu_data() is only called at boot time,
there is no data race and spinlock is not needed.
Signed-off-by: Lihao Liang <lianglihao@huawei.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
If audit is disabled, who cares if there is a bug indicating syscall in
process or names already recorded. Bail immediately on audit disabled.
Signed-off-by: Richard Guy Briggs <rgb@redhat.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
The audit entry filter has been long deprecated with userspace support
finally removed in audit-v2.6.7 and plans to remove kernel support have
existed since kernel-v2.6.31.
Remove it.
Since removing the audit entry filter, test for early return before
setting up any context state.
Passes audit-testsuite.
See: https://github.com/linux-audit/audit-kernel/issues/6
Signed-off-by: Richard Guy Briggs <rgb@redhat.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Pull locking fixes from Ingo Molnar:
"This contains two qspinlock fixes and three documentation and comment
fixes"
* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
locking/semaphore: Update the file path in documentation
locking/atomic/bitops: Document and clarify ordering semantics for failed test_and_{}_bit()
locking/qspinlock: Ensure node->count is updated before initialising node
locking/qspinlock: Ensure node is initialised before updating prev->next
Documentation/locking/mutex-design: Update to reflect latest changes
