| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Delay accounting already measures the time a task spends in direct reclaim
and waiting for swapin, but in low memory situations tasks spend can spend
a significant amount of their time waiting on thrashing page cache. This
isn't tracked right now.
To know the full impact of memory contention on an individual task,
measure the delay when waiting for a recently evicted active cache page to
read back into memory.
Also update tools/accounting/getdelays.c:
[hannes@computer accounting]$ sudo ./getdelays -d -p 1
print delayacct stats ON
PID 1
CPU count real total virtual total delay total delay average
50318 745000000 847346785 400533713 0.008ms
IO count delay total delay average
435 122601218 0ms
SWAP count delay total delay average
0 0 0ms
RECLAIM count delay total delay average
0 0 0ms
THRASHING count delay total delay average
19 12621439 0ms
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Johannes Weiner <[email protected]>
Acked-by: Peter Zijlstra (Intel) <[email protected]>
Tested-by: Daniel Drake <[email protected]>
Tested-by: Suren Baghdasaryan <[email protected]>
Cc: Christopher Lameter <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Johannes Weiner <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Peter Enderborg <[email protected]>
Cc: Randy Dunlap <[email protected]>
Cc: Shakeel Butt <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: Vinayak Menon <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
try_to_wake_up() might invoke delayacct_blkio_end() while holding the
pi_lock (which is a raw_spinlock_t). delayacct_blkio_end() acquires
task_delay_info.lock which is a spinlock_t. This causes a might sleep splat
on -RT where non raw spinlocks are converted to 'sleeping' spinlocks.
task_delay_info.lock is only held for a short amount of time so it's not a
problem latency wise to make convert it to a raw spinlock.
Signed-off-by: Sebastian Andrzej Siewior <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Balbir Singh <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]
|
|
Before commit:
e33a9bba85a8 ("sched/core: move IO scheduling accounting from io_schedule_timeout() into scheduler")
delayacct_blkio_end() was called after context-switching into the task which
completed I/O.
This resulted in double counting: the task would account a delay both waiting
for I/O and for time spent in the runqueue.
With e33a9bba85a8, delayacct_blkio_end() is called by try_to_wake_up().
In ttwu, we have not yet context-switched. This is more correct, in that
the delay accounting ends when the I/O is complete.
But delayacct_blkio_end() relies on 'get_current()', and we have not yet
context-switched into the task whose I/O completed. This results in the
wrong task having its delay accounting statistics updated.
Instead of doing that, pass the task_struct being woken to delayacct_blkio_end(),
so that it can update the statistics of the correct task.
Signed-off-by: Josh Snyder <[email protected]>
Acked-by: Tejun Heo <[email protected]>
Acked-by: Balbir Singh <[email protected]>
Cc: <[email protected]>
Cc: Brendan Gregg <[email protected]>
Cc: Jens Axboe <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Fixes: e33a9bba85a8 ("sched/core: move IO scheduling accounting from io_schedule_timeout() into scheduler")
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
into <linux/sched/cputime.h>
Introduce a trivial, mostly empty <linux/sched/cputime.h> header
to prepare for the moving of cputime functionality out of sched.h.
Update all code that relies on these facilities.
Acked-by: Linus Torvalds <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
<linux/sched.h> to <linux/sched/task.h>
Update all usage sites first.
Acked-by: Linus Torvalds <[email protected]>
Cc: Mike Galbraith <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Use the new nsec based cputime accessors as part of the whole cputime
conversion from cputime_t to nsecs.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Cc: Fenghua Yu <[email protected]>
Cc: Heiko Carstens <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Martin Schwidefsky <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Stanislaw Gruszka <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Tony Luck <[email protected]>
Cc: Wanpeng Li <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
cputime
This API returns a task's cputime in cputime_t in order to ease the
conversion of cputime internals to use nsecs units instead. Blindly
converting all cputime readers to use this API now will later let us
convert more smoothly and step by step all these places to use the
new nsec based cputime.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Cc: Fenghua Yu <[email protected]>
Cc: Heiko Carstens <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Martin Schwidefsky <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Stanislaw Gruszka <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Tony Luck <[email protected]>
Cc: Wanpeng Li <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Mark those kmem allocations that are known to be easily triggered from
userspace as __GFP_ACCOUNT/SLAB_ACCOUNT, which makes them accounted to
memcg. For the list, see below:
- threadinfo
- task_struct
- task_delay_info
- pid
- cred
- mm_struct
- vm_area_struct and vm_region (nommu)
- anon_vma and anon_vma_chain
- signal_struct
- sighand_struct
- fs_struct
- files_struct
- fdtable and fdtable->full_fds_bits
- dentry and external_name
- inode for all filesystems. This is the most tedious part, because
most filesystems overwrite the alloc_inode method.
The list is far from complete, so feel free to add more objects.
Nevertheless, it should be close to "account everything" approach and
keep most workloads within bounds. Malevolent users will be able to
breach the limit, but this was possible even with the former "account
everything" approach (simply because it did not account everything in
fact).
[[email protected]: coding-style fixes]
Signed-off-by: Vladimir Davydov <[email protected]>
Acked-by: Johannes Weiner <[email protected]>
Acked-by: Michal Hocko <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: Greg Thelen <[email protected]>
Cc: Christoph Lameter <[email protected]>
Cc: Pekka Enberg <[email protected]>
Cc: David Rientjes <[email protected]>
Cc: Joonsoo Kim <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Converting cputime to timespec and timespec to nanoseconds makes no
sense. Use cputime_to_ns() and be done with it.
Signed-off-by: Thomas Gleixner <[email protected]>
Signed-off-by: John Stultz <[email protected]>
|
|
Kill the timespec juggling and calculate with plain nanoseconds.
Signed-off-by: Thomas Gleixner <[email protected]>
Signed-off-by: John Stultz <[email protected]>
|
|
do_posix_clock_monotonic_gettime() is a leftover from the initial
posix timer implementation which maps to ktime_get_ts(). Remove the
silly wrapper while at it.
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: John Stultz <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
|
|
The wrapper function delayacct_add_tsk() already checked 'tsk->delays',
and __delayacct_add_tsk() has no another direct callers, so can remove the
redundancy checking code.
And the label 'done' is also useless, so remove it, too.
Signed-off-by: Chen Gang <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
This is in preparation for the full dynticks feature. While
remotely reading the cputime of a task running in a full
dynticks CPU, we'll need to do some extra-computation. This
way we can account the time it spent tickless in userspace
since its last cputime snapshot.
Signed-off-by: Frederic Weisbecker <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Li Zhong <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Paul Gortmaker <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: Thomas Gleixner <[email protected]>
|
|
To implement steal time, we need the hypervisor to pass the guest
information about how much time was spent running other processes
outside the VM, while the vcpu had meaningful work to do - halt
time does not count.
This information is acquired through the run_delay field of
delayacct/schedstats infrastructure, that counts time spent in a
runqueue but not running.
Steal time is a per-cpu information, so the traditional MSR-based
infrastructure is used. A new msr, KVM_MSR_STEAL_TIME, holds the
memory area address containing information about steal time
This patch contains the hypervisor part of the steal time infrasructure,
and can be backported independently of the guest portion.
[avi, yongjie: export delayacct_on, to avoid build failures in some configs]
Signed-off-by: Glauber Costa <[email protected]>
Tested-by: Eric B Munson <[email protected]>
CC: Rik van Riel <[email protected]>
CC: Jeremy Fitzhardinge <[email protected]>
CC: Peter Zijlstra <[email protected]>
CC: Anthony Liguori <[email protected]>
Signed-off-by: Yongjie Ren <[email protected]>
Signed-off-by: Avi Kivity <[email protected]>
|
|
Remove net/genetlink.h inclusion, now sched.c won't be recompiled
because of some networking changes.
Signed-off-by: Alexey Dobriyan <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Impact: simplify code
When we turn on CONFIG_SCHEDSTATS, per-task cpu runtime is accumulated
twice. Once in task->se.sum_exec_runtime and once in sched_info.cpu_time.
These two stats are exactly the same.
Given that task->se.sum_exec_runtime is always accumulated by the core
scheduler, sched_info can reuse that data instead of duplicate the accounting.
Signed-off-by: Ken Chen <[email protected]>
Acked-by: Peter Zijlstra <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
|
|
Add members for memory reclaim delay to taskstats, and accumulate them in
__delayacct_add_tsk() .
Signed-off-by: Keika Kobayashi <[email protected]>
Cc: Hiroshi Shimamoto <[email protected]>
Cc: Balbir Singh <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Sometimes, application responses become bad under heavy memory load.
Applications take a bit time to reclaim memory. The statistics, how long
memory reclaim takes, will be useful to measure memory usage.
This patch adds accounting memory reclaim to per-task-delay-accounting for
accounting the time of do_try_to_free_pages().
<i.e>
- When System is under low memory load,
memory reclaim may not occur.
$ free
total used free shared buffers cached
Mem: 8197800 1577300 6620500 0 4808 1516724
-/+ buffers/cache: 55768 8142032
Swap: 16386292 0 16386292
$ vmstat 1
procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
0 0 0 5069748 10612 3014060 0 0 0 0 3 26 0 0 100 0
0 0 0 5069748 10612 3014060 0 0 0 0 4 22 0 0 100 0
0 0 0 5069748 10612 3014060 0 0 0 0 3 18 0 0 100 0
Measure the time of tar command.
$ ls -s test.dat
1501472 test.dat
$ time tar cvf test.tar test.dat
real 0m13.388s
user 0m0.116s
sys 0m5.304s
$ ./delayget -d -p <pid>
CPU count real total virtual total delay total
428 5528345500 5477116080 62749891
IO count delay total
338 8078977189
SWAP count delay total
0 0
RECLAIM count delay total
0 0
- When system is under heavy memory load
memory reclaim may occur.
$ vmstat 1
procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
0 0 7159032 49724 1812 3012 0 0 0 0 3 24 0 0 100 0
0 0 7159032 49724 1812 3012 0 0 0 0 4 24 0 0 100 0
0 0 7159032 49848 1812 3012 0 0 0 0 3 22 0 0 100 0
In this case, one process uses more 8G memory
by execution of malloc() and memset().
$ time tar cvf test.tar test.dat
real 1m38.563s <- increased by 85 sec
user 0m0.140s
sys 0m7.060s
$ ./delayget -d -p <pid>
CPU count real total virtual total delay total
9021 7140446250 7315277975 923201824
IO count delay total
8965 90466349669
SWAP count delay total
3 21036367
RECLAIM count delay total
740 61011951153
In the later case, the value of RECLAIM is increasing.
So, taskstats can show how much memory reclaim influences TAT.
Signed-off-by: Keika Kobayashi <[email protected]>
Acked-by: Balbir Singh <[email protected]>
Acked-by: KOSAKI Motohiro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
This adds items to the taststats struct to account for user and system
time based on scaling the CPU frequency and instruction issue rates.
Adds account_(user|system)_time_scaled callbacks which architectures
can use to account for time using this mechanism.
Signed-off-by: Michael Neuling <[email protected]>
Cc: Balbir Singh <[email protected]>
Cc: Jay Lan <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
rename all 'cnt' fields and variables to the less yucky 'count' name.
yuckage noticed by Andrew Morton.
no change in code, other than the /proc/sched_debug bkl_count string got
a bit larger:
text data bss dec hex filename
38236 3506 24 41766 a326 sched.o.before
38240 3506 24 41770 a32a sched.o.after
Signed-off-by: Ingo Molnar <[email protected]>
Reviewed-by: Thomas Gleixner <[email protected]>
|
|
update delay-accounting to use CFS's precise stats.
Signed-off-by: Ingo Molnar <[email protected]>
|
|
This patch provides a new macro
KMEM_CACHE(<struct>, <flags>)
to simplify slab creation. KMEM_CACHE creates a slab with the name of the
struct, with the size of the struct and with the alignment of the struct.
Additional slab flags may be specified if necessary.
Example
struct test_slab {
int a,b,c;
struct list_head;
} __cacheline_aligned_in_smp;
test_slab_cache = KMEM_CACHE(test_slab, SLAB_PANIC)
will create a new slab named "test_slab" of the size sizeof(struct
test_slab) and aligned to the alignment of test slab. If it fails then we
panic.
Signed-off-by: Christoph Lameter <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Replace all uses of kmem_cache_t with struct kmem_cache.
The patch was generated using the following script:
#!/bin/sh
#
# Replace one string by another in all the kernel sources.
#
set -e
for file in `find * -name "*.c" -o -name "*.h"|xargs grep -l $1`; do
quilt add $file
sed -e "1,\$s/$1/$2/g" $file >/tmp/$$
mv /tmp/$$ $file
quilt refresh
done
The script was run like this
sh replace kmem_cache_t "struct kmem_cache"
Signed-off-by: Christoph Lameter <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
SLAB_KERNEL is an alias of GFP_KERNEL.
Signed-off-by: Christoph Lameter <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Make the delayacct lock irqsave; this avoids the possible deadlock where
an interrupt is taken while holding the delayacct lock which needs to
take the delayacct lock.
Signed-off-by: Peter Zijlstra <[email protected]>
Acked-by: Oleg Nesterov <[email protected]>
Cc: Balbir Singh <[email protected]>
Cc: Shailabh Nagar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Cleanup allocation and freeing of tsk->delays used by delay accounting.
This solves two problems reported for delay accounting:
1. oops in __delayacct_blkio_ticks
http://www.uwsg.indiana.edu/hypermail/linux/kernel/0608.2/1844.html
Currently tsk->delays is getting freed too early in task exit which can
cause a NULL tsk->delays to get accessed via reading of /proc/<tgid>/stats.
The patch fixes this problem by freeing tsk->delays closer to when
task_struct itself is freed up. As a result, it also eliminates the use of
tsk->delays_lock which was only being used (inadequately) to safeguard
access to tsk->delays while a task was exiting.
2. Possible memory leak in kernel/delayacct.c
http://www.uwsg.indiana.edu/hypermail/linux/kernel/0608.2/1389.html
The patch cleans up tsk->delays allocations after a bad fork which was
missing earlier.
The patch has been tested to fix the problems listed above and stress
tested with rapid calls to delay accounting's taskstats command interface
(which is the other path that can access the same data, besides the /proc
interface causing the oops above).
Signed-off-by: Shailabh Nagar <[email protected]>
Cc: Balbir Singh <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Enable delay accounting by default so that feature gets coverage testing
without requiring special measures.
Earlier, it was off by default and had to be enabled via a boot time param.
This patch reverses the default behaviour to improve coverage testing. It
can be removed late in the kernel development cycle if its believed users
shouldn't have to incur any cost if they don't want delay accounting. Or
it can be retained forever if the utility of the stats is deemed common
enough to warrant keeping the feature on.
Signed-off-by: Shailabh Nagar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Export I/O delays seen by a task through /proc/<tgid>/stats for use in top
etc.
Note that delays for I/O done for swapping in pages (swapin I/O) is clubbed
together with all other I/O here (this is not the case in the netlink
interface where the swapin I/O is kept distinct)
[[email protected]: printk warning fix]
Signed-off-by: Shailabh Nagar <[email protected]>
Signed-off-by: Balbir Singh <[email protected]>
Cc: Jes Sorensen <[email protected]>
Cc: Peter Chubb <[email protected]>
Cc: Erich Focht <[email protected]>
Cc: Levent Serinol <[email protected]>
Cc: Jay Lan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Usage of taskstats interface by delay accounting.
Signed-off-by: Shailabh Nagar <[email protected]>
Signed-off-by: Balbir Singh <[email protected]>
Cc: Jes Sorensen <[email protected]>
Cc: Peter Chubb <[email protected]>
Cc: Erich Focht <[email protected]>
Cc: Levent Serinol <[email protected]>
Cc: Jay Lan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Unlike earlier iterations of the delay accounting patches, now delays are only
collected for the actual I/O waits rather than try and cover the delays seen
in I/O submission paths.
Account separately for block I/O delays incurred as a result of swapin page
faults whose frequency can be affected by the task/process' rss limit. Hence
swapin delays can act as feedback for rss limit changes independent of I/O
priority changes.
Signed-off-by: Shailabh Nagar <[email protected]>
Signed-off-by: Balbir Singh <[email protected]>
Cc: Jes Sorensen <[email protected]>
Cc: Peter Chubb <[email protected]>
Cc: Erich Focht <[email protected]>
Cc: Levent Serinol <[email protected]>
Cc: Jay Lan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|
|
Initialization code related to collection of per-task "delay" statistics which
measure how long it had to wait for cpu, sync block io, swapping etc. The
collection of statistics and the interface are in other patches. This patch
sets up the data structures and allows the statistics collection to be
disabled through a kernel boot parameter.
Signed-off-by: Shailabh Nagar <[email protected]>
Signed-off-by: Balbir Singh <[email protected]>
Cc: Jes Sorensen <[email protected]>
Cc: Peter Chubb <[email protected]>
Cc: Erich Focht <[email protected]>
Cc: Levent Serinol <[email protected]>
Cc: Jay Lan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
|