[PATCH 0/3] sched/numa: Update numa_balancing stats in /proc

e=1344 group_shared=0 numa_faults node=1 task_private=641 task_shared=65 group_private=641 group_shared=65 numa_faults node=2 task_private=512 task_shared=0 group_private=512 group_shared=0 numa_faults node=3 task_private=64 task_shared=1 group_private=64 group_shared=1 Srikar Dronamraju (3):

[PATCH 3/3] sched/numa: Fix numabalancing stats in /proc/pid/sched

o have absolute numbers since differential migrations between two accesses can be easily calculated. Signed-off-by: Srikar Dronamraju --- kernel/sched/debug.c | 38 +- kernel/sched/fair.c | 22 +- kernel/sched/sched.h | 10 +- 3

[PATCH 2/3] sched/numa: Show numa_group id in sched_debug task listings

Having the numa group id in /proc/sched_debug helps to see how the numa groups have spread across the system. Signed-off-by: Srikar Dronamraju --- kernel/sched/debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c index 315c68e

[PATCH 2/3] sched/numa: Show numa_group id in sched_debug task listings

Having the numa group id in /proc/sched_debug helps to see how the numa groups have spread across the system. Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/sched/debug.c b/kernel

[PATCH 3/3] sched/numa: Fix numabalancing stats in /proc/pid/sched

absolute numbers since differential migrations between two accesses can be easily calculated. Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/debug.c | 38 +- kernel/sched/fair.c | 22 +- kernel/sched/sched.h

[PATCH 0/3] sched/numa: Update numa_balancing stats in /proc

group_shared=0 numa_faults node=1 task_private=641 task_shared=65 group_private=641 group_shared=65 numa_faults node=2 task_private=512 task_shared=0 group_private=512 group_shared=0 numa_faults node=3 task_private=64 task_shared=1 group_private=64 group_shared=1 Srikar Dronamraju (3): sched

[PATCH 1/3] sched: Move print_cfs_rq declaration to kernel/sched/sched.h

Currently print_cfs_rq() is declared in include/linux/sched.h. However its not used outside kernel/sched. Hence move the declaration to kernel/sched/sched.h Also some functions are only available for CONFIG_SCHED_DEBUG. Hence move the declarations within #ifdef. Signed-off-by: Srikar Dronamraju

Regression in perf bench numa convergence stats

1e455f on top of tip/master seems to avoid the problem. The below patch fixes the problem. -- Thanks and Regards Srikar Dronamraju >8 >From 88199ad8a3d6495080eaa016b87a612bc742b1c4 Mon Sep 17 00:00:00 2001 From: Srikar Dronamraju Date: Wed,

Regression in perf bench numa convergence stats

, total 1.173 nsecs/byte/thread runtime 0.853 GB/sec/thread speed 54.562 GB/sec total speed Even reverting e1e455f on top of tip/master seems to avoid the problem. The below patch fixes the problem. -- Thanks and Regards Srikar Dronamraju 8

Re: [PATCH v2 2/4] sched:Consider imbalance_pct when comparing loads in numa_has_capacity

the workload 3 times (doing only time measurement) and report > the > stddev in a human readable form. > Thanks again for this hint. Wouldnt system time/ user time also matter? I guess once Mel did point out that it was important to make sure that system time and user time dont increase when e

Re: [PATCH v2 2/4] sched:Consider imbalance_pct when comparing loads in numa_has_capacity

form. Thanks again for this hint. Wouldnt system time/ user time also matter? I guess once Mel did point out that it was important to make sure that system time and user time dont increase when elapsed time decreases. But I cant find the email though. -- Thanks and Regards Srikar Dronamraju

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

+ numa_has_capacity fix + Rik's modified patch. (Rik's modified patch == I removed node_isset check before setting nid as the preferred node) -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in Please read the FA

Re: [PATCH v2 2/4] sched:Consider imbalance_pct when comparing loads in numa_has_capacity

* Rik van Riel [2015-06-16 10:39:13]: > On 06/16/2015 07:56 AM, Srikar Dronamraju wrote: > > This is consistent with all other load balancing instances where we > > absorb unfairness upto env->imbalance_pct. Absorbing unfairness upto > > env->imbalance_pct all

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

e. In such a case, we shouldnt be ruling out migrating the task to the node. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

dont regress in numa02 (probably a little less with modified patch) -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

) -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

+ numa_has_capacity fix + Rik's modified patch. (Rik's modified patch == I removed node_isset check before setting nid as the preferred node) -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in Please read the FAQ at http

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH v2 2/4] sched:Consider imbalance_pct when comparing loads in numa_has_capacity

* Rik van Riel r...@redhat.com [2015-06-16 10:39:13]: On 06/16/2015 07:56 AM, Srikar Dronamraju wrote: This is consistent with all other load balancing instances where we absorb unfairness upto env-imbalance_pct. Absorbing unfairness upto env-imbalance_pct allows to pull and retain task

[tip:sched/core] sched/debug: Add sum_sleep_runtime to /proc//sched

Commit-ID: 82a0d2762699b95d6ce4114d00dc1865df9b0df3 Gitweb: http://git.kernel.org/tip/82a0d2762699b95d6ce4114d00dc1865df9b0df3 Author: Srikar Dronamraju AuthorDate: Mon, 8 Jun 2015 13:40:41 +0530 Committer: Ingo Molnar CommitDate: Fri, 19 Jun 2015 10:03:11 +0200 sched/debug: Add

[tip:sched/core] sched/debug: Replace vruntime with wait_sum in / proc/sched_debug

Commit-ID: c5f3ab1c3b2e277cca6462415038dab02b4ad396 Gitweb: http://git.kernel.org/tip/c5f3ab1c3b2e277cca6462415038dab02b4ad396 Author: Srikar Dronamraju AuthorDate: Mon, 8 Jun 2015 13:40:40 +0530 Committer: Ingo Molnar CommitDate: Fri, 19 Jun 2015 10:03:11 +0200 sched/debug: Replace

[tip:sched/core] sched/debug: Properly format runnable tasks in / proc/sched_debug

Commit-ID: 33d6176eb12d1b0ae6d2f672b47367fd90726b91 Gitweb: http://git.kernel.org/tip/33d6176eb12d1b0ae6d2f672b47367fd90726b91 Author: Srikar Dronamraju AuthorDate: Mon, 8 Jun 2015 13:40:39 +0530 Committer: Ingo Molnar CommitDate: Fri, 19 Jun 2015 10:03:10 +0200 sched/debug: Properly

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

Max Avg StdDev %Change bopsperJVM: 266774.00 272434.00 269839.60 2083.19 +0.17% So fix for numa_has_capacity and always setting preferred node based on fault stats seems to help autonuma and specjbb. -- Thanks and Regards Srikar Dronamraju -- To un

[tip:sched/core] sched/debug: Replace vruntime with wait_sum in / proc/sched_debug

Commit-ID: c5f3ab1c3b2e277cca6462415038dab02b4ad396 Gitweb: http://git.kernel.org/tip/c5f3ab1c3b2e277cca6462415038dab02b4ad396 Author: Srikar Dronamraju sri...@linux.vnet.ibm.com AuthorDate: Mon, 8 Jun 2015 13:40:40 +0530 Committer: Ingo Molnar mi...@kernel.org CommitDate: Fri, 19 Jun

[tip:sched/core] sched/debug: Add sum_sleep_runtime to /proc/pid /sched

Commit-ID: 82a0d2762699b95d6ce4114d00dc1865df9b0df3 Gitweb: http://git.kernel.org/tip/82a0d2762699b95d6ce4114d00dc1865df9b0df3 Author: Srikar Dronamraju sri...@linux.vnet.ibm.com AuthorDate: Mon, 8 Jun 2015 13:40:41 +0530 Committer: Ingo Molnar mi...@kernel.org CommitDate: Fri, 19 Jun

[tip:sched/core] sched/debug: Properly format runnable tasks in / proc/sched_debug

Commit-ID: 33d6176eb12d1b0ae6d2f672b47367fd90726b91 Gitweb: http://git.kernel.org/tip/33d6176eb12d1b0ae6d2f672b47367fd90726b91 Author: Srikar Dronamraju sri...@linux.vnet.ibm.com AuthorDate: Mon, 8 Jun 2015 13:40:39 +0530 Committer: Ingo Molnar mi...@kernel.org CommitDate: Fri, 19 Jun

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

%Change bopsperJVM: 266774.00 272434.00 269839.60 2083.19 +0.17% So fix for numa_has_capacity and always setting preferred node based on fault stats seems to help autonuma and specjbb. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

gether. > -- Okay, I will do the needful and come back to you. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/major

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

t, on a two node system, both nodes end up in the numa_group's > active_mask. > Just to add this was on a 4 node machine. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

2155.02 1916.12 132.57 -4.38% Testcase: Min Max Avg StdDev %Change total_numa01:56940.5077128.2065935.92 7993.49 17.91% total_numa02: 1834.97 2227.03 1980.76 136.51 -3.99% -- Thanks and Rega

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

up in the numa_group's active_mask. Just to add this was on a 4 node machine. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

: Min Max Avg StdDev %Change total_numa01:56940.5077128.2065935.92 7993.49 17.91% total_numa02: 1834.97 2227.03 1980.76 136.51 -3.99% -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send

Re: [PATCH] sched,numa: document and fix numa_preferred_nid setting

the needful and come back to you. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http

Re: [PATCH v2 4/4] sched:Use correct nid while evaluating task weights

* Rik van Riel [2015-06-16 11:00:18]: > On 06/16/2015 07:56 AM, Srikar Dronamraju wrote: > > In task_numa_migrate(), while evaluating other nodes for group > > consolidation, env.dst_nid is used instead of using the iterator nid. > > Using env.dst_nid would mean dist is al

Re: [PATCH v2 3/4] sched:Fix task_numa_migrate to always update preferred node

sk from moving > to its preferred nid at a later time (when the good > nodes are no longer overloaded). > > Have you tested this patch with any workload that does > not consist of tasks that are running at 100% cpu time > for the duration of the test? > Okay, I will try to se

[PATCH v2 2/4] sched:Consider imbalance_pct when comparing loads in numa_has_capacity

This is consistent with all other load balancing instances where we absorb unfairness upto env->imbalance_pct. Absorbing unfairness upto env->imbalance_pct allows to pull and retain task to their preferred nodes. Signed-off-by: Srikar Dronamraju --- kernel/sched/fair.c | 5 +++-- 1 file c

[PATCH v2 1/4] sched/tip:Prefer numa hotness over cache hotness

the above, this commit merges migrate_degrades_locality() and migrate_improves_locality(). It also replaces 3 sched features NUMA, NUMA_FAVOUR_HIGHER and NUMA_RESIST_LOWER by a single sched feature NUMA. Acked-by: Rik van Riel Signed-off-by: Srikar Dronamraju --- Changes from previous version: - Rebased

[PATCH v2 3/4] sched:Fix task_numa_migrate to always update preferred node

e. Signed-off-by: Srikar Dronamraju --- kernel/sched/fair.c | 11 ++- 1 file changed, 2 insertions(+), 9 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 7b23efa..d1aa374 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -1503,7 +1503,7 @@

[PATCH v2 4/4] sched:Use correct nid while evaluating task weights

the iterator nid. Also the task/group weights from the src_nid should be calculated irrespective of numa topology type. Signed-off-by: Srikar Dronamraju --- kernel/sched/fair.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index d1aa374

[PATCH v2 0/4] Improve numa load balancing

20 10 Srikar Dronamraju (4): sched/tip:Prefer numa hotness over cache hotness sched: Consider imbalance_pct when comparing loads in numa_has_capacity sched: Fix task_numa_migrate to always update preferred node sched: Use correct nid while evaluating task weights kernel/sched/fair.c

[PATCH v2 0/4] Improve numa load balancing

20 10 Srikar Dronamraju (4): sched/tip:Prefer numa hotness over cache hotness sched: Consider imbalance_pct when comparing loads in numa_has_capacity sched: Fix task_numa_migrate to always update preferred node sched: Use correct nid while evaluating task weights kernel/sched/fair.c

[PATCH v2 2/4] sched:Consider imbalance_pct when comparing loads in numa_has_capacity

This is consistent with all other load balancing instances where we absorb unfairness upto env-imbalance_pct. Absorbing unfairness upto env-imbalance_pct allows to pull and retain task to their preferred nodes. Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/fair.c

[PATCH v2 1/4] sched/tip:Prefer numa hotness over cache hotness

the above, this commit merges migrate_degrades_locality() and migrate_improves_locality(). It also replaces 3 sched features NUMA, NUMA_FAVOUR_HIGHER and NUMA_RESIST_LOWER by a single sched feature NUMA. Acked-by: Rik van Riel r...@redhat.com Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com

[PATCH v2 3/4] sched:Fix task_numa_migrate to always update preferred node

-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/fair.c | 11 ++- 1 file changed, 2 insertions(+), 9 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 7b23efa..d1aa374 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -1503,7 +1503,7

[PATCH v2 4/4] sched:Use correct nid while evaluating task weights

the iterator nid. Also the task/group weights from the src_nid should be calculated irrespective of numa topology type. Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/fair.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/kernel/sched/fair.c b/kernel

Re: [PATCH v2 4/4] sched:Use correct nid while evaluating task weights

* Rik van Riel r...@redhat.com [2015-06-16 11:00:18]: On 06/16/2015 07:56 AM, Srikar Dronamraju wrote: In task_numa_migrate(), while evaluating other nodes for group consolidation, env.dst_nid is used instead of using the iterator nid. Using env.dst_nid would mean dist is always the same

Re: [PATCH v2 3/4] sched:Fix task_numa_migrate to always update preferred node

nodes are no longer overloaded). Have you tested this patch with any workload that does not consist of tasks that are running at 100% cpu time for the duration of the test? Okay, I will try to see if I can run some workloads and get back to you. -- Thanks and Regards Srikar Dronamraju

[PATCH] sched:Prefer numa hotness over cache hotness

the above, this commit merges migrate_degrades_locality() and migrate_improves_locality(). It also replaces 3 sched features NUMA, NUMA_FAVOUR_HIGHER and NUMA_RESIST_LOWER by a single sched feature NUMA. Signed-off-by: Srikar Dronamraju --- kernel/sched/fair.c | 96

Performance numbers with prefer numa hotness over cache hotness

: 31455 MB node distances: node 0 1 2 3 0: 10 20 40 40 1: 20 10 40 40 2: 40 40 10 20 3: 40 40 20 10 Srikar Dronamraju (1): sched:Prefer numa hotness over cache hotness kernel/sched/fair.c | 96 ++--- kernel/sched

[PATCH] sched:Prefer numa hotness over cache hotness

the above, this commit merges migrate_degrades_locality() and migrate_improves_locality(). It also replaces 3 sched features NUMA, NUMA_FAVOUR_HIGHER and NUMA_RESIST_LOWER by a single sched feature NUMA. Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/fair.c | 96

Performance numbers with prefer numa hotness over cache hotness

: 31455 MB node distances: node 0 1 2 3 0: 10 20 40 40 1: 20 10 40 40 2: 40 40 10 20 3: 40 40 20 10 Srikar Dronamraju (1): sched:Prefer numa hotness over cache hotness kernel/sched/fair.c | 96 ++--- kernel/sched

[PATCH 2/3] sched:Replace vruntime with wait_sum in /proc/sched_debug

Within runnable tasks in /proc/sched_debug, vruntime is printed twice, once as tree-key and again as exec-runtime. Since exec-runtime isnt populated in !CONFIG_SCHEDSTATS, use this field to print wait_sum. Signed-off-by: Srikar Dronamraju --- kernel/sched/debug.c | 4 ++-- 1 file changed, 2

[PATCH 0/3] : Fixes to make /proc/sched_debug better

120 0.012962 0.163522 17495.280072 0 / kworker/31:2 712 10136.506903 158 120 5.358470 2.269444 81162.182416 0 / kworker/31:1H 5904 10047.528224 6 100 0.028400 0.207288 20745.582406 0 / Srikar

[PATCH 3/3] sched:Add sum_sleep_runtime to /proc//sched

When CONFIG_SCHEDSTATS is enabled, /proc//sched prints almost all sched statistics except sum_sleep_runtime. Since sum_sleep_runtime is a good info to collect, add this it to /proc//sched. Signed-off-by: Srikar Dronamraju --- kernel/sched/debug.c | 1 + 1 file changed, 1 insertion(+) diff

[PATCH 1/3] sched:Properly format runnable tasks in /proc/sched_debug

With !CONFIG_SCHEDSTATS, runnable tasks in /proc/sched_debug has too many columns than required. Fix this by printing appropriate columns. While at this, print sum_exec_runtime, since this information is available even in !CONFIG_SCHEDSTATS case. Signed-off-by: Srikar Dronamraju --- kernel

Re: [PATCH 0/3] : Fixes to make /proc/sched_debug better

* Peter Zijlstra [2015-06-08 09:06:40]: > On Thu, May 28, 2015 at 03:56:16PM +0530, Srikar Dronamraju wrote: > > > > Srikar Dronamraju (3): > > sched:Properly format runnable tasks in /proc/sched_debug > > sched:Replace vruntime with wait_sum in /proc/

[PATCH 2/3] sched:Replace vruntime with wait_sum in /proc/sched_debug

Within runnable tasks in /proc/sched_debug, vruntime is printed twice, once as tree-key and again as exec-runtime. Since exec-runtime isnt populated in !CONFIG_SCHEDSTATS, use this field to print wait_sum. Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/debug.c | 4

[PATCH 0/3] : Fixes to make /proc/sched_debug better

120 0.012962 0.163522 17495.280072 0 / kworker/31:2 712 10136.506903 158 120 5.358470 2.269444 81162.182416 0 / kworker/31:1H 5904 10047.528224 6 100 0.028400 0.207288 20745.582406 0 / Srikar

Re: [PATCH 0/3] : Fixes to make /proc/sched_debug better

* Peter Zijlstra pet...@infradead.org [2015-06-08 09:06:40]: On Thu, May 28, 2015 at 03:56:16PM +0530, Srikar Dronamraju wrote: Srikar Dronamraju (3): sched:Properly format runnable tasks in /proc/sched_debug sched:Replace vruntime with wait_sum in /proc/sched_debug sched:Add

[PATCH 1/3] sched:Properly format runnable tasks in /proc/sched_debug

With !CONFIG_SCHEDSTATS, runnable tasks in /proc/sched_debug has too many columns than required. Fix this by printing appropriate columns. While at this, print sum_exec_runtime, since this information is available even in !CONFIG_SCHEDSTATS case. Signed-off-by: Srikar Dronamraju sri

[PATCH 3/3] sched:Add sum_sleep_runtime to /proc/pid/sched

When CONFIG_SCHEDSTATS is enabled, /proc/pid/sched prints almost all sched statistics except sum_sleep_runtime. Since sum_sleep_runtime is a good info to collect, add this it to /proc/pid/sched. Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- kernel/sched/debug.c | 1 + 1 file

Re: [PATCH RESEND] sched: prefer an idle cpu vs an idle sibling for BALANCE_WAKE

> At Facebook we have a pretty heavily multi-threaded application that is > sensitive to latency. We have been pulling forward the old SD_WAKE_IDLE code > because it gives us a pretty significant performance gain (like 20%). It > turns > out this is because there are cases where the scheduler

Re: [PATCH 2/2] numa,sched: only consider less busy nodes as numa balancing destination

if (src->load * dst->compute_capacity > > + dst->load * src->compute_capacity) > + return true; > + > + return false; > +} > + > -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-

[PATCH 0/3] : Fixes to make /proc/sched_debug better

120 0.012962 0.163522 17495.280072 0 / kworker/31:2 712 10136.506903 158 120 5.358470 2.269444 81162.182416 0 / kworker/31:1H 5904 10047.528224 6 100 0.028400 0.207288 20745.582406 0 / Srikar

[PATCH 1/3] sched:Properly format runnable tasks in /proc/sched_debug

With !CONFIG_SCHEDSTATS, runnable tasks in /proc/sched_debug has too many columns than required. Fix this by printing appropriate columns. While at this, print sum_exec_runtime, since this information is available even in !CONFIG_SCHEDSTATS case. --- kernel/sched/debug.c | 6 -- 1 file

[PATCH 2/3] sched:Replace vruntime with wait_sum in /proc/sched_debug

Within runnable tasks in /proc/sched_debug, vruntime is printed twice, once as tree-key and again as exec-runtime. Since exec-runtime isnt populated in !CONFIG_SCHEDSTATS, use this field to print wait_sum. --- kernel/sched/debug.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff

[PATCH 3/3] sched:Add sum_sleep_runtime to /proc//sched

When CONFIG_SCHEDSTATS is enabled, /proc//sched prints almost all sched statistics except sum_sleep_runtime. Since sum_sleep_runtime is a good info to collect, add this it to /proc//sched. --- kernel/sched/debug.c | 1 + 1 file changed, 1 insertion(+) diff --git a/kernel/sched/debug.c

[PATCH 0/3] : Fixes to make /proc/sched_debug better

120 0.012962 0.163522 17495.280072 0 / kworker/31:2 712 10136.506903 158 120 5.358470 2.269444 81162.182416 0 / kworker/31:1H 5904 10047.528224 6 100 0.028400 0.207288 20745.582406 0 / Srikar

[PATCH 1/3] sched:Properly format runnable tasks in /proc/sched_debug

With !CONFIG_SCHEDSTATS, runnable tasks in /proc/sched_debug has too many columns than required. Fix this by printing appropriate columns. While at this, print sum_exec_runtime, since this information is available even in !CONFIG_SCHEDSTATS case. --- kernel/sched/debug.c | 6 -- 1 file

[PATCH 3/3] sched:Add sum_sleep_runtime to /proc/pid/sched

When CONFIG_SCHEDSTATS is enabled, /proc/pid/sched prints almost all sched statistics except sum_sleep_runtime. Since sum_sleep_runtime is a good info to collect, add this it to /proc/pid/sched. --- kernel/sched/debug.c | 1 + 1 file changed, 1 insertion(+) diff --git a/kernel/sched/debug.c

[PATCH 2/3] sched:Replace vruntime with wait_sum in /proc/sched_debug

Within runnable tasks in /proc/sched_debug, vruntime is printed twice, once as tree-key and again as exec-runtime. Since exec-runtime isnt populated in !CONFIG_SCHEDSTATS, use this field to print wait_sum. --- kernel/sched/debug.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff

Re: [PATCH RESEND] sched: prefer an idle cpu vs an idle sibling for BALANCE_WAKE

At Facebook we have a pretty heavily multi-threaded application that is sensitive to latency. We have been pulling forward the old SD_WAKE_IDLE code because it gives us a pretty significant performance gain (like 20%). It turns out this is because there are cases where the scheduler puts

Re: [PATCH 2/2] numa,sched: only consider less busy nodes as numa balancing destination

; +} + -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/

Re: [RFC][PATCH 4/4] sched, numa: Ignore pinned tasks

* Peter Zijlstra [2015-05-18 15:06:58]: > On Mon, 2015-05-18 at 18:30 +0530, Srikar Dronamraju wrote: > > > > > > static void account_numa_dequeue(struct rq *rq, struct task_struct *p) > > > { > > > + if (p->nr_cpus_allowed ==

Re: [RFC][PATCH 4/4] sched, numa: Ignore pinned tasks

+ * - movable: there are (migratable) tasks > + * - all: there are tasks >* >* In order to avoid migrating ideally placed numa tasks, >* ignore those when there's better options. -- Thanks and Regards Srik

Re: [RFC][PATCH 4/4] sched, numa: Ignore pinned tasks

its numa faults but then is pinned to a different cpu, then we can see this warning.:w! -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at

Re: [RFC][PATCH 4/4] sched, numa: Ignore pinned tasks

* Peter Zijlstra pet...@infradead.org [2015-05-18 15:06:58]: On Mon, 2015-05-18 at 18:30 +0530, Srikar Dronamraju wrote: static void account_numa_dequeue(struct rq *rq, struct task_struct *p) { + if (p-nr_cpus_allowed == 1) { + rq-nr_pinned_running

Re: [RFC][PATCH 4/4] sched, numa: Ignore pinned tasks

this warning.:w! -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org

Re: [RFC][PATCH 4/4] sched, numa: Ignore pinned tasks

+ * - all: there are tasks * * In order to avoid migrating ideally placed numa tasks, * ignore those when there's better options. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

we can always cleanup > this later, this is trivial. Right now I'd like to ensure that if the > same or similar logic can work on powerpc, it only needs to touch the > code in arch/powerpc. > > Oleg. > -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

with (say) powerpc, we can always cleanup this later, this is trivial. Right now I'd like to ensure that if the same or similar logic can work on powerpc, it only needs to touch the code in arch/powerpc. Oleg. -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

even seem to use this assumption when kprobe_tracer/uprobe_tracer fetch arguments from stack. See fetch_kernel_stack_address() / fetch_user_stack_address() and get_user_stack_nth(). -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linu

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

fetch_kernel_stack_address() / fetch_user_stack_address() and get_user_stack_nth(). -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo

Re: [PATCH 09/10] uprobes: Change prepare_uretprobe() to (try to) flush the dead frames

() can be false positive, > the stack can grow after longjmp(). Unfortunately, the kernel can't > 100% solve this problem, but see the next patch. > > Signed-off-by: Oleg Nesterov > --- > kernel/events/uprobes.c | 13 +++++ Looks good to me. Acked-by: Srikar Dronamra

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

(regs) <= sp; } Am I missing something? > -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH 08/10] uprobes: Change handle_trampoline() to flush the frames invalidated by longjmp()

e() > can be fooled by sigaltstack/etc. > > Signed-off-by: Oleg Nesterov > --- Looks good to me. Acked-by: Srikar Dronamraju -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message t

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

know that the probed func > has already returned. > > TODO: this assumes that the probed app can't use multiple stacks (say > sigaltstack). We will try to improve this logic later. > > Signed-off-by: Oleg Nesterov Looks good to me. Acked-by: Srikar Dronamraju >

Re: [PATCH 06/10] uprobes: Introduce struct arch_uretprobe

rov > --- Looks good to me. Acked-by: Srikar Dronamraju -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info

Re: [PATCH 05/10] uprobes: Change handle_trampoline() to find the next chain beforehand

erov Looks good to me. Acked-by: Srikar Dronamraju -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH 04/10] uprobes: Change prepare_uretprobe() to use uprobe_warn()

esterov Looks good to me. Acked-by: Srikar Dronamraju > --- -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

that the probed func has already returned. TODO: this assumes that the probed app can't use multiple stacks (say sigaltstack). We will try to improve this logic later. Signed-off-by: Oleg Nesterov o...@redhat.com Looks good to me. Acked-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- -- Thanks

Re: [PATCH 09/10] uprobes: Change prepare_uretprobe() to (try to) flush the dead frames

be false positive, the stack can grow after longjmp(). Unfortunately, the kernel can't 100% solve this problem, but see the next patch. Signed-off-by: Oleg Nesterov o...@redhat.com --- kernel/events/uprobes.c | 13 + Looks good to me. Acked-by: Srikar Dronamraju sri

Re: [PATCH 04/10] uprobes: Change prepare_uretprobe() to use uprobe_warn()

Looks good to me. Acked-by: Srikar Dronamraju sri...@linux.vnet.ibm.com --- -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org

Re: [PATCH 05/10] uprobes: Change handle_trampoline() to find the next chain beforehand

-by: Srikar Dronamraju sri...@linux.vnet.ibm.com -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ

Re: [PATCH 07/10] uprobes/x86: Introduce arch_uretprobe_is_alive()

and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/

Re: [PATCH 08/10] uprobes: Change handle_trampoline() to flush the frames invalidated by longjmp()

to me. Acked-by: Srikar Dronamraju sri...@linux.vnet.ibm.com -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: [PATCH 06/10] uprobes: Introduce struct arch_uretprobe

to me. Acked-by: Srikar Dronamraju sri...@linux.vnet.ibm.com -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: [PATCH 03/10] uprobes: Send SIGILL if handle_trampoline() fails

It is pointless to return to user mode >with the corrupted instruction_pointer() which we can't restore. I agree > > Signed-off-by: Oleg Nesterov > --- Acked-by: Srikar Dronamraju -- Thanks and Regards Srikar Dronamraju -- To unsubscribe from this list: send the line "unsub

Re: [PATCH 02/10] uprobes: Introduce free_ret_instance()

* Oleg Nesterov [2015-05-04 14:48:54]: > We can simplify uprobe_free_utask() and handle_uretprobe_chain() > if we add a simple helper which does put_uprobe/kfree and returns > the ->next return_instance. > > Signed-off-by: Oleg Nesterov > --- Looks good to me. Acked-

Re: [PATCH 01/10] uprobes: Introduce get_uprobe()

* Oleg Nesterov [2015-05-04 14:48:50]: > Cosmetic. Add the new trivial helper, get_uprobe(). It matches > put_uprobe() we already have and we can simplify a couple of its > users. > > Signed-off-by: Oleg Nesterov Looks good to me. Acked-by: Srikar Dronamraju -- To u

<    3   4   5   6   7   8   9   10   11   12   >