Re: [tip:sched/core] sched/fair: Fix group power_orig computation
On Tue, Nov 12, 2013 at 04:25:47PM +0530, Srikar Dronamraju wrote:
> From: Srikar Dronamraju
> Date: Tue, 12 Nov 2013 03:05:31 -0500
> Subject: [PATCH] sched: Check sched_domain before computing group power.
>
> After Commit-id 863bffc80898 (sched/fair: Fix group power_orig
> computation), we might end up computing group power before the
> sched_domain for a cpu is updated.
>
> Check for rq->sd before updating group power.
>
> Signed-off-by: Srikar Dronamraju
Thanks! My head hurts a bit from going through the overlap init paths
but this does indeed revert to the previous behaviour.
I'm not entirely sure if either are fully correct, but given we
initialize the sgp->power to some 'reasonable' default we can rely on
runtime updates to correct any funnies.
> ---
> kernel/sched/fair.c |7 ++-
> 1 files changed, 6 insertions(+), 1 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index df77c60..f86f704 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -5354,8 +5354,13 @@ void update_group_power(struct sched_domain *sd, int
> cpu)
>*/
>
> for_each_cpu(cpu, sched_group_cpus(sdg)) {
> - struct sched_group *sg = cpu_rq(cpu)->sd->groups;
> + struct rq *rq = cpu_rq(cpu);
> + struct sched_group *sg;
>
> + if (!rq->sd)
> + continue;
> +
> + sg = rq->sd->groups;
> power_orig += sg->sgp->power_orig;
> power += sg->sgp->power;
> }
> --
> 1.7.1
>
--
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: [tip:sched/core] sched/fair: Fix group power_orig computation
With Commit-id 863bffc80898 (sched/fair: Fix group power_orig computation) which is part of latest tip/master, numa machine may fail to boot both on powerpc and x86_64. On powerpc [0.710162] Unable to handle kernel paging request for data at address 0x0010 [0.710170] Faulting instruction address: 0xc00f3db4 [0.710177] Oops: Kernel access of bad area, sig: 11 [#1] [0.710182] SMP NR_CPUS=1024 NUMA pSeries [0.710190] Modules linked in: [0.710199] CPU: 53 PID: 1 Comm: swapper/53 Not tainted 3.12.0-tip_master+ #1 [0.710205] task: c0171398 ti: c01713a0 task.ti: c01713a0 [0.710211] NIP: c00f3db4 LR: c00f3de0 CTR: [0.710217] REGS: c01713a036d0 TRAP: 0300 Not tainted (3.12.0-tip_master+) [0.710223] MSR: 80009032 CR: 48002044 XER: 2003 [0.710243] SOFTE: 1 [0.710246] CFAR: c000908c [0.710250] DAR: 0010, DSISR: 4000 [0.710254] GPR00: c0e85e00 c01713a03950 c15a7a58 0030 GPR04: 0030 0400 1dcd6500 GPR08: c160c958 GPR12: c7edb980 c007e2e374a8 c1607a58 GPR16: c01ef38f8f58 0180 c1770258 0001 GPR20: 005f c01ef38f8f40 0001 c14b9acf GPR24: c01712884a40 c1607a58 c0e87a58 GPR28: c160fc54 c01712884a58 [0.710349] NIP [c00f3db4] .update_group_power+0xd4/0x2e0 [0.710355] LR [c00f3de0] .update_group_power+0x100/0x2e0 [0.710360] Call Trace: [0.710364] [c01713a03950] [c00f3d28] .update_group_power+0x48/0x2e0 (unreliable) [0.710375] [c01713a03a00] [c00ed73c] .build_sched_domains+0xadc/0xd90 [0.710385] [c01713a03b70] [c0bf39b0] .sched_init_smp+0x528/0x66c [0.710394] [c01713a03ce0] [c0bd46a8] .kernel_init_freeable+0x200/0x398 [0.710405] [c01713a03db0] [c000bc04] .kernel_init+0x24/0x140 [0.710413] [c01713a03e30] [c000a16c] .ret_from_kernel_thread+0x5c/0x70 [0.710419] Instruction dump: [0.710424] 3bb90018 3bc0 3be0 3860 3f62ff8e 4834 6000 397a4f00 [0.710439] 381be3a8 7d2b482a 7d204a14 e9290950 e9290010 81690004 80090008 [0.710465] ---[ end trace b5091a0959b24fe3 ]--- [0.710469] [2.710530] Kernel panic - not syncing: Fatal exception [2.710943] Rebooting in 10 seconds.. On x86_64 CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.12.0-tip_v312+ #5 Hardware name: IBM System x3750 M4 -[8722C1A]-/00D1432, BIOS -[KOE116JUS-1.10]- 09/25/2012 task: 8810373a6040 ti: 8810373a8000 task.ti: 8810373a8000 RIP: 0010:[] [] update_group_power+0xa3/0x130 RSP: :8810373a9db8 EFLAGS: 00010283 RAX: 0008 RBX: RCX: RDX: RSI: 0040 RDI: 0040 RBP: 8810373a9de8 R08: 88203632f818 R09: R10: 0001 R11: 0001 R12: R13: 001d3cc0 R14: 88203632f818 R15: 88203632f800 FS: () GS:88103de0() knlGS: CS: 0010 DS: ES: CR0: 80050033 CR2: 0010 CR3: 01a0b000 CR4: 000407f0 Stack: 8810373a9dd8 881035aa0c00 88203632f800 8820362fa800 0008 0008 8810373a9e58 8108627b 8810373a9e18 8129fcf8 373a76e0 003f Call Trace: [] build_sched_domains+0x37b/0x3f0 [] ? alloc_cpumask_var_node+0x58/0x80 [] sched_init_smp+0x8f/0x13c [] kernel_init_freeable+0x27b/0x303 [] ? kernel_init+0xe/0xf0 [] ? trace_hardirqs_on_caller+0xfd/0x1c0 [] ? rest_init+0xd0/0xd0 [] kernel_init+0xe/0xf0 [] ret_from_fork+0x7c/0xb0 [] ? rest_init+0xd0/0xd0 Code: ad 00 4d 8d 77 18 45 31 e4 31 db b8 ff ff ff ff eb 2d 66 0f 1f 44 00 00 48 63 d0 48 8b 14 d5 c0 9d b4 81 49 8b 94 15 50 09 00 00 <48> 8b 52 10 48 8b 52 10 8b 4a 08 8b 52 04 49 01 cc 48 01 d3 8 3 RIP [] update_group_power+0xa3/0x130 RSP CR2: 0010 ---[ end trace cd8cb7fb261d7bea ]--- Kernel panic - not syncing: Fatal exception This can be fixed by a simple check below. -- Thanks and Regards Srikar Dronamraju >8- >From bfc5dced04472c6c499aa3c6773ddef42d83fefc Mon Sep 17 00:00:00 2001 From: Srikar Dronamraju Date: Tue, 12 Nov 2013 03:05:31 -0500 Subject: [PATCH] sched: Check sched_domain before computing group power. After Commit-id 863bffc80898 (sched/fair: Fix group power_orig computation), we might end up computing group power before the sched_domain for a cpu is updated. Check for rq->sd before updating group power. Signed-off-by: Srikar Dronamraju
Re: [tip:sched/core] sched/fair: Fix group power_orig computation
With Commit-id 863bffc80898 (sched/fair: Fix group power_orig computation) which is part of latest tip/master, numa machine may fail to boot both on powerpc and x86_64. On powerpc [0.710162] Unable to handle kernel paging request for data at address 0x0010 [0.710170] Faulting instruction address: 0xc00f3db4 [0.710177] Oops: Kernel access of bad area, sig: 11 [#1] [0.710182] SMP NR_CPUS=1024 NUMA pSeries [0.710190] Modules linked in: [0.710199] CPU: 53 PID: 1 Comm: swapper/53 Not tainted 3.12.0-tip_master+ #1 [0.710205] task: c0171398 ti: c01713a0 task.ti: c01713a0 [0.710211] NIP: c00f3db4 LR: c00f3de0 CTR: [0.710217] REGS: c01713a036d0 TRAP: 0300 Not tainted (3.12.0-tip_master+) [0.710223] MSR: 80009032 SF,EE,ME,IR,DR,RI CR: 48002044 XER: 2003 [0.710243] SOFTE: 1 [0.710246] CFAR: c000908c [0.710250] DAR: 0010, DSISR: 4000 [0.710254] GPR00: c0e85e00 c01713a03950 c15a7a58 0030 GPR04: 0030 0400 1dcd6500 GPR08: c160c958 GPR12: c7edb980 c007e2e374a8 c1607a58 GPR16: c01ef38f8f58 0180 c1770258 0001 GPR20: 005f c01ef38f8f40 0001 c14b9acf GPR24: c01712884a40 c1607a58 c0e87a58 GPR28: c160fc54 c01712884a58 [0.710349] NIP [c00f3db4] .update_group_power+0xd4/0x2e0 [0.710355] LR [c00f3de0] .update_group_power+0x100/0x2e0 [0.710360] Call Trace: [0.710364] [c01713a03950] [c00f3d28] .update_group_power+0x48/0x2e0 (unreliable) [0.710375] [c01713a03a00] [c00ed73c] .build_sched_domains+0xadc/0xd90 [0.710385] [c01713a03b70] [c0bf39b0] .sched_init_smp+0x528/0x66c [0.710394] [c01713a03ce0] [c0bd46a8] .kernel_init_freeable+0x200/0x398 [0.710405] [c01713a03db0] [c000bc04] .kernel_init+0x24/0x140 [0.710413] [c01713a03e30] [c000a16c] .ret_from_kernel_thread+0x5c/0x70 [0.710419] Instruction dump: [0.710424] 3bb90018 3bc0 3be0 3860 3f62ff8e 4834 6000 397a4f00 [0.710439] 381be3a8 7d2b482a 7d204a14 e9290950 e9290010 e9290010 81690004 80090008 [0.710465] ---[ end trace b5091a0959b24fe3 ]--- [0.710469] [2.710530] Kernel panic - not syncing: Fatal exception [2.710943] Rebooting in 10 seconds.. On x86_64 CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.12.0-tip_v312+ #5 Hardware name: IBM System x3750 M4 -[8722C1A]-/00D1432, BIOS -[KOE116JUS-1.10]- 09/25/2012 task: 8810373a6040 ti: 8810373a8000 task.ti: 8810373a8000 RIP: 0010:[8108aa53] [8108aa53] update_group_power+0xa3/0x130 RSP: :8810373a9db8 EFLAGS: 00010283 RAX: 0008 RBX: RCX: RDX: RSI: 0040 RDI: 0040 RBP: 8810373a9de8 R08: 88203632f818 R09: R10: 0001 R11: 0001 R12: R13: 001d3cc0 R14: 88203632f818 R15: 88203632f800 FS: () GS:88103de0() knlGS: CS: 0010 DS: ES: CR0: 80050033 CR2: 0010 CR3: 01a0b000 CR4: 000407f0 Stack: 8810373a9dd8 881035aa0c00 88203632f800 8820362fa800 0008 0008 8810373a9e58 8108627b 8810373a9e18 8129fcf8 373a76e0 003f Call Trace: [8108627b] build_sched_domains+0x37b/0x3f0 [8129fcf8] ? alloc_cpumask_var_node+0x58/0x80 [81d59783] sched_init_smp+0x8f/0x13c [81d3db78] kernel_init_freeable+0x27b/0x303 [81593c6e] ? kernel_init+0xe/0xf0 [810a0ead] ? trace_hardirqs_on_caller+0xfd/0x1c0 [81593c60] ? rest_init+0xd0/0xd0 [81593c6e] kernel_init+0xe/0xf0 [815a9eac] ret_from_fork+0x7c/0xb0 [81593c60] ? rest_init+0xd0/0xd0 Code: ad 00 4d 8d 77 18 45 31 e4 31 db b8 ff ff ff ff eb 2d 66 0f 1f 44 00 00 48 63 d0 48 8b 14 d5 c0 9d b4 81 49 8b 94 15 50 09 00 00 48 8b 52 10 48 8b 52 10 8b 4a 08 8b 52 04 49 01 cc 48 01 d3 8 3 RIP [8108aa53] update_group_power+0xa3/0x130 RSP 8810373a9db8 CR2: 0010 ---[ end trace cd8cb7fb261d7bea ]--- Kernel panic - not syncing: Fatal exception This can be fixed by a simple check below. -- Thanks and Regards Srikar Dronamraju 8- From bfc5dced04472c6c499aa3c6773ddef42d83fefc Mon Sep 17 00:00:00 2001 From: Srikar Dronamraju sri...@linux.vnet.ibm.com Date: Tue, 12 Nov 2013 03:05:31 -0500 Subject: [PATCH] sched: Check sched_domain
Re: [tip:sched/core] sched/fair: Fix group power_orig computation
On Tue, Nov 12, 2013 at 04:25:47PM +0530, Srikar Dronamraju wrote:
From: Srikar Dronamraju sri...@linux.vnet.ibm.com
Date: Tue, 12 Nov 2013 03:05:31 -0500
Subject: [PATCH] sched: Check sched_domain before computing group power.
After Commit-id 863bffc80898 (sched/fair: Fix group power_orig
computation), we might end up computing group power before the
sched_domain for a cpu is updated.
Check for rq-sd before updating group power.
Signed-off-by: Srikar Dronamraju sri...@linux.vnet.ibm.com
Thanks! My head hurts a bit from going through the overlap init paths
but this does indeed revert to the previous behaviour.
I'm not entirely sure if either are fully correct, but given we
initialize the sgp-power to some 'reasonable' default we can rely on
runtime updates to correct any funnies.
---
kernel/sched/fair.c |7 ++-
1 files changed, 6 insertions(+), 1 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index df77c60..f86f704 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5354,8 +5354,13 @@ void update_group_power(struct sched_domain *sd, int
cpu)
*/
for_each_cpu(cpu, sched_group_cpus(sdg)) {
- struct sched_group *sg = cpu_rq(cpu)-sd-groups;
+ struct rq *rq = cpu_rq(cpu);
+ struct sched_group *sg;
+ if (!rq-sd)
+ continue;
+
+ sg = rq-sd-groups;
power_orig += sg-sgp-power_orig;
power += sg-sgp-power;
}
--
1.7.1
--
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: [tip:sched/core] sched/fair: Fix group power_orig computation
tip-bot for Peter Zijlstra wrote:
> Commit-ID: 863bffc80898b8df295ebac111af2335ec05f85d
> Gitweb: http://git.kernel.org/tip/863bffc80898b8df295ebac111af2335ec05f85d
> Author: Peter Zijlstra
> AuthorDate: Wed, 28 Aug 2013 11:44:39 +0200
> Committer: Ingo Molnar
> CommitDate: Thu, 12 Sep 2013 19:14:43 +0200
>
> sched/fair: Fix group power_orig computation
>
> When looking at the code I noticed we don't actually compute
> sgp->power_orig correctly for groups, fix that.
>
> Currently the only consumer of that value is fix_small_capacity()
> which is only used on POWER7+ and that code excludes this case by
> being limited to SD_SHARE_CPUPOWER which is only ever set on the SMT
> domain which must be the lowest domain and this has singleton groups.
>
> So nothing should be affected by this change.
>
> Cc: Michael Neuling
> Signed-off-by: Peter Zijlstra
> Link: http://lkml.kernel.org/n/tip-db2pe0vxwunv37plc7onn...@git.kernel.org
> Signed-off-by: Ingo Molnar
FWIW, this doesn't seem to break POWER7.
Thanks!
Mikey
> ---
> kernel/sched/fair.c | 16 +++-
> 1 file changed, 11 insertions(+), 5 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index f9f4385..baba313 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -4450,7 +4450,7 @@ void update_group_power(struct sched_domain *sd, int
> cpu)
> {
> struct sched_domain *child = sd->child;
> struct sched_group *group, *sdg = sd->groups;
> - unsigned long power;
> + unsigned long power, power_orig;
> unsigned long interval;
>
> interval = msecs_to_jiffies(sd->balance_interval);
> @@ -4462,7 +4462,7 @@ void update_group_power(struct sched_domain *sd, int
> cpu)
> return;
> }
>
> - power = 0;
> + power_orig = power = 0;
>
> if (child->flags & SD_OVERLAP) {
> /*
> @@ -4470,8 +4470,12 @@ void update_group_power(struct sched_domain *sd, int
> cpu)
>* span the current group.
>*/
>
> - for_each_cpu(cpu, sched_group_cpus(sdg))
> - power += power_of(cpu);
> + for_each_cpu(cpu, sched_group_cpus(sdg)) {
> + struct sched_group *sg = cpu_rq(cpu)->sd->groups;
> +
> + power_orig += sg->sgp->power_orig;
> + power += sg->sgp->power;
> + }
> } else {
> /*
>* !SD_OVERLAP domains can assume that child groups
> @@ -4480,12 +4484,14 @@ void update_group_power(struct sched_domain *sd, int
> cpu)
>
> group = child->groups;
> do {
> + power_orig += group->sgp->power_orig;
> power += group->sgp->power;
> group = group->next;
> } while (group != child->groups);
> }
>
> - sdg->sgp->power_orig = sdg->sgp->power = power;
> + sdg->sgp->power_orig = power_orig;
> + sdg->sgp->power = power;
> }
>
> /*
>
--
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/
[tip:sched/core] sched/fair: Fix group power_orig computation
Commit-ID: 863bffc80898b8df295ebac111af2335ec05f85d
Gitweb: http://git.kernel.org/tip/863bffc80898b8df295ebac111af2335ec05f85d
Author: Peter Zijlstra
AuthorDate: Wed, 28 Aug 2013 11:44:39 +0200
Committer: Ingo Molnar
CommitDate: Thu, 12 Sep 2013 19:14:43 +0200
sched/fair: Fix group power_orig computation
When looking at the code I noticed we don't actually compute
sgp->power_orig correctly for groups, fix that.
Currently the only consumer of that value is fix_small_capacity()
which is only used on POWER7+ and that code excludes this case by
being limited to SD_SHARE_CPUPOWER which is only ever set on the SMT
domain which must be the lowest domain and this has singleton groups.
So nothing should be affected by this change.
Cc: Michael Neuling
Signed-off-by: Peter Zijlstra
Link: http://lkml.kernel.org/n/tip-db2pe0vxwunv37plc7onn...@git.kernel.org
Signed-off-by: Ingo Molnar
---
kernel/sched/fair.c | 16 +++-
1 file changed, 11 insertions(+), 5 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index f9f4385..baba313 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4450,7 +4450,7 @@ void update_group_power(struct sched_domain *sd, int cpu)
{
struct sched_domain *child = sd->child;
struct sched_group *group, *sdg = sd->groups;
- unsigned long power;
+ unsigned long power, power_orig;
unsigned long interval;
interval = msecs_to_jiffies(sd->balance_interval);
@@ -4462,7 +4462,7 @@ void update_group_power(struct sched_domain *sd, int cpu)
return;
}
- power = 0;
+ power_orig = power = 0;
if (child->flags & SD_OVERLAP) {
/*
@@ -4470,8 +4470,12 @@ void update_group_power(struct sched_domain *sd, int cpu)
* span the current group.
*/
- for_each_cpu(cpu, sched_group_cpus(sdg))
- power += power_of(cpu);
+ for_each_cpu(cpu, sched_group_cpus(sdg)) {
+ struct sched_group *sg = cpu_rq(cpu)->sd->groups;
+
+ power_orig += sg->sgp->power_orig;
+ power += sg->sgp->power;
+ }
} else {
/*
* !SD_OVERLAP domains can assume that child groups
@@ -4480,12 +4484,14 @@ void update_group_power(struct sched_domain *sd, int
cpu)
group = child->groups;
do {
+ power_orig += group->sgp->power_orig;
power += group->sgp->power;
group = group->next;
} while (group != child->groups);
}
- sdg->sgp->power_orig = sdg->sgp->power = power;
+ sdg->sgp->power_orig = power_orig;
+ sdg->sgp->power = power;
}
/*
--
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/
[tip:sched/core] sched/fair: Fix group power_orig computation
Commit-ID: 863bffc80898b8df295ebac111af2335ec05f85d
Gitweb: http://git.kernel.org/tip/863bffc80898b8df295ebac111af2335ec05f85d
Author: Peter Zijlstra pet...@infradead.org
AuthorDate: Wed, 28 Aug 2013 11:44:39 +0200
Committer: Ingo Molnar mi...@kernel.org
CommitDate: Thu, 12 Sep 2013 19:14:43 +0200
sched/fair: Fix group power_orig computation
When looking at the code I noticed we don't actually compute
sgp-power_orig correctly for groups, fix that.
Currently the only consumer of that value is fix_small_capacity()
which is only used on POWER7+ and that code excludes this case by
being limited to SD_SHARE_CPUPOWER which is only ever set on the SMT
domain which must be the lowest domain and this has singleton groups.
So nothing should be affected by this change.
Cc: Michael Neuling mi...@neuling.org
Signed-off-by: Peter Zijlstra pet...@infradead.org
Link: http://lkml.kernel.org/n/tip-db2pe0vxwunv37plc7onn...@git.kernel.org
Signed-off-by: Ingo Molnar mi...@kernel.org
---
kernel/sched/fair.c | 16 +++-
1 file changed, 11 insertions(+), 5 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index f9f4385..baba313 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4450,7 +4450,7 @@ void update_group_power(struct sched_domain *sd, int cpu)
{
struct sched_domain *child = sd-child;
struct sched_group *group, *sdg = sd-groups;
- unsigned long power;
+ unsigned long power, power_orig;
unsigned long interval;
interval = msecs_to_jiffies(sd-balance_interval);
@@ -4462,7 +4462,7 @@ void update_group_power(struct sched_domain *sd, int cpu)
return;
}
- power = 0;
+ power_orig = power = 0;
if (child-flags SD_OVERLAP) {
/*
@@ -4470,8 +4470,12 @@ void update_group_power(struct sched_domain *sd, int cpu)
* span the current group.
*/
- for_each_cpu(cpu, sched_group_cpus(sdg))
- power += power_of(cpu);
+ for_each_cpu(cpu, sched_group_cpus(sdg)) {
+ struct sched_group *sg = cpu_rq(cpu)-sd-groups;
+
+ power_orig += sg-sgp-power_orig;
+ power += sg-sgp-power;
+ }
} else {
/*
* !SD_OVERLAP domains can assume that child groups
@@ -4480,12 +4484,14 @@ void update_group_power(struct sched_domain *sd, int
cpu)
group = child-groups;
do {
+ power_orig += group-sgp-power_orig;
power += group-sgp-power;
group = group-next;
} while (group != child-groups);
}
- sdg-sgp-power_orig = sdg-sgp-power = power;
+ sdg-sgp-power_orig = power_orig;
+ sdg-sgp-power = power;
}
/*
--
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: [tip:sched/core] sched/fair: Fix group power_orig computation
tip-bot for Peter Zijlstra tip...@zytor.com wrote:
Commit-ID: 863bffc80898b8df295ebac111af2335ec05f85d
Gitweb: http://git.kernel.org/tip/863bffc80898b8df295ebac111af2335ec05f85d
Author: Peter Zijlstra pet...@infradead.org
AuthorDate: Wed, 28 Aug 2013 11:44:39 +0200
Committer: Ingo Molnar mi...@kernel.org
CommitDate: Thu, 12 Sep 2013 19:14:43 +0200
sched/fair: Fix group power_orig computation
When looking at the code I noticed we don't actually compute
sgp-power_orig correctly for groups, fix that.
Currently the only consumer of that value is fix_small_capacity()
which is only used on POWER7+ and that code excludes this case by
being limited to SD_SHARE_CPUPOWER which is only ever set on the SMT
domain which must be the lowest domain and this has singleton groups.
So nothing should be affected by this change.
Cc: Michael Neuling mi...@neuling.org
Signed-off-by: Peter Zijlstra pet...@infradead.org
Link: http://lkml.kernel.org/n/tip-db2pe0vxwunv37plc7onn...@git.kernel.org
Signed-off-by: Ingo Molnar mi...@kernel.org
FWIW, this doesn't seem to break POWER7.
Thanks!
Mikey
---
kernel/sched/fair.c | 16 +++-
1 file changed, 11 insertions(+), 5 deletions(-)
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index f9f4385..baba313 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4450,7 +4450,7 @@ void update_group_power(struct sched_domain *sd, int
cpu)
{
struct sched_domain *child = sd-child;
struct sched_group *group, *sdg = sd-groups;
- unsigned long power;
+ unsigned long power, power_orig;
unsigned long interval;
interval = msecs_to_jiffies(sd-balance_interval);
@@ -4462,7 +4462,7 @@ void update_group_power(struct sched_domain *sd, int
cpu)
return;
}
- power = 0;
+ power_orig = power = 0;
if (child-flags SD_OVERLAP) {
/*
@@ -4470,8 +4470,12 @@ void update_group_power(struct sched_domain *sd, int
cpu)
* span the current group.
*/
- for_each_cpu(cpu, sched_group_cpus(sdg))
- power += power_of(cpu);
+ for_each_cpu(cpu, sched_group_cpus(sdg)) {
+ struct sched_group *sg = cpu_rq(cpu)-sd-groups;
+
+ power_orig += sg-sgp-power_orig;
+ power += sg-sgp-power;
+ }
} else {
/*
* !SD_OVERLAP domains can assume that child groups
@@ -4480,12 +4484,14 @@ void update_group_power(struct sched_domain *sd, int
cpu)
group = child-groups;
do {
+ power_orig += group-sgp-power_orig;
power += group-sgp-power;
group = group-next;
} while (group != child-groups);
}
- sdg-sgp-power_orig = sdg-sgp-power = power;
+ sdg-sgp-power_orig = power_orig;
+ sdg-sgp-power = power;
}
/*
--
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/
