Basic perf PMU support for Haswell v12
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 18 lines · Andi Kleen
All outstanding issues fixed I hope. And I added mem-loads/stores support.
Contains support for:
- Basic Haswell PMU and PEBS support
- Late unmasking of the PMI
- mem-loads/stores supportv2: Addressed Stephane's feedback. See individual patches for details.
v3: now even more bite-sized. Qualifier constraints merged earlier.
v4: Rename some variables, add some comments and other minor changes. Add some Reviewed/Tested-bys.
v5: Address some minor review feedback. Port to latest perf/core
v6: Add just some variable names, add comments, edit descriptions, some more testing, rebased to latest perf/core
v7: Expand comment
v8: Rename structure field.
v9: No wide counters, but add basic LBRs. Add some more constraints. Rebase to 3.9rc1
Quoted 8 lines
v10: Change some whitespace. Rebase to 3.9rc3
v11: Rebase to perf/core. Fix extra regs. Rename INTX.
v12: Rebase to 3.10-rc2
Add mem-loads/stores support for parity with Sandy Bridge. Fix fixed counters (Thanks Ingo!)
Make late ack optional
Export new config bits in sysfs.
Minor changes
I reported a pretty nasty regression with the previous version (v10) which made this series break default 'perf top' on non-Haswell systems - but it's unclear from this changelog to what extent you managed to reproduce the bug and fix it, and what the fix was?
If it's fixed then please don't hide fixes - you need to follow up in the original threads where they got reported, not just some obscure changelog entry in the next version post ...
I'd really like to make progress with this feature - 11 iterations is ridiculous really.
Thanks,
Ingo
Quoted 10 lines · Ingo Molnar
Quoted 7 lines · Andi Kleen
All outstanding issues fixed I hope. And I added mem-loads/stores support.
Contains support for:
- Basic Haswell PMU and PEBS support
- Late unmasking of the PMI
- mem-loads/stores support
Quoted 16 lines
Quoted 8 lines
v2: Addressed Stephane's feedback. See individual patches for details.
v3: now even more bite-sized. Qualifier constraints merged earlier.
v4: Rename some variables, add some comments and other minor changes. Add some Reviewed/Tested-bys.
v5: Address some minor review feedback. Port to latest perf/core
v6: Add just some variable names, add comments, edit descriptions, some more testing, rebased to latest perf/core
v7: Expand commentQuoted 8 lines
v8: Rename structure field.
v9: No wide counters, but add basic LBRs. Add some more constraints. Rebase to 3.9rc1
v10: Change some whitespace. Rebase to 3.9rc3
v11: Rebase to perf/core. Fix extra regs. Rename INTX.
v12: Rebase to 3.10-rc2
Add mem-loads/stores support for parity with Sandy Bridge. Fix fixed counters (Thanks Ingo!)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Quoted 8 lines
Quoted 3 lines
Make late ack optional
Export new config bits in sysfs.
Minor changesI reported a pretty nasty regression with the previous version (v10) which made this series break default 'perf top' on non-Haswell systems - but it's unclear from this changelog to what extent you managed to reproduce the bug and fix it, and what the fix was?
Thanks for checking.
I didn't reproduce it, but I found a problem by code review with the fixed counter constraints.
I think I fixed it by adding this hunk:
@@ -2227,7 +2313,7 @@ __init int intel_pmu_init(void)
* counter, so do not extend mask to generic counters
*/
for_each_event_constraint(c, x86_pmu.event_constraints) {
- if (c->cmask != X86_RAW_EVENT_MASK
+ if (c->cmask != FIXED_EVENT_FLAGS
|| c->idxmsk64 == INTEL_PMC_MSK_FIXED_REF_CYCLES) {
continue;
}It would be cleaner to detect the fixed counters in some other way, but that was the simplest fix I could find.
Testing appreciated
Quoted 2 lines
I'd really like to make progress with this feature - 11 iterations is ridiculous really.
Thanks.
There are actually more patches unfortunately, this is just a subset. I'll send the others once that one is in, probably split into less and more important ones.
https://git.kernel.org/cgit/linux/kernel/git/ak/linux-misc.git/log/?h=hsw/pmu6
-Andi
Signature · 1 line
[email protected] -- Speaking for myself only.
Quoted 53 lines · Andi Kleen
Quoted 10 lines · Ingo Molnar
Quoted 7 lines · Andi Kleen
All outstanding issues fixed I hope. And I added mem-loads/stores support.
Contains support for:
- Basic Haswell PMU and PEBS support
- Late unmasking of the PMI
- mem-loads/stores supportQuoted 16 lines
Quoted 8 lines
v2: Addressed Stephane's feedback. See individual patches for details.
v3: now even more bite-sized. Qualifier constraints merged earlier.
v4: Rename some variables, add some comments and other minor changes. Add some Reviewed/Tested-bys.
v5: Address some minor review feedback. Port to latest perf/core
v6: Add just some variable names, add comments, edit descriptions, some more testing, rebased to latest perf/core
v7: Expand commentQuoted 8 lines
v8: Rename structure field.
v9: No wide counters, but add basic LBRs. Add some more constraints. Rebase to 3.9rc1
v10: Change some whitespace. Rebase to 3.9rc3
v11: Rebase to perf/core. Fix extra regs. Rename INTX.
v12: Rebase to 3.10-rc2
Add mem-loads/stores support for parity with Sandy Bridge. Fix fixed counters (Thanks Ingo!)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Quoted 8 lines
Quoted 3 lines
Make late ack optional
Export new config bits in sysfs.
Minor changesI reported a pretty nasty regression with the previous version (v10) which made this series break default 'perf top' on non-Haswell systems - but it's unclear from this changelog to what extent you managed to reproduce the bug and fix it, and what the fix was?
Thanks for checking.
I didn't reproduce it, but I found a problem by code review with the fixed counter constraints.I think I fixed it by adding this hunk:
(quoted diff) +1
-1@@ -2227,7 +2313,7 @@ __init int intel_pmu_init(void) * counter, so do not extend mask to generic counters */ for_each_event_constraint(c, x86_pmu.event_constraints) { - if (c->cmask != X86_RAW_EVENT_MASK + if (c->cmask != FIXED_EVENT_FLAGS || c->idxmsk64 == INTEL_PMC_MSK_FIXED_REF_CYCLES) { continue; }
Quoted 4 lines
It would be cleaner to detect the fixed counters in some other way, but that was the simplest fix I could find.
Testing appreciated
Fair enough - I'll give it a whirl - that hunk does indeed look like it could make a difference.
Thanks,
Ingo
Quoted 55 lines · Ingo Molnar
Quoted 53 lines · Andi Kleen
Quoted 10 lines · Ingo Molnar
Quoted 7 lines · Andi Kleen
All outstanding issues fixed I hope. And I added mem-loads/stores support.
Contains support for:
- Basic Haswell PMU and PEBS support
- Late unmasking of the PMI
- mem-loads/stores supportQuoted 16 lines
Quoted 8 lines
v2: Addressed Stephane's feedback. See individual patches for details.
v3: now even more bite-sized. Qualifier constraints merged earlier.
v4: Rename some variables, add some comments and other minor changes. Add some Reviewed/Tested-bys.
v5: Address some minor review feedback. Port to latest perf/core
v6: Add just some variable names, add comments, edit descriptions, some more testing, rebased to latest perf/core
v7: Expand commentQuoted 8 lines
v8: Rename structure field.
v9: No wide counters, but add basic LBRs. Add some more constraints. Rebase to 3.9rc1
v10: Change some whitespace. Rebase to 3.9rc3
v11: Rebase to perf/core. Fix extra regs. Rename INTX.
v12: Rebase to 3.10-rc2
Add mem-loads/stores support for parity with Sandy Bridge. Fix fixed counters (Thanks Ingo!)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Quoted 8 lines
Quoted 3 lines
Make late ack optional
Export new config bits in sysfs.
Minor changesI reported a pretty nasty regression with the previous version (v10) which made this series break default 'perf top' on non-Haswell systems - but it's unclear from this changelog to what extent you managed to reproduce the bug and fix it, and what the fix was?
Thanks for checking.
I didn't reproduce it, but I found a problem by code review with the fixed counter constraints.I think I fixed it by adding this hunk:
(quoted diff) +1
-1@@ -2227,7 +2313,7 @@ __init int intel_pmu_init(void) * counter, so do not extend mask to generic counters */ for_each_event_constraint(c, x86_pmu.event_constraints) { - if (c->cmask != X86_RAW_EVENT_MASK + if (c->cmask != FIXED_EVENT_FLAGS || c->idxmsk64 == INTEL_PMC_MSK_FIXED_REF_CYCLES) { continue; }
Quoted 7 lines
Quoted 4 lines
It would be cleaner to detect the fixed counters in some other way, but that was the simplest fix I could find.
Testing appreciated
Fair enough - I'll give it a whirl - that hunk does indeed look like it could make a difference.
Ok, I can confirm that this fixed the perf top and perf record regression I saw on Intel systems.
Thanks,
Ingo
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 4/6] perf, x86: Move NMI clearing to end of PMI handler v2
Quoted 70 lines · Andi Kleen
From: Andi Kleen <[email protected]>
This avoids some problems with spurious PMIs on Haswell. Haswell seems to behave more like P4 in this regard. Do the same thing as the P4 perf handler by unmasking the NMI only at the end. Shouldn't make any difference for earlier family 6 cores.
Tested on Haswell, IvyBridge, Westmere, Saltwell (Atom)
v2: Enable only for Haswell
Signed-off-by Andi Kleen <[email protected]>--- arch/x86/kernel/cpu/perf_event.h | 1 + arch/x86/kernel/cpu/perf_event_intel.c | 22 +++++++++++++--------- 2 files changed, 14 insertions(+), 9 deletions(-)arch/x86/kernel/cpu/perf_event.h +1
-0@@ -378,6 +378,7 @@ struct x86_pmu {struct event_constraint *event_constraints;struct x86_pmu_quirk *quirks;int perfctr_second_write; +bool late_ack;/* * sysfs attrs arch/x86/kernel/cpu/perf_event_intel.c +12
-9@@ -1184,16 +1184,12 @@ static int intel_pmu_handle_irq(struct pt_regs *regs) cpuc = &__get_cpu_var (cpu_hw_events); -/* -* Some chipsets need to unmask the LVTPC in a particular spot -* inside the nmi handler. As a result, the unmasking was pushed -* into all the nmi handlers. -* -* This handler doesn't seem to have any issues with the unmasking -* so it was left at the top. +/* +* No known reason to not always do late ACK, +* but just in case do it opt-in. */ -apic_write (APIC_LVTPC ,APIC_DM_NMI ); - +if (!x86_pmu.late_ack) +apic_write (APIC_LVTPC ,APIC_DM_NMI );intel_pmu_disable_all (); handled =intel_pmu_drain_bts_buffer (); status =intel_pmu_get_status ();@@ -1253,6 +1249,13 @@ again: done:intel_pmu_enable_all (0 ); +/* +* Only unmask the NMI after the overflow counters +* have been reset. This avoids spurious NMIs on +* Haswell CPUs. +*/ +if (x86_pmu.late_ack) +apic_write (APIC_LVTPC ,APIC_DM_NMI );return handled; }
arch/x86/kernel/cpu/perf_event_intel.c __init int intel_pmu_init(void) · line 2266
Quoted 7 lines
arch/x86/kernel/cpu/perf_event_intel.c +1
-0@@ -2257,6 +2260,7 @@ __init int intel_pmu_init(void)case 70 :case 71 :case 63 : + x86_pmu.late_ack = true;memcpy (hw_cache_event_ids, snb_hw_cache_event_ids,sizeof (hw_cache_event_ids)); memcpy(hw_cache_extra_regs, snb_hw_cache_extra_regs,
Ok - this is a lot less intrusive solution.
Once the dust has settled we can try setting late_ack for all models, and if that works out without regressing, we can switch to the late ack method altogether.
Thanks,
Ingo
Quoted 6 lines · Andi Kleen
v12: Rebase to 3.10-rc2
Add mem-loads/stores support for parity with Sandy Bridge. Fix fixed counters (Thanks Ingo!)
Make late ack optional
Export new config bits in sysfs.
Minor changes
Sigh, what you have not fixed in your patches are the basic stylistic mistakes I pointed out to the past:
https://lkml.org/lkml/2013/5/1/78
(my previous feedback is also quoted below.)
Here checkpatch.pl says this about your series:
total: 6 errors, 10 warnings, 662 lines checked
and a handful of those checkpatch.pl complaints are for valid, real problems.
Furthermore, you have not replied to any of those two mails of mine, nor have you fixed the stylistic problems I pointed out, in these latest patches!
To fix it simply follow the advice I gave you twice before: run scripts/checkpatch.pl against your patches and address any valid complaints it gives _BEFORE YOU RESUBMIT THEM_!
Andi, what the heck is going on here? Your behavior makes no sense to me. You are pretty much the only contributor I know who makes a habit out of willfully ignoring maintainer feedback...
Thanks,
Ingo
--------------------->
Quoted 46 lines · Ingo Molnar
Quoted 22 lines · Ingo Molnar
Quoted 6 lines · Ingo Molnar
You say it's barebones, yet it does not work :-( How well was this patch-set tested on non-Haswell hardware, which makes up 99.99% of our installed base?
In particular, after applying your patches, 'perf top' stopped working on an Intel testbox of mine:
The other problem I noticed was stylistic: when I applied your patches for testing even Git complained about their cleanliness ...
To quote from
Documentation/SubmittingPatches:4) Style check your changes.
Check your patch for basic style violations, details of which can be found in
Documentation/CodingStyle. Failure to do so simply wastes the reviewers time and will get your patch rejected, probably without even being read.Quoted 5 lines
At a minimum you should check your patches with the patch style checker prior to submission (
scripts/checkpatch.pl). You should be able to justify all violations that remain in your patch.Please make your patches less sloppy!
Andi, you have not replied to this mail of mine.
What new measures are you taking to avoid such annoying stylistic problems to creep into your patches?
These problems are regular in your patches and that has been going on for years - causing maintenance overhead for many maintainers, not just me.
Apparently you are not using proper tooling (checkpatch.pl for example) to check your patches. If you refuse to take action I will have to stop dealing with your patches directly altogether - the overhead just does not justify the effort. You'll need to get your patches reviewed by and signed off by a more experienced kernel hacker who knows how to submit patches.
Thanks,