perf/x86: Only expose userspace rdpmc for events on current CPU
Subsystems (2)
RFC 1/1 perf/x86: Only expose userspace rdpmc for events on current CPU
Userspace access using rdpmc only makes sense if the event is valid for the current CPU. However, cap_user_rdpmc is currently set no matter which CPU the event is associated with. The result is userspace reading another CPU's event thinks it can use rdpmc to read the counter. In doing so, the wrong counter will be read.
---
I'm working on adding userspace counter access for arm64 along with a common access routine in libperf. I found this issue when testing per CPU events. More details are here[1].
I'm going to need the same thing for arm64. This could possibly go into perf_event_update_userpage() instead, but perhaps there could be an arch where userspace can read other cpu's counters.
What's the ABI between libperf and kernel versions? This change will only help the libperf implementation if libperf doesn't need to support old kernels.
[1] https://lore.kernel.org/r/CAL_JsqJzeCebq4VP+xBtfh=fbomvaJoVMp35AQQDGTYD-fRWgw@mail.gmail.com
--- arch/x86/events/core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
arch/x86/events/core.c +2 -1
@@ -2490,7 +2490,8 @@ void arch_perf_update_userpage(struct perf_event *event, userpg->cap_user_time =0 ; userpg->cap_user_time_zero =0 ; userpg->cap_user_rdpmc = - !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ); + !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ) && + (event->oncpu ==smp_processor_id ()); userpg->pmc_width = x86_pmu.cntval_bits;if (!using_native_sched_clock () || !sched_clock_stable ())
Signature · 1 line
2.27.0
Peter Zijlstra
maintainerQuoted 5 lines · Rob Herring
Userspace access using rdpmc only makes sense if the event is valid for the current CPU. However,
cap_user_rdpmcis currently set no matter which CPU the event is associated with. The result is userspace reading another CPU's event thinks it can use rdpmc to read the counter. In doing so, the wrong counter will be read.
Don't do that then?
Quoted 5 lines
arch/x86/events/core.c +0
-0
arch/x86/events/core.c void arch_perf_update_userpage(struct perf_event *event, · line 2497
Quoted 9 lines
arch/x86/events/core.c +2
-1@@ -2490,7 +2490,8 @@ void arch_perf_update_userpage(struct perf_event *event, userpg->cap_user_time =0 ; userpg->cap_user_time_zero =0 ; userpg->cap_user_rdpmc = - !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ); + !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ) && + (event->oncpu ==smp_processor_id ()); userpg->pmc_width = x86_pmu.cntval_bits;if (!using_native_sched_clock () || !sched_clock_stable ())
Isn't that a nop? That is, from the few sites I checked, we're always calling this on the event's CPU.
Quoted 9 lines · Peter Zijlstra
Quoted 5 lines · Rob Herring
Userspace access using rdpmc only makes sense if the event is valid for the current CPU. However,
cap_user_rdpmcis currently set no matter which CPU the event is associated with. The result is userspace reading another CPU's event thinks it can use rdpmc to read the counter. In doing so, the wrong counter will be read.Don't do that then?
I could check this in userspace I suppose, but then it's yet another thing the rdpmc loop has to check. I think it's better to not add more overhead there.
Or we just ignore this. This issue came up in the referenced thread with Jiri. I'm not all that convinced that per CPU events and userspace rdpmc is too useful of a combination. It would only reduce the overhead on 1 out of N cpus. In this case, we'd have to limit the userspace read to per thread events (and ones on any cpu I suppose).
Quoted 14 lines
Quoted 14 lines
arch/x86/events/core.c +2
-1@@ -2490,7 +2490,8 @@ void arch_perf_update_userpage(struct perf_event *event, userpg->cap_user_time =0 ; userpg->cap_user_time_zero =0 ; userpg->cap_user_rdpmc = - !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ); + !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ) && + (event->oncpu ==smp_processor_id ()); userpg->pmc_width = x86_pmu.cntval_bits;
Quoted 4 lines
arch/x86/events/core.c void arch_perf_update_userpage(struct perf_event *event, · line 2497
Quoted 1 line
arch/x86/events/core.c +0
-0if (!using_native_sched_clock () || !sched_clock_stable ())Isn't that a nop? That is, from the few sites I checked, we're always calling this on the event's CPU.
If cpu0 opens and mmaps an event for cpu1, then cpu0 will see cap_user_rdpmc set and think it can use rdpmc.
Rob
Peter Zijlstra
maintainerQuoted 9 lines · Rob Herring
Quoted 8 lines · Peter Zijlstra
Quoted 5 lines · Rob Herring
Userspace access using rdpmc only makes sense if the event is valid for the current CPU. However,
cap_user_rdpmcis currently set no matter which CPU the event is associated with. The result is userspace reading another CPU's event thinks it can use rdpmc to read the counter. In doing so, the wrong counter will be read.
Quoted 5 lines
Quoted 1 line
Don't do that then?
I could check this in userspace I suppose, but then it's yet another thing the rdpmc loop has to check. I think it's better to not add more overhead there.
So all this was designed for self monitoring; attempting rdpmc on an event not for yourself is out of spec.
Quoted 13 lines
Quoted 13 lines
Quoted 13 lines
arch/x86/events/core.c +2
-1@@ -2490,7 +2490,8 @@ void arch_perf_update_userpage(struct perf_event *event, userpg->cap_user_time =0 ; userpg->cap_user_time_zero =0 ; userpg->cap_user_rdpmc = - !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ); + !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ) && + (event->oncpu ==smp_processor_id ()); userpg->pmc_width = x86_pmu.cntval_bits;
Quoted 7 lines
Quoted 4 lines
arch/x86/events/core.c void arch_perf_update_userpage(struct perf_event *event, · line 2497
Quoted 1 line
arch/x86/events/core.c +0
-0if (!using_native_sched_clock () || !sched_clock_stable ())Isn't that a nop? That is, from the few sites I checked, we're always calling this on the event's CPU.
If cpu0 opens and mmaps an event for cpu1, then cpu0 will see
cap_user_rdpmcset and think it can use rdpmc.
I don't think your check helps with that. IIRC we always call arch_perf_update_userpage() on the CPU the event actually runs on. So it's always true.
Quoted 38 lines · Peter Zijlstra
Quoted 9 lines · Rob Herring
Quoted 8 lines · Peter Zijlstra
Quoted 5 lines · Rob Herring
Userspace access using rdpmc only makes sense if the event is valid for the current CPU. However,
cap_user_rdpmcis currently set no matter which CPU the event is associated with. The result is userspace reading another CPU's event thinks it can use rdpmc to read the counter. In doing so, the wrong counter will be read.Quoted 5 lines
Quoted 1 line
Don't do that then?
I could check this in userspace I suppose, but then it's yet another thing the rdpmc loop has to check. I think it's better to not add more overhead there.
So all this was designed for self monitoring; attempting rdpmc on an event not for yourself is out of spec.
Quoted 18 lines
Quoted 17 lines
Quoted 5 lines
arch/x86/events/core.c +0
-0arch/x86/events/core.c void arch_perf_update_userpage(struct perf_event *event, · line 2497
Quoted 9 lines
arch/x86/events/core.c +2
-1@@ -2490,7 +2490,8 @@ void arch_perf_update_userpage(struct perf_event *event, userpg->cap_user_time =0 ; userpg->cap_user_time_zero =0 ; userpg->cap_user_rdpmc = - !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ); + !!(event->hw.flags &PERF_X86_EVENT_RDPMC_ALLOWED ) && + (event->oncpu ==smp_processor_id ()); userpg->pmc_width = x86_pmu.cntval_bits;if (!using_native_sched_clock () || !sched_clock_stable ())Isn't that a nop? That is, from the few sites I checked, we're always calling this on the event's CPU.
Quoted 6 lines
Quoted 2 lines
If cpu0 opens and mmaps an event for cpu1, then cpu0 will see
cap_user_rdpmcset and think it can use rdpmc.I don't think your check helps with that. IIRC we always call
arch_perf_update_userpage()on the CPU the event actually runs on. So it's always true.
My testing says otherwise. I tested this change on the arm64 version of arch_perf_update_userpage, but I don't think x86 should be any different here.
I'm testing with libperf test_stat_cpu() modified to mmap each cpu event. Without the change I get the following result:
# taskset 2 test-evsel-a -v - running test-evsel.c... mmap base 0xffff9fd77000 userspace counter access enabled on cpu0 <<<<< Reflects cap_user_rdpmc state cpu0: count = 0x72cf, ena = 0x1a838, run = 0x1a838 <<<<< count is from rdpmc mmap base 0xffff9fd76000 userspace counter access enabled on cpu1 cpu1: count = 0xc978, ena = 0x163e6, run = 0x163e6 <<<<< count is from rdpmc
cpu0 is idle here, so we'd expect count to be near zero, but it's not.
Then with the change, I get the following:
# taskset 2 test-evsel-a -v - running test-evsel.c... mmap base 0xffffa742d000 cpu0: count = 0xddb, ena = 0x3f8d6, run = 0x3f8d6 <<<<< count is from read() mmap base 0xffffa742c000 userspace counter access enabled on cpu1 cpu1: count = 0xb538, ena = 0x154f0, run = 0x154f0 <<<<< count is from rdpmc
# taskset 1 test-evsel-a -v - running test-evsel.c... mmap base 0xffff8b008000 userspace counter access enabled on cpu0 cpu0: count = 0x7c21, ena = 0x18574, run = 0x18574 mmap base 0xffff8b007000 cpu1: count = 0xb3c, ena = 0x61aa8, run = 0x61aa8
As you can see, count tracks the idle and not idle cpu, and cap_user_rdpmc is only set for the cpu event matching the cpu we are running on.
Rob