Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Ingo writes:
"perf fixes:
- fix a CPU#0 hot unplug bug and a PCI enumeration bug in the x86 Intel uncore PMU driver
- fix a CPU event enumeration bug in the x86 AMD PMU driver
- fix a perf ring-buffer corruption bug when using tracepoints
- fix a PMU unregister locking bug"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf/x86/amd/uncore: Set ThreadMask and SliceMask for L3 Cache perf events
perf/x86/intel/uncore: Fix PCI BDF address of M3UPI on SKX
perf/ring_buffer: Prevent concurent ring buffer access
perf/x86/intel/uncore: Use boot_cpu_data.phys_proc_id instead of hardcorded physical package ID 0
perf/core: Fix perf_pmu_unregister() locking
+29
-14
+10
arch/x86/events/amd/uncore.c
+10
arch/x86/events/amd/uncore.c
···
36
36
37
37
static int num_counters_llc;
38
38
static int num_counters_nb;
39
+
static bool l3_mask;
39
40
40
41
static HLIST_HEAD(uncore_unused_list);
41
42
···
209
208
/* and we do not enable counter overflow interrupts */
210
209
hwc->config = event->attr.config & AMD64_RAW_EVENT_MASK_NB;
211
210
hwc->idx = -1;
211
+
212
+
/*
213
+
* SliceMask and ThreadMask need to be set for certain L3 events in
214
+
* Family 17h. For other events, the two fields do not affect the count.
215
+
*/
216
+
if (l3_mask)
217
+
hwc->config |= (AMD64_L3_SLICE_MASK | AMD64_L3_THREAD_MASK);
212
218
213
219
if (event->cpu < 0)
214
220
return -EINVAL;
···
533
525
amd_llc_pmu.name = "amd_l3";
534
526
format_attr_event_df.show = &event_show_df;
535
527
format_attr_event_l3.show = &event_show_l3;
528
+
l3_mask = true;
536
529
} else {
537
530
num_counters_nb = NUM_COUNTERS_NB;
538
531
num_counters_llc = NUM_COUNTERS_L2;
···
541
532
amd_llc_pmu.name = "amd_l2";
542
533
format_attr_event_df = format_attr_event;
543
534
format_attr_event_l3 = format_attr_event;
535
+
l3_mask = false;
544
536
}
545
537
546
538
amd_nb_pmu.attr_groups = amd_uncore_attr_groups_df;
+7
-7
arch/x86/events/intel/uncore_snbep.c
+7
-7
arch/x86/events/intel/uncore_snbep.c
···
3061
3061
3062
3062
void bdx_uncore_cpu_init(void)
3063
3063
{
3064
-
int pkg = topology_phys_to_logical_pkg(0);
3064
+
int pkg = topology_phys_to_logical_pkg(boot_cpu_data.phys_proc_id);
3065
3065
3066
3066
if (bdx_uncore_cbox.num_boxes > boot_cpu_data.x86_max_cores)
3067
3067
bdx_uncore_cbox.num_boxes = boot_cpu_data.x86_max_cores;
···
3931
3931
.driver_data = UNCORE_PCI_DEV_FULL_DATA(21, 5, SKX_PCI_UNCORE_M2PCIE, 3),
3932
3932
},
3933
3933
{ /* M3UPI0 Link 0 */
3934
-
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x204C),
3935
-
.driver_data = UNCORE_PCI_DEV_FULL_DATA(18, 0, SKX_PCI_UNCORE_M3UPI, 0),
3934
+
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x204D),
3935
+
.driver_data = UNCORE_PCI_DEV_FULL_DATA(18, 1, SKX_PCI_UNCORE_M3UPI, 0),
3936
3936
},
3937
3937
{ /* M3UPI0 Link 1 */
3938
-
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x204D),
3939
-
.driver_data = UNCORE_PCI_DEV_FULL_DATA(18, 1, SKX_PCI_UNCORE_M3UPI, 1),
3938
+
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x204E),
3939
+
.driver_data = UNCORE_PCI_DEV_FULL_DATA(18, 2, SKX_PCI_UNCORE_M3UPI, 1),
3940
3940
},
3941
3941
{ /* M3UPI1 Link 2 */
3942
-
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x204C),
3943
-
.driver_data = UNCORE_PCI_DEV_FULL_DATA(18, 4, SKX_PCI_UNCORE_M3UPI, 2),
3942
+
PCI_DEVICE(PCI_VENDOR_ID_INTEL, 0x204D),
3943
+
.driver_data = UNCORE_PCI_DEV_FULL_DATA(18, 5, SKX_PCI_UNCORE_M3UPI, 2),
3944
3944
},
3945
3945
{ /* end: all zeroes */ }
3946
3946
};
+8
arch/x86/include/asm/perf_event.h
+8
arch/x86/include/asm/perf_event.h
···
46
46
#define INTEL_ARCH_EVENT_MASK \
47
47
(ARCH_PERFMON_EVENTSEL_UMASK | ARCH_PERFMON_EVENTSEL_EVENT)
48
48
49
+
#define AMD64_L3_SLICE_SHIFT 48
50
+
#define AMD64_L3_SLICE_MASK \
51
+
((0xFULL) << AMD64_L3_SLICE_SHIFT)
52
+
53
+
#define AMD64_L3_THREAD_SHIFT 56
54
+
#define AMD64_L3_THREAD_MASK \
55
+
((0xFFULL) << AMD64_L3_THREAD_SHIFT)
56
+
49
57
#define X86_RAW_EVENT_MASK \
50
58
(ARCH_PERFMON_EVENTSEL_EVENT | \
51
59
ARCH_PERFMON_EVENTSEL_UMASK | \
+4
-7
kernel/events/core.c
+4
-7
kernel/events/core.c
···
8314
8314
goto unlock;
8315
8315
8316
8316
list_for_each_entry_rcu(event, &ctx->event_list, event_entry) {
8317
+
if (event->cpu != smp_processor_id())
8318
+
continue;
8317
8319
if (event->attr.type != PERF_TYPE_TRACEPOINT)
8318
8320
continue;
8319
8321
if (event->attr.config != entry->type)
···
9433
9431
if (pmu->task_ctx_nr > perf_invalid_context)
9434
9432
return;
9435
9433
9436
-
mutex_lock(&pmus_lock);
9437
9434
free_percpu(pmu->pmu_cpu_context);
9438
-
mutex_unlock(&pmus_lock);
9439
9435
}
9440
9436
9441
9437
/*
···
9689
9689
9690
9690
void perf_pmu_unregister(struct pmu *pmu)
9691
9691
{
9692
-
int remove_device;
9693
-
9694
9692
mutex_lock(&pmus_lock);
9695
-
remove_device = pmu_bus_running;
9696
9693
list_del_rcu(&pmu->entry);
9697
-
mutex_unlock(&pmus_lock);
9698
9694
9699
9695
/*
9700
9696
* We dereference the pmu list under both SRCU and regular RCU, so
···
9702
9706
free_percpu(pmu->pmu_disable_count);
9703
9707
if (pmu->type >= PERF_TYPE_MAX)
9704
9708
idr_remove(&pmu_idr, pmu->type);
9705
-
if (remove_device) {
9709
+
if (pmu_bus_running) {
9706
9710
if (pmu->nr_addr_filters)
9707
9711
device_remove_file(pmu->dev, &dev_attr_nr_addr_filters);
9708
9712
device_del(pmu->dev);
9709
9713
put_device(pmu->dev);
9710
9714
}
9711
9715
free_pmu_context(pmu);
9716
+
mutex_unlock(&pmus_lock);
9712
9717
}
9713
9718
EXPORT_SYMBOL_GPL(perf_pmu_unregister);
9714
9719