perf/x86: Disable non generic regs for software/probe events
Subsystems (2)
PATCH v2 1/3 perf/x86: Disable non generic regs for software/probe events
From: Kan Liang <[email protected]>
The perf fuzzer caused skylake machine to crash.
[ 9680.085831] Call Trace: [ 9680.088301] <IRQ> [ 9680.090363] perf_output_sample_regs+0x43/0xa0 [ 9680.094928] perf_output_sample+0x3aa/0x7a0 [ 9680.099181] perf_event_output_forward+0x53/0x80 [ 9680.103917] __perf_event_overflow+0x52/0xf0 [ 9680.108266] ? perf_trace_run_bpf_submit+0xc0/0xc0 [ 9680.113108] perf_swevent_hrtimer+0xe2/0x150 [ 9680.117475] ? check_preempt_wakeup+0x181/0x230 [ 9680.122091] ? check_preempt_curr+0x62/0x90 [ 9680.126361] ? ttwu_do_wakeup+0x19/0x140 [ 9680.130355] ? try_to_wake_up+0x54/0x460 [ 9680.134366] ? reweight_entity+0x15b/0x1a0 [ 9680.138559] ? __queue_work+0x103/0x3f0 [ 9680.142472] ? update_dl_rq_load_avg+0x1cd/0x270 [ 9680.147194] ? timerqueue_del+0x1e/0x40 [ 9680.151092] ? __remove_hrtimer+0x35/0x70 [ 9680.155191] __hrtimer_run_queues+0x100/0x280 [ 9680.159658] hrtimer_interrupt+0x100/0x220 [ 9680.163835] smp_apic_timer_interrupt+0x6a/0x140 [ 9680.168555] apic_timer_interrupt+0xf/0x20 [ 9680.172756] </IRQ>
The XMM registers can only be collected by hardware events, not software/probe events.
Add has_non_generic_regs() to check if non-generic registers, e.g. XMM on X86, are applied for software/probe events. If yes, return -EOPNOTSUPP.
The generic code define the mask of non-generic registers as 0 if arch headers haven't overridden it.
---
Changes since V1:
- Use macro PERF_REG_NON_GENERIC_MASK to replace function
- Avoid unnecessary abbreviations in comments
- Remove unnecessary bracket for return.
arch/x86/include/uapi/asm/perf_regs.h | 3 +++ include/linux/perf_regs.h | 8 ++++++++ kernel/events/core.c | 30 ++++++++++++++++++++++++++++++ 3 files changed, 41 insertions(+)
arch/x86/include/uapi/asm/perf_regs.h +3 -0
@@ -52,4 +52,7 @@ enum perf_event_x86_regs {/* These include both GPRs and XMMX registers */ PERF_REG_X86_XMM_MAX = PERF_REG_X86_XMM15 + 2, }; + +#define PERF_REG_NON_GENERIC_MASK (~((1ULL << PERF_REG_X86_XMM0) - 1)) +#endif /* _ASM_X86_PERF_REGS_H */
include/linux/perf_regs.h +8 -0
@@ -11,6 +11,11 @@ struct perf_regs {#ifdef CONFIG_HAVE_PERF_REGS #include <asm/perf_regs.h> + +#ifndef PERF_REG_NON_GENERIC_MASK +#define PERF_REG_NON_GENERIC_MASK 0 +#endif +u64 perf_reg_value (struct pt_regs *regs,int idx);int perf_reg_validate (u64 mask);u64 perf_reg_abi (struct task_struct *task);
@@ -18,6 +23,9 @@ void perf_get_regs_user(struct perf_regs *regs_user,struct pt_regs *regs, struct pt_regs *regs_user_copy);#else + +#define PERF_REG_NON_GENERIC_MASK 0 +static inline u64 perf_reg_value (struct pt_regs *regs,int idx) {return 0 ;
kernel/events/core.c +30 -0
@@ -8457,6 +8457,12 @@ static void sw_perf_event_destroy(struct perf_event *event)swevent_hlist_put (); } +static inline bool has_non_generic_regs (struct perf_event *event) +{ +return (event->attr.sample_regs_user &PERF_REG_NON_GENERIC_MASK ) || + (event->attr.sample_regs_intr &PERF_REG_NON_GENERIC_MASK ); +} +static int perf_swevent_init (struct perf_event *event) {u64 event_id = event->attr.config;
@@ -8470,6 +8476,10 @@ static int perf_swevent_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +switch (event_id) {case PERF_COUNT_SW_CPU_CLOCK :case PERF_COUNT_SW_TASK_CLOCK :
@@ -8633,6 +8643,10 @@ static int perf_tp_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + err =perf_trace_init (event);if (err)return err;
@@ -8722,6 +8736,10 @@ static int perf_kprobe_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + is_retprobe = event->attr.config &PERF_PROBE_CONFIG_IS_RETPROBE ; err =perf_kprobe_init (event, is_retprobe);if (err)
@@ -8782,6 +8800,10 @@ static int perf_uprobe_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + is_retprobe = event->attr.config &PERF_PROBE_CONFIG_IS_RETPROBE ; ref_ctr_offset = event->attr.config >>PERF_UPROBE_REF_CTR_OFFSET_SHIFT ; err =perf_uprobe_init (event, ref_ctr_offset, is_retprobe);
@@ -9562,6 +9584,10 @@ static int cpu_clock_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +perf_swevent_init_hrtimer (event);return 0 ;
@@ -9643,6 +9669,10 @@ static int task_clock_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +perf_swevent_init_hrtimer (event);return 0 ;
Signature · 1 line
2.7.4
PATCH v2 2/3 perf/x86/regs: Check reserved bits
From: Kan Liang <[email protected]>
The perf fuzzer triggers a warning which map to:
if (WARN_ON_ONCE(idx >= ARRAY_SIZE(pt_regs_offset)))
return 0;
The bits between XMM registers and generic registers are reserved. But perf_reg_validate() doesn't check these bits.
Add PERF_REG_X86_RESERVED for reserved bits on X86. Check the reserved bits in perf_reg_validate().
---
Changes since V1:
- Rename REG_RESERVED to PERF_REG_X86_RESERVED
arch/x86/kernel/perf_regs.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
arch/x86/kernel/perf_regs.c +5 -2
@@ -74,6 +74,9 @@ u64 perf_reg_value(struct pt_regs *regs, int idx)return regs_get_register (regs, pt_regs_offset[idx]); } +#define PERF_REG_X86_RESERVED (((1ULL << PERF_REG_X86_XMM0) - 1) & \ + ~((1ULL << PERF_REG_X86_MAX) - 1)) +#ifdef CONFIG_X86_32 #define REG_NOSUPPORT ((1ULL << PERF_REG_X86_R8) | \ (1ULL << PERF_REG_X86_R9) | \
@@ -86,7 +89,7 @@ u64 perf_reg_value(struct pt_regs *regs, int idx)int perf_reg_validate (u64 mask) { -if (!mask || (mask ®_NOSUPPORT )) +if (!mask || (mask & (REG_NOSUPPORT |PERF_REG_X86_RESERVED )))return -EINVAL ;return 0 ;
@@ -112,7 +115,7 @@ void perf_get_regs_user(struct perf_regs *regs_user,int perf_reg_validate (u64 mask) { -if (!mask || (mask ®_NOSUPPORT )) +if (!mask || (mask & (REG_NOSUPPORT |PERF_REG_X86_RESERVED )))return -EINVAL ;return 0 ;
Signature · 1 line
2.7.4
PATCH v2 3/3 perf regs x86: Use PERF_REG_NON_GENERIC_MASK
In reply to [email protected]From: Kan Liang <[email protected]>
Use the macro defined in kernel ABI header to replace the local name.
No functional change.
---
New for V2
tools/arch/x86/include/uapi/asm/perf_regs.h | 3 +++ tools/perf/arch/x86/include/perf_regs.h | 1 - tools/perf/arch/x86/util/perf_regs.c | 4 ++-- 3 files changed, 5 insertions(+), 3 deletions(-)
tools/arch/x86/include/uapi/asm/perf_regs.h +3 -0
@@ -52,4 +52,7 @@ enum perf_event_x86_regs {/* These include both GPRs and XMMX registers */ PERF_REG_X86_XMM_MAX = PERF_REG_X86_XMM15 + 2, }; + +#define PERF_REG_NON_GENERIC_MASK (~((1ULL << PERF_REG_X86_XMM0) - 1)) +#endif /* _ASM_X86_PERF_REGS_H */
tools/perf/arch/x86/include/perf_regs.h +0 -1
@@ -9,7 +9,6 @@void perf_regs_load (u64 *regs);#define PERF_REGS_MAX PERF_REG_X86_XMM_MAX -#define PERF_XMM_REGS_MASK (~((1ULL << PERF_REG_X86_XMM0) - 1))#ifndef HAVE_ARCH_X86_64_SUPPORT #define PERF_REGS_MASK ((1ULL << PERF_REG_X86_32_MAX) - 1)#define PERF_SAMPLE_REGS_ABI PERF_SAMPLE_REGS_ABI_32
tools/perf/arch/x86/util/perf_regs.c +2 -2
@@ -277,7 +277,7 @@ uint64_t arch__intr_reg_mask(void) .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES, .sample_type = PERF_SAMPLE_REGS_INTR, - .sample_regs_intr = PERF_XMM_REGS_MASK, + .sample_regs_intr = PERF_REG_NON_GENERIC_MASK, .precise_ip = 1, .disabled = 1, .exclude_kernel = 1,
@@ -293,7 +293,7 @@ uint64_t arch__intr_reg_mask(void) fd =sys_perf_event_open (&attr,0 ,-1 ,-1 ,0 );if (fd !=-1 ) {close (fd); -return (PERF_XMM_REGS_MASK |PERF_REGS_MASK ); +return (PERF_REG_NON_GENERIC_MASK |PERF_REGS_MASK ); }return PERF_REGS_MASK ;
Signature · 1 line
2.7.4
Peter Zijlstra
maintainerQuoted 29 lines · On Mon, May 27, 2019 at 12:07:55PM -0700, [email protected] wrote:
arch/x86/include/uapi/asm/perf_regs.h +3
-0@@ -52,4 +52,7 @@ enum perf_event_x86_regs {/* These include both GPRs and XMMX registers */ PERF_REG_X86_XMM_MAX = PERF_REG_X86_XMM15 + 2, }; + +#define PERF_REG_NON_GENERIC_MASK (~((1ULL << PERF_REG_X86_XMM0) - 1)) +#endif /* _ASM_X86_PERF_REGS_H */ include/linux/perf_regs.h +5
-0@@ -11,6 +11,11 @@ struct perf_regs {#ifdef CONFIG_HAVE_PERF_REGS #include <asm/perf_regs.h> + +#ifndef PERF_REG_NON_GENERIC_MASK +#define PERF_REG_NON_GENERIC_MASK 0 +#endif +u64 perf_reg_value (struct pt_regs *regs,int idx);int perf_reg_validate (u64 mask);u64 perf_reg_abi (struct task_struct *task);
include/linux/perf_regs.h void perf_get_regs_user(struct perf_regs *regs_user, · line 31
Quoted 9 lines
include/linux/perf_regs.h +3
-0@@ -18,6 +23,9 @@ void perf_get_regs_user(struct perf_regs *regs_user,struct pt_regs *regs, struct pt_regs *regs_user_copy);#else + +#define PERF_REG_NON_GENERIC_MASK 0 +static inline u64 perf_reg_value (struct pt_regs *regs,int idx) {return 0 ;
Much better than the last version; however..
Quoted 73 lines
kernel/events/core.c +26
-0@@ -8457,6 +8457,12 @@ static void sw_perf_event_destroy(struct perf_event *event)swevent_hlist_put (); } +static inline bool has_non_generic_regs (struct perf_event *event) +{ +return (event->attr.sample_regs_user &PERF_REG_NON_GENERIC_MASK ) || + (event->attr.sample_regs_intr &PERF_REG_NON_GENERIC_MASK ); +} +static int perf_swevent_init (struct perf_event *event) {u64 event_id = event->attr.config;@@ -8470,6 +8476,10 @@ static int perf_swevent_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +switch (event_id) {case PERF_COUNT_SW_CPU_CLOCK :case PERF_COUNT_SW_TASK_CLOCK :@@ -8633,6 +8643,10 @@ static int perf_tp_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + err =perf_trace_init (event);if (err)return err;@@ -8722,6 +8736,10 @@ static int perf_kprobe_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + is_retprobe = event->attr.config &PERF_PROBE_CONFIG_IS_RETPROBE ; err =perf_kprobe_init (event, is_retprobe);if (err)@@ -8782,6 +8800,10 @@ static int perf_uprobe_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + is_retprobe = event->attr.config &PERF_PROBE_CONFIG_IS_RETPROBE ; ref_ctr_offset = event->attr.config >>PERF_UPROBE_REF_CTR_OFFSET_SHIFT ; err =perf_uprobe_init (event, ref_ctr_offset, is_retprobe);@@ -9562,6 +9584,10 @@ static int cpu_clock_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +perf_swevent_init_hrtimer (event);return 0 ;
kernel/events/core.c static int task_clock_event_init(struct perf_event *event) · line 9678
Quoted 10 lines
kernel/events/core.c +4
-0@@ -9643,6 +9669,10 @@ static int task_clock_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +perf_swevent_init_hrtimer (event);return 0 ;
I don't think this is anywhere near sufficient. What happens if we request XMM regs for an uncore PMU ?
I'm thinking you want something along these lines...
---
arch/x86/events/intel/core.c +7 -1
@@ -3281,7 +3281,13 @@ static int intel_pmu_hw_config(struct perf_event *event)if (event->attr.sample_type &PERF_SAMPLE_CALLCHAIN ) event->attr.sample_type |= __PERF_SAMPLE_CALLCHAIN_EARLY; - } + +/* we only support extended (XMM) registers for sample_regs_intr */ +if (event->attr.sample_regs_user &PERF_REGS_EXTENDED_MASK ) +return -EOPNOTSUPP ; + + }else if (has_extended_regs (event)) +return -EOPNOTSUPP ;if (needs_branch_stack (event)) { ret =intel_pmu_setup_lbr_filter (event);
arch/x86/events/intel/ds.c +1 -0
@@ -2020,6 +2020,7 @@ void __init intel_ds_init(void)PERF_SAMPLE_TIME ; x86_pmu.flags |=PMU_FL_PEBS_ALL ; pebs_qual ="-baseline" ; +x86_get_pmu ()->capabilities |=PERF_PMU_CAP_EXTENDED_REGS ; }else {/* Only basic record supported */ x86_pmu.pebs_no_xmm_regs =1 ;
include/linux/perf_event.h +1 -0
@@ -241,6 +241,7 @@ struct perf_event;#define PERF_PMU_CAP_NO_INTERRUPT 0x01#define PERF_PMU_CAP_NO_NMI 0x02#define PERF_PMU_CAP_AUX_NO_SG 0x04 +#define PERF_PMU_CAP_EXTENDED_REGS 0x08#define PERF_PMU_CAP_EXCLUSIVE 0x10#define PERF_PMU_CAP_ITRACE 0x20#define PERF_PMU_CAP_HETEROGENEOUS_CPUS 0x40
kernel/events/core.c +8 -4
@@ -10064,12 +10064,16 @@ static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)perf_event_ctx_unlock (event->group_leader, ctx);if (!ret) { +if (!(pmu->capabilities &PERF_PMU_CAP_EXTENDED_REGS ) && +has_extended_regs (event)) + ret = -EOPNOTSUPP ; +if (pmu->capabilities &PERF_PMU_CAP_NO_EXCLUDE && -event_has_any_exclude_flag (event)) { -if (event->destroy) - event->destroy (event); +event_has_any_exclude_flag (event)) ret = -EINVAL ; - } + +if (ret && event->destroy) + event->destroy (event); }if (ret)
Quoted 124 lines · Peter Zijlstra
Quoted 29 lines · On Mon, May 27, 2019 at 12:07:55PM -0700, [email protected] wrote:
arch/x86/include/uapi/asm/perf_regs.h +3
-0@@ -52,4 +52,7 @@ enum perf_event_x86_regs {/* These include both GPRs and XMMX registers */ PERF_REG_X86_XMM_MAX = PERF_REG_X86_XMM15 + 2, }; + +#define PERF_REG_NON_GENERIC_MASK (~((1ULL << PERF_REG_X86_XMM0) - 1)) +#endif /* _ASM_X86_PERF_REGS_H */ include/linux/perf_regs.h +5
-0@@ -11,6 +11,11 @@ struct perf_regs {#ifdef CONFIG_HAVE_PERF_REGS #include <asm/perf_regs.h> + +#ifndef PERF_REG_NON_GENERIC_MASK +#define PERF_REG_NON_GENERIC_MASK 0 +#endif +u64 perf_reg_value (struct pt_regs *regs,int idx);int perf_reg_validate (u64 mask);u64 perf_reg_abi (struct task_struct *task);include/linux/perf_regs.h void perf_get_regs_user(struct perf_regs *regs_user, · line 31
Quoted 9 lines
include/linux/perf_regs.h +3
-0@@ -18,6 +23,9 @@ void perf_get_regs_user(struct perf_regs *regs_user,struct pt_regs *regs, struct pt_regs *regs_user_copy);#else + +#define PERF_REG_NON_GENERIC_MASK 0 +static inline u64 perf_reg_value (struct pt_regs *regs,int idx) {return 0 ;Much better than the last version; however..
Quoted 82 lines
kernel/events/core.c +30
-0@@ -8457,6 +8457,12 @@ static void sw_perf_event_destroy(struct perf_event *event)swevent_hlist_put (); } +static inline bool has_non_generic_regs (struct perf_event *event) +{ +return (event->attr.sample_regs_user &PERF_REG_NON_GENERIC_MASK ) || + (event->attr.sample_regs_intr &PERF_REG_NON_GENERIC_MASK ); +} +static int perf_swevent_init (struct perf_event *event) {u64 event_id = event->attr.config;@@ -8470,6 +8476,10 @@ static int perf_swevent_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +switch (event_id) {case PERF_COUNT_SW_CPU_CLOCK :case PERF_COUNT_SW_TASK_CLOCK :@@ -8633,6 +8643,10 @@ static int perf_tp_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + err =perf_trace_init (event);if (err)return err;@@ -8722,6 +8736,10 @@ static int perf_kprobe_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + is_retprobe = event->attr.config &PERF_PROBE_CONFIG_IS_RETPROBE ; err =perf_kprobe_init (event, is_retprobe);if (err)@@ -8782,6 +8800,10 @@ static int perf_uprobe_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; + is_retprobe = event->attr.config &PERF_PROBE_CONFIG_IS_RETPROBE ; ref_ctr_offset = event->attr.config >>PERF_UPROBE_REF_CTR_OFFSET_SHIFT ; err =perf_uprobe_init (event, ref_ctr_offset, is_retprobe);@@ -9562,6 +9584,10 @@ static int cpu_clock_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +perf_swevent_init_hrtimer (event);return 0 ;@@ -9643,6 +9669,10 @@ static int task_clock_event_init(struct perf_event *event)if (has_branch_stack (event))return -EOPNOTSUPP ; +/* Only support generic registers */ +if (has_non_generic_regs (event)) +return -EOPNOTSUPP ; +perf_swevent_init_hrtimer (event);
Quoted 5 lines
kernel/events/core.c static int task_clock_event_init(struct perf_event *event) · line 9678
Quoted 1 line
kernel/events/core.c +0
-0return 0 ;I don't think this is anywhere near sufficient. What happens if we request XMM regs for an uncore PMU ?
Uncore PMU doesn't support sampling. It will return -EINVAL. There is no regs support for counting. The request will be ignored.
I think current check for uncore is good enough.
Thanks,
Kan
Quoted 72 lines
I'm thinking you want something along these lines...
---
arch/x86/events/intel/core.c +7
-1@@ -3281,7 +3281,13 @@ static int intel_pmu_hw_config(struct perf_event *event)if (event->attr.sample_type &PERF_SAMPLE_CALLCHAIN ) event->attr.sample_type |= __PERF_SAMPLE_CALLCHAIN_EARLY; - } + +/* we only support extended (XMM) registers for sample_regs_intr */ +if (event->attr.sample_regs_user &PERF_REGS_EXTENDED_MASK ) +return -EOPNOTSUPP ; + + }else if (has_extended_regs (event)) +return -EOPNOTSUPP ;if (needs_branch_stack (event)) { ret =intel_pmu_setup_lbr_filter (event);arch/x86/events/intel/ds.c +1
-0@@ -2020,6 +2020,7 @@ void __init intel_ds_init(void)PERF_SAMPLE_TIME ; x86_pmu.flags |=PMU_FL_PEBS_ALL ; pebs_qual ="-baseline" ; +x86_get_pmu ()->capabilities |=PERF_PMU_CAP_EXTENDED_REGS ; }else {/* Only basic record supported */ x86_pmu.pebs_no_xmm_regs =1 ;include/linux/perf_event.h +1
-0@@ -241,6 +241,7 @@ struct perf_event;#define PERF_PMU_CAP_NO_INTERRUPT 0x01#define PERF_PMU_CAP_NO_NMI 0x02#define PERF_PMU_CAP_AUX_NO_SG 0x04 +#define PERF_PMU_CAP_EXTENDED_REGS 0x08#define PERF_PMU_CAP_EXCLUSIVE 0x10#define PERF_PMU_CAP_ITRACE 0x20#define PERF_PMU_CAP_HETEROGENEOUS_CPUS 0x40kernel/events/core.c +8
-4@@ -10064,12 +10064,16 @@ static int perf_try_init_event(struct pmu *pmu, struct perf_event *event)perf_event_ctx_unlock (event->group_leader, ctx);if (!ret) { +if (!(pmu->capabilities &PERF_PMU_CAP_EXTENDED_REGS ) && +has_extended_regs (event)) + ret = -EOPNOTSUPP ; +if (pmu->capabilities &PERF_PMU_CAP_NO_EXCLUDE && -event_has_any_exclude_flag (event)) { -if (event->destroy) - event->destroy (event); +event_has_any_exclude_flag (event)) ret = -EINVAL ; - } + +if (ret && event->destroy) + event->destroy (event); }if (ret)
Peter Zijlstra
maintainerQuoted 4 lines · Kan
Uncore PMU doesn't support sampling. It will return
-EINVAL. There is no regs support for counting. The request will be ignored.I think current check for uncore is good enough.
breakpoints then.. There's also no guarantee you covered all software events, and the core rewrite will allow other per-task/sampling PMUs too.
The approach you take is just not complete, don't do that.
Quoted 9 lines · Peter Zijlstra
Quoted 4 lines · Kan
Uncore PMU doesn't support sampling. It will return
-EINVAL. There is no regs support for counting. The request will be ignored.I think current check for uncore is good enough.
breakpoints then.. There's also no guarantee you covered all software events, and the core rewrite will allow other per-task/sampling PMUs too.
possibly related, even with the patches applied, the skylake machine eventually did still crash while fuzzing:
[133621.333101] BUG: unable to handle page fault for address: 00000001000000c8 [133621.333102] #PF: supervisor read access in kernel mode [133621.333103] #PF: error_code(0x0000) - not-present page [133621.333104] PGD 0 P4D 0 [133621.333106] Oops: 0000 [#1] SMP PTI [133621.333108] CPU: 4 PID: 22203 Comm: perf_fuzzer Tainted: G W 5.2.0-rc1+ #39 [133621.333109] Hardware name: LENOVO 10FY0017US/SKYBAY, BIOS FWKT53A 06/06/2016 [133621.333109] RIP: 0010:perf_reg_value+0x1e/0x50 [133621.333111] Code: 00 48 b8 00 00 00 00 ff ff ff ff c3 0f 1f 44 00 00 8d 46 e0 83 f8 1f 77 1d 48 8b 97 a8 00 00 00 31 c0 48 85 d2 74 0e 48 63 f6 <48> 8b 84 f2 00 ff ff ff c3 31 c0 c3 83 fe 17 77 16 48 63 f6 8b 04 [133621.333112] RSP: 0000:fffffe00000d5a80 EFLAGS: 00010006 [133621.333113] RAX: 0000000000000000 RBX: 0000000000000039 RCX: 0000000000000039 [133621.333114] RDX: 0000000100000000 RSI: 0000000000000039 RDI: fffffe00000d5c88 [133621.333115] RBP: fffffe00000d5b38 R08: 0000000000000000 R09: 0000000000000000 [133621.333116] R10: 00000000bffffff0 R11: 0000000000000012 R12: fffffe00000d5c88 [133621.333117] R13: ffff99883253ed10 R14: 0000000000000050 R15: 0000000000000000 [133621.333118] FS: 00007fb9741d3540(0000) GS:ffff998835b00000(0000) knlGS:0000000000000000 [133621.333119] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [133621.333119] CR2: 00000001000000c8 CR3: 000000023333c004 CR4: 00000000003607e0 [133621.333120] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [133621.333121] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000600 [133621.333122] Call Trace: [133621.333122] <NMI> [133621.333123] perf_output_sample_regs+0x43/0xa0 [133621.333124] perf_output_sample+0x3aa/0x7a0 [133621.333125] perf_event_output_forward+0x53/0x80 [133621.333125] __perf_event_overflow+0x52/0xf0 [133621.333126] handle_pmi_common+0x1b3/0x240 [133621.333127] ? visit_groups_merge+0xeb/0x180 [133621.333127] ? native_write_msr+0xb/0x20 [133621.333128] ? native_write_msr+0x1a/0x20 [133621.333129] ? native_write_msr+0xc/0x20 [133621.333129] ? intel_pmu_lbr_read+0x29f/0x3d0 [133621.333130] ? intel_pmu_lbr_filter+0x7f/0x1f0 [133621.333131] intel_pmu_handle_irq+0xbf/0x160 [133621.333132] perf_event_nmi_handler+0x2d/0x50 [133621.333132] nmi_handle+0x63/0x110 [133621.333133] default_do_nmi+0x4e/0x100 [133621.333134] do_nmi+0x14d/0x1b0 [133621.333134] end_repeat_nmi+0x16/0x50 [133621.333135] RIP: 0010:visit_groups_merge+0xeb/0x180 [133621.333137] Code: c0 75 73 48 8d 7b 30 e8 c3 de 55 00 48 85 c0 0f 84 9a 00 00 00 48 89 c2 48 83 ea 30 74 10 8b b3 74 02 00 00 39 b0 44 02 00 00 <49> 0f 45 d7 48 89 55 00 48 8b 04 24 48 8b 5c 24 08 48 85 c0 48 89 [133621.333137] RSP: 0000:ffffb28f4c897e10 EFLAGS: 00000046 [133621.333139] RAX: ffff998833315830 RBX: ffff99882a17a000 RCX: 0000000000000001 [133621.333140] RDX: ffff998833315800 RSI: 0000000000000004 RDI: ffff99882a17a030 [133621.333140] RBP: ffffb28f4c897e18 R08: 0000000000000000 R09: ffff998835b26a80 [133621.333141] R10: ffff99882a17a000 R11: 0000000000000001 R12: ffffffffb6bb93b0 [133621.333142] R13: ffffb28f4c897e68 R14: ffffb28f4c897e10 R15: 0000000000000000 [133621.333143] ? __perf_event_disable+0x160/0x160 [133621.333144] ? visit_groups_merge+0xeb/0x180 [133621.333144] ? visit_groups_merge+0xeb/0x180 [133621.333145] </NMI> [133621.333145] ctx_sched_in+0xb7/0x180 [133621.333146] __perf_event_task_sched_in+0x16e/0x1c0 [133621.333147] ? __switch_to_asm+0x40/0x70 [133621.333147] ? __switch_to_asm+0x34/0x70 [133621.333148] ? __switch_to_asm+0x40/0x70 [133621.333149] ? __switch_to_asm+0x34/0x70 [133621.333149] finish_task_switch+0xcd/0x270 [133621.333150] schedule_tail+0xb/0x50 [133621.333151] ret_from_fork+0x8/0x40 [133621.333151] Modules linked in: intel_rapl x86_pkg_temp_thermal intel_powerclamp coretemp kvm irqbypass snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic crct10dif_pclmul crc32_pclmul ledtrig_audio ghash_clmulni_intel snd_hda_intel snd_hda_codec aesni_intel snd_hda_core aes_x86_64 crypto_simd snd_hwdep cryptd snd_pcm mei_me glue_helper snd_timer snd sg mei iTCO_wdt iTCO_vendor_support soundcore wmi_bmof tpm_tis evdev tpm_tis_core acpi_pad tpm rng_core pcspkr pcc_cpufreq fuse parport_pc sunrpc ppdev lp parport ip_tables x_tables autofs4 ext4 crc32c_generic crc16 mbcache jbd2 sr_mod sd_mod cdrom i915 i2c_algo_bit ahci libahci crc32c_intel xhci_pci libata drm_kms_helper i2c_i801 xhci_hcd e1000e drm scsi_mod usbcore fan thermal wmi video button [133621.333183] CR2: 00000001000000c8 [133621.743913] ---[ end trace 7a151c3de6b000fc ]--- [133621.743913] RIP: 0010:perf_reg_value+0x1e/0x50 [133621.743913] Code: 00 48 b8 00 00 00 00 ff ff ff ff c3 0f 1f 44 00 00 8d 46 e0 83 f8 1f 77 1d 48 8b 97 a8 00 00 00 31 c0 48 85 d2 74 0e 48 63 f6 <48> 8b 84 f2 00 ff ff ff c3 31 c0 c3 83 fe 17 77 16 48 63 f6 8b 04 [133621.743913] RSP: 0000:fffffe00000d5a80 EFLAGS: 00010006 [133621.743914] RAX: 0000000000000000 RBX: 0000000000000039 RCX: 0000000000000039 [133621.743914] RDX: 0000000100000000 RSI: 0000000000000039 RDI: fffffe00000d5c88 [133621.743914] RBP: fffffe00000d5b38 R08: 0000000000000000 R09: 0000000000000000 [133621.743914] R10: 00000000bffffff0 R11: 0000000000000012 R12: fffffe00000d5c88 [133621.743915] R13: ffff99883253ed10 R14: 0000000000000050 R15: 0000000000000000 [133621.743915] FS: 00007fb9741d3540(0000) GS:ffff998835b00000(0000) knlGS:0000000000000000 [133621.743915] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [133621.743915] CR2: 00000001000000c8 CR3: 000000023333c004 CR4: 00000000003607e0 [133621.743915] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [133621.743916] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000600 [133621.743916] Kernel panic - not syncing: Fatal exception in interrupt [133622.777346] Shutting down cpus with NMI [133622.777347] Kernel Offset: 0x35a00000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
Quoted 4 lines · Peter Zijlstra
Quoted 3 lines · Kan
Uncore PMU doesn't support sampling. It will return
-EINVAL. There is no regs support for counting. The request will be ignored.
Quoted 8 lines
Quoted 1 line
I think current check for uncore is good enough.
breakpoints then.. There's also no guarantee you covered all software events, and the core rewrite will allow other per-task/sampling PMUs too.
The approach you take is just not complete, don't do that.
OK. I will send V3 base on your proposed patches.
Thanks,
Kan