From e2d371484653ac83b970d3ebcf343383f39f8b6b Mon Sep 17 00:00:00 2001 From: Ravi Bangoria Date: Fri, 18 Nov 2022 10:45:39 +0530 Subject: perf core: Return error pointer if inherit_event() fails to find pmu_ctx inherit_event() returns NULL only when it finds orphaned events otherwise it returns either valid child_event pointer or an error pointer. Follow the same when it fails to find pmu_ctx. Fixes: bd2756811766 ("perf: Rewrite core context handling") Reported-by: Dan Carpenter Signed-off-by: Ravi Bangoria Signed-off-by: Peter Zijlstra (Intel) Link: https://lkml.kernel.org/r/20221118051539.820-1-ravi.bangoria@amd.com --- kernel/events/core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index eacc3702654d..4bd2434251f0 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -13231,7 +13231,7 @@ inherit_event(struct perf_event *parent_event, pmu_ctx = find_get_pmu_context(child_event->pmu, child_ctx, child_event); if (IS_ERR(pmu_ctx)) { free_event(child_event); - return NULL; + return ERR_CAST(pmu_ctx); } child_event->pmu_ctx = pmu_ctx; -- cgit v1.2.3-70-g09d2 From f841b682baef90ee144df8b12e2c76aa460717c1 Mon Sep 17 00:00:00 2001 From: Chengming Zhou Date: Wed, 7 Dec 2022 20:40:23 +0800 Subject: perf/core: Fix cgroup events tracking We encounter perf warnings when using cgroup events like: cd /sys/fs/cgroup mkdir test perf stat -e cycles -a -G test Which then triggers: WARNING: CPU: 0 PID: 690 at kernel/events/core.c:849 perf_cgroup_switch+0xb2/0xc0 Call Trace: __schedule+0x4ae/0x9f0 ? _raw_spin_unlock_irqrestore+0x23/0x40 ? __cond_resched+0x18/0x20 preempt_schedule_common+0x2d/0x70 __cond_resched+0x18/0x20 wait_for_completion+0x2f/0x160 ? cpu_stop_queue_work+0x9e/0x130 affine_move_task+0x18a/0x4f0 WARNING: CPU: 0 PID: 690 at kernel/events/core.c:829 ctx_sched_in+0x1cf/0x1e0 Call Trace: ? ctx_sched_out+0xb7/0x1b0 perf_cgroup_switch+0x88/0xc0 __schedule+0x4ae/0x9f0 ? _raw_spin_unlock_irqrestore+0x23/0x40 ? __cond_resched+0x18/0x20 preempt_schedule_common+0x2d/0x70 __cond_resched+0x18/0x20 wait_for_completion+0x2f/0x160 ? cpu_stop_queue_work+0x9e/0x130 affine_move_task+0x18a/0x4f0 The above two warnings are not complete here since I remove other unimportant information. The problem is caused by the perf cgroup events tracking: CPU0 CPU1 perf_event_open() perf_event_alloc() account_event() account_event_cpu() atomic_inc(perf_cgroup_events) __perf_event_task_sched_out() if (atomic_read(perf_cgroup_events)) perf_cgroup_switch() // kernel/events/core.c:849 WARN_ON_ONCE(cpuctx->ctx.nr_cgroups == 0) if (READ_ONCE(cpuctx->cgrp) == cgrp) // false return perf_ctx_lock() ctx_sched_out() cpuctx->cgrp = cgrp ctx_sched_in() perf_cgroup_set_timestamp() // kernel/events/core.c:829 WARN_ON_ONCE(!ctx->nr_cgroups) perf_ctx_unlock() perf_install_in_context() cpu_function_call() __perf_install_in_context() add_event_to_ctx() list_add_event() perf_cgroup_event_enable() ctx->nr_cgroups++ cpuctx->cgrp = X We can see from above that we wrongly use percpu atomic perf_cgroup_events to check if we need to perf_cgroup_switch(), which should only be used when we know this CPU has cgroup events enabled. The commit bd2756811766 ("perf: Rewrite core context handling") change to have only one context per-CPU, so we can just use cpuctx->cgrp to check if this CPU has cgroup events enabled. So percpu atomic perf_cgroup_events is not needed. Fixes: bd2756811766 ("perf: Rewrite core context handling") Signed-off-by: Chengming Zhou Signed-off-by: Peter Zijlstra (Intel) Tested-by: Ravi Bangoria Link: https://lkml.kernel.org/r/20221207124023.66252-1-zhouchengming@bytedance.com --- kernel/events/core.c | 42 ++++++++++-------------------------------- 1 file changed, 10 insertions(+), 32 deletions(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index 4bd2434251f0..37c0f04d7a00 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -380,7 +380,6 @@ enum event_type_t { /* * perf_sched_events : >0 events exist - * perf_cgroup_events: >0 per-cpu cgroup events exist on this cpu */ static void perf_sched_delayed(struct work_struct *work); @@ -389,7 +388,6 @@ static DECLARE_DELAYED_WORK(perf_sched_work, perf_sched_delayed); static DEFINE_MUTEX(perf_sched_mutex); static atomic_t perf_sched_count; -static DEFINE_PER_CPU(atomic_t, perf_cgroup_events); static DEFINE_PER_CPU(struct pmu_event_list, pmu_sb_events); static atomic_t nr_mmap_events __read_mostly; @@ -844,9 +842,16 @@ static void perf_cgroup_switch(struct task_struct *task) struct perf_cpu_context *cpuctx = this_cpu_ptr(&perf_cpu_context); struct perf_cgroup *cgrp; - cgrp = perf_cgroup_from_task(task, NULL); + /* + * cpuctx->cgrp is set when the first cgroup event enabled, + * and is cleared when the last cgroup event disabled. + */ + if (READ_ONCE(cpuctx->cgrp) == NULL) + return; WARN_ON_ONCE(cpuctx->ctx.nr_cgroups == 0); + + cgrp = perf_cgroup_from_task(task, NULL); if (READ_ONCE(cpuctx->cgrp) == cgrp) return; @@ -3631,8 +3636,7 @@ void __perf_event_task_sched_out(struct task_struct *task, * to check if we have to switch out PMU state. * cgroup event are system-wide mode only */ - if (atomic_read(this_cpu_ptr(&perf_cgroup_events))) - perf_cgroup_switch(next); + perf_cgroup_switch(next); } static bool perf_less_group_idx(const void *l, const void *r) @@ -4974,15 +4978,6 @@ static void unaccount_pmu_sb_event(struct perf_event *event) detach_sb_event(event); } -static void unaccount_event_cpu(struct perf_event *event, int cpu) -{ - if (event->parent) - return; - - if (is_cgroup_event(event)) - atomic_dec(&per_cpu(perf_cgroup_events, cpu)); -} - #ifdef CONFIG_NO_HZ_FULL static DEFINE_SPINLOCK(nr_freq_lock); #endif @@ -5048,8 +5043,6 @@ static void unaccount_event(struct perf_event *event) schedule_delayed_work(&perf_sched_work, HZ); } - unaccount_event_cpu(event, event->cpu); - unaccount_pmu_sb_event(event); } @@ -11679,15 +11672,6 @@ static void account_pmu_sb_event(struct perf_event *event) attach_sb_event(event); } -static void account_event_cpu(struct perf_event *event, int cpu) -{ - if (event->parent) - return; - - if (is_cgroup_event(event)) - atomic_inc(&per_cpu(perf_cgroup_events, cpu)); -} - /* Freq events need the tick to stay alive (see perf_event_task_tick). */ static void account_freq_event_nohz(void) { @@ -11775,8 +11759,6 @@ static void account_event(struct perf_event *event) } enabled: - account_event_cpu(event, event->cpu); - account_pmu_sb_event(event); } @@ -12822,13 +12804,11 @@ static void __perf_pmu_remove(struct perf_event_context *ctx, perf_event_groups_for_cpu_pmu(event, groups, cpu, pmu) { perf_remove_from_context(event, 0); - unaccount_event_cpu(event, cpu); put_pmu_ctx(event->pmu_ctx); list_add(&event->migrate_entry, events); for_each_sibling_event(sibling, event) { perf_remove_from_context(sibling, 0); - unaccount_event_cpu(sibling, cpu); put_pmu_ctx(sibling->pmu_ctx); list_add(&sibling->migrate_entry, events); } @@ -12847,7 +12827,6 @@ static void __perf_pmu_install_event(struct pmu *pmu, if (event->state >= PERF_EVENT_STATE_OFF) event->state = PERF_EVENT_STATE_INACTIVE; - account_event_cpu(event, cpu); perf_install_in_context(ctx, event, cpu); } @@ -13742,8 +13721,7 @@ static int __perf_cgroup_move(void *info) struct task_struct *task = info; preempt_disable(); - if (atomic_read(this_cpu_ptr(&perf_cgroup_events))) - perf_cgroup_switch(task); + perf_cgroup_switch(task); preempt_enable(); return 0; -- cgit v1.2.3-70-g09d2 From 08245672cdc6505550d1a5020603b0a8d4a6dcc7 Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Fri, 2 Dec 2022 13:51:49 +0000 Subject: perf/x86/amd: fix potential integer overflow on shift of a int The left shift of int 32 bit integer constant 1 is evaluated using 32 bit arithmetic and then passed as a 64 bit function argument. In the case where i is 32 or more this can lead to an overflow. Avoid this by shifting using the BIT_ULL macro instead. Fixes: 471af006a747 ("perf/x86/amd: Constrain Large Increment per Cycle events") Signed-off-by: Colin Ian King Signed-off-by: Peter Zijlstra (Intel) Acked-by: Ian Rogers Acked-by: Kim Phillips Link: https://lore.kernel.org/r/20221202135149.1797974-1-colin.i.king@gmail.com --- arch/x86/events/amd/core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/events/amd/core.c b/arch/x86/events/amd/core.c index d6f3703e4119..4386b10682ce 100644 --- a/arch/x86/events/amd/core.c +++ b/arch/x86/events/amd/core.c @@ -1387,7 +1387,7 @@ static int __init amd_core_pmu_init(void) * numbered counter following it. */ for (i = 0; i < x86_pmu.num_counters - 1; i += 2) - even_ctr_mask |= 1 << i; + even_ctr_mask |= BIT_ULL(i); pair_constraint = (struct event_constraint) __EVENT_CONSTRAINT(0, even_ctr_mask, 0, -- cgit v1.2.3-70-g09d2 From a551844e345ba2a1c533dee4b55cb0efddb1bcda Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 19 Dec 2022 15:40:04 +0100 Subject: perf: Fix use-after-free in error path The syscall error path has a use-after-free; put_pmu_ctx() will reference ctx, therefore we must ensure ctx is destroyed after pmu_ctx is. Fixes: bd2756811766 ("perf: Rewrite core context handling") Reported-by: syzbot+b8e8c01c8ade4fe6e48f@syzkaller.appspotmail.com Signed-off-by: Peter Zijlstra (Intel) Tested-by: Chengming Zhou Link: https://lkml.kernel.org/r/Y6B3xEgkbmFUCeni@hirez.programming.kicks-ass.net --- kernel/events/core.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index 37c0f04d7a00..63d674c9b70e 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -12671,7 +12671,8 @@ SYSCALL_DEFINE5(perf_event_open, return event_fd; err_context: - /* event->pmu_ctx freed by free_event() */ + put_pmu_ctx(event->pmu_ctx); + event->pmu_ctx = NULL; /* _free_event() */ err_locked: mutex_unlock(&ctx->mutex); perf_unpin_context(ctx); @@ -12784,6 +12785,7 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu, err_pmu_ctx: put_pmu_ctx(pmu_ctx); + event->pmu_ctx = NULL; /* _free_event() */ err_unlock: mutex_unlock(&ctx->mutex); perf_unpin_context(ctx); -- cgit v1.2.3-70-g09d2 From 0a041ebca4956292cadfb14a63ace3a9c1dcb0a3 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 20 Dec 2022 14:31:40 -0800 Subject: perf/core: Call LSM hook after copying perf_event_attr It passes the attr struct to the security_perf_event_open() but it's not initialized yet. Fixes: da97e18458fb ("perf_event: Add support for LSM and SELinux checks") Signed-off-by: Namhyung Kim Signed-off-by: Peter Zijlstra (Intel) Reviewed-by: Joel Fernandes (Google) Cc: stable@vger.kernel.org Link: https://lkml.kernel.org/r/20221220223140.4020470-1-namhyung@kernel.org --- kernel/events/core.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/kernel/events/core.c b/kernel/events/core.c index 63d674c9b70e..d56328e5080e 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -12321,12 +12321,12 @@ SYSCALL_DEFINE5(perf_event_open, if (flags & ~PERF_FLAG_ALL) return -EINVAL; - /* Do we allow access to perf_event_open(2) ? */ - err = security_perf_event_open(&attr, PERF_SECURITY_OPEN); + err = perf_copy_attr(attr_uptr, &attr); if (err) return err; - err = perf_copy_attr(attr_uptr, &attr); + /* Do we allow access to perf_event_open(2) ? */ + err = security_perf_event_open(&attr, PERF_SECURITY_OPEN); if (err) return err; -- cgit v1.2.3-70-g09d2