From 34adc8062227f41b04ade0ff3fbd1dbe3002669e Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 20 May 2009 20:13:28 +0200 Subject: [PATCH] perf_counter: Fix context removal deadlock Disable the PMU globally before removing a counter from a context. This fixes the following lockup: [22081.741922] ------------[ cut here ]------------ [22081.746668] WARNING: at arch/x86/kernel/cpu/perf_counter.c:803 intel_pmu_handle_irq+0x9b/0x24e() [22081.755624] Hardware name: X8DTN [22081.758903] perfcounters: irq loop stuck! [22081.762985] Modules linked in: [22081.766136] Pid: 11082, comm: perf Not tainted 2.6.30-rc6-tip #226 [22081.772432] Call Trace: [22081.774940] [] ? intel_pmu_handle_irq+0x9b/0x24e [22081.781993] [] ? intel_pmu_handle_irq+0x9b/0x24e [22081.788368] [] ? warn_slowpath_common+0x77/0xa3 [22081.794649] [] ? warn_slowpath_fmt+0x40/0x45 [22081.800696] [] ? intel_pmu_handle_irq+0x9b/0x24e [22081.807080] [] ? perf_counter_nmi_handler+0x3f/0x4a [22081.813751] [] ? notifier_call_chain+0x58/0x86 [22081.819951] [] ? notify_die+0x2d/0x32 [22081.825392] [] ? do_nmi+0x8e/0x242 [22081.830538] [] ? nmi+0x1a/0x20 [22081.835342] [] ? selinux_file_free_security+0x0/0x1a [22081.842105] [] ? x86_pmu_disable_counter+0x15/0x41 [22081.848673] <> [] ? x86_pmu_disable+0x86/0x103 [22081.855512] [] ? __perf_counter_remove_from_context+0x0/0xfe [22081.862926] [] ? counter_sched_out+0x30/0xce [22081.868909] [] ? __perf_counter_remove_from_context+0x59/0xfe [22081.876382] [] ? smp_call_function_single+0x6c/0xe6 [22081.882955] [] ? perf_release+0x86/0x14c [22081.888600] [] ? __fput+0xe7/0x195 [22081.893718] [] ? filp_close+0x5b/0x62 [22081.899107] [] ? put_files_struct+0x64/0xc2 [22081.905031] [] ? do_exit+0x1e2/0x6ef [22081.910360] [] ? _spin_lock_irqsave+0x9/0xe [22081.916292] [] ? do_group_exit+0x67/0x93 [22081.921953] [] ? sys_exit_group+0x12/0x16 [22081.927759] [] ? system_call_fastpath+0x16/0x1b [22081.934076] ---[ end trace 3a3936ce3e1b4505 ]--- And could potentially also fix the lockup reported by Marcelo Tosatti. Also, print more debug info in case of a detected lockup. [ Impact: fix lockup ] Reported-by: Marcelo Tosatti Acked-by: Peter Zijlstra Cc: Paul Mackerras Cc: Corey Ashford Cc: Thomas Gleixner LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 1 + kernel/perf_counter.c | 12 ++++++------ 2 files changed, 7 insertions(+), 6 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index c109819c2cb..6cc1660db8d 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -740,6 +740,7 @@ static int intel_pmu_handle_irq(struct pt_regs *regs, int nmi) again: if (++loops > 100) { WARN_ONCE(1, "perfcounters: irq loop stuck!\n"); + perf_counter_print_debug(); return 1; } diff --git a/kernel/perf_counter.c b/kernel/perf_counter.c index 69d4de81596..08584c16049 100644 --- a/kernel/perf_counter.c +++ b/kernel/perf_counter.c @@ -208,18 +208,17 @@ static void __perf_counter_remove_from_context(void *info) return; spin_lock_irqsave(&ctx->lock, flags); + /* + * Protect the list operation against NMI by disabling the + * counters on a global level. + */ + perf_disable(); counter_sched_out(counter, cpuctx, ctx); counter->task = NULL; - /* - * Protect the list operation against NMI by disabling the - * counters on a global level. NOP for non NMI based counters. - */ - perf_disable(); list_del_counter(counter, ctx); - perf_enable(); if (!ctx->task) { /* @@ -231,6 +230,7 @@ static void __perf_counter_remove_from_context(void *info) perf_max_counters - perf_reserved_percpu); } + perf_enable(); spin_unlock_irqrestore(&ctx->lock, flags); } -- 2.11.4.GIT