From 97f7f8189fe54e3cfe324ef9ad35064f3d2d3bff Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Mon, 10 Oct 2011 16:21:10 +0200 Subject: oprofile, x86: Fix crash when unloading module (nmi timer mode) If oprofile uses the nmi timer interrupt there is a crash while unloading the module. The bug can be triggered with oprofile build as module and kernel parameter nolapic set. This patch fixes this. oprofile: using NMI timer interrupt. BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 IP: [] unregister_syscore_ops+0x41/0x58 PGD 42dbca067 PUD 41da6a067 PMD 0 Oops: 0002 [#1] PREEMPT SMP CPU 5 Modules linked in: oprofile(-) [last unloaded: oprofile] Pid: 2518, comm: modprobe Not tainted 3.1.0-rc7-00019-gb2fb49d #19 Advanced Micro Device Anaheim/Anaheim RIP: 0010:[] [] unregister_syscore_ops+0x41/0x58 RSP: 0018:ffff88041ef71e98 EFLAGS: 00010296 RAX: 0000000000000000 RBX: ffffffffa0017100 RCX: dead000000200200 RDX: 0000000000000000 RSI: dead000000100100 RDI: ffffffff8178c620 RBP: ffff88041ef71ea8 R08: 0000000000000001 R09: 0000000000000082 R10: 0000000000000000 R11: ffff88041ef71de8 R12: 0000000000000080 R13: fffffffffffffff5 R14: 0000000000000001 R15: 0000000000610210 FS: 00007fc902f20700(0000) GS:ffff88042fd40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 0000000000000008 CR3: 000000041cdb6000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process modprobe (pid: 2518, threadinfo ffff88041ef70000, task ffff88041d348040) Stack: ffff88041ef71eb8 ffffffffa0017790 ffff88041ef71eb8 ffffffffa0013532 ffff88041ef71ec8 ffffffffa00132d6 ffff88041ef71ed8 ffffffffa00159b2 ffff88041ef71f78 ffffffff81073115 656c69666f72706f 0000000000610200 Call Trace: [] op_nmi_exit+0x15/0x17 [oprofile] [] oprofile_arch_exit+0xe/0x10 [oprofile] [] oprofile_exit+0x1e/0x20 [oprofile] [] sys_delete_module+0x1c3/0x22f [] ? trace_hardirqs_on_thunk+0x3a/0x3f [] system_call_fastpath+0x16/0x1b Code: 20 c6 78 81 e8 c5 cc 23 00 48 8b 13 48 8b 43 08 48 be 00 01 10 00 00 00 ad de 48 b9 00 02 20 00 00 00 ad de 48 c7 c7 20 c6 78 81 89 42 08 48 89 10 48 89 33 48 89 4b 08 e8 a6 c0 23 00 5a 5b RIP [] unregister_syscore_ops+0x41/0x58 RSP CR2: 0000000000000008 ---[ end trace 43a541a52956b7b0 ]--- CC: stable@kernel.org # 2.6.37+ Signed-off-by: Robert Richter --- arch/x86/oprofile/init.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/oprofile/init.c b/arch/x86/oprofile/init.c index cdfe4c54deca..f148cf652678 100644 --- a/arch/x86/oprofile/init.c +++ b/arch/x86/oprofile/init.c @@ -21,6 +21,7 @@ extern int op_nmi_timer_init(struct oprofile_operations *ops); extern void op_nmi_exit(void); extern void x86_backtrace(struct pt_regs * const regs, unsigned int depth); +static int nmi_timer; int __init oprofile_arch_init(struct oprofile_operations *ops) { @@ -31,8 +32,9 @@ int __init oprofile_arch_init(struct oprofile_operations *ops) #ifdef CONFIG_X86_LOCAL_APIC ret = op_nmi_init(ops); #endif + nmi_timer = (ret != 0); #ifdef CONFIG_X86_IO_APIC - if (ret < 0) + if (nmi_timer) ret = op_nmi_timer_init(ops); #endif ops->backtrace = x86_backtrace; @@ -44,6 +46,7 @@ int __init oprofile_arch_init(struct oprofile_operations *ops) void oprofile_arch_exit(void) { #ifdef CONFIG_X86_LOCAL_APIC - op_nmi_exit(); + if (!nmi_timer) + op_nmi_exit(); #endif } -- cgit v1.2.3 From 57d1c0c03c6b48b2b96870d831b9ce6b917f53ac Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 7 Oct 2011 13:36:40 +0200 Subject: perf/x86: Fix PEBS instruction unwind Masami spotted that we always try to decode the instruction stream as 64bit instructions when running a 64bit kernel, this doesn't work for ia32-compat proglets. Use TIF_IA32 to detect if we need to use the 32bit instruction decoder. Reported-by: Masami Hiramatsu Cc: stable@kernel.org Signed-off-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel_ds.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_event_intel_ds.c b/arch/x86/kernel/cpu/perf_event_intel_ds.c index c0d238f49db8..73da6b64f5b7 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_ds.c +++ b/arch/x86/kernel/cpu/perf_event_intel_ds.c @@ -493,6 +493,7 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs) unsigned long from = cpuc->lbr_entries[0].from; unsigned long old_to, to = cpuc->lbr_entries[0].to; unsigned long ip = regs->ip; + int is_64bit = 0; /* * We don't need to fixup if the PEBS assist is fault like @@ -544,7 +545,10 @@ static int intel_pmu_pebs_fixup_ip(struct pt_regs *regs) } else kaddr = (void *)to; - kernel_insn_init(&insn, kaddr); +#ifdef CONFIG_X86_64 + is_64bit = kernel_ip(to) || !test_thread_flag(TIF_IA32); +#endif + insn_init(&insn, kaddr, is_64bit); insn_get_length(&insn); to += insn.length; } while (to < ip); -- cgit v1.2.3 From aa2bc1ade59003a379ffc485d6da2d92ea3370a6 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 9 Nov 2011 17:56:37 +0100 Subject: perf: Don't use -ENOSPC for out of PMU resources People (Linus) objected to using -ENOSPC to signal not having enough resources on the PMU to satisfy the request. Use -EINVAL. Requested-by: Linus Torvalds Cc: Stephane Eranian Cc: Will Deacon Cc: Deng-Cheng Zhu Cc: David Daney Cc: Ralf Baechle Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-xv8geaz2zpbjhlx0svmpp28n@git.kernel.org [ merged to newer kernel, fixed up MIPS impact ] Signed-off-by: Ingo Molnar --- arch/arm/kernel/perf_event.c | 6 +++--- arch/mips/kernel/perf_event_mipsxx.c | 8 ++++---- arch/x86/kernel/cpu/perf_event.c | 10 +++++----- arch/x86/kernel/cpu/perf_event_p4.c | 2 +- 4 files changed, 13 insertions(+), 13 deletions(-) (limited to 'arch') diff --git a/arch/arm/kernel/perf_event.c b/arch/arm/kernel/perf_event.c index 24e2347be6b1..ff17b17b668e 100644 --- a/arch/arm/kernel/perf_event.c +++ b/arch/arm/kernel/perf_event.c @@ -347,15 +347,15 @@ validate_group(struct perf_event *event) memset(&fake_pmu, 0, sizeof(fake_pmu)); if (!validate_event(&fake_pmu, leader)) - return -ENOSPC; + return -EINVAL; list_for_each_entry(sibling, &leader->sibling_list, group_entry) { if (!validate_event(&fake_pmu, sibling)) - return -ENOSPC; + return -EINVAL; } if (!validate_event(&fake_pmu, event)) - return -ENOSPC; + return -EINVAL; return 0; } diff --git a/arch/mips/kernel/perf_event_mipsxx.c b/arch/mips/kernel/perf_event_mipsxx.c index 4f2971bcf8e5..315fc0b250f8 100644 --- a/arch/mips/kernel/perf_event_mipsxx.c +++ b/arch/mips/kernel/perf_event_mipsxx.c @@ -623,7 +623,7 @@ static int mipspmu_event_init(struct perf_event *event) if (!atomic_inc_not_zero(&active_events)) { if (atomic_read(&active_events) > MIPS_MAX_HWEVENTS) { atomic_dec(&active_events); - return -ENOSPC; + return -EINVAL; } mutex_lock(&pmu_reserve_mutex); @@ -732,15 +732,15 @@ static int validate_group(struct perf_event *event) memset(&fake_cpuc, 0, sizeof(fake_cpuc)); if (!validate_event(&fake_cpuc, leader)) - return -ENOSPC; + return -EINVAL; list_for_each_entry(sibling, &leader->sibling_list, group_entry) { if (!validate_event(&fake_cpuc, sibling)) - return -ENOSPC; + return -EINVAL; } if (!validate_event(&fake_cpuc, event)) - return -ENOSPC; + return -EINVAL; return 0; } diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 640891014b2a..ff0e8d498750 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -588,7 +588,7 @@ done: x86_pmu.put_event_constraints(cpuc, cpuc->event_list[i]); } } - return num ? -ENOSPC : 0; + return num ? -EINVAL : 0; } /* @@ -607,7 +607,7 @@ static int collect_events(struct cpu_hw_events *cpuc, struct perf_event *leader, if (is_x86_event(leader)) { if (n >= max_count) - return -ENOSPC; + return -EINVAL; cpuc->event_list[n] = leader; n++; } @@ -620,7 +620,7 @@ static int collect_events(struct cpu_hw_events *cpuc, struct perf_event *leader, continue; if (n >= max_count) - return -ENOSPC; + return -EINVAL; cpuc->event_list[n] = event; n++; @@ -1316,7 +1316,7 @@ static int validate_event(struct perf_event *event) c = x86_pmu.get_event_constraints(fake_cpuc, event); if (!c || !c->weight) - ret = -ENOSPC; + ret = -EINVAL; if (x86_pmu.put_event_constraints) x86_pmu.put_event_constraints(fake_cpuc, event); @@ -1341,7 +1341,7 @@ static int validate_group(struct perf_event *event) { struct perf_event *leader = event->group_leader; struct cpu_hw_events *fake_cpuc; - int ret = -ENOSPC, n; + int ret = -EINVAL, n; fake_cpuc = allocate_fake_cpuc(); if (IS_ERR(fake_cpuc)) diff --git a/arch/x86/kernel/cpu/perf_event_p4.c b/arch/x86/kernel/cpu/perf_event_p4.c index 492bf1358a7c..ef484d9d0a25 100644 --- a/arch/x86/kernel/cpu/perf_event_p4.c +++ b/arch/x86/kernel/cpu/perf_event_p4.c @@ -1268,7 +1268,7 @@ reserve: } done: - return num ? -ENOSPC : 0; + return num ? -EINVAL : 0; } static __initconst const struct x86_pmu p4_pmu = { -- cgit v1.2.3 From ed13ec58bfe0d5dc95f748e6118432cb0fa283cb Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 14 Nov 2011 10:03:25 +0100 Subject: perf/x86: Enable raw event access to Intel offcore events Now that the core offcore support is fixed up (thanks Stephane) and we have sane generic events utilizing them, re-enable the raw access to the feature as well. Note that it doesn't matter if you use event 0x1b7 or 0x1bb to specify an offcore event, either one works and neither guarantees you'll end up on a particular offcore MSR. Based on original patch from: Vince Weaver . Signed-off-by: Peter Zijlstra Cc: Vince Weaver . Cc: Stephane Eranian Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1108031200390.703@cl320.eecs.utk.edu Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index ff0e8d498750..2bda212a0010 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -312,12 +312,8 @@ int x86_setup_perfctr(struct perf_event *event) return -EOPNOTSUPP; } - /* - * Do not allow config1 (extended registers) to propagate, - * there's no sane user-space generalization yet: - */ if (attr->type == PERF_TYPE_RAW) - return 0; + return x86_pmu_extra_regs(event->attr.config, event); if (attr->type == PERF_TYPE_HW_CACHE) return set_ext_hw_attr(hwc, event); -- cgit v1.2.3 From 6a600a8b8749566a7d81ad75dcb8bf5342b5a39a Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 15 Nov 2011 10:51:15 +0100 Subject: perf, x86: Disable PEBS on SandyBridge chips Cc: Stephane Eranian Cc: stable@kernel.org Signed-off-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 2be5ebe99872..8d601b18bf9f 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -1545,6 +1545,13 @@ static void intel_clovertown_quirks(void) x86_pmu.pebs_constraints = NULL; } +static void intel_sandybridge_quirks(void) +{ + printk(KERN_WARNING "PEBS disabled due to CPU errata.\n"); + x86_pmu.pebs = 0; + x86_pmu.pebs_constraints = NULL; +} + __init int intel_pmu_init(void) { union cpuid10_edx edx; @@ -1694,6 +1701,7 @@ __init int intel_pmu_init(void) break; case 42: /* SandyBridge */ + x86_pmu.quirks = intel_sandybridge_quirks; case 45: /* SandyBridge, "Romely-EP" */ memcpy(hw_cache_event_ids, snb_hw_cache_event_ids, sizeof(hw_cache_event_ids)); -- cgit v1.2.3 From 16e5294e5f8303756a179cf218e37dfb9ed34417 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Tue, 8 Nov 2011 19:20:44 +0100 Subject: perf, x86: Force IBS LVT offset assignment for family 10h On AMD family 10h we see firmware bug messages like the following: [Firmware Bug]: cpu 6, try to use APIC500 (LVT offset 0) for vector 0x10400, but the register is already in use for vector 0xf9 on another cpu [Firmware Bug]: cpu 6, IBS interrupt offset 0 not available (MSRC001103A=0x0000000000000100) [Firmware Bug]: using offset 1 for IBS interrupts [Firmware Bug]: workaround enabled for IBS LVT offset perf: AMD IBS detected (0x00000007) We always see this, since the offsets are not assigned by the BIOS for this family. Force LVT offset assignment in this case. If the OS assignment fails, fallback to BIOS settings and try to setup this. The fallback to BIOS settings weakens the family check since force_ibs_eilvt_setup() may fail e.g. in case of virtual machines. But setup may still succeed if BIOS offsets are correct. Other families don't have a workaround implemented that assigns LVT offsets. It's ok, to drop calling force_ibs_eilvt_setup() for that families. With the patch the [Firmware Bug] messages vanish. We see now: IBS: LVT offset 1 assigned perf: AMD IBS detected (0x00000007) Signed-off-by: Robert Richter Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/r/20111109162225.GO12451@erda.amd.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_amd_ibs.c | 29 ++++++++++++++++++----------- 1 file changed, 18 insertions(+), 11 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_event_amd_ibs.c b/arch/x86/kernel/cpu/perf_event_amd_ibs.c index ab6343d21825..3b8a2d30d14e 100644 --- a/arch/x86/kernel/cpu/perf_event_amd_ibs.c +++ b/arch/x86/kernel/cpu/perf_event_amd_ibs.c @@ -199,8 +199,7 @@ static int force_ibs_eilvt_setup(void) goto out; } - pr_err(FW_BUG "using offset %d for IBS interrupts\n", offset); - pr_err(FW_BUG "workaround enabled for IBS LVT offset\n"); + pr_info("IBS: LVT offset %d assigned\n", offset); return 0; out: @@ -265,19 +264,23 @@ perf_ibs_cpu_notifier(struct notifier_block *self, unsigned long action, void *h static __init int amd_ibs_init(void) { u32 caps; - int ret; + int ret = -EINVAL; caps = __get_ibs_caps(); if (!caps) return -ENODEV; /* ibs not supported by the cpu */ - if (!ibs_eilvt_valid()) { - ret = force_ibs_eilvt_setup(); - if (ret) { - pr_err("Failed to setup IBS, %d\n", ret); - return ret; - } - } + /* + * Force LVT offset assignment for family 10h: The offsets are + * not assigned by the BIOS for this family, so the OS is + * responsible for doing it. If the OS assignment fails, fall + * back to BIOS settings and try to setup this. + */ + if (boot_cpu_data.x86 == 0x10) + force_ibs_eilvt_setup(); + + if (!ibs_eilvt_valid()) + goto out; get_online_cpus(); ibs_caps = caps; @@ -287,7 +290,11 @@ static __init int amd_ibs_init(void) smp_call_function(setup_APIC_ibs, NULL, 1); put_online_cpus(); - return perf_event_ibs_init(); + ret = perf_event_ibs_init(); +out: + if (ret) + pr_err("Failed to setup IBS, %d\n", ret); + return ret; } /* Since we need the pci subsystem to init ibs we can't do this earlier: */ -- cgit v1.2.3