From b23ff0e9330e4b11e18af984d50573598e10e7f9 Mon Sep 17 00:00:00 2001 From: Mahesh Salgaonkar Date: Thu, 21 Jan 2010 18:25:16 +0530 Subject: hw_breakpoints: Release the bp slot if arch_validate_hwbkpt_settings() fails. On a given architecture, when hardware breakpoint registration fails due to un-supported access type (read/write/execute), we lose the bp slot since register_perf_hw_breakpoint() does not release the bp slot on failure. Hence, any subsequent hardware breakpoint registration starts failing with 'no space left on device' error. This patch introduces error handling in register_perf_hw_breakpoint() function and releases bp slot on error. Signed-off-by: Mahesh Salgaonkar Cc: Ananth N Mavinakayanahalli Cc: K. Prasad Cc: Maneesh Soni LKML-Reference: <20100121125516.GA32521@in.ibm.com> Signed-off-by: Frederic Weisbecker --- kernel/hw_breakpoint.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'kernel/hw_breakpoint.c') diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 50dbd599958..c030ae657f2 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -296,6 +296,10 @@ int register_perf_hw_breakpoint(struct perf_event *bp) if (!bp->attr.disabled || !bp->overflow_handler) ret = arch_validate_hwbkpt_settings(bp, bp->ctx->task); + /* if arch_validate_hwbkpt_settings() fails then release bp slot */ + if (ret) + release_bp_slot(bp); + return ret; } -- cgit v1.2.3-70-g09d2 From 5352ae638e2d7d5c9b2e4d528676bbf2af6fd6f3 Mon Sep 17 00:00:00 2001 From: Jason Wessel Date: Thu, 28 Jan 2010 17:04:43 -0600 Subject: perf, hw_breakpoint, kgdb: Do not take mutex for kernel debugger This patch fixes the regression in functionality where the kernel debugger and the perf API do not nicely share hw breakpoint reservations. The kernel debugger cannot use any mutex_lock() calls because it can start the kernel running from an invalid context. A mutex free version of the reservation API needed to get created for the kernel debugger to safely update hw breakpoint reservations. The possibility for a breakpoint reservation to be concurrently processed at the time that kgdb interrupts the system is improbable. Should this corner case occur the end user is warned, and the kernel debugger will prohibit updating the hardware breakpoint reservations. Any time the kernel debugger reserves a hardware breakpoint it will be a system wide reservation. Signed-off-by: Jason Wessel Acked-by: Frederic Weisbecker Cc: kgdb-bugreport@lists.sourceforge.net Cc: K.Prasad Cc: Peter Zijlstra Cc: Alan Stern Cc: torvalds@linux-foundation.org LKML-Reference: <1264719883-7285-3-git-send-email-jason.wessel@windriver.com> Signed-off-by: Ingo Molnar --- arch/x86/kernel/kgdb.c | 51 ++++++++++++++++++++++++++++++++++++++++++ include/linux/hw_breakpoint.h | 2 ++ kernel/hw_breakpoint.c | 52 ++++++++++++++++++++++++++++++++++--------- 3 files changed, 95 insertions(+), 10 deletions(-) (limited to 'kernel/hw_breakpoint.c') diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c index 62bea7307ea..bfba6019d76 100644 --- a/arch/x86/kernel/kgdb.c +++ b/arch/x86/kernel/kgdb.c @@ -239,6 +239,49 @@ static void kgdb_correct_hw_break(void) hw_breakpoint_restore(); } +static int hw_break_reserve_slot(int breakno) +{ + int cpu; + int cnt = 0; + struct perf_event **pevent; + + for_each_online_cpu(cpu) { + cnt++; + pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu); + if (dbg_reserve_bp_slot(*pevent)) + goto fail; + } + + return 0; + +fail: + for_each_online_cpu(cpu) { + cnt--; + if (!cnt) + break; + pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu); + dbg_release_bp_slot(*pevent); + } + return -1; +} + +static int hw_break_release_slot(int breakno) +{ + struct perf_event **pevent; + int cpu; + + for_each_online_cpu(cpu) { + pevent = per_cpu_ptr(breakinfo[breakno].pev, cpu); + if (dbg_release_bp_slot(*pevent)) + /* + * The debugger is responisble for handing the retry on + * remove failure. + */ + return -1; + } + return 0; +} + static int kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype) { @@ -250,6 +293,10 @@ kgdb_remove_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype) if (i == 4) return -1; + if (hw_break_release_slot(i)) { + printk(KERN_ERR "Cannot remove hw breakpoint at %lx\n", addr); + return -1; + } breakinfo[i].enabled = 0; return 0; @@ -316,6 +363,10 @@ kgdb_set_hw_break(unsigned long addr, int len, enum kgdb_bptype bptype) return -1; } breakinfo[i].addr = addr; + if (hw_break_reserve_slot(i)) { + breakinfo[i].addr = 0; + return -1; + } breakinfo[i].enabled = 1; return 0; diff --git a/include/linux/hw_breakpoint.h b/include/linux/hw_breakpoint.h index 41235c93e4e..070ba062173 100644 --- a/include/linux/hw_breakpoint.h +++ b/include/linux/hw_breakpoint.h @@ -75,6 +75,8 @@ extern int __register_perf_hw_breakpoint(struct perf_event *bp); extern void unregister_hw_breakpoint(struct perf_event *bp); extern void unregister_wide_hw_breakpoint(struct perf_event **cpu_events); +extern int dbg_reserve_bp_slot(struct perf_event *bp); +extern int dbg_release_bp_slot(struct perf_event *bp); extern int reserve_bp_slot(struct perf_event *bp); extern void release_bp_slot(struct perf_event *bp); diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index c030ae657f2..8a5c7d55ac9 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -243,38 +243,70 @@ static void toggle_bp_slot(struct perf_event *bp, bool enable) * ((per_cpu(nr_bp_flexible, *) > 1) + max(per_cpu(nr_cpu_bp_pinned, *)) * + max(per_cpu(nr_task_bp_pinned, *))) < HBP_NUM */ -int reserve_bp_slot(struct perf_event *bp) +static int __reserve_bp_slot(struct perf_event *bp) { struct bp_busy_slots slots = {0}; - int ret = 0; - - mutex_lock(&nr_bp_mutex); fetch_bp_busy_slots(&slots, bp); /* Flexible counters need to keep at least one slot */ - if (slots.pinned + (!!slots.flexible) == HBP_NUM) { - ret = -ENOSPC; - goto end; - } + if (slots.pinned + (!!slots.flexible) == HBP_NUM) + return -ENOSPC; toggle_bp_slot(bp, true); -end: + return 0; +} + +int reserve_bp_slot(struct perf_event *bp) +{ + int ret; + + mutex_lock(&nr_bp_mutex); + + ret = __reserve_bp_slot(bp); + mutex_unlock(&nr_bp_mutex); return ret; } +static void __release_bp_slot(struct perf_event *bp) +{ + toggle_bp_slot(bp, false); +} + void release_bp_slot(struct perf_event *bp) { mutex_lock(&nr_bp_mutex); - toggle_bp_slot(bp, false); + __release_bp_slot(bp); mutex_unlock(&nr_bp_mutex); } +/* + * Allow the kernel debugger to reserve breakpoint slots without + * taking a lock using the dbg_* variant of for the reserve and + * release breakpoint slots. + */ +int dbg_reserve_bp_slot(struct perf_event *bp) +{ + if (mutex_is_locked(&nr_bp_mutex)) + return -1; + + return __reserve_bp_slot(bp); +} + +int dbg_release_bp_slot(struct perf_event *bp) +{ + if (mutex_is_locked(&nr_bp_mutex)) + return -1; + + __release_bp_slot(bp); + + return 0; +} int register_perf_hw_breakpoint(struct perf_event *bp) { -- cgit v1.2.3-70-g09d2 From cd757645fbdc34a8343c04bb0e74e06fccc2cb10 Mon Sep 17 00:00:00 2001 From: Mahesh Salgaonkar Date: Sat, 30 Jan 2010 10:25:18 +0530 Subject: perf: Make bp_len type to u64 generic across the arch Change 'bp_len' type to __u64 to make it work across archs as the s390 architecture watch point length can be upto 2^64. reference: http://lkml.org/lkml/2010/1/25/212 This is an ABI change that is not backward compatible with the previous hardware breakpoint info layout integrated in this development cycle, a rebuilt of perf tools is necessary for versions based on 2.6.33-rc1 - 2.6.33-rc6 to work with a kernel based on this patch. Signed-off-by: Mahesh Salgaonkar Acked-by: Peter Zijlstra Cc: Ananth N Mavinakayanahalli Cc: "K. Prasad" Cc: Maneesh Soni Cc: Heiko Carstens Cc: Martin LKML-Reference: <20100130045518.GA20776@in.ibm.com> Signed-off-by: Frederic Weisbecker --- include/linux/hw_breakpoint.h | 2 +- include/linux/perf_event.h | 6 ++---- kernel/hw_breakpoint.c | 2 +- kernel/perf_event.c | 2 +- 4 files changed, 5 insertions(+), 7 deletions(-) (limited to 'kernel/hw_breakpoint.c') diff --git a/include/linux/hw_breakpoint.h b/include/linux/hw_breakpoint.h index 070ba062173..5977b724f7c 100644 --- a/include/linux/hw_breakpoint.h +++ b/include/linux/hw_breakpoint.h @@ -44,7 +44,7 @@ static inline int hw_breakpoint_type(struct perf_event *bp) return bp->attr.bp_type; } -static inline int hw_breakpoint_len(struct perf_event *bp) +static inline unsigned long hw_breakpoint_len(struct perf_event *bp) { return bp->attr.bp_len; } diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 8fa71874113..a177698d95e 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -211,11 +211,9 @@ struct perf_event_attr { __u32 wakeup_watermark; /* bytes before wakeup */ }; - __u32 __reserved_2; - - __u64 bp_addr; __u32 bp_type; - __u32 bp_len; + __u64 bp_addr; + __u64 bp_len; }; /* diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 8a5c7d55ac9..967e66143e1 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -360,8 +360,8 @@ EXPORT_SYMBOL_GPL(register_user_hw_breakpoint); int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr) { u64 old_addr = bp->attr.bp_addr; + u64 old_len = bp->attr.bp_len; int old_type = bp->attr.bp_type; - int old_len = bp->attr.bp_len; int err = 0; perf_event_disable(bp); diff --git a/kernel/perf_event.c b/kernel/perf_event.c index d27746bd3a0..2b19297742c 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -4580,7 +4580,7 @@ static int perf_copy_attr(struct perf_event_attr __user *uattr, if (attr->type >= PERF_TYPE_MAX) return -EINVAL; - if (attr->__reserved_1 || attr->__reserved_2) + if (attr->__reserved_1) return -EINVAL; if (attr->sample_type & ~(PERF_SAMPLE_MAX-1)) -- cgit v1.2.3-70-g09d2 From 44ee63587dce85593c22497140db16f4e5027860 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Wed, 17 Feb 2010 10:50:50 +0900 Subject: percpu: Add __percpu sparse annotations to hw_breakpoint Add __percpu sparse annotations to hw_breakpoint. These annotations are to make sparse consider percpu variables to be in a different address space and warn if accessed without going through percpu accessors. This patch doesn't affect normal builds. In kernel/hw_breakpoint.c, per_cpu(nr_task_bp_pinned, cpu)'s will trigger spurious noderef related warnings from sparse. Changing it to &per_cpu(nr_task_bp_pinned[0], cpu) will work around the problem but deemed to ugly by the maintainer. Leave it alone until better solution can be found. Signed-off-by: Tejun Heo Cc: Stephen Rothwell Cc: K.Prasad LKML-Reference: <4B7B4B7A.9050902@kernel.org> Signed-off-by: Frederic Weisbecker --- include/linux/hw_breakpoint.h | 8 ++++---- kernel/hw_breakpoint.c | 10 +++++----- samples/hw_breakpoint/data_breakpoint.c | 6 +++--- 3 files changed, 12 insertions(+), 12 deletions(-) (limited to 'kernel/hw_breakpoint.c') diff --git a/include/linux/hw_breakpoint.h b/include/linux/hw_breakpoint.h index 5977b724f7c..c70d27af03f 100644 --- a/include/linux/hw_breakpoint.h +++ b/include/linux/hw_breakpoint.h @@ -66,14 +66,14 @@ register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr, perf_overflow_handler_t triggered, int cpu); -extern struct perf_event ** +extern struct perf_event * __percpu * register_wide_hw_breakpoint(struct perf_event_attr *attr, perf_overflow_handler_t triggered); extern int register_perf_hw_breakpoint(struct perf_event *bp); extern int __register_perf_hw_breakpoint(struct perf_event *bp); extern void unregister_hw_breakpoint(struct perf_event *bp); -extern void unregister_wide_hw_breakpoint(struct perf_event **cpu_events); +extern void unregister_wide_hw_breakpoint(struct perf_event * __percpu *cpu_events); extern int dbg_reserve_bp_slot(struct perf_event *bp); extern int dbg_release_bp_slot(struct perf_event *bp); @@ -100,7 +100,7 @@ static inline struct perf_event * register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr, perf_overflow_handler_t triggered, int cpu) { return NULL; } -static inline struct perf_event ** +static inline struct perf_event * __percpu * register_wide_hw_breakpoint(struct perf_event_attr *attr, perf_overflow_handler_t triggered) { return NULL; } static inline int @@ -109,7 +109,7 @@ static inline int __register_perf_hw_breakpoint(struct perf_event *bp) { return -ENOSYS; } static inline void unregister_hw_breakpoint(struct perf_event *bp) { } static inline void -unregister_wide_hw_breakpoint(struct perf_event **cpu_events) { } +unregister_wide_hw_breakpoint(struct perf_event * __percpu *cpu_events) { } static inline int reserve_bp_slot(struct perf_event *bp) {return -ENOSYS; } static inline void release_bp_slot(struct perf_event *bp) { } diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 967e66143e1..6542eacb3fa 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -413,17 +413,17 @@ EXPORT_SYMBOL_GPL(unregister_hw_breakpoint); * * @return a set of per_cpu pointers to perf events */ -struct perf_event ** +struct perf_event * __percpu * register_wide_hw_breakpoint(struct perf_event_attr *attr, perf_overflow_handler_t triggered) { - struct perf_event **cpu_events, **pevent, *bp; + struct perf_event * __percpu *cpu_events, **pevent, *bp; long err; int cpu; cpu_events = alloc_percpu(typeof(*cpu_events)); if (!cpu_events) - return ERR_PTR(-ENOMEM); + return (void __percpu __force *)ERR_PTR(-ENOMEM); get_online_cpus(); for_each_online_cpu(cpu) { @@ -451,7 +451,7 @@ fail: put_online_cpus(); free_percpu(cpu_events); - return ERR_PTR(err); + return (void __percpu __force *)ERR_PTR(err); } EXPORT_SYMBOL_GPL(register_wide_hw_breakpoint); @@ -459,7 +459,7 @@ EXPORT_SYMBOL_GPL(register_wide_hw_breakpoint); * unregister_wide_hw_breakpoint - unregister a wide breakpoint in the kernel * @cpu_events: the per cpu set of events to unregister */ -void unregister_wide_hw_breakpoint(struct perf_event **cpu_events) +void unregister_wide_hw_breakpoint(struct perf_event * __percpu *cpu_events) { int cpu; struct perf_event **pevent; diff --git a/samples/hw_breakpoint/data_breakpoint.c b/samples/hw_breakpoint/data_breakpoint.c index c69cbe9b242..bd0f337afca 100644 --- a/samples/hw_breakpoint/data_breakpoint.c +++ b/samples/hw_breakpoint/data_breakpoint.c @@ -34,7 +34,7 @@ #include #include -struct perf_event **sample_hbp; +struct perf_event * __percpu *sample_hbp; static char ksym_name[KSYM_NAME_LEN] = "pid_max"; module_param_string(ksym, ksym_name, KSYM_NAME_LEN, S_IRUGO); @@ -61,8 +61,8 @@ static int __init hw_break_module_init(void) attr.bp_type = HW_BREAKPOINT_W | HW_BREAKPOINT_R; sample_hbp = register_wide_hw_breakpoint(&attr, sample_hbp_handler); - if (IS_ERR(sample_hbp)) { - ret = PTR_ERR(sample_hbp); + if (IS_ERR((void __force *)sample_hbp)) { + ret = PTR_ERR((void __force *)sample_hbp); goto fail; } -- cgit v1.2.3-70-g09d2 From 1e259e0a9982078896f3404240096cbea01daca4 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sun, 28 Feb 2010 20:51:15 +0100 Subject: hw-breakpoints: Remove stub unthrottle callback We support event unthrottling in breakpoint events. It means that if we have more than sysctl_perf_event_sample_rate/HZ, perf will throttle, ignoring subsequent events until the next tick. So if ptrace exceeds this max rate, it will omit events, which breaks the ptrace determinism that is supposed to report every triggered breakpoints. This is likely to happen if we set sysctl_perf_event_sample_rate to 1. This patch removes support for unthrottling in breakpoint events to break throttling and restore ptrace determinism. Signed-off-by: Frederic Weisbecker Cc: 2.6.33.x Cc: Peter Zijlstra Cc: K.Prasad Cc: Paul Mackerras --- arch/x86/kernel/hw_breakpoint.c | 5 ----- kernel/hw_breakpoint.c | 1 - 2 files changed, 6 deletions(-) (limited to 'kernel/hw_breakpoint.c') diff --git a/arch/x86/kernel/hw_breakpoint.c b/arch/x86/kernel/hw_breakpoint.c index bb6006e3e29..1e8ceadc0d6 100644 --- a/arch/x86/kernel/hw_breakpoint.c +++ b/arch/x86/kernel/hw_breakpoint.c @@ -531,8 +531,3 @@ void hw_breakpoint_pmu_read(struct perf_event *bp) { /* TODO */ } - -void hw_breakpoint_pmu_unthrottle(struct perf_event *bp) -{ - /* TODO */ -} diff --git a/kernel/hw_breakpoint.c b/kernel/hw_breakpoint.c index 967e66143e1..4d99512ee14 100644 --- a/kernel/hw_breakpoint.c +++ b/kernel/hw_breakpoint.c @@ -489,5 +489,4 @@ struct pmu perf_ops_bp = { .enable = arch_install_hw_breakpoint, .disable = arch_uninstall_hw_breakpoint, .read = hw_breakpoint_pmu_read, - .unthrottle = hw_breakpoint_pmu_unthrottle }; -- cgit v1.2.3-70-g09d2