]> git.baikalelectronics.ru Git - kernel.git/commitdiff
hw-breakpoints: Use overflow handler instead of the event callback
authorFrederic Weisbecker <fweisbec@gmail.com>
Sat, 5 Dec 2009 08:44:31 +0000 (09:44 +0100)
committerFrederic Weisbecker <fweisbec@gmail.com>
Sun, 6 Dec 2009 07:27:18 +0000 (08:27 +0100)
struct perf_event::event callback was called when a breakpoint
triggers. But this is a rather opaque callback, pretty
tied-only to the breakpoint API and not really integrated into perf
as it triggers even when we don't overflow.

We prefer to use overflow_handler() as it fits into the perf events
rules, being called only when we overflow.

Reported-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: "K. Prasad" <prasad@linux.vnet.ibm.com>
arch/x86/kernel/hw_breakpoint.c
arch/x86/kernel/ptrace.c
include/linux/hw_breakpoint.h
include/linux/perf_event.h
kernel/hw_breakpoint.c
kernel/perf_event.c
kernel/trace/trace_ksym.c
samples/hw_breakpoint/data_breakpoint.c

index d42f65ac492766ecaeb60944b4ae573f5bf6f0bf..05d5fec64a9475855e224012dbb54a9d82b1ee4c 100644 (file)
@@ -362,8 +362,7 @@ int arch_validate_hwbkpt_settings(struct perf_event *bp,
                return ret;
        }
 
-       if (bp->callback)
-               ret = arch_store_info(bp);
+       ret = arch_store_info(bp);
 
        if (ret < 0)
                return ret;
@@ -519,7 +518,7 @@ static int __kprobes hw_breakpoint_handler(struct die_args *args)
                        break;
                }
 
-               (bp->callback)(bp, args->regs);
+               perf_bp_event(bp, args->regs);
 
                rcu_read_unlock();
        }
index dbb395572ae2b33d7cec13f54bebb6d1e65ec361..b361d28061d0f4e364a1e234b101c4dbcded94cf 100644 (file)
@@ -555,7 +555,9 @@ static int genregs_set(struct task_struct *target,
        return ret;
 }
 
-static void ptrace_triggered(struct perf_event *bp, void *data)
+static void ptrace_triggered(struct perf_event *bp, int nmi,
+                            struct perf_sample_data *data,
+                            struct pt_regs *regs)
 {
        int i;
        struct thread_struct *thread = &(current->thread);
@@ -599,7 +601,7 @@ ptrace_modify_breakpoint(struct perf_event *bp, int len, int type,
 {
        int err;
        int gen_len, gen_type;
-       DEFINE_BREAKPOINT_ATTR(attr);
+       struct perf_event_attr attr;
 
        /*
         * We shoud have at least an inactive breakpoint at this
@@ -721,9 +723,10 @@ static int ptrace_set_breakpoint_addr(struct task_struct *tsk, int nr,
 {
        struct perf_event *bp;
        struct thread_struct *t = &tsk->thread;
-       DEFINE_BREAKPOINT_ATTR(attr);
+       struct perf_event_attr attr;
 
        if (!t->ptrace_bps[nr]) {
+               hw_breakpoint_init(&attr);
                /*
                 * Put stub len and type to register (reserve) an inactive but
                 * correct bp
index d33096e0dbd491eb6d1ce1cb9394b59ad89bbd0c..4d14a384a01eee25deb5f346a2725c0c9a1939b4 100644 (file)
@@ -20,19 +20,16 @@ enum {
 
 #ifdef CONFIG_HAVE_HW_BREAKPOINT
 
-/* As it's for in-kernel or ptrace use, we want it to be pinned */
-#define DEFINE_BREAKPOINT_ATTR(name)   \
-struct perf_event_attr name = {                \
-       .type = PERF_TYPE_BREAKPOINT,   \
-       .size = sizeof(name),           \
-       .pinned = 1,                    \
-};
-
 static inline void hw_breakpoint_init(struct perf_event_attr *attr)
 {
        attr->type = PERF_TYPE_BREAKPOINT;
        attr->size = sizeof(*attr);
+       /*
+        * As it's for in-kernel or ptrace use, we want it to be pinned
+        * and to call its callback every hits.
+        */
        attr->pinned = 1;
+       attr->sample_period = 1;
 }
 
 static inline unsigned long hw_breakpoint_addr(struct perf_event *bp)
@@ -52,7 +49,7 @@ static inline int hw_breakpoint_len(struct perf_event *bp)
 
 extern struct perf_event *
 register_user_hw_breakpoint(struct perf_event_attr *attr,
-                           perf_callback_t triggered,
+                           perf_overflow_handler_t triggered,
                            struct task_struct *tsk);
 
 /* FIXME: only change from the attr, and don't unregister */
@@ -64,12 +61,12 @@ modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr);
  */
 extern struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,
-                               perf_callback_t triggered,
+                               perf_overflow_handler_t triggered,
                                int cpu);
 
 extern struct perf_event **
 register_wide_hw_breakpoint(struct perf_event_attr *attr,
-                           perf_callback_t triggered);
+                           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);
@@ -90,18 +87,18 @@ static inline struct arch_hw_breakpoint *counter_arch_bp(struct perf_event *bp)
 
 static inline struct perf_event *
 register_user_hw_breakpoint(struct perf_event_attr *attr,
-                           perf_callback_t triggered,
+                           perf_overflow_handler_t triggered,
                            struct task_struct *tsk)    { return NULL; }
 static inline struct perf_event *
 modify_user_hw_breakpoint(struct perf_event *bp,
                          struct perf_event_attr *attr) { return NULL; }
 static inline struct perf_event *
 register_wide_hw_breakpoint_cpu(struct perf_event_attr *attr,
-                               perf_callback_t triggered,
+                               perf_overflow_handler_t  triggered,
                                int cpu)                { return NULL; }
 static inline struct perf_event **
 register_wide_hw_breakpoint(struct perf_event_attr *attr,
-                           perf_callback_t triggered)  { return NULL; }
+                           perf_overflow_handler_t triggered)  { return NULL; }
 static inline int
 register_perf_hw_breakpoint(struct perf_event *bp)     { return -ENOSYS; }
 static inline int
index 84bd28a0ffab8ca35ca1d8d81996593a1f56053e..d2f2667430da9167520eb6a486fb90ce7f4aa8d9 100644 (file)
@@ -565,10 +565,13 @@ struct perf_pending_entry {
        void (*func)(struct perf_pending_entry *);
 };
 
-typedef void (*perf_callback_t)(struct perf_event *, void *);
-
 struct perf_sample_data;
 
+typedef void (*perf_callback_t)(struct perf_event *, void *);
+typedef void (*perf_overflow_handler_t)(struct perf_event *, int,
+                                       struct perf_sample_data *,
+                                       struct pt_regs *regs);
+
 /**
  * struct perf_event - performance event kernel representation:
  */
@@ -660,9 +663,7 @@ struct perf_event {
        struct pid_namespace            *ns;
        u64                             id;
 
-       void (*overflow_handler)(struct perf_event *event,
-                       int nmi, struct perf_sample_data *data,
-                       struct pt_regs *regs);
+       perf_overflow_handler_t         overflow_handler;
 
 #ifdef CONFIG_EVENT_PROFILE
        struct event_filter             *filter;
@@ -779,7 +780,7 @@ extern struct perf_event *
 perf_event_create_kernel_counter(struct perf_event_attr *attr,
                                int cpu,
                                pid_t pid,
-                               perf_callback_t callback);
+                               perf_overflow_handler_t callback);
 extern u64 perf_event_read_value(struct perf_event *event,
                                 u64 *enabled, u64 *running);
 
index 2d10b012828f0ebbf9918e7724b5db7ecc80766f..b600fc27f1613e386ee4397a84e65cb623ea6b26 100644 (file)
@@ -259,7 +259,7 @@ void release_bp_slot(struct perf_event *bp)
 }
 
 
-int __register_perf_hw_breakpoint(struct perf_event *bp)
+int register_perf_hw_breakpoint(struct perf_event *bp)
 {
        int ret;
 
@@ -276,19 +276,12 @@ int __register_perf_hw_breakpoint(struct perf_event *bp)
         * This is a quick hack that will be removed soon, once we remove
         * the tmp breakpoints from ptrace
         */
-       if (!bp->attr.disabled || bp->callback == perf_bp_event)
+       if (!bp->attr.disabled || !bp->overflow_handler)
                ret = arch_validate_hwbkpt_settings(bp, bp->ctx->task);
 
        return ret;
 }
 
-int register_perf_hw_breakpoint(struct perf_event *bp)
-{
-       bp->callback = perf_bp_event;
-
-       return __register_perf_hw_breakpoint(bp);
-}
-
 /**
  * register_user_hw_breakpoint - register a hardware breakpoint for user space
  * @attr: breakpoint attributes
@@ -297,7 +290,7 @@ int register_perf_hw_breakpoint(struct perf_event *bp)
  */
 struct perf_event *
 register_user_hw_breakpoint(struct perf_event_attr *attr,
-                           perf_callback_t triggered,
+                           perf_overflow_handler_t triggered,
                            struct task_struct *tsk)
 {
        return perf_event_create_kernel_counter(attr, -1, tsk->pid, triggered);
@@ -322,7 +315,7 @@ modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *attr)
        unregister_hw_breakpoint(bp);
 
        return perf_event_create_kernel_counter(attr, -1, bp->ctx->task->pid,
-                                               bp->callback);
+                                               bp->overflow_handler);
 }
 EXPORT_SYMBOL_GPL(modify_user_hw_breakpoint);
 
@@ -347,7 +340,7 @@ EXPORT_SYMBOL_GPL(unregister_hw_breakpoint);
  */
 struct perf_event **
 register_wide_hw_breakpoint(struct perf_event_attr *attr,
-                           perf_callback_t triggered)
+                           perf_overflow_handler_t triggered)
 {
        struct perf_event **cpu_events, **pevent, *bp;
        long err;
index 6b7ddba1dd640cc94f2af66fd163961e944a94f2..fd43ff4ac860077150b0ecfdd0566e1a99c50082 100644 (file)
@@ -4286,15 +4286,8 @@ static void bp_perf_event_destroy(struct perf_event *event)
 static const struct pmu *bp_perf_event_init(struct perf_event *bp)
 {
        int err;
-       /*
-        * The breakpoint is already filled if we haven't created the counter
-        * through perf syscall
-        * FIXME: manage to get trigerred to NULL if it comes from syscalls
-        */
-       if (!bp->callback)
-               err = register_perf_hw_breakpoint(bp);
-       else
-               err = __register_perf_hw_breakpoint(bp);
+
+       err = register_perf_hw_breakpoint(bp);
        if (err)
                return ERR_PTR(err);
 
@@ -4390,7 +4383,7 @@ perf_event_alloc(struct perf_event_attr *attr,
                   struct perf_event_context *ctx,
                   struct perf_event *group_leader,
                   struct perf_event *parent_event,
-                  perf_callback_t callback,
+                  perf_overflow_handler_t overflow_handler,
                   gfp_t gfpflags)
 {
        const struct pmu *pmu;
@@ -4433,10 +4426,10 @@ perf_event_alloc(struct perf_event_attr *attr,
 
        event->state            = PERF_EVENT_STATE_INACTIVE;
 
-       if (!callback && parent_event)
-               callback = parent_event->callback;
+       if (!overflow_handler && parent_event)
+               overflow_handler = parent_event->overflow_handler;
        
-       event->callback = callback;
+       event->overflow_handler = overflow_handler;
 
        if (attr->disabled)
                event->state = PERF_EVENT_STATE_OFF;
@@ -4776,7 +4769,8 @@ err_put_context:
  */
 struct perf_event *
 perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
-                                pid_t pid, perf_callback_t callback)
+                                pid_t pid,
+                                perf_overflow_handler_t overflow_handler)
 {
        struct perf_event *event;
        struct perf_event_context *ctx;
@@ -4793,7 +4787,7 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
        }
 
        event = perf_event_alloc(attr, cpu, ctx, NULL,
-                                    NULL, callback, GFP_KERNEL);
+                                NULL, overflow_handler, GFP_KERNEL);
        if (IS_ERR(event)) {
                err = PTR_ERR(event);
                goto err_put_context;
index ddfa0fd43bc0ebab1fa4953918a41323f409e08a..acb87d4a4ac15e8127fc64138918111a1e948936 100644 (file)
@@ -79,11 +79,12 @@ void ksym_collect_stats(unsigned long hbp_hit_addr)
 }
 #endif /* CONFIG_PROFILE_KSYM_TRACER */
 
-void ksym_hbp_handler(struct perf_event *hbp, void *data)
+void ksym_hbp_handler(struct perf_event *hbp, int nmi,
+                     struct perf_sample_data *data,
+                     struct pt_regs *regs)
 {
        struct ring_buffer_event *event;
        struct ksym_trace_entry *entry;
-       struct pt_regs *regs = data;
        struct ring_buffer *buffer;
        int pc;
 
index 29525500df00e51c5630e7da0e79a8db2b8c99f0..c69cbe9b2426a287bf754a82344957f9a2ead426 100644 (file)
@@ -41,7 +41,9 @@ module_param_string(ksym, ksym_name, KSYM_NAME_LEN, S_IRUGO);
 MODULE_PARM_DESC(ksym, "Kernel symbol to monitor; this module will report any"
                        " write operations on the kernel symbol");
 
-static void sample_hbp_handler(struct perf_event *temp, void *data)
+static void sample_hbp_handler(struct perf_event *bp, int nmi,
+                              struct perf_sample_data *data,
+                              struct pt_regs *regs)
 {
        printk(KERN_INFO "%s value is changed\n", ksym_name);
        dump_stack();
@@ -51,8 +53,9 @@ static void sample_hbp_handler(struct perf_event *temp, void *data)
 static int __init hw_break_module_init(void)
 {
        int ret;
-       DEFINE_BREAKPOINT_ATTR(attr);
+       struct perf_event_attr attr;
 
+       hw_breakpoint_init(&attr);
        attr.bp_addr = kallsyms_lookup_name(ksym_name);
        attr.bp_len = HW_BREAKPOINT_LEN_4;
        attr.bp_type = HW_BREAKPOINT_W | HW_BREAKPOINT_R;