/*
* perf counter paranoia level:
- * 0 - not paranoid
- * 1 - disallow cpu counters to unpriv
- * 2 - disallow kernel profiling to unpriv
+ * -1 - not paranoid at all
+ * 0 - disallow raw tracepoint access for unpriv
+ * 1 - disallow cpu counters for unpriv
+ * 2 - disallow kernel profiling for unpriv
*/
int sysctl_perf_counter_paranoid __read_mostly = 1;
+static inline bool perf_paranoid_tracepoint_raw(void)
+{
+ return sysctl_perf_counter_paranoid > -1;
+}
+
static inline bool perf_paranoid_cpu(void)
{
return sysctl_perf_counter_paranoid > 0;
void __weak perf_counter_print_debug(void) { }
-static DEFINE_PER_CPU(int, disable_count);
+static DEFINE_PER_CPU(int, perf_disable_count);
void __perf_disable(void)
{
- __get_cpu_var(disable_count)++;
+ __get_cpu_var(perf_disable_count)++;
}
bool __perf_enable(void)
{
- return !--__get_cpu_var(disable_count);
+ return !--__get_cpu_var(perf_disable_count);
}
void perf_disable(void)
struct perf_counter_context *ctx = counter->ctx;
u64 run_end;
- if (counter->state < PERF_COUNTER_STATE_INACTIVE)
+ if (counter->state < PERF_COUNTER_STATE_INACTIVE ||
+ counter->group_leader->state < PERF_COUNTER_STATE_INACTIVE)
return;
counter->total_time_enabled = ctx->time - counter->tstamp_enabled;
*/
if (counter->state >= PERF_COUNTER_STATE_INACTIVE) {
update_context_time(ctx);
- update_counter_times(counter);
+ update_group_times(counter);
if (counter == counter->group_leader)
group_sched_out(counter, cpuctx, ctx);
else
* in, so we can change the state safely.
*/
if (counter->state == PERF_COUNTER_STATE_INACTIVE) {
- update_counter_times(counter);
+ update_group_times(counter);
counter->state = PERF_COUNTER_STATE_OFF;
}
spin_unlock_irq(&ctx->lock);
}
+/*
+ * Put a counter into inactive state and update time fields.
+ * Enabling the leader of a group effectively enables all
+ * the group members that aren't explicitly disabled, so we
+ * have to update their ->tstamp_enabled also.
+ * Note: this works for group members as well as group leaders
+ * since the non-leader members' sibling_lists will be empty.
+ */
+static void __perf_counter_mark_enabled(struct perf_counter *counter,
+ struct perf_counter_context *ctx)
+{
+ struct perf_counter *sub;
+
+ counter->state = PERF_COUNTER_STATE_INACTIVE;
+ counter->tstamp_enabled = ctx->time - counter->total_time_enabled;
+ list_for_each_entry(sub, &counter->sibling_list, list_entry)
+ if (sub->state >= PERF_COUNTER_STATE_INACTIVE)
+ sub->tstamp_enabled =
+ ctx->time - sub->total_time_enabled;
+}
+
/*
* Cross CPU call to enable a performance counter
*/
if (counter->state >= PERF_COUNTER_STATE_INACTIVE)
goto unlock;
- counter->state = PERF_COUNTER_STATE_INACTIVE;
- counter->tstamp_enabled = ctx->time - counter->total_time_enabled;
+ __perf_counter_mark_enabled(counter, ctx);
/*
* If the counter is in a group and isn't the group leader,
* Since we have the lock this context can't be scheduled
* in, so we can change the state safely.
*/
- if (counter->state == PERF_COUNTER_STATE_OFF) {
- counter->state = PERF_COUNTER_STATE_INACTIVE;
- counter->tstamp_enabled =
- ctx->time - counter->total_time_enabled;
- }
+ if (counter->state == PERF_COUNTER_STATE_OFF)
+ __perf_counter_mark_enabled(counter, ctx);
+
out:
spin_unlock_irq(&ctx->lock);
}
counter->attr.enable_on_exec = 0;
if (counter->state >= PERF_COUNTER_STATE_INACTIVE)
continue;
- counter->state = PERF_COUNTER_STATE_INACTIVE;
- counter->tstamp_enabled =
- ctx->time - counter->total_time_enabled;
+ __perf_counter_mark_enabled(counter, ctx);
enabled = 1;
}
atomic_dec(&nr_task_counters);
}
+ if (counter->output) {
+ fput(counter->output->filp);
+ counter->output = NULL;
+ }
+
if (counter->destroy)
counter->destroy(counter);
return ret;
}
+int perf_counter_set_output(struct perf_counter *counter, int output_fd);
+
static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
struct perf_counter *counter = file->private_data;
case PERF_COUNTER_IOC_PERIOD:
return perf_counter_period(counter, (u64 __user *)arg);
+ case PERF_COUNTER_IOC_SET_OUTPUT:
+ return perf_counter_set_output(counter, arg);
+
default:
return -ENOTTY;
}
data->nr_pages = nr_pages;
atomic_set(&data->lock, -1);
+ if (counter->attr.watermark) {
+ data->watermark = min_t(long, PAGE_SIZE * nr_pages,
+ counter->attr.wakeup_watermark);
+ }
+ if (!data->watermark)
+ data->watermark = max(PAGE_SIZE, PAGE_SIZE * nr_pages / 4);
+
rcu_assign_pointer(counter->data, data);
return 0;
WARN_ON_ONCE(counter->ctx->parent_ctx);
mutex_lock(&counter->mmap_mutex);
+ if (counter->output) {
+ ret = -EINVAL;
+ goto unlock;
+ }
+
if (atomic_inc_not_zero(&counter->mmap_count)) {
if (nr_pages != counter->data->nr_pages)
ret = -EINVAL;
lock_limit >>= PAGE_SHIFT;
locked = vma->vm_mm->locked_vm + extra;
- if ((locked > lock_limit) && !capable(CAP_IPC_LOCK)) {
+ if ((locked > lock_limit) && perf_paranoid_tracepoint_raw() &&
+ !capable(CAP_IPC_LOCK)) {
ret = -EPERM;
goto unlock;
}
/*
* Output
*/
-
-struct perf_output_handle {
- struct perf_counter *counter;
- struct perf_mmap_data *data;
- unsigned long head;
- unsigned long offset;
- int nmi;
- int sample;
- int locked;
- unsigned long flags;
-};
-
-static bool perf_output_space(struct perf_mmap_data *data,
- unsigned int offset, unsigned int head)
+static bool perf_output_space(struct perf_mmap_data *data, unsigned long tail,
+ unsigned long offset, unsigned long head)
{
- unsigned long tail;
unsigned long mask;
if (!data->writable)
return true;
mask = (data->nr_pages << PAGE_SHIFT) - 1;
- /*
- * Userspace could choose to issue a mb() before updating the tail
- * pointer. So that all reads will be completed before the write is
- * issued.
- */
- tail = ACCESS_ONCE(data->user_page->data_tail);
- smp_rmb();
offset = (offset - tail) & mask;
head = (head - tail) & mask;
local_irq_restore(handle->flags);
}
-static void perf_output_copy(struct perf_output_handle *handle,
- const void *buf, unsigned int len)
+void perf_output_copy(struct perf_output_handle *handle,
+ const void *buf, unsigned int len)
{
unsigned int pages_mask;
unsigned int offset;
WARN_ON_ONCE(((long)(handle->head - handle->offset)) < 0);
}
-#define perf_output_put(handle, x) \
- perf_output_copy((handle), &(x), sizeof(x))
-
-static int perf_output_begin(struct perf_output_handle *handle,
- struct perf_counter *counter, unsigned int size,
- int nmi, int sample)
+int perf_output_begin(struct perf_output_handle *handle,
+ struct perf_counter *counter, unsigned int size,
+ int nmi, int sample)
{
+ struct perf_counter *output_counter;
struct perf_mmap_data *data;
- unsigned int offset, head;
+ unsigned long tail, offset, head;
int have_lost;
struct {
struct perf_event_header header;
u64 lost;
} lost_event;
+ rcu_read_lock();
/*
* For inherited counters we send all the output towards the parent.
*/
if (counter->parent)
counter = counter->parent;
- rcu_read_lock();
+ output_counter = rcu_dereference(counter->output);
+ if (output_counter)
+ counter = output_counter;
+
data = rcu_dereference(counter->data);
if (!data)
goto out;
perf_output_lock(handle);
do {
+ /*
+ * Userspace could choose to issue a mb() before updating the
+ * tail pointer. So that all reads will be completed before the
+ * write is issued.
+ */
+ tail = ACCESS_ONCE(data->user_page->data_tail);
+ smp_rmb();
offset = head = atomic_long_read(&data->head);
head += size;
- if (unlikely(!perf_output_space(data, offset, head)))
+ if (unlikely(!perf_output_space(data, tail, offset, head)))
goto fail;
} while (atomic_long_cmpxchg(&data->head, offset, head) != offset);
handle->offset = offset;
handle->head = head;
- if ((offset >> PAGE_SHIFT) != (head >> PAGE_SHIFT))
+ if (head - tail > data->watermark)
atomic_set(&data->wakeup, 1);
if (have_lost) {
return -ENOSPC;
}
-static void perf_output_end(struct perf_output_handle *handle)
+void perf_output_end(struct perf_output_handle *handle)
{
struct perf_counter *counter = handle->counter;
struct perf_mmap_data *data = handle->data;
perf_output_read_one(handle, counter);
}
-void perf_counter_output(struct perf_counter *counter, int nmi,
- struct perf_sample_data *data)
+void perf_output_sample(struct perf_output_handle *handle,
+ struct perf_event_header *header,
+ struct perf_sample_data *data,
+ struct perf_counter *counter)
{
- int ret;
- u64 sample_type = counter->attr.sample_type;
- struct perf_output_handle handle;
- struct perf_event_header header;
- u64 ip;
- struct {
- u32 pid, tid;
- } tid_entry;
- struct perf_callchain_entry *callchain = NULL;
- int callchain_size = 0;
- u64 time;
- struct {
- u32 cpu, reserved;
- } cpu_entry;
-
- header.type = PERF_EVENT_SAMPLE;
- header.size = sizeof(header);
-
- header.misc = 0;
- header.misc |= perf_misc_flags(data->regs);
-
- if (sample_type & PERF_SAMPLE_IP) {
- ip = perf_instruction_pointer(data->regs);
- header.size += sizeof(ip);
- }
+ u64 sample_type = data->type;
- if (sample_type & PERF_SAMPLE_TID) {
- /* namespace issues */
- tid_entry.pid = perf_counter_pid(counter, current);
- tid_entry.tid = perf_counter_tid(counter, current);
+ perf_output_put(handle, *header);
- header.size += sizeof(tid_entry);
- }
+ if (sample_type & PERF_SAMPLE_IP)
+ perf_output_put(handle, data->ip);
- if (sample_type & PERF_SAMPLE_TIME) {
- /*
- * Maybe do better on x86 and provide cpu_clock_nmi()
- */
- time = sched_clock();
+ if (sample_type & PERF_SAMPLE_TID)
+ perf_output_put(handle, data->tid_entry);
- header.size += sizeof(u64);
- }
+ if (sample_type & PERF_SAMPLE_TIME)
+ perf_output_put(handle, data->time);
if (sample_type & PERF_SAMPLE_ADDR)
- header.size += sizeof(u64);
+ perf_output_put(handle, data->addr);
if (sample_type & PERF_SAMPLE_ID)
- header.size += sizeof(u64);
+ perf_output_put(handle, data->id);
if (sample_type & PERF_SAMPLE_STREAM_ID)
- header.size += sizeof(u64);
+ perf_output_put(handle, data->stream_id);
- if (sample_type & PERF_SAMPLE_CPU) {
- header.size += sizeof(cpu_entry);
-
- cpu_entry.cpu = raw_smp_processor_id();
- cpu_entry.reserved = 0;
- }
+ if (sample_type & PERF_SAMPLE_CPU)
+ perf_output_put(handle, data->cpu_entry);
if (sample_type & PERF_SAMPLE_PERIOD)
- header.size += sizeof(u64);
+ perf_output_put(handle, data->period);
if (sample_type & PERF_SAMPLE_READ)
- header.size += perf_counter_read_size(counter);
+ perf_output_read(handle, counter);
if (sample_type & PERF_SAMPLE_CALLCHAIN) {
- callchain = perf_callchain(data->regs);
+ if (data->callchain) {
+ int size = 1;
- if (callchain) {
- callchain_size = (1 + callchain->nr) * sizeof(u64);
- header.size += callchain_size;
- } else
- header.size += sizeof(u64);
+ if (data->callchain)
+ size += data->callchain->nr;
+
+ size *= sizeof(u64);
+
+ perf_output_copy(handle, data->callchain, size);
+ } else {
+ u64 nr = 0;
+ perf_output_put(handle, nr);
+ }
}
if (sample_type & PERF_SAMPLE_RAW) {
- int size = sizeof(u32);
+ if (data->raw) {
+ perf_output_put(handle, data->raw->size);
+ perf_output_copy(handle, data->raw->data,
+ data->raw->size);
+ } else {
+ struct {
+ u32 size;
+ u32 data;
+ } raw = {
+ .size = sizeof(u32),
+ .data = 0,
+ };
+ perf_output_put(handle, raw);
+ }
+ }
+}
- if (data->raw)
- size += data->raw->size;
- else
- size += sizeof(u32);
+void perf_prepare_sample(struct perf_event_header *header,
+ struct perf_sample_data *data,
+ struct perf_counter *counter,
+ struct pt_regs *regs)
+{
+ u64 sample_type = counter->attr.sample_type;
- WARN_ON_ONCE(size & (sizeof(u64)-1));
- header.size += size;
- }
+ data->type = sample_type;
- ret = perf_output_begin(&handle, counter, header.size, nmi, 1);
- if (ret)
- return;
+ header->type = PERF_EVENT_SAMPLE;
+ header->size = sizeof(*header);
- perf_output_put(&handle, header);
+ header->misc = 0;
+ header->misc |= perf_misc_flags(regs);
- if (sample_type & PERF_SAMPLE_IP)
- perf_output_put(&handle, ip);
+ if (sample_type & PERF_SAMPLE_IP) {
+ data->ip = perf_instruction_pointer(regs);
- if (sample_type & PERF_SAMPLE_TID)
- perf_output_put(&handle, tid_entry);
+ header->size += sizeof(data->ip);
+ }
- if (sample_type & PERF_SAMPLE_TIME)
- perf_output_put(&handle, time);
+ if (sample_type & PERF_SAMPLE_TID) {
+ /* namespace issues */
+ data->tid_entry.pid = perf_counter_pid(counter, current);
+ data->tid_entry.tid = perf_counter_tid(counter, current);
+
+ header->size += sizeof(data->tid_entry);
+ }
+
+ if (sample_type & PERF_SAMPLE_TIME) {
+ data->time = perf_clock();
+
+ header->size += sizeof(data->time);
+ }
if (sample_type & PERF_SAMPLE_ADDR)
- perf_output_put(&handle, data->addr);
+ header->size += sizeof(data->addr);
if (sample_type & PERF_SAMPLE_ID) {
- u64 id = primary_counter_id(counter);
+ data->id = primary_counter_id(counter);
- perf_output_put(&handle, id);
+ header->size += sizeof(data->id);
}
- if (sample_type & PERF_SAMPLE_STREAM_ID)
- perf_output_put(&handle, counter->id);
+ if (sample_type & PERF_SAMPLE_STREAM_ID) {
+ data->stream_id = counter->id;
- if (sample_type & PERF_SAMPLE_CPU)
- perf_output_put(&handle, cpu_entry);
+ header->size += sizeof(data->stream_id);
+ }
+
+ if (sample_type & PERF_SAMPLE_CPU) {
+ data->cpu_entry.cpu = raw_smp_processor_id();
+ data->cpu_entry.reserved = 0;
+
+ header->size += sizeof(data->cpu_entry);
+ }
if (sample_type & PERF_SAMPLE_PERIOD)
- perf_output_put(&handle, data->period);
+ header->size += sizeof(data->period);
if (sample_type & PERF_SAMPLE_READ)
- perf_output_read(&handle, counter);
+ header->size += perf_counter_read_size(counter);
if (sample_type & PERF_SAMPLE_CALLCHAIN) {
- if (callchain)
- perf_output_copy(&handle, callchain, callchain_size);
- else {
- u64 nr = 0;
- perf_output_put(&handle, nr);
- }
+ int size = 1;
+
+ data->callchain = perf_callchain(regs);
+
+ if (data->callchain)
+ size += data->callchain->nr;
+
+ header->size += size * sizeof(u64);
}
if (sample_type & PERF_SAMPLE_RAW) {
- if (data->raw) {
- perf_output_put(&handle, data->raw->size);
- perf_output_copy(&handle, data->raw->data, data->raw->size);
- } else {
- struct {
- u32 size;
- u32 data;
- } raw = {
- .size = sizeof(u32),
- .data = 0,
- };
- perf_output_put(&handle, raw);
- }
+ int size = sizeof(u32);
+
+ if (data->raw)
+ size += data->raw->size;
+ else
+ size += sizeof(u32);
+
+ WARN_ON_ONCE(size & (sizeof(u64)-1));
+ header->size += size;
}
+}
+
+static void perf_counter_output(struct perf_counter *counter, int nmi,
+ struct perf_sample_data *data,
+ struct pt_regs *regs)
+{
+ struct perf_output_handle handle;
+ struct perf_event_header header;
+
+ perf_prepare_sample(&header, data, counter, regs);
+
+ if (perf_output_begin(&handle, counter, header.size, nmi, 1))
+ return;
+
+ perf_output_sample(&handle, &header, data, counter);
perf_output_end(&handle);
}
u32 ppid;
u32 tid;
u32 ptid;
+ u64 time;
} event;
};
struct perf_task_event *task_event)
{
struct perf_output_handle handle;
- int size = task_event->event.header.size;
+ int size;
struct task_struct *task = task_event->task;
- int ret = perf_output_begin(&handle, counter, size, 0, 0);
+ int ret;
+
+ size = task_event->event.header.size;
+ ret = perf_output_begin(&handle, counter, size, 0, 0);
if (ret)
return;
task_event->event.tid = perf_counter_tid(counter, task);
task_event->event.ptid = perf_counter_tid(counter, current);
+ task_event->event.time = perf_clock();
+
perf_output_put(&handle, task_event->event);
+
perf_output_end(&handle);
}
.misc = 0,
.size = sizeof(throttle_event),
},
- .time = sched_clock(),
+ .time = perf_clock(),
.id = primary_counter_id(counter),
.stream_id = counter->id,
};
* Generic counter overflow handling, sampling.
*/
-int perf_counter_overflow(struct perf_counter *counter, int nmi,
- struct perf_sample_data *data)
+static int __perf_counter_overflow(struct perf_counter *counter, int nmi,
+ int throttle, struct perf_sample_data *data,
+ struct pt_regs *regs)
{
int events = atomic_read(&counter->event_limit);
- int throttle = counter->pmu->unthrottle != NULL;
struct hw_perf_counter *hwc = &counter->hw;
int ret = 0;
+ throttle = (throttle && counter->pmu->unthrottle != NULL);
+
if (!throttle) {
hwc->interrupts++;
} else {
}
if (counter->attr.freq) {
- u64 now = sched_clock();
+ u64 now = perf_clock();
s64 delta = now - hwc->freq_stamp;
hwc->freq_stamp = now;
perf_counter_disable(counter);
}
- perf_counter_output(counter, nmi, data);
+ perf_counter_output(counter, nmi, data, regs);
return ret;
}
+int perf_counter_overflow(struct perf_counter *counter, int nmi,
+ struct perf_sample_data *data,
+ struct pt_regs *regs)
+{
+ return __perf_counter_overflow(counter, nmi, 1, data, regs);
+}
+
/*
* Generic software counter infrastructure
*/
}
static void perf_swcounter_overflow(struct perf_counter *counter,
- int nmi, struct perf_sample_data *data)
+ int nmi, struct perf_sample_data *data,
+ struct pt_regs *regs)
{
struct hw_perf_counter *hwc = &counter->hw;
+ int throttle = 0;
u64 overflow;
data->period = counter->hw.last_period;
return;
for (; overflow; overflow--) {
- if (perf_counter_overflow(counter, nmi, data)) {
+ if (__perf_counter_overflow(counter, nmi, throttle,
+ data, regs)) {
/*
* We inhibit the overflow from happening when
* hwc->interrupts == MAX_INTERRUPTS.
*/
break;
}
+ throttle = 1;
}
}
}
static void perf_swcounter_add(struct perf_counter *counter, u64 nr,
- int nmi, struct perf_sample_data *data)
+ int nmi, struct perf_sample_data *data,
+ struct pt_regs *regs)
{
struct hw_perf_counter *hwc = &counter->hw;
if (!hwc->sample_period)
return;
- if (!data->regs)
+ if (!regs)
return;
if (!atomic64_add_negative(nr, &hwc->period_left))
- perf_swcounter_overflow(counter, nmi, data);
+ perf_swcounter_overflow(counter, nmi, data, regs);
}
static int perf_swcounter_is_counting(struct perf_counter *counter)
static void perf_swcounter_ctx_event(struct perf_counter_context *ctx,
enum perf_type_id type,
u32 event, u64 nr, int nmi,
- struct perf_sample_data *data)
+ struct perf_sample_data *data,
+ struct pt_regs *regs)
{
struct perf_counter *counter;
rcu_read_lock();
list_for_each_entry_rcu(counter, &ctx->event_list, event_entry) {
- if (perf_swcounter_match(counter, type, event, data->regs))
- perf_swcounter_add(counter, nr, nmi, data);
+ if (perf_swcounter_match(counter, type, event, regs))
+ perf_swcounter_add(counter, nr, nmi, data, regs);
}
rcu_read_unlock();
}
static void do_perf_swcounter_event(enum perf_type_id type, u32 event,
u64 nr, int nmi,
- struct perf_sample_data *data)
+ struct perf_sample_data *data,
+ struct pt_regs *regs)
{
struct perf_cpu_context *cpuctx = &get_cpu_var(perf_cpu_context);
int *recursion = perf_swcounter_recursion_context(cpuctx);
barrier();
perf_swcounter_ctx_event(&cpuctx->ctx, type, event,
- nr, nmi, data);
+ nr, nmi, data, regs);
rcu_read_lock();
/*
* doesn't really matter which of the child contexts the
*/
ctx = rcu_dereference(current->perf_counter_ctxp);
if (ctx)
- perf_swcounter_ctx_event(ctx, type, event, nr, nmi, data);
+ perf_swcounter_ctx_event(ctx, type, event, nr, nmi, data, regs);
rcu_read_unlock();
barrier();
struct pt_regs *regs, u64 addr)
{
struct perf_sample_data data = {
- .regs = regs,
.addr = addr,
};
- do_perf_swcounter_event(PERF_TYPE_SOFTWARE, event, nr, nmi, &data);
+ do_perf_swcounter_event(PERF_TYPE_SOFTWARE, event, nr, nmi,
+ &data, regs);
}
static void perf_swcounter_read(struct perf_counter *counter)
{
enum hrtimer_restart ret = HRTIMER_RESTART;
struct perf_sample_data data;
+ struct pt_regs *regs;
struct perf_counter *counter;
u64 period;
counter->pmu->read(counter);
data.addr = 0;
- data.regs = get_irq_regs();
+ regs = get_irq_regs();
/*
* In case we exclude kernel IPs or are somehow not in interrupt
* context, provide the next best thing, the user IP.
*/
- if ((counter->attr.exclude_kernel || !data.regs) &&
+ if ((counter->attr.exclude_kernel || !regs) &&
!counter->attr.exclude_user)
- data.regs = task_pt_regs(current);
+ regs = task_pt_regs(current);
- if (data.regs) {
- if (perf_counter_overflow(counter, 0, &data))
+ if (regs) {
+ if (perf_counter_overflow(counter, 0, &data, regs))
ret = HRTIMER_NORESTART;
}
};
struct perf_sample_data data = {
- .regs = get_irq_regs(),
.addr = addr,
.raw = &raw,
};
- if (!data.regs)
- data.regs = task_pt_regs(current);
+ struct pt_regs *regs = get_irq_regs();
+
+ if (!regs)
+ regs = task_pt_regs(current);
- do_perf_swcounter_event(PERF_TYPE_TRACEPOINT, event_id, count, 1, &data);
+ do_perf_swcounter_event(PERF_TYPE_TRACEPOINT, event_id, count, 1,
+ &data, regs);
}
EXPORT_SYMBOL_GPL(perf_tpcounter_event);
* have these.
*/
if ((counter->attr.sample_type & PERF_SAMPLE_RAW) &&
+ perf_paranoid_tracepoint_raw() &&
!capable(CAP_SYS_ADMIN))
return ERR_PTR(-EPERM);
static int perf_copy_attr(struct perf_counter_attr __user *uattr,
struct perf_counter_attr *attr)
{
- int ret;
u32 size;
+ int ret;
if (!access_ok(VERIFY_WRITE, uattr, PERF_ATTR_SIZE_VER0))
return -EFAULT;
/*
* If we're handed a bigger struct than we know of,
- * ensure all the unknown bits are 0.
+ * ensure all the unknown bits are 0 - i.e. new
+ * user-space does not rely on any kernel feature
+ * extensions we dont know about yet.
*/
if (size > sizeof(*attr)) {
- unsigned long val;
- unsigned long __user *addr;
- unsigned long __user *end;
+ unsigned char __user *addr;
+ unsigned char __user *end;
+ unsigned char val;
- addr = PTR_ALIGN((void __user *)uattr + sizeof(*attr),
- sizeof(unsigned long));
- end = PTR_ALIGN((void __user *)uattr + size,
- sizeof(unsigned long));
+ addr = (void __user *)uattr + sizeof(*attr);
+ end = (void __user *)uattr + size;
- for (; addr < end; addr += sizeof(unsigned long)) {
+ for (; addr < end; addr++) {
ret = get_user(val, addr);
if (ret)
return ret;
if (val)
goto err_size;
}
+ size = sizeof(*attr);
}
ret = copy_from_user(attr, uattr, size);
goto out;
}
+int perf_counter_set_output(struct perf_counter *counter, int output_fd)
+{
+ struct perf_counter *output_counter = NULL;
+ struct file *output_file = NULL;
+ struct perf_counter *old_output;
+ int fput_needed = 0;
+ int ret = -EINVAL;
+
+ if (!output_fd)
+ goto set;
+
+ output_file = fget_light(output_fd, &fput_needed);
+ if (!output_file)
+ return -EBADF;
+
+ if (output_file->f_op != &perf_fops)
+ goto out;
+
+ output_counter = output_file->private_data;
+
+ /* Don't chain output fds */
+ if (output_counter->output)
+ goto out;
+
+ /* Don't set an output fd when we already have an output channel */
+ if (counter->data)
+ goto out;
+
+ atomic_long_inc(&output_file->f_count);
+
+set:
+ mutex_lock(&counter->mmap_mutex);
+ old_output = counter->output;
+ rcu_assign_pointer(counter->output, output_counter);
+ mutex_unlock(&counter->mmap_mutex);
+
+ if (old_output) {
+ /*
+ * we need to make sure no existing perf_output_*()
+ * is still referencing this counter.
+ */
+ synchronize_rcu();
+ fput(old_output->filp);
+ }
+
+ ret = 0;
+out:
+ fput_light(output_file, fput_needed);
+ return ret;
+}
+
/**
* sys_perf_counter_open - open a performance counter, associate it to a task/cpu
*
struct file *group_file = NULL;
int fput_needed = 0;
int fput_needed2 = 0;
- int ret;
+ int err;
/* for future expandability... */
- if (flags)
+ if (flags & ~(PERF_FLAG_FD_NO_GROUP | PERF_FLAG_FD_OUTPUT))
return -EINVAL;
- ret = perf_copy_attr(attr_uptr, &attr);
- if (ret)
- return ret;
+ err = perf_copy_attr(attr_uptr, &attr);
+ if (err)
+ return err;
if (!attr.exclude_kernel) {
if (perf_paranoid_kernel() && !capable(CAP_SYS_ADMIN))
* Look up the group leader (we will attach this counter to it):
*/
group_leader = NULL;
- if (group_fd != -1) {
- ret = -EINVAL;
+ if (group_fd != -1 && !(flags & PERF_FLAG_FD_NO_GROUP)) {
+ err = -EINVAL;
group_file = fget_light(group_fd, &fput_needed);
if (!group_file)
goto err_put_context;
counter = perf_counter_alloc(&attr, cpu, ctx, group_leader,
NULL, GFP_KERNEL);
- ret = PTR_ERR(counter);
+ err = PTR_ERR(counter);
if (IS_ERR(counter))
goto err_put_context;
- ret = anon_inode_getfd("[perf_counter]", &perf_fops, counter, 0);
- if (ret < 0)
+ err = anon_inode_getfd("[perf_counter]", &perf_fops, counter, 0);
+ if (err < 0)
goto err_free_put_context;
- counter_file = fget_light(ret, &fput_needed2);
+ counter_file = fget_light(err, &fput_needed2);
if (!counter_file)
goto err_free_put_context;
+ if (flags & PERF_FLAG_FD_OUTPUT) {
+ err = perf_counter_set_output(counter, group_fd);
+ if (err)
+ goto err_fput_free_put_context;
+ }
+
counter->filp = counter_file;
WARN_ON_ONCE(ctx->parent_ctx);
mutex_lock(&ctx->mutex);
list_add_tail(&counter->owner_entry, ¤t->perf_counter_list);
mutex_unlock(¤t->perf_counter_mutex);
+err_fput_free_put_context:
fput_light(counter_file, fput_needed2);
-out_fput:
- fput_light(group_file, fput_needed);
-
- return ret;
-
err_free_put_context:
- kfree(counter);
+ if (err < 0)
+ kfree(counter);
err_put_context:
- put_ctx(ctx);
+ if (err < 0)
+ put_ctx(ctx);
+
+ fput_light(group_file, fput_needed);
- goto out_fput;
+ return err;
}
/*