[GIT PULL] perf fixes

From: Ingo Molnar
Date: Thu Sep 13 2012 - 10:39:31 EST

Next message: Arnaldo Carvalho de Melo: "Re: [tip:perf/core] perf hists: Introduce perf_hpp for hist periodprinting"
Previous message: Arnd Bergmann: "Re: Implement devicetree support for AB8500 Btemp"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Linus,

Please pull the latest perf-urgent-for-linus git tree from:

git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git perf-urgent-for-linus

HEAD: e4390fa632d7c592e68e8106b7daea923ac995f5 Merge branch 'urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/rric/oprofile into perf/urgent

This tree includes various fixes.

Thanks,

Ingo

------------------>
Al Viro (1):
perf_event: Switch to internal refcount, fix race with close()

K.Prasad (1):
perf/hwpb: Invoke __perf_event_disable() if interrupts are already disabled

Robert Richter (1):
oprofile, s390: Fix uninitialized memory access when writing to oprofilefs

Stephane Eranian (2):
perf/x86: Fix microcode revision check for SNB-PEBS
perf/x86: Enable Intel Cedarview Atom suppport

arch/s390/oprofile/init.c | 10 ++---
arch/x86/kernel/cpu/perf_event_intel.c | 1 +
arch/x86/kernel/cpu/perf_event_intel_lbr.c | 3 +-
arch/x86/kernel/microcode_core.c | 3 ++
include/linux/perf_event.h | 4 +-
kernel/events/core.c | 64 ++++++++++++++++--------------
kernel/events/hw_breakpoint.c | 11 ++++-
7 files changed, 58 insertions(+), 38 deletions(-)

diff --git a/arch/s390/oprofile/init.c b/arch/s390/oprofile/init.c
index a1e9d69..584b936 100644
--- a/arch/s390/oprofile/init.c
+++ b/arch/s390/oprofile/init.c
@@ -169,7 +169,7 @@ static ssize_t hw_interval_write(struct file *file, char const __user *buf,
if (*offset)
return -EINVAL;
retval = oprofilefs_ulong_from_user(&val, buf, count);
- if (retval)
+ if (retval <= 0)
return retval;
if (val < oprofile_min_interval)
oprofile_hw_interval = oprofile_min_interval;
@@ -212,7 +212,7 @@ static ssize_t hwsampler_zero_write(struct file *file, char const __user *buf,
return -EINVAL;

retval = oprofilefs_ulong_from_user(&val, buf, count);
- if (retval)
+ if (retval <= 0)
return retval;
if (val != 0)
return -EINVAL;
@@ -243,7 +243,7 @@ static ssize_t hwsampler_kernel_write(struct file *file, char const __user *buf,
return -EINVAL;

retval = oprofilefs_ulong_from_user(&val, buf, count);
- if (retval)
+ if (retval <= 0)
return retval;

if (val != 0 && val != 1)
@@ -278,7 +278,7 @@ static ssize_t hwsampler_user_write(struct file *file, char const __user *buf,
return -EINVAL;

retval = oprofilefs_ulong_from_user(&val, buf, count);
- if (retval)
+ if (retval <= 0)
return retval;

if (val != 0 && val != 1)
@@ -317,7 +317,7 @@ static ssize_t timer_enabled_write(struct file *file, char const __user *buf,
return -EINVAL;

retval = oprofilefs_ulong_from_user(&val, buf, count);
- if (retval)
+ if (retval <= 0)
return retval;

if (val != 0 && val != 1)
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 7f2739e..0d3d63a 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -2008,6 +2008,7 @@ __init int intel_pmu_init(void)
break;

case 28: /* Atom */
+ case 54: /* Cedariew */
memcpy(hw_cache_event_ids, atom_hw_cache_event_ids,
sizeof(hw_cache_event_ids));

diff --git a/arch/x86/kernel/cpu/perf_event_intel_lbr.c b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
index 520b426..da02e9c 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_lbr.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_lbr.c
@@ -686,7 +686,8 @@ void intel_pmu_lbr_init_atom(void)
* to have an operational LBR which can freeze
* on PMU interrupt
*/
- if (boot_cpu_data.x86_mask < 10) {
+ if (boot_cpu_data.x86_model == 28
+ && boot_cpu_data.x86_mask < 10) {
pr_cont("LBR disabled due to erratum");
return;
}
diff --git a/arch/x86/kernel/microcode_core.c b/arch/x86/kernel/microcode_core.c
index 4873e62..9e5bcf1 100644
--- a/arch/x86/kernel/microcode_core.c
+++ b/arch/x86/kernel/microcode_core.c
@@ -225,6 +225,9 @@ static ssize_t microcode_write(struct file *file, const char __user *buf,
if (do_microcode_update(buf, len) == 0)
ret = (ssize_t)len;

+ if (ret > 0)
+ perf_check_microcode();
+
mutex_unlock(&microcode_mutex);
put_online_cpus();

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 7602ccb..33ed9d6 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -926,7 +926,7 @@ struct perf_event {
struct hw_perf_event hw;

struct perf_event_context *ctx;
- struct file *filp;
+ atomic_long_t refcount;

/*
* These accumulate total time (in nanoseconds) that children
@@ -1296,6 +1296,7 @@ extern int perf_swevent_get_recursion_context(void);
extern void perf_swevent_put_recursion_context(int rctx);
extern void perf_event_enable(struct perf_event *event);
extern void perf_event_disable(struct perf_event *event);
+extern int __perf_event_disable(void *info);
extern void perf_event_task_tick(void);
#else
static inline void
@@ -1334,6 +1335,7 @@ static inline int perf_swevent_get_recursion_context(void) { return -1; }
static inline void perf_swevent_put_recursion_context(int rctx) { }
static inline void perf_event_enable(struct perf_event *event) { }
static inline void perf_event_disable(struct perf_event *event) { }
+static inline int __perf_event_disable(void *info) { return -1; }
static inline void perf_event_task_tick(void) { }
#endif

diff --git a/kernel/events/core.c b/kernel/events/core.c
index b7935fc..7fee567 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -1253,7 +1253,7 @@ retry:
/*
* Cross CPU call to disable a performance event
*/
-static int __perf_event_disable(void *info)
+int __perf_event_disable(void *info)
{
struct perf_event *event = info;
struct perf_event_context *ctx = event->ctx;
@@ -2935,12 +2935,12 @@ EXPORT_SYMBOL_GPL(perf_event_release_kernel);
/*
* Called when the last reference to the file is gone.
*/
-static int perf_release(struct inode *inode, struct file *file)
+static void put_event(struct perf_event *event)
{
- struct perf_event *event = file->private_data;
struct task_struct *owner;

- file->private_data = NULL;
+ if (!atomic_long_dec_and_test(&event->refcount))
+ return;

rcu_read_lock();
owner = ACCESS_ONCE(event->owner);
@@ -2975,7 +2975,13 @@ static int perf_release(struct inode *inode, struct file *file)
put_task_struct(owner);
}

- return perf_event_release_kernel(event);
+ perf_event_release_kernel(event);
+}
+
+static int perf_release(struct inode *inode, struct file *file)
+{
+ put_event(file->private_data);
+ return 0;
}

u64 perf_event_read_value(struct perf_event *event, u64 *enabled, u64 *running)
@@ -3227,7 +3233,7 @@ unlock:

static const struct file_operations perf_fops;

-static struct perf_event *perf_fget_light(int fd, int *fput_needed)
+static struct file *perf_fget_light(int fd, int *fput_needed)
{
struct file *file;

@@ -3241,7 +3247,7 @@ static struct perf_event *perf_fget_light(int fd, int *fput_needed)
return ERR_PTR(-EBADF);
}

- return file->private_data;
+ return file;
}

static int perf_event_set_output(struct perf_event *event,
@@ -3273,19 +3279,21 @@ static long perf_ioctl(struct file *file, unsigned int cmd, unsigned long arg)

case PERF_EVENT_IOC_SET_OUTPUT:
{
+ struct file *output_file = NULL;
struct perf_event *output_event = NULL;
int fput_needed = 0;
int ret;

if (arg != -1) {
- output_event = perf_fget_light(arg, &fput_needed);
- if (IS_ERR(output_event))
- return PTR_ERR(output_event);
+ output_file = perf_fget_light(arg, &fput_needed);
+ if (IS_ERR(output_file))
+ return PTR_ERR(output_file);
+ output_event = output_file->private_data;
}

ret = perf_event_set_output(event, output_event);
if (output_event)
- fput_light(output_event->filp, fput_needed);
+ fput_light(output_file, fput_needed);

return ret;
}
@@ -5950,6 +5958,7 @@ perf_event_alloc(struct perf_event_attr *attr, int cpu,

mutex_init(&event->mmap_mutex);

+ atomic_long_set(&event->refcount, 1);
event->cpu = cpu;
event->attr = *attr;
event->group_leader = group_leader;
@@ -6260,12 +6269,12 @@ SYSCALL_DEFINE5(perf_event_open,
return event_fd;

if (group_fd != -1) {
- group_leader = perf_fget_light(group_fd, &fput_needed);
- if (IS_ERR(group_leader)) {
- err = PTR_ERR(group_leader);
+ group_file = perf_fget_light(group_fd, &fput_needed);
+ if (IS_ERR(group_file)) {
+ err = PTR_ERR(group_file);
goto err_fd;
}
- group_file = group_leader->filp;
+ group_leader = group_file->private_data;
if (flags & PERF_FLAG_FD_OUTPUT)
output_event = group_leader;
if (flags & PERF_FLAG_FD_NO_GROUP)
@@ -6402,7 +6411,6 @@ SYSCALL_DEFINE5(perf_event_open,
put_ctx(gctx);
}

- event->filp = event_file;
WARN_ON_ONCE(ctx->parent_ctx);
mutex_lock(&ctx->mutex);

@@ -6496,7 +6504,6 @@ perf_event_create_kernel_counter(struct perf_event_attr *attr, int cpu,
goto err_free;
}

- event->filp = NULL;
WARN_ON_ONCE(ctx->parent_ctx);
mutex_lock(&ctx->mutex);
perf_install_in_context(ctx, event, cpu);
@@ -6578,7 +6585,7 @@ static void sync_child_event(struct perf_event *child_event,
* Release the parent event, if this was the last
* reference to it.
*/
- fput(parent_event->filp);
+ put_event(parent_event);
}

static void
@@ -6654,9 +6661,8 @@ static void perf_event_exit_task_context(struct task_struct *child, int ctxn)
*
* __perf_event_exit_task()
* sync_child_event()
- * fput(parent_event->filp)
- * perf_release()
- * mutex_lock(&ctx->mutex)
+ * put_event()
+ * mutex_lock(&ctx->mutex)
*
* But since its the parent context it won't be the same instance.
*/
@@ -6724,7 +6730,7 @@ static void perf_free_event(struct perf_event *event,
list_del_init(&event->child_list);
mutex_unlock(&parent->child_mutex);

- fput(parent->filp);
+ put_event(parent);

perf_group_detach(event);
list_del_event(event, ctx);
@@ -6804,6 +6810,12 @@ inherit_event(struct perf_event *parent_event,
NULL, NULL);
if (IS_ERR(child_event))
return child_event;
+
+ if (!atomic_long_inc_not_zero(&parent_event->refcount)) {
+ free_event(child_event);
+ return NULL;
+ }
+
get_ctx(child_ctx);

/*
@@ -6845,14 +6857,6 @@ inherit_event(struct perf_event *parent_event,
raw_spin_unlock_irqrestore(&child_ctx->lock, flags);

/*
- * Get a reference to the parent filp - we will fput it
- * when the child event exits. This is safe to do because
- * we are in the parent and we know that the filp still
- * exists and has a nonzero count:
- */
- atomic_long_inc(&parent_event->filp->f_count);
-
- /*
* Link this into the parent event's child list
*/
WARN_ON_ONCE(parent_event->ctx->parent_ctx);
diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c
index bb38c4d..9a7b487 100644
--- a/kernel/events/hw_breakpoint.c
+++ b/kernel/events/hw_breakpoint.c
@@ -453,7 +453,16 @@ int modify_user_hw_breakpoint(struct perf_event *bp, struct perf_event_attr *att
int old_type = bp->attr.bp_type;
int err = 0;

- perf_event_disable(bp);
+ /*
+ * modify_user_hw_breakpoint can be invoked with IRQs disabled and hence it
+ * will not be possible to raise IPIs that invoke __perf_event_disable.
+ * So call the function directly after making sure we are targeting the
+ * current task.
+ */
+ if (irqs_disabled() && bp->ctx && bp->ctx->task == current)
+ __perf_event_disable(bp);
+ else
+ perf_event_disable(bp);

bp->attr.bp_addr = attr->bp_addr;
bp->attr.bp_type = attr->bp_type;
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Arnaldo Carvalho de Melo: "Re: [tip:perf/core] perf hists: Introduce perf_hpp for hist periodprinting"
Previous message: Arnd Bergmann: "Re: Implement devicetree support for AB8500 Btemp"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]