2019-10-08 17:35:38

by Eric Dumazet

Subject: [PATCH] hrtimer: annotate lockless access to timer->base

Followup to commit dd2261ed45aa ("hrtimer: Protect lockless access
to timer->base")

lock_hrtimer_base() fetches timer->base without lock exclusion.

Compiler is allowed to read timer->base twice (even if considered dumb)
and we could end up trying to lock migration_base and
return &migration_base.

base = timer->base;
if (likely(base != &migration_base)) {

/* compiler reads timer->base again, and now (base == &migration_base)

raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
if (likely(base == timer->base))
return base; /* == &migration_base ! */

Similarly the write sides should use WRITE_ONCE() to avoid
store tearing.

Signed-off-by: Eric Dumazet <[email protected]>
Cc: Julien Grall <[email protected]>
Cc: Thomas Gleixner <[email protected]>
---
kernel/time/hrtimer.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
index 0d4dc241c0fb498036c91a571e65cb00f5d19ba6..65605530ee349c9682690c4fccb43aa9284d4287 100644
--- a/kernel/time/hrtimer.c
+++ b/kernel/time/hrtimer.c
@@ -164,7 +164,7 @@ struct hrtimer_clock_base *lock_hrtimer_base(const struct hrtimer *timer,
struct hrtimer_clock_base *base;

for (;;) {
- base = timer->base;
+ base = READ_ONCE(timer->base);
if (likely(base != &migration_base)) {
raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
if (likely(base == timer->base))
@@ -244,7 +244,7 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base,
return base;

/* See the comment in lock_hrtimer_base() */
- timer->base = &migration_base;
+ WRITE_ONCE(timer->base, &migration_base);
raw_spin_unlock(&base->cpu_base->lock);
raw_spin_lock(&new_base->cpu_base->lock);

@@ -253,10 +253,10 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base,
raw_spin_unlock(&new_base->cpu_base->lock);
raw_spin_lock(&base->cpu_base->lock);
new_cpu_base = this_cpu_base;
- timer->base = base;
+ WRITE_ONCE(timer->base, base);
goto again;
}
- timer->base = new_base;
+ WRITE_ONCE(timer->base, new_base);
} else {
if (new_cpu_base != this_cpu_base &&
hrtimer_check_target(timer, new_base)) {
--
2.23.0.581.g78d2f28ef7-goog

2019-10-14 04:16:12

by Eric Dumazet

Subject: Re: [PATCH] hrtimer: annotate lockless access to timer->base

On Tue, Oct 8, 2019 at 10:32 AM Eric Dumazet <[email protected]> wrote:
>
> Followup to commit dd2261ed45aa ("hrtimer: Protect lockless access
> to timer->base")
>
> lock_hrtimer_base() fetches timer->base without lock exclusion.
>
> Compiler is allowed to read timer->base twice (even if considered dumb)
> and we could end up trying to lock migration_base and
> return &migration_base.
>
> base = timer->base;
> if (likely(base != &migration_base)) {
>
> /* compiler reads timer->base again, and now (base == &migration_base)
>
> raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
> if (likely(base == timer->base))
> return base; /* == &migration_base ! */
>
> Similarly the write sides should use WRITE_ONCE() to avoid
> store tearing.
>
> Signed-off-by: Eric Dumazet <[email protected]>
> Cc: Julien Grall <[email protected]>
> Cc: Thomas Gleixner <[email protected]>
> ---
> kernel/time/hrtimer.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
> index 0d4dc241c0fb498036c91a571e65cb00f5d19ba6..65605530ee349c9682690c4fccb43aa9284d4287 100644
> --- a/kernel/time/hrtimer.c
> +++ b/kernel/time/hrtimer.c
> @@ -164,7 +164,7 @@ struct hrtimer_clock_base *lock_hrtimer_base(const struct hrtimer *timer,
> struct hrtimer_clock_base *base;
>
> for (;;) {
> - base = timer->base;
> + base = READ_ONCE(timer->base);
> if (likely(base != &migration_base)) {
> raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
> if (likely(base == timer->base))
> @@ -244,7 +244,7 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base,
> return base;
>
> /* See the comment in lock_hrtimer_base() */
> - timer->base = &migration_base;
> + WRITE_ONCE(timer->base, &migration_base);
> raw_spin_unlock(&base->cpu_base->lock);
> raw_spin_lock(&new_base->cpu_base->lock);
>
> @@ -253,10 +253,10 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base,
> raw_spin_unlock(&new_base->cpu_base->lock);
> raw_spin_lock(&base->cpu_base->lock);
> new_cpu_base = this_cpu_base;
> - timer->base = base;
> + WRITE_ONCE(timer->base, base);
> goto again;
> }
> - timer->base = new_base;
> + WRITE_ONCE(timer->base, new_base);
> } else {
> if (new_cpu_base != this_cpu_base &&
> hrtimer_check_target(timer, new_base)) {
> --
> 2.23.0.581.g78d2f28ef7-goog
>

Any news on this patch ?

If more information is needed, let me know.

Maybe I need to point to :

commit b831275a3553c32091222ac619cfddd73a5553fb timers: Plug locking
race vs. timer migration

Thanks

2019-10-14 14:00:47

by tip-bot2 for Alexey Makhalov

Subject: [tip: timers/urgent] hrtimer: Annotate lockless access to timer->base

The following commit has been merged into the timers/urgent branch of tip:

Commit-ID: ff229eee3d897f52bd001c841f2d3cce8853ecdc
Gitweb: https://git.kernel.org/tip/ff229eee3d897f52bd001c841f2d3cce8853ecdc
Author: Eric Dumazet <[email protected]>
AuthorDate: Tue, 08 Oct 2019 10:32:04 -07:00
Committer: Thomas Gleixner <[email protected]>
CommitterDate: Mon, 14 Oct 2019 15:51:49 +02:00

hrtimer: Annotate lockless access to timer->base

Followup to commit dd2261ed45aa ("hrtimer: Protect lockless access
to timer->base")

lock_hrtimer_base() fetches timer->base without lock exclusion.

Compiler is allowed to read timer->base twice (even if considered dumb)
which could end up trying to lock migration_base and return
&migration_base.

base = timer->base;
if (likely(base != &migration_base)) {

/* compiler reads timer->base again, and now (base == &migration_base)

raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
if (likely(base == timer->base))
return base; /* == &migration_base ! */

Similarly the write sides must use WRITE_ONCE() to avoid store tearing.

Signed-off-by: Eric Dumazet <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]

---
kernel/time/hrtimer.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
index 0d4dc24..6560553 100644
--- a/kernel/time/hrtimer.c
+++ b/kernel/time/hrtimer.c
@@ -164,7 +164,7 @@ struct hrtimer_clock_base *lock_hrtimer_base(const struct hrtimer *timer,
struct hrtimer_clock_base *base;

for (;;) {
- base = timer->base;
+ base = READ_ONCE(timer->base);
if (likely(base != &migration_base)) {
raw_spin_lock_irqsave(&base->cpu_base->lock, *flags);
if (likely(base == timer->base))
@@ -244,7 +244,7 @@ again:
return base;

/* See the comment in lock_hrtimer_base() */
- timer->base = &migration_base;
+ WRITE_ONCE(timer->base, &migration_base);
raw_spin_unlock(&base->cpu_base->lock);
raw_spin_lock(&new_base->cpu_base->lock);

@@ -253,10 +253,10 @@ again:
raw_spin_unlock(&new_base->cpu_base->lock);
raw_spin_lock(&base->cpu_base->lock);
new_cpu_base = this_cpu_base;
- timer->base = base;
+ WRITE_ONCE(timer->base, base);
goto again;
}
- timer->base = new_base;
+ WRITE_ONCE(timer->base, new_base);
} else {
if (new_cpu_base != this_cpu_base &&
hrtimer_check_target(timer, new_base)) {