2019-02-21 20:28:19

by Dennis Zhou

Subject: [PATCH] btrfs: zstd ensure reclaim timer is properly cleaned up

The timer function, zstd_reclaim_timer_fn(), reschedules itself under
certain conditions. Switch to del_timer_sync() to ensure that the timer
function hasn't rescheduled itself.

Signed-off-by: Dennis Zhou <[email protected]>
---
fs/btrfs/zstd.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/btrfs/zstd.c b/fs/btrfs/zstd.c
index 3e418a3aeb11..62de9a211321 100644
--- a/fs/btrfs/zstd.c
+++ b/fs/btrfs/zstd.c
@@ -195,7 +195,7 @@ static void zstd_cleanup_workspace_manager(void)
struct workspace *workspace;
int i;

- del_timer(&wsm.timer);
+ del_timer_sync(&wsm.timer);

for (i = 0; i < ZSTD_BTRFS_MAX_LEVEL; i++) {
while (!list_empty(&wsm.idle_ws[i])) {
--
2.17.1

2019-02-22 12:20:03

by Nikolay Borisov

Subject: Re: [PATCH] btrfs: zstd ensure reclaim timer is properly cleaned up

On 21.02.19 г. 22:25 ч., Dennis Zhou wrote:
> The timer function, zstd_reclaim_timer_fn(), reschedules itself under
> certain conditions. Switch to del_timer_sync() to ensure that the timer
> function hasn't rescheduled itself.

According to del_timer_sync it just waits for any concurrent invocation
to finish. But it's responsibility of the caller to ensure the timer
cannot really be restarted. It's not obvious how
zstd_cleanup_workspace_manager ensures that the timer won't be restared.
Looking at the timer function for it to not restart wsm.lru_list has to
be empty. And seeing that workspace->lru_list is deleted afterwards I'm
not sure this invariant holds.

>
> Signed-off-by: Dennis Zhou <[email protected]>
> ---
> fs/btrfs/zstd.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/btrfs/zstd.c b/fs/btrfs/zstd.c
> index 3e418a3aeb11..62de9a211321 100644
> --- a/fs/btrfs/zstd.c
> +++ b/fs/btrfs/zstd.c
> @@ -195,7 +195,7 @@ static void zstd_cleanup_workspace_manager(void)
> struct workspace *workspace;
> int i;
>
> - del_timer(&wsm.timer);
> + del_timer_sync(&wsm.timer);
>
> for (i = 0; i < ZSTD_BTRFS_MAX_LEVEL; i++) {
> while (!list_empty(&wsm.idle_ws[i])) {
>

2019-02-22 16:54:40

by Dennis Zhou

Subject: Re: [PATCH] btrfs: zstd ensure reclaim timer is properly cleaned up

On Fri, Feb 22, 2019 at 02:17:41PM +0200, Nikolay Borisov wrote:
>
>
> On 21.02.19 г. 22:25 ч., Dennis Zhou wrote:
> > The timer function, zstd_reclaim_timer_fn(), reschedules itself under
> > certain conditions. Switch to del_timer_sync() to ensure that the timer
> > function hasn't rescheduled itself.
>
> According to del_timer_sync it just waits for any concurrent invocation
> to finish. But it's responsibility of the caller to ensure the timer
> cannot really be restarted. It's not obvious how
> zstd_cleanup_workspace_manager ensures that the timer won't be restared.
> Looking at the timer function for it to not restart wsm.lru_list has to
> be empty. And seeing that workspace->lru_list is deleted afterwards I'm
> not sure this invariant holds.
>

Hmmm, I see. I misunderstood the invariant. I'll spin v2 and handle it
another way.

Thanks,
Dennis

> >
> > Signed-off-by: Dennis Zhou <[email protected]>
> > ---
> > fs/btrfs/zstd.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/fs/btrfs/zstd.c b/fs/btrfs/zstd.c
> > index 3e418a3aeb11..62de9a211321 100644
> > --- a/fs/btrfs/zstd.c
> > +++ b/fs/btrfs/zstd.c
> > @@ -195,7 +195,7 @@ static void zstd_cleanup_workspace_manager(void)
> > struct workspace *workspace;
> > int i;
> >
> > - del_timer(&wsm.timer);
> > + del_timer_sync(&wsm.timer);
> >
> > for (i = 0; i < ZSTD_BTRFS_MAX_LEVEL; i++) {
> > while (!list_empty(&wsm.idle_ws[i])) {
> >