2023-08-03 14:03:04

by Yu Kuai

Subject: [PATCH -next 3/7] md: don't rely on 'mddev->pers' to be set in mddev_suspend()

From: Yu Kuai <[email protected]>

Now that active io is initialized when mddev is allocated, it's safe to
call mddev_suspend() before 'mddev->pers' is set.

This also prevent null-ptr-def in some cases that caller doesn't
guarantee 'mddev->pers' to be set.

Signed-off-by: Yu Kuai <[email protected]>
---
drivers/md/md.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/md/md.c b/drivers/md/md.c
index 897e94a9e47d..f14f2f0a9484 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -448,7 +448,7 @@ void mddev_suspend(struct mddev *mddev)
set_bit(MD_ALLOW_SB_UPDATE, &mddev->flags);
percpu_ref_kill(&mddev->active_io);

- if (mddev->pers->prepare_suspend)
+ if (mddev->pers && mddev->pers->prepare_suspend)
mddev->pers->prepare_suspend(mddev);

wait_event(mddev->sb_wait, percpu_ref_is_zero(&mddev->active_io));
--
2.39.2

2023-08-22 01:16:56

by Song Liu

Subject: Re: [PATCH -next 3/7] md: don't rely on 'mddev->pers' to be set in mddev_suspend()

On Thu, Aug 3, 2023 at 6:30 AM Yu Kuai <[email protected]> wrote:
>
> From: Yu Kuai <[email protected]>
>
> Now that active io is initialized when mddev is allocated, it's safe to
> call mddev_suspend() before 'mddev->pers' is set.
>
> This also prevent null-ptr-def in some cases that caller doesn't
> guarantee 'mddev->pers' to be set.

This description is a little confusing (to me at least). Please revise it.

Other than that, the set looks good to me.

Thanks,
Song

>
> Signed-off-by: Yu Kuai <[email protected]>
> ---
> drivers/md/md.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index 897e94a9e47d..f14f2f0a9484 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -448,7 +448,7 @@ void mddev_suspend(struct mddev *mddev)
> set_bit(MD_ALLOW_SB_UPDATE, &mddev->flags);
> percpu_ref_kill(&mddev->active_io);
>
> - if (mddev->pers->prepare_suspend)
> + if (mddev->pers && mddev->pers->prepare_suspend)
> mddev->pers->prepare_suspend(mddev);
>
> wait_event(mddev->sb_wait, percpu_ref_is_zero(&mddev->active_io));
> --
> 2.39.2
>

2023-08-22 05:13:58

by Yu Kuai

Subject: Re: [PATCH -next 3/7] md: don't rely on 'mddev->pers' to be set in mddev_suspend()

Hi,

在 2023/08/22 8:13, Song Liu 写道:
> On Thu, Aug 3, 2023 at 6:30 AM Yu Kuai <[email protected]> wrote:
>>
>> From: Yu Kuai <[email protected]>
>>
>> Now that active io is initialized when mddev is allocated, it's safe to
>> call mddev_suspend() before 'mddev->pers' is set.
>>
>> This also prevent null-ptr-def in some cases that caller doesn't
>> guarantee 'mddev->pers' to be set.
>
> This description is a little confusing (to me at least). Please revise it.

Sorry about this, how about this:

'active_io' used to be initialized while the array is running, and
'mddev->pers' is set while the array is running as well. Hence caller
must hold 'reconfig_mutex' and guarantee 'mddev->pers' is set before
calling mddev_suspend().

Now that 'active_io' is initialized when mddev is allocated, such
restriction doesn't exist anymore. In the meantime, follow up patches
will refactor mddev_suspend(), hence add checking for 'mddev->pers' to
prevent null-ptr-deref.

Thanks,
Kuai

>
> Other than that, the set looks good to me.
>
> Thanks,
> Song
>
>>
>> Signed-off-by: Yu Kuai <[email protected]>
>> ---
>> drivers/md/md.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/md/md.c b/drivers/md/md.c
>> index 897e94a9e47d..f14f2f0a9484 100644
>> --- a/drivers/md/md.c
>> +++ b/drivers/md/md.c
>> @@ -448,7 +448,7 @@ void mddev_suspend(struct mddev *mddev)
>> set_bit(MD_ALLOW_SB_UPDATE, &mddev->flags);
>> percpu_ref_kill(&mddev->active_io);
>>
>> - if (mddev->pers->prepare_suspend)
>> + if (mddev->pers && mddev->pers->prepare_suspend)
>> mddev->pers->prepare_suspend(mddev);
>>
>> wait_event(mddev->sb_wait, percpu_ref_is_zero(&mddev->active_io));
>> --
>> 2.39.2
>>
> .
>

2023-08-22 14:04:02

by Song Liu

Subject: Re: [PATCH -next 3/7] md: don't rely on 'mddev->pers' to be set in mddev_suspend()

On Mon, Aug 21, 2023 at 7:08 PM Yu Kuai <[email protected]> wrote:
>
> Hi,
>
> 在 2023/08/22 8:13, Song Liu 写道:
> > On Thu, Aug 3, 2023 at 6:30 AM Yu Kuai <[email protected]> wrote:
> >>
> >> From: Yu Kuai <[email protected]>
> >>
> >> Now that active io is initialized when mddev is allocated, it's safe to
> >> call mddev_suspend() before 'mddev->pers' is set.
> >>
> >> This also prevent null-ptr-def in some cases that caller doesn't
> >> guarantee 'mddev->pers' to be set.
> >
> > This description is a little confusing (to me at least). Please revise it.
>
> Sorry about this, how about this:
>
> 'active_io' used to be initialized while the array is running, and
> 'mddev->pers' is set while the array is running as well. Hence caller
> must hold 'reconfig_mutex' and guarantee 'mddev->pers' is set before
> calling mddev_suspend().
>
> Now that 'active_io' is initialized when mddev is allocated, such
> restriction doesn't exist anymore. In the meantime, follow up patches
> will refactor mddev_suspend(), hence add checking for 'mddev->pers' to
> prevent null-ptr-deref.

This is much better. Thanks!

Song