Re: Scrub resume regression

[Graham Cobb <g.btrfs@cobb.uk.net>]

On 15/01/2020 09:33, Filipe Manana wrote:
> On Wed, Jan 15, 2020 at 9:04 AM Graham Cobb <g.btrfs@cobb.uk.net> wrote:
>>
>> OK, I have bisected the problem with scrub resume being broken by the
>> scrub ioctl ABI being changed.
[snip]
> 
> Try this:
> 
> diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
> index 3a4bd5cd67fa..611dfe8cdbb1 100644
> --- a/fs/btrfs/ioctl.c
> +++ b/fs/btrfs/ioctl.c
> @@ -4253,8 +4253,10 @@ static long btrfs_ioctl_scrub(struct file
> *file, void __user *arg)
>                               &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
>                               0);
> 
> -       if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
> -               ret = -EFAULT;
> +       if (copy_to_user(arg, sa, sizeof(*sa))) {
> +               if (!ret)
> +                       ret = -EFAULT;
> +       }
> 
>         if (!(sa->flags & BTRFS_SCRUB_READONLY))
>                 mnt_drop_write_file(file);
> 
> I'll later send a patch with a changelog to the list.
> Thanks.

Thanks Filipe. I have tested that the patch builds and works with my
reproducer on a 5.4 kernel.

Later I will build a distro kernel with the patch and try my monthly
scrub-with-interruptions script. That will take a day or two.