On Mon, Nov 05 2018, Gioh Kim wrote:
> Before updating superblock of slave disks, desired_state value
> is set for the target state of the slave disks. But it forgets
> to check MD_DISK_FAILFAST flag. Then start_arrays() calls
> ADD_NEW_DISK ioctl-call and pass the state without MD_DISK_FAILFAST.
>
> Currenlty it does not generate any problem because kernel does not
> care MD_DISK_FAILFAST flag.
>
> Signed-off-by: Gioh Kim <gi-oh.kim@xxxxxxxxxxxxxxxx>
> ---
> Assemble.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/Assemble.c b/Assemble.c
> index a79466c..fc0d464 100644
> --- a/Assemble.c
> +++ b/Assemble.c
> @@ -1704,6 +1704,9 @@ try_again:
> else
> desired_state = (1<<MD_DISK_ACTIVE) | (1<<MD_DISK_SYNC);
>
> + if (devices[j].i.disk.state & (1<<MD_DISK_FAILFAST))
> + desired_state |= (1<<MD_DISK_FAILFAST);
> +
> if (!devices[j].uptodate)
> continue;
>
This looks right, but I think we should preserve MD_DISK_WRITEMOSTLY as
well.
So maybe
desired_state |= devices[j].i.disk_state & ((1<<MD_DISK_FAILFAST) |
(1<<MD_DISK_WRITEMOSTLY));
NeilBrown
> --
> 2.14.1
Attachment:
signature.asc
Description: PGP signature
