[dm-devel] [PATCH v2 1/2] md raid0/linear: Introduce new array state 'broken'

[Song Liu]

> On Aug 21, 2019, at 9:14 AM, Song Liu <songliubraving at fb.com> wrote:
> 
> 
> 
>> On Aug 21, 2019, at 7:16 AM, Guilherme G. Piccoli <gpiccoli at canonical.com> wrote:
>> 
>> On 19/08/2019 18:57, Song Liu wrote:
>>> [...]
>>> How about we test this when we do clear_bit(Faulty..)? And maybe also in 
>>> add_new_disk()?
>>> 
>>> Thanks,
>>> Song
>>> 
>> 
>> Song, thanks for the suggestions. I've been working in the refactor, so
>> far it's working fine. But I cannot re-add a member to raid0/linear
>> without performing a full stop (with "mdadm --stop"), and in this case
>> md_clean() will clear the flag. Restarting array this way works fine.
>> 
>> If I try writing 'inactive' to array_state, I cannot reinsert the member
>> to the array. That said, I don't think we need to worry in clearing
>> MD_BROKEN for RAID0/LINEAR, and it makes things far easier.
>> Are you ok with that? I'll submit V3 after our discussion.
>> 
> 
> What do you mean by "not clear MD_BROKEN"? Do you mean we need to restart
> the array? 
> 
> IOW, the following won't work:
> 
>  mdadm --fail /dev/md0 /dev/sdx
>  mdadm --remove /dev/md0 /dev/sdx
>  mdadm --add /dev/md0 /dev/sdx
> 
> And we need the following instead:
> 
>  mdadm --fail /dev/md0 /dev/sdx
>  mdadm --remove /dev/md0 /dev/sdx
>  mdadm --stop /dev/md0 /dev/sdx
>  mdadm --add /dev/md0 /dev/sdx
>  mdadm --run /dev/md0 /dev/sdx

Btw, the MD_BROKEN patch conflicts with one of Neil's patches. Please 
rebase your work on top of

https://git.kernel.org/pub/scm/linux/kernel/git/song/md.git/tree/?h=md-next

Thanks,
Song