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

From: Song Liu <songliubraving@fb.com>
To: "Guilherme G. Piccoli" <gpiccoli@canonical.com>
Cc: Song Liu <liu.song.a23@gmail.com>,
	linux-raid <linux-raid@vger.kernel.org>,
	"linux-block@vger.kernel.org" <linux-block@vger.kernel.org>,
	"dm-devel@redhat.com" <dm-devel@redhat.com>,
	Jay Vosburgh <jay.vosburgh@canonical.com>,
	NeilBrown <neilb@suse.com>
Subject: Re: [PATCH v2 1/2] md raid0/linear: Introduce new array state 'broken'
Date: Mon, 19 Aug 2019 21:57:11 +0000	[thread overview]
Message-ID: <8E880472-67DA-4597-AFAD-0DAFFD223620@fb.com> (raw)
In-Reply-To: <4c95f76c-dfbc-150c-2950-d34521d1e39d@canonical.com>

> On Aug 19, 2019, at 12:11 PM, Guilherme G. Piccoli <gpiccoli@canonical.com> wrote:
> 
> On 19/08/2019 15:57, Song Liu wrote:
>> [...] 
>> 
>> I was thinking, if we can set MD_BROKEN when the device fails, we can 
>> just test MD_BROKEN in array_state_show() (instead of iterating through 
>> all devices). 
>> 
>> Would this work?
>> 
>> Thanks,
>> Song
>> 
> 
> This could work, but will require some refactors; it'll simplify the
> check for a healthy array (no need for is_missing_dev() function) but
> will complicate the criteria to clear MD_BROKEN and the detection of
> more members failing (we would stop the detection after the 1st failure
> if we only test MD_BROKEN).
> 
> We will always need to test the GENHD_FL_UP in some code path, to be
> able to show users device recovered from a failure.
> Where do you suggest to test this flag?

How about we test this when we do clear_bit(Faulty..)? And maybe also in 
add_new_disk()?

Thanks,
Song