> On Nov 17, 2017, at 11:21 AM, Shaohua Li <shli@xxxxxxxxxx> wrote:
>
> On Thu, Nov 16, 2017 at 10:12:44AM -0800, Song Liu wrote:
>> In do_md_run(), md threads should not wake up until the array is fully
>> initialized in md_run(). However, in raid5_run(), raid5-cache may wake
>> up mddev->thread to flush stripes that need to be written back. This
>> design doesn't break badly right now. But it could lead to bad bug in
>> the future.
>>
>> This patch tries to resolve this problem by splitting start up work
>> into two personality functions, run() and start(). Tasks that do not
>> require the md threads should go into run(), while task that require
>> the md threads go into start().
>>
>> r5l_load_log() is moved to raid5_start(), so it is not called until
>> the md threads are started in do_md_run().
>>
>> Signed-off-by: Song Liu <songliubraving@xxxxxx>
>> ---
>> drivers/md/md.c | 13 ++++++++-----
>> drivers/md/md.h | 5 +++++
>> drivers/md/raid5-cache.c | 22 +++++++++++++++++-----
>> drivers/md/raid5-log.h | 1 +
>> drivers/md/raid5.c | 10 ++++++++++
>> 5 files changed, 41 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/md/md.c b/drivers/md/md.c
>> index e014d39..afeab7b 100644
>> --- a/drivers/md/md.c
>> +++ b/drivers/md/md.c
>> @@ -5561,11 +5561,6 @@ int md_run(struct mddev *mddev)
>> if (start_readonly && mddev->ro == 0)
>> mddev->ro = 2; /* read-only, but switch on first write */
>>
>> - /*
>> - * NOTE: some pers->run(), for example r5l_recovery_log(), wakes
>> - * up mddev->thread. It is important to initialize critical
>> - * resources for mddev->thread BEFORE calling pers->run().
>> - */
>> err = pers->run(mddev);
>> if (err)
>> pr_warn("md: pers->run() failed ...\n");
>> @@ -5680,6 +5675,14 @@ static int do_md_run(struct mddev *mddev)
>> md_allow_write(mddev);
>>
>> md_wakeup_thread(mddev->thread);
>> + /* run start up tasks that require md_thread */
>> + if (mddev->pers->start) {
>> + err = mddev->pers->start(mddev);
>> + if (err) {
>> + bitmap_destroy(mddev);
>> + goto out;
>> + }
>> + }
>
> Looks there is one thing left. dm-raid.c calls md_run but not do_md_run,
> dm-raid supports raid5-cache too. So we need export a wrap for ->start and let
> dm-raid use it.
I can add this.
> This doesn't seem to prevent reshape happening. mddev->thread calls
> check_recovery which can create and wakeup sync_thread. Probably we should
> preserve ->recovery and set it 0, do log load and then set it back, this way we
> should prevent the reshape for sure.
Maybe we need some systematic way to gate both threads (md_thread and
sync_thread). Maybe some extra flags like:
MD_THREAD_ENABLED
SYNC_THREAD_ENABLED
Thanks,
Song
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
