On Thu, 28 May 2015 17:33:45 -0700 Shaohua Li <shli@xxxxxx> wrote:
> The run time stripe allocation is done at raid5d. When we set the
> R5_ALLOC_MORE flag, we should notify raid5d to handle it
>
> Signed-off-by: Shaohua Li <shli@xxxxxx>
> ---
> drivers/md/raid5.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
> index 61e8e04..bfa2042 100644
> --- a/drivers/md/raid5.c
> +++ b/drivers/md/raid5.c
> @@ -674,9 +674,11 @@ get_active_stripe(struct r5conf *conf, sector_t sector,
> if (!test_bit(R5_INACTIVE_BLOCKED, &conf->cache_state)) {
> sh = get_free_stripe(conf, hash);
> if (!sh && llist_empty(&conf->released_stripes) &&
> - !test_bit(R5_DID_ALLOC, &conf->cache_state))
> + !test_bit(R5_DID_ALLOC, &conf->cache_state)) {
> set_bit(R5_ALLOC_MORE,
> &conf->cache_state);
> + md_wakeup_thread(conf->mddev->thread);
> + }
> }
> if (noblock && sh == NULL)
> break;
Thanks for reviewing my code !!!
I'm not exactly against this patch, but I wonder if it is really needed.
R5_ALLOC_MORE is really just a hint - "You can allocate another stripe if you
like". If the array is at all busy, raid5d will be called fairly often and
the allocation will happen. If the array is idle, it doesn't matter if
memory is allocated for a while. Does it?
Thanks,
NeilBrown
Attachment:
pgpXSyt1yxzY4.pgp
Description: OpenPGP digital signature
