On 2020/8/21 13:04, Song Liu wrote:
On Thu, Aug 20, 2020 at 6:21 AM Yufen Yu <yuyufen@xxxxxxxxxx> wrote:
[...]
Patch 8 ~ 10 actually implement shared page between multiple devices of
stripe_head. But they only make sense for PAGE_SIZE != 4096, likely, 64KB arm64
system. It doesn't make any difference for PAGE_SIZE == 4096 system, likely x86.
We have run tests of mdadm for raid456 and test raid6test module.
Not found obvious errors.
Applied series to md-next.
I noticed there is another case where we allocate page for
sh->dev[i].orig_page in
handle_stripe_dirtying(). This only happens with journal devices.
Please run tests
with journal device.
For now, raid5 journal and PPL feature are both required PAGE_SIZE == 4096.
(Seen r5l_init_log() and ppl_init_log() return -EINVAL when PAGE_SIZE is not
4096). So, in theory, this patch set should not make any different for them.
But, run tests is also necessary. Tests under mdadm git repository have two
journal cases: 21raid5cache and 20raid5journal. To test PPL, I write a simple
test case: create raid5 array with option '-k ppl'. mkfs and mount the array.
Then, some read and write data test on mount directory. Concurrently remove disk
from array and re-add it. Not found obvious errors.
Thanks,
Yufen
