All of lore.kernel.org
 help / color / mirror / Atom feed
From: Yufen Yu <yuyufen@huawei.com>
To: Song Liu <song@kernel.org>
Cc: linux-raid <linux-raid@vger.kernel.org>,
	NeilBrown <neilb@suse.com>,
	Guoqing Jiang <guoqing.jiang@cloud.ionos.com>,
	Hou Tao <houtao1@huawei.com>
Subject: Re: [PATCH 00/12] Save memory for stripe_head buffer
Date: Thu, 20 Aug 2020 14:30:52 +0800	[thread overview]
Message-ID: <bfe3b715-e6d4-4754-dc4f-c4441123e7c7@huawei.com> (raw)
In-Reply-To: <CAPhsuW6rB-+APsw77CqOz8ipdneRyCDBdgx_rxi0ep22xC5gVw@mail.gmail.com>



On 2020/8/20 8:25, Song Liu wrote:
> On Wed, Aug 12, 2020 at 5:48 AM Yufen Yu <yuyufen@huawei.com> wrote:
>>
>> Hi, all
>>
>>   In current implementation, grow_buffers() uses alloc_page() to allocate
>>   the buffers for each stripe_head, i.e. allocate a page for each dev[i]
>>   in stripe_head.
>>
>>   After setting stripe_size as a configurable value by writing sysfs entry,
>>   it means that we always allocate 64K buffers, but just use 4K of them when
>>   stripe_size is 4K in 64KB arm64.
>>
>>   To save memory, we try to let multiple buffers of stripe_head to share only
>>   one real page when page size is bigger than stripe_size. Detail can be
>>   seen in patch #10.
>>
>>   This patch set is subsequent optimization for configurable stripe_size,
>>   which based on the origin patches[1] but reorganized them.
>>
>>   Patch 1 ~ 2 try to replace current page offset '0' with dev[i].offset.
>>   Patch 3 ~ 5 let xor compute functions support different page offset for raid5.
>>   Patch 6 ~ 9 let syndrome and recovery function support different page offset for raid6.
>>   All of these patch are preparing for shared page. There is no functional change.
>>
>>   Patch 10 ~ 11 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.
> 
> Thanks for the patches.
> 
> I went through the first half of the set, most of the code looks fine.
> However, there is
> one issue: the way you split the changes into 12 patches is not ideal. The most
> significant issue is that build failed after 5/12 or 6/12 (and was
> fixed after 9/12). We
> would like every commit build successfully. I think the proper fix is
> to break 9/12 and
> merge changes in raid5.c to proper patches. Also, I think we can merge
> 1/12 and 2/12.
> 
> Please resubmit after these changes.
> 
> Thanks,
> Song
> .
> 

Thanks a lot for your review. These changes can make patch set more reasonable.

Thanks,
Yufen

      reply	other threads:[~2020-08-20  6:31 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-08-12 12:49 [PATCH 00/12] Save memory for stripe_head buffer Yufen Yu
2020-08-12 12:49 ` [PATCH 01/12] md/raid5: add offset array in scribble buffer Yufen Yu
2020-08-12 12:49 ` [PATCH 02/12] md/raid5: add a new member of offset into r5dev Yufen Yu
2020-08-12 12:49 ` [PATCH 03/12] md/raid5: make async_copy_data() to support different page offset Yufen Yu
2020-08-12 12:49 ` [PATCH 04/12] md/raid5: add new xor function " Yufen Yu
2020-08-12 12:49 ` [PATCH 05/12] md/raid5: convert to new xor compution interface Yufen Yu
2020-08-12 12:49 ` [PATCH 06/12] md/raid6: let syndrome computor support different page offset Yufen Yu
2020-08-12 12:49 ` [PATCH 07/12] md/raid6: let async recovery function " Yufen Yu
2020-08-12 12:49 ` [PATCH 08/12] md/raid6: convert to new syndrome and recovery interface Yufen Yu
2020-08-12 12:49 ` [PATCH 09/12] raid6test: adaptation with new syndrome function Yufen Yu
2020-08-12 12:49 ` [PATCH 10/12] md/raid5: let multiple devices of stripe_head share page Yufen Yu
2020-08-12 12:49 ` [PATCH 11/12] md/raid5: resize stripe_head when reshape array Yufen Yu
2020-08-12 12:49 ` [PATCH 12/12] md/raid5: reallocate page array after setting new stripe_size Yufen Yu
2020-08-20  0:25 ` [PATCH 00/12] Save memory for stripe_head buffer Song Liu
2020-08-20  6:30   ` Yufen Yu [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bfe3b715-e6d4-4754-dc4f-c4441123e7c7@huawei.com \
    --to=yuyufen@huawei.com \
    --cc=guoqing.jiang@cloud.ionos.com \
    --cc=houtao1@huawei.com \
    --cc=linux-raid@vger.kernel.org \
    --cc=neilb@suse.com \
    --cc=song@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.