[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAGWkznHdE_zJ1pnxf-0-D3M=8aBwn-gFNte333WbKPioygGdLA@mail.gmail.com>
Date: Fri, 26 Jan 2024 09:14:30 +0800
From: Zhaoyang Huang <huangzhaoyang@...il.com>
To: Damien Le Moal <dlemoal@...nel.org>
Cc: "zhaoyang.huang" <zhaoyang.huang@...soc.com>, Andrew Morton <akpm@...ux-foundation.org>,
Jens Axboe <axboe@...nel.dk>, Yu Zhao <yuzhao@...gle.com>,
Niklas Cassel <niklas.cassel@....com>, "Martin K . Petersen" <martin.petersen@...cle.com>,
Hannes Reinecke <hare@...e.de>, Linus Walleij <linus.walleij@...aro.org>, linux-mm@...ck.org,
linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
steve.kang@...soc.com
Subject: Re: [PATCHv3 1/1] block: introduce content activity based ioprio
On Thu, Jan 25, 2024 at 6:10 PM Zhaoyang Huang <huangzhaoyang@...ilcom> wrote:
>
> On Thu, Jan 25, 2024 at 5:54 PM Damien Le Moal <dlemoal@...nel.org> wrote:
> >
> > On 1/25/24 18:37, Zhaoyang Huang wrote:
> > > Actually, the activity information comes from page's history (recorded
> > > at page cache's slot) instead of user space in step(1) and can be
> > > associate with bio in step(2) or iterate the bio in step(3)
> > > page fault(or vfs)(1)
> > > |
> > > alloc_pages
> > > |
> > > add_page_to_pagecache(get the page's activity information)
> > > |
> > > fs_readpage
> > > |
> > > bio_add_page(2)
> > > |
> >
> > set your prio here. No change to the block layer needed. That is FS land.
> Yes, that is what I suggested in patchv3 which keeps the block layer
> clean by introducing a set of helper functions which wrap up the
> activity things for FS utilization.
Maybe I misunderstand Damien's opinion which suggests iterating the
bio->bvec here via a new API. I think it is more costly than counting
activities when adding pages to bio and need more modifications on
each fs.
> >
> > > submit_bio(3)
> >
> >
> > --
> > Damien Le Moal
> > Western Digital Research
> >
Powered by blists - more mailing lists