[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <YVQQ6mNhdMVQOhZN@infradead.org>
Date: Wed, 29 Sep 2021 08:08:26 +0100
From: Christoph Hellwig <hch@...radead.org>
To: Po-Wen Kao <powen.kao@...iatek.com>
Cc: linux-block@...r.kernel.org, axboe@...nel.dk,
linux-kernel@...r.kernel.org, stanley.chu@...iatek.com,
linux-scsi@...r.kernel.org, martin.petersen@...cle.com,
avri.altman@....com, alim.akhtar@...sung.com, jejb@...ux.ibm.com,
peter.wang@...iatek.com, chun-hung.wu@...iatek.com,
alice.chao@...iatek.com, jonathan.hsu@...iatek.com,
cc.chou@...iatek.com, chaotian.jing@...iatek.com,
jiajie.hao@...iatek.com, wsd_upstream@...iatek.com,
ed.tsai@...iatek.com
Subject: Re: [PATCH 1/2] blk-mq: new busy request iterator for driver
On Wed, Sep 29, 2021 at 03:00:46PM +0800, Po-Wen Kao wrote:
> Driver occasionally execute allocated request directly without
> dispatching to block layer, thus request never appears in tags->rqs.
> To allow driver to iterate through requests in static_rqs, a new
> interface blk_mq_drv_tagset_busy_iter() is introduced.
Don't do that. All requests must be dispatched to blk-mq. Let's not
even get started on these hacks that will make our life painful forever.
Powered by blists - more mailing lists