Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   "Kiwoong Kim" <kwmad.kim@samsung.com>
To:     <linux-scsi@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
        <alim.akhtar@samsung.com>, <avri.altman@wdc.com>,
        <jejb@linux.ibm.com>, <martin.petersen@oracle.com>,
        <beanhuo@micron.com>, <cang@codeaurora.org>,
        <adrian.hunter@intel.com>, <sc.suh@samsung.com>,
        <hy50.seo@samsung.com>, <sh425.lee@samsung.com>,
        <bhoon95.kim@samsung.com>
Subject: Question about ufs_bsg
Date:   Tue, 14 Sep 2021 14:59:06 +0900
Message-ID: <000001d7a92d$a0edcb00$e2c96100$@samsung.com>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Thread-Index: AdepKfDft4+mL617SwWrELRDClUMdw==
Content-Language: ko
Content-Type: text/plain; charset="utf-8"
CMS-TYPE: 102P
DLP-Filter: Pass
References: <CGME20210914055907epcas2p1c9d5d91c61592a67b9ce4b2a88d8f279@epcas2p1.samsung.com>
Precedence: bulk

Hi,

ufs_bsg was introduced nearly three years ago and it allocates its own requ=
est queue.
I faced a sytmpom with this and want to ask something about it.

That is, sometimes queue depth for ufs is limited to half of the its maximu=
m value
even in a situation with many IO requests from filesystem.
It turned out that it only occurs when a query is being processed at the sa=
me time.
Regarding my tracing, when the query process starts, users for the hctx tha=
t represents
a ufs host increase to two and with this, some pathes calling 'hctx_may_que=
ue'
function in blk-mq seems to throttle dispatches, technically with 16 becaus=
e the number of
ufs slots (32 in my case) is dividend by two (users).

I found that it happened when a query for write booster is processed
because write booster only turns on in some conditions in my base that is d=
ifferent
from kernel mainline. But when an exceptional event or others that could le=
ad to a query occurs,
it can happen even in mainline.

I think the throttling is a little bit excessive,
so the question: is there any way to assign queue depth per user on an asym=
metric basis?

Thanks.
Kiwoong Kim