From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: From: Shinichiro Kawasaki Subject: Re: [PATCH v3 18/38] zbd: handle conventional start zone in zbd_convert_to_open_zone() Date: Fri, 22 Jan 2021 04:36:40 +0000 Message-ID: <20210122043639.2eek4ghpycmeyaoq@shindev.dhcp.fujisawa.hgst.com> References: <20210106215739.264524-1-dmitry.fomichev@wdc.com> <20210106215739.264524-19-dmitry.fomichev@wdc.com> In-Reply-To: <20210106215739.264524-19-dmitry.fomichev@wdc.com> Content-Language: en-US Content-Type: text/plain; charset="us-ascii" Content-ID: <8D96FFAFB437224AB2C28A6968BAA3E4@namprd04.prod.outlook.com> Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 To: Dmitry Fomichev Cc: Jens Axboe , "fio@vger.kernel.org" , Aravind Ramesh , Bart Van Assche , Naohiro Aota , Niklas Cassel , Damien Le Moal List-ID: On Jan 07, 2021 / 06:57, Dmitry Fomichev wrote: > At the beginning of zbd_convert_to_open_zone() function, a zone > is picked in semi-random manner to become a candidate zone for > redirecting the incoming write. In some circumstances, such as > unlimited MaxOpen or i/o range that spans the boundary between > conventional and sequential zones, a conventional zone may be > selected. >=20 > This may create problems in the subsequent for (;;) loop in the > same function. Failed assertions were observed during the execution > of newly introduced test #51 that showed that the code in that loop > was trying to lock and unlock conventional zones. >=20 > Check if the zone which has been initially picked is conventional. > If yes, force the zone selection to be re-tried until a sequential > zone is selected for further processing. >=20 > Signed-off-by: Dmitry Fomichev Looks good to me. Reviewed-by: Shin'ichiro Kawasaki --=20 Best Regards, Shin'ichiro Kawasaki=