linux-kernel - Re: [RFC PATCH v2 4/5] scsi: ufs: L2P map management for HPB read

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <231786897.01592275801931.JavaMail.epsvc@epcpadp1>
Date:   Tue, 16 Jun 2020 10:17:44 +0900
From:   Daejun Park <daejun7.park@...sung.com>
To:     Bean Huo <huobean@...il.com>,
        Daejun Park <daejun7.park@...sung.com>,
        ALIM AKHTAR <alim.akhtar@...sung.com>,
        "avri.altman@....com" <avri.altman@....com>,
        "jejb@...ux.ibm.com" <jejb@...ux.ibm.com>,
        "martin.petersen@...cle.com" <martin.petersen@...cle.com>,
        "asutoshd@...eaurora.org" <asutoshd@...eaurora.org>,
        "beanhuo@...ron.com" <beanhuo@...ron.com>,
        "stanley.chu@...iatek.com" <stanley.chu@...iatek.com>,
        "cang@...eaurora.org" <cang@...eaurora.org>,
        "bvanassche@....org" <bvanassche@....org>,
        "tomas.winkler@...el.com" <tomas.winkler@...el.com>
CC:     "linux-scsi@...r.kernel.org" <linux-scsi@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Sang-yoon Oh <sangyoon.oh@...sung.com>,
        Sung-Jun Park <sungjun07.park@...sung.com>,
        yongmyung lee <ymhungry.lee@...sung.com>,
        Jinyoung CHOI <j-young.choi@...sung.com>,
        Adel Choi <adel.choi@...sung.com>,
        BoRam Shin <boram.shin@...sung.com>
Subject: Re: [RFC PATCH v2 4/5] scsi: ufs: L2P map management for HPB read

Hi Bean
> 
> 
> On Mon, 2020-06-15 at 18:30 +0900, Daejun Park wrote:
> > +static int ufshpb_execute_map_req(struct ufshpb_lu *hpb,
> > +                                 struct ufshpb_req *map_req)
> > +{
> > +       struct request_queue *q;
> > +       struct request *req;
> > +       struct scsi_request *rq;
> > +       int ret = 0;
> > +
> > +       q = hpb->sdev_ufs_lu->request_queue;
> > +       ret = ufshpb_map_req_add_bio_page(hpb, q, map_req->bio,
> > +                                         map_req->mctx);
> > +       if (ret) {
> > +               dev_notice(&hpb->hpb_lu_dev,
> > +                          "map_req_add_bio_page fail %d - %d\n",
> > +                          map_req->rgn_idx, map_req->srgn_idx);
> > +               return ret;
> > +       }
> > +
> > +       req = map_req->req;
> > +
> > +       blk_rq_append_bio(req, &map_req->bio);
> > +
> > +       req->timeout = 0;
> > +       req->end_io_data = (void *)map_req;
> > +
> > +       rq = scsi_req(req);
> > +       ufshpb_set_read_buf_cmd(rq->cmd, map_req->rgn_idx,
> > +                               map_req->srgn_idx, hpb-
> > >srgn_mem_size);
> > +       rq->cmd_len = HPB_READ_BUFFER_CMD_LENGTH;
> > +
> > +       blk_execute_rq_nowait(q, NULL, req, 1,
> > ufshpb_map_req_compl_fn);
> 
> 
> HPB map_req is now being en-queued in sdev->request_queue.
> This is ok for the HPB v1.0. Have you ever been thinking about
> changing this way to directly issue HPB requests to UFS?
I think it is enough to support HPB.

> Actually, there are two reasons for this way:
> 
> 1. Latency of loading mapping entries is lower comparing to your curret
> approach.
Our map request style utilizes block layer API. It makes the codes keep
simple and reliable. Also, I think the latency of loading mapping entries 
is not critical for HPB. Futhermore, I think the other requests from
user is important than loading mapping entries. If the map_work issues map
request directly, the user requests can be disturbed by map requests.

> 2. Also, it is preparing for the HPB v2.0.  After all HPB 1.0 only
> supports 4KB read, this is useless, I am looking for the HPB v2.0.
I don't understand this comment. In HPB v2.0, this code can be used without
changing.

Thanks,
Daejun