linux-kernel - Re: [RFC PATCH v3 1/4] ufs: core: prepare ufs for multi circular queue support

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: <df32bbd5-4109-903b-c7ba-7777a298f5aa@quicinc.com>
Date:   Wed, 7 Sep 2022 20:00:56 -0700
From:   "Asutosh Das (asd)" <quic_asutoshd@...cinc.com>
To:     Bean Huo <huobean@...il.com>, <quic_nguyenb@...cinc.com>,
        <quic_xiaosenh@...cinc.com>, <stanley.chu@...iatek.com>,
        <adrian.hunter@...el.com>, <bvanassche@....org>,
        <avri.altman@....com>, <mani@...nel.org>, <quic_cang@...cinc.com>,
        <beanhuo@...ron.com>, <martin.petersen@...cle.com>,
        <linux-scsi@...r.kernel.org>
CC:     <linux-arm-msm@...r.kernel.org>,
        Alim Akhtar <alim.akhtar@...sung.com>,
        "James E.J. Bottomley" <jejb@...ux.ibm.com>,
        open list <linux-kernel@...r.kernel.org>
Subject: Re: [RFC PATCH v3 1/4] ufs: core: prepare ufs for multi circular
 queue support

Hello Bean
Thanks for your comments. Responses inline.

On 9/5/2022 5:26 AM, Bean Huo wrote:
> Asutosh,
> 
> 
[...]
> 
>>   static inline
>> -void ufshcd_send_command(struct ufs_hba *hba, unsigned int task_tag)
>> +void ufshcd_send_command(struct ufs_hba *hba, struct ufshcd_lrb
>> *lrbp)
>>   {
>> -       struct ufshcd_lrb *lrbp = &hba->lrb[task_tag];
>> +       int task_tag = lrbp->task_tag;
>>          unsigned long flags;
>>   
>>          lrbp->issue_time_stamp = ktime_get();
>>          lrbp->compl_time_stamp = ktime_set(0, 0);
>> -       ufshcd_add_command_trace(hba, task_tag, UFS_CMD_SEND);
>> +       ufshcd_add_command_trace(hba, lrbp, UFS_CMD_SEND);
>>          ufshcd_clk_scaling_start_busy(hba);
>>          if (unlikely(ufshcd_should_inform_monitor(hba, lrbp)))
>>                  ufshcd_start_monitor(hba, lrbp);
>> @@ -2553,9 +2554,10 @@ void ufshcd_prepare_utp_scsi_cmd_upiu(struct
>> ufshcd_lrb *lrbp, u8 upiu_flags)
>>          /* command descriptor fields */
>>          ucd_req_ptr->header.dword_0 = UPIU_HEADER_DWORD(
>>                                  UPIU_TRANSACTION_COMMAND, upiu_flags,
>> -                               lrbp->lun, lrbp->task_tag);
>> +                               lrbp->lun, lrbp->task_tag & 0xff);
>>          ucd_req_ptr->header.dword_1 = UPIU_HEADER_DWORD(
>> -                               UPIU_COMMAND_SET_TYPE_SCSI, 0, 0, 0);
>> +                               UPIU_COMMAND_SET_TYPE_SCSI, 0, 0,
>> +                               (lrbp->task_tag & 0xf00) << 4);
>>   
> 
> Are you sure here "(lrbp->task_tag & 0xf00) << 4" is correct?
> 
EXT_IID is the higher nibble in DWORD1. So this looks correct to me.

COMMAND UPIU
0 	    1 	  2 	3
xx000001b Flags LUN Task Tag
4 			5 	6 	7
IID Command Set Type Reserved Reserved EXT_IID | Reserved
> 
> this will overwrite other fields, see UPIU_HEADER_DWORD:
> 
> #define UPIU_HEADER_DWORD(byte3, byte2, byte1, byte0)\
>                          cpu_to_be32((byte3 << 24) | (byte2 << 16) |\
>                           (byte1 << 8) | (byte0))
> 
> 
> 
>>
>>   
>> -       ufshcd_send_command(hba, tag);
>> +       ufshcd_send_command(hba, lrbp);
>>          err = ufshcd_wait_for_dev_cmd(hba, lrbp, timeout);
>>          ufshcd_add_query_upiu_trace(hba, err ? UFS_QUERY_ERR :
>> UFS_QUERY_COMP,
>>                                      (struct utp_upiu_req *)lrbp-
>>> ucd_rsp_ptr);
>> @@ -4513,6 +4515,7 @@ int ufshcd_make_hba_operational(struct ufs_hba
>> *hba)
>>                          REG_UTP_TRANSFER_REQ_LIST_BASE_L);
>>          ufshcd_writel(hba, upper_32_bits(hba->utrdl_dma_addr),
>>                          REG_UTP_TRANSFER_REQ_LIST_BASE_H);
>> +
>>          ufshcd_writel(hba, lower_32_bits(hba->utmrdl_dma_addr),
>>                          REG_UTP_TASK_REQ_LIST_BASE_L);
>>          ufshcd_writel(hba, upper_32_bits(hba->utmrdl_dma_addr),
>> @@ -5320,6 +5323,32 @@ static void ufshcd_release_scsi_cmd(struct
>> ufs_hba *hba,
>>          ufshcd_clk_scaling_update_busy(hba);
>>   }
>>   
>> +void ufshcd_compl_one_cqe(struct ufs_hba *hba, int task_tag)
> 
> This function does only complete one task. What does cqe stand for?
> 
CQE - Completion Queue Entry. The term 'task' is used for TM commands in 
scsi, hence there was a comment to change it to CQE.

>> +{
>> +       struct ufshcd_lrb *lrbp;
>> +       struct scsi_cmnd *cmd;
>> +
>> +       lrbp = &hba->lrb[task_tag];
>> +       lrbp->compl_time_stamp = ktime_get();
>> +       cmd = lrbp->cmd;
>> +       if (cmd) {
>> +               if (unlikely(ufshcd_should_inform_monitor(hba,
>> lrbp)))
>> +                       ufshcd_update_monitor(hba, lrbp);
>> +               ufshcd_add_command_trace(hba, lrbp, UFS_CMD_COMP);
>> +               cmd->result = ufshcd_transfer_rsp_status(hba, lrbp);
>> +               ufshcd_release_scsi_cmd(hba, lrbp);
>> +               /* Do not touch lrbp after scsi done */
>> +               scsi_done(cmd);
>> +       } else if (lrbp->command_type == UTP_CMD_TYPE_DEV_MANAGE ||
>> +                  lrbp->command_type == UTP_CMD_TYPE_UFS_STORAGE) {
>> +               if (hba->dev_cmd.complete) {
>> +                       ufshcd_add_command_trace(hba, lrbp,
>> UFS_DEV_COMP);
>> +                       complete(hba->dev_cmd.complete);
>> +                       ufshcd_clk_scaling_update_busy(hba);
>> +               }
>> +       }
>> +}
>> +
> 
> 
> Kind regards,
> Bean
>