Re: [PATCH v2 2/2] scsi: ufs: core: Cleanup ufshcd_slave_alloc()

[Bean Huo <beanhuo@iokpp.de>]

Hi Bart,

I took all your suggestions and will send new version patch tomorrow
when the autotest is complete.

thanks,
Bean

On Fri, 2022-10-14 at 14:05 -0700, Bart Van Assche wrote:
> On 10/10/22 02:29, Bean Huo wrote:
> > From: Bean Huo <beanhuo@micron.com>
> > 
> > Combine ufshcd_get_lu_power_on_wp_status() and
> > ufshcd_set_queue_depth()
> > into one single ufshcd_lu_init(), so that we only need to read the
> > LUN
> > descriptor once to replace the original twice.
> 
> The following part can probably be left out from the patch
> description 
> without reducing clarity: " to replace the original twice".
> 
> > +/**
> > + * ufshcd_lu_power_on_wp_init - Initialize LU's power on write
> > protect state
> > + * @hba: per-adapter instance
> > + * @sdev: pointer to SCSI device
> > + * @b_lu_write_protect: bLUWriteProtect value read from LU
> > descriptor
> > + */
> > +static inline void ufshcd_lu_power_on_wp_init(struct ufs_hba *hba,
> > const struct scsi_device *sdev,
> > +                                             u8
> > b_lu_write_protect)
> > +{
> > +       if (hba->dev_info.f_power_on_wp_en && !hba-
> > >dev_info.is_lu_power_on_wp &&
> > +           b_lu_write_protect == UFS_LU_POWER_ON_WP)
> > +               hba->dev_info.is_lu_power_on_wp = true;
> > +}
> 
> The body of this function is only three lines long and this function
> is 
> only called once. Are you sure that you want a separate function
> instead 
> of inlining this function in its only caller?
> 
> > +static void ufshcd_lu_init(struct ufs_hba *hba, struct scsi_device
> > *sdev)
> > +{
> > +       int ret;
> > +       int len;
> > +       u8 lun;
> > +       u8 lun_qdepth;
> > +       u8 *desc_buf;
> 
> Most kernel developers these days order local variable declarations
> from 
> longest to shortest line ("reverse Christmas tree").
> 
> > +       lun_qdepth = hba->nutrs;
> > +       lun = ufshcd_scsi_to_upiu_lun(sdev->lun);
> > +       len = hba->desc_size[QUERY_DESC_IDN_UNIT];
> > +
> > +       desc_buf = kmalloc(len, GFP_KERNEL);
> > +       if (!desc_buf)
> > +               goto set_qdepth;
> > +
> > +       ret = ufshcd_read_unit_desc_param(hba, lun, 0, desc_buf,
> > len);
> > +       if (ret == -EOPNOTSUPP)
> > +               /* If LU doesn't support unit descriptor, its queue
> > depth is set to 1 */
> > +               lun_qdepth = 1;
> > +       else if (desc_buf[UNIT_DESC_PARAM_LU_Q_DEPTH])
> > +               lun_qdepth = min_t(int,
> > desc_buf[UNIT_DESC_PARAM_LU_Q_DEPTH], hba->nutrs);
> 
> ufshcd_read_unit_desc_param() can return fewer bytes than requested.
> How 
> about modifying ufshcd_read_unit_desc_param() such that it returns
> the 
> number of bytes that has been copied and using that return value
> above 
> to check whether at least UNIT_DESC_PARAM_LU_Q_DEPTH bytes have been 
> initialized in desc_buf?
> 
> > +       /*
> > +        * According to UFS device spec, The write protection mode
> > is only supported by normal LU,
> > +        * not supported by WLUN.
> > +        */
> > +       if (!ret && lun < hba->dev_info.max_lu_supported)
> > +               ufshcd_lu_power_on_wp_init(hba, sdev,
> > desc_buf[UNIT_DESC_PARAM_LU_WR_PROTECT]);
> 
> Please insert an if (ret < 0) check after the 
> ufshcd_read_unit_desc_param() call and jump to the kfree() statement
> if 
> ret < 0 instead of checking several times whether or not ret < 0.
> 
> Thanks,
> 
> Bart.