Re: [PATCH v7 4/6] scsi: add BLIST_NO_LUN_1F blacklist flag

Re: [PATCH v7 4/6] scsi: add BLIST_NO_LUN_1F blacklist flag

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Unsynchronized read-modify-write of bitfields in `struct scsi_target` (`pdt_1f_for_no_lun`) causes a data race that can corrupt concurrent updates to `expecting_lun_change`.
--

commit 22aca9342fbe7619010508e9209d9f91ff554f8a
Author: Phil Pemberton <philpem@philpem.me.uk>

scsi: add BLIST_NO_LUN_1F blacklist flag

This commit adds a new blacklist flag, BLIST_NO_LUN_1F, to handle multi-LUN
devices that respond to INQUIRY on unpopulated LUNs with PQ=0 / PDT=0x1f. It
wires this flag up in scsi_probe_and_add_lun() to set the pdt_1f_for_no_lun
field on the scsi_target.

> diff --git a/drivers/scsi/scsi_scan.c b/drivers/scsi/scsi_scan.c
> index ef22a4228b855..ef0f5ee6be87d 100644
> --- a/drivers/scsi/scsi_scan.c
> +++ b/drivers/scsi/scsi_scan.c
> @@ -1285,6 +1285,9 @@ static int scsi_probe_and_add_lun(struct scsi_target *starget,
>  	 * PDT=00h Direct-access device (floppy)
>  	 * PDT=1Fh none (no FDD connected to the requested logical unit)
>  	 */
> +	if (bflags & BLIST_NO_LUN_1F)
> +		starget->pdt_1f_for_no_lun = 1;

[Severity: High]
Does this unprotected bitfield assignment risk a data race?

In include/scsi/scsi_device.h, pdt_1f_for_no_lun shares a memory word with
expecting_lun_change in struct scsi_target.

If this function processes LUN 1 while LUN 0 has already been added and is
processing I/O, a concurrent UNIT_ATTENTION could trigger an update to
expecting_lun_change from softirq context in
drivers/scsi/scsi_error.c:scsi_report_lun_change():

	sdev->sdev_target->expecting_lun_change = 1;

Since bitfield assignments are implemented as non-atomic read-modify-write
sequences by the compiler, could this process-context write to
pdt_1f_for_no_lun silently overwrite and drop the softirq's update to
expecting_lun_change?

> +
>  	if (((result[0] >> 5) == 1 || starget->pdt_1f_for_no_lun) &&
>  	    (result[0] & 0x1f) == 0x1f &&
>  	    !scsi_is_wlun(lun)) {

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260611024356.2769320-1-philpem@philpem.me.uk?part=4