public inbox for linux-block@vger.kernel.org
 help / color / mirror / Atom feed
From: Ming Lei <ming.lei@redhat.com>
To: Alexander Atanasov <alex@zazolabs.com>
Cc: Jens Axboe <axboe@kernel.dk>,
	linux-block@vger.kernel.org, Yoav Cohen <yoav@nvidia.com>
Subject: Re: [PATCH] ublk: cleanup flag handling and fix UBLK_F_NO_AUTO_PART_SCAN
Date: Fri, 13 Feb 2026 11:48:18 +0800	[thread overview]
Message-ID: <aY6fAlXmzQZCKwmM@fedora> (raw)
In-Reply-To: <20260212152553.1737373-1-alex@zazolabs.com>

On Thu, Feb 12, 2026 at 03:25:52PM +0000, Alexander Atanasov wrote:
> Currently the code misuse GD_SUPPRESS_PART_SCAN flag
> as it tries to use it as a switch for the auto partition scan.
> When UBLK_F_NO_AUTO_PART_SCAN is not passed a proper
> auto partition scanning is not done because the first
> call to ublk_partition_scan_work will clear the bit
> GD_SUPPRESS_PART_SCAN but not actually load the partitions.
> 
> How it should work:
> Unprivileged daemons - never scan partitions, they have
> GD_SUPPRESS_PART_SCAN always set - they are marked as devices
> without partitions. While ioctl(BLKRRPART) does require
> CAP_SYS_ADMIN, block layer have its own a check in
> disk_has_partscan(...) if partitions are supported.
> 
> Trusted daemons - auto scan partitions by default
> and they do not have the bit set, so they do have partitions.
> 
> For trusted daemons auto partitions scan is performed but it can
> be skipped if user have set UBLK_F_NO_AUTO_PART_SCAN flag.
> 
> Rework the code to work as described above:
> - remove checks in ublk_partition_scan_work and rely on
> the caller to schedule the work only if it has to be done.
> - keep GD_SUPPRESS_PART_SCAN always set on unprivileged
> - keep GD_SUPPRESS_PART_SCAN always clear on trusted
> - Properly handle user request to skip auto partitioning scanning
> 
> Fixes: 8443e2087e70 ("ublk: add UBLK_F_NO_AUTO_PART_SCAN feature flag")
> Signed-off-by: Alexander Atanasov <alex@zazolabs.com>
> ---
>  drivers/block/ublk_drv.c | 14 ++++----------
>  1 file changed, 4 insertions(+), 10 deletions(-)
> 
> Cc: Yoav Cohen <yoav@nvidia.com>
> 
> diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
> index 3c918db4905c..e662c6e8e722 100644
> --- a/drivers/block/ublk_drv.c
> +++ b/drivers/block/ublk_drv.c
> @@ -2361,14 +2361,9 @@ static void ublk_partition_scan_work(struct work_struct *work)
>  	if (!disk)
>  		return;
>  
> -	if (WARN_ON_ONCE(!test_and_clear_bit(GD_SUPPRESS_PART_SCAN,
> -					     &disk->state)))
> -		goto out;
> -

It is wrong to remove clearing GD_SUPPRESS_PART_SCAN, otherwise ioctl(BLKRRPART)
can't succeed.

>  	mutex_lock(&disk->open_mutex);
>  	bdev_disk_changed(disk, false);
>  	mutex_unlock(&disk->open_mutex);
> -out:
>  	ublk_put_disk(disk);
>  }
>  
> @@ -4429,12 +4424,11 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub,
>  
>  	set_bit(UB_STATE_USED, &ub->state);
>  
> -	/* Skip partition scan if disabled by user */
> -	if (ub->dev_info.flags & UBLK_F_NO_AUTO_PART_SCAN) {
> +	if (!ub->unprivileged_daemons) {
> +		/* Enable partition scanning for trusted daemons */
>  		clear_bit(GD_SUPPRESS_PART_SCAN, &disk->state);
> -	} else {
> -		/* Schedule async partition scan for trusted daemons */
> -		if (!ub->unprivileged_daemons)
> +		/* Skip auto partition scan if requested by user */
> +		if (!(ub->dev_info.flags & UBLK_F_NO_AUTO_PART_SCAN))
>  			schedule_work(&ub->partition_scan_work);

We may switch to disallow partition scan permanently for UBLK_F_NO_AUTO_PART_SCAN
per discussion in the following thread:

https://lore.kernel.org/linux-block/0535f4dd-ada3-414a-84c6-7abc232aa670@kernel.dk/

So the change may can be simplified as below, meantime
selftests/ublk/test_part_01.sh need to be updated.

Alexander, please let us know if you'd like work this way?

diff --git a/include/uapi/linux/ublk_cmd.h b/include/uapi/linux/ublk_cmd.h
index a88876756805..d6f479fff75b 100644
--- a/include/uapi/linux/ublk_cmd.h
+++ b/include/uapi/linux/ublk_cmd.h
@@ -367,8 +367,8 @@
  */
 #define UBLK_F_SAFE_STOP_DEV   (1ULL << 17)

-/* Disable automatic partition scanning when device is started */
-#define UBLK_F_NO_AUTO_PART_SCAN (1ULL << 18)
+/* Disable partition scanning */
+#define UBLK_F_NO_PART_SCAN    (1ULL << 18)

 /* device state */
 #define UBLK_S_DEV_DEAD        0
[root@ktest-40 linux]# git diff
diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c
index c13cda58a7c6..f1a104222fb1 100644
--- a/drivers/block/ublk_drv.c
+++ b/drivers/block/ublk_drv.c
@@ -81,7 +81,7 @@
                | (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) ? UBLK_F_INTEGRITY : 0) \
                | UBLK_F_SAFE_STOP_DEV \
                | UBLK_F_BATCH_IO \
-               | UBLK_F_NO_AUTO_PART_SCAN)
+               | UBLK_F_NO_PART_SCAN)

 #define UBLK_F_ALL_RECOVERY_FLAGS (UBLK_F_USER_RECOVERY \
                | UBLK_F_USER_RECOVERY_REISSUE \
@@ -4438,14 +4438,13 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub,

        set_bit(UB_STATE_USED, &ub->state);

-       /* Skip partition scan if disabled by user */
-       if (ub->dev_info.flags & UBLK_F_NO_AUTO_PART_SCAN) {
-               clear_bit(GD_SUPPRESS_PART_SCAN, &disk->state);
-       } else {
-               /* Schedule async partition scan for trusted daemons */
-               if (!ub->unprivileged_daemons)
-                       schedule_work(&ub->partition_scan_work);
-       }
+       /*
+        * Only schedule async partition scan in case of
+        * !UBLK_F_NO_PART_SCAN and trusted daemons
+        */
+       if (!(ub->dev_info.flags & UBLK_F_NO_PART_SCAN) &&
+                       !ub->unprivileged_daemons)
+               schedule_work(&ub->partition_scan_work);

 out_put_cdev:
        if (ret) {
diff --git a/include/uapi/linux/ublk_cmd.h b/include/uapi/linux/ublk_cmd.h
index a88876756805..d6f479fff75b 100644
--- a/include/uapi/linux/ublk_cmd.h
+++ b/include/uapi/linux/ublk_cmd.h
@@ -367,8 +367,8 @@
  */
 #define UBLK_F_SAFE_STOP_DEV   (1ULL << 17)

-/* Disable automatic partition scanning when device is started */
-#define UBLK_F_NO_AUTO_PART_SCAN (1ULL << 18)
+/* Disable partition scanning */
+#define UBLK_F_NO_PART_SCAN    (1ULL << 18)

 /* device state */
 #define UBLK_S_DEV_DEAD        0



Thanks,
Ming


  reply	other threads:[~2026-02-13  3:48 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-02-12 15:25 [PATCH] ublk: cleanup flag handling and fix UBLK_F_NO_AUTO_PART_SCAN Alexander Atanasov
2026-02-13  3:48 ` Ming Lei [this message]
2026-02-13  6:56   ` Hannes Reinecke
2026-02-13 11:05   ` Alexander Atanasov
2026-02-20  3:58     ` Ming Lei
2026-02-21 13:00       ` Alexander Atanasov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=aY6fAlXmzQZCKwmM@fedora \
    --to=ming.lei@redhat.com \
    --cc=alex@zazolabs.com \
    --cc=axboe@kernel.dk \
    --cc=linux-block@vger.kernel.org \
    --cc=yoav@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox