mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: John Garry <john.g.garry@oracle.com>
To: Luis Chamberlain <mcgrof@kernel.org>,
	James.Bottomley@HansenPartnership.com,
	martin.petersen@oracle.com, linux-scsi@vger.kernel.org
Cc: gost.dev@samsung.com, s.prabhu@samsung.com, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] scsi_debug: Fix shared UUID issue when creating multiple hosts
Date: Wed, 2 Jul 2025 10:57:35 +0100	[thread overview]
Message-ID: <a44a8b7d-c8c8-4bc1-b9f1-b8becb966db2@oracle.com> (raw)
In-Reply-To: <20250701220234.1605559-1-mcgrof@kernel.org>

On 01/07/2025 23:02, Luis Chamberlain wrote:
> When creating multiple scsi_debug hosts (e.g., modprobe scsi_debug
> add_host=2), all devices share the same backing storage by default.
> This causes filesystem UUIDs to be shared between devices like /dev/sda
> and /dev/sdb, leading to confusion and potential mount issues.
> 
> For example:
>    # modprobe scsi_debug add_host=2 dev_size_mb=1024
>    # mkfs.xfs -f /dev/sda
>    # wipefs /dev/sda /dev/sdb
> Both devices show the same UUID, which is incorrect behavior.
> 
> Fix this by automatically enabling separate per-host stores when
> multiple hosts are created, while preserving backward compatibility
> for single-host scenarios. Users can still explicitly control this
> behavior via the per_host_store parameter.
> 
> After this fix:
> - Single host: Uses shared store (unchanged behavior)

Does phs even have any affect for only a single scsi_debug host?

I thought that phs means that each host has its own backing storage 
which is shared among all devices for that host.

> - Multiple hosts: Automatically uses separate stores (fixes UUID issue)
> - Explicit per_host_store=1: Always separate stores (unchanged behavior)
 > > Reported-by: Swarna Prabhu <s.prabhu@samsung.com>
> Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
> ---
>   drivers/scsi/scsi_debug.c | 17 ++++++++++++++---
>   1 file changed, 14 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c
> index f0eec4708ddd..4d0b17861adb 100644
> --- a/drivers/scsi/scsi_debug.c
> +++ b/drivers/scsi/scsi_debug.c
> @@ -8034,12 +8034,18 @@ static ssize_t add_host_store(struct device_driver *ddp, const char *buf,
>   	bool found;
>   	unsigned long idx;
>   	struct sdeb_store_info *sip;
> +	bool uniq_uuid_required = false;
>   	bool want_phs = (sdebug_fake_rw == 0) && sdebug_per_host_store;
>   	int delta_hosts;
>   
>   	if (sscanf(buf, "%d", &delta_hosts) != 1)
>   		return -EINVAL;
>   	if (delta_hosts > 0) {
> +		if (delta_hosts > 1) {
> +			uniq_uuid_required = true;
> +			if (sdebug_fake_rw == 0)
> +				want_phs = true;
> +		}
>   		do {
>   			found = false;
>   			if (want_phs) {
> @@ -8054,7 +8060,7 @@ static ssize_t add_host_store(struct device_driver *ddp, const char *buf,
>   				else
>   					sdebug_do_add_host(true);
>   			} else {
> -				sdebug_do_add_host(false);
> +				sdebug_do_add_host(uniq_uuid_required);
>   			}
>   		} while (--delta_hosts);
>   	} else if (delta_hosts < 0) {
> @@ -8383,6 +8389,7 @@ static struct device *pseudo_primary;
>   static int __init scsi_debug_init(void)
>   {
>   	bool want_store = (sdebug_fake_rw == 0);
> +	bool uniq_uuid_required = false;
>   	unsigned long sz;
>   	int k, ret, hosts_to_add;
>   	int idx = -1;
> @@ -8578,6 +8585,9 @@ static int __init scsi_debug_init(void)
>   	}
>   
>   	hosts_to_add = sdebug_add_host;
> +	if (hosts_to_add > 1)
> +		uniq_uuid_required = true;
> +
>   	sdebug_add_host = 0;
>   
>   	sdebug_debugfs_root = debugfs_create_dir("scsi_debug", NULL);
> @@ -8593,8 +8603,9 @@ static int __init scsi_debug_init(void)
>   				break;
>   			}
>   		} else {
> -			ret = sdebug_do_add_host(want_store &&
> -						 sdebug_per_host_store);
> +			bool use_store = want_store &&
> +				(sdebug_per_host_store || uniq_uuid_required);
> +			ret = sdebug_do_add_host(use_store);
>   			if (ret < 0) {
>   				pr_err("add_host k=%d error=%d\n", k, -ret);
>   				break;


      reply	other threads:[~2025-07-02  9:57 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-07-01 22:02 Luis Chamberlain
2025-07-02  9:57 ` John Garry [this message]

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=a44a8b7d-c8c8-4bc1-b9f1-b8becb966db2@oracle.com \
    --to=john.g.garry@oracle.com \
    --cc=James.Bottomley@HansenPartnership.com \
    --cc=gost.dev@samsung.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=martin.petersen@oracle.com \
    --cc=mcgrof@kernel.org \
    --cc=s.prabhu@samsung.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

all inboxes | Powered by JetHome®