Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   "Marc Hartmayer" <mhartmay@linux.ibm.com>
To:     Yu Kuai <yukuai1@huaweicloud.com>
Cc:     linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org,
        yi.zhang@huawei.com, yangerkun@huawei.com, hch@lst.de,
        chaitanyak@nvidia.com, shinichiro.kawasaki@wdc.com,
        dgilbert@interlog.com, jejb@linux.ibm.com,
        martin.petersen@oracle.com, yukuai3@huawei.com, axboe@kernel.dk,
        Benjamin Block <bblock@linux.ibm.com>
Subject: Re: [PATCH] scsi/sg: don't grab scsi host module reference
In-Reply-To: <20230621160111.1433521-1-yukuai1@huaweicloud.com>
References: <20230621160111.1433521-1-yukuai1@huaweicloud.com>
Date:   Tue, 04 Jul 2023 19:04:00 +0200
Message-ID: <87lefv622n.fsf@linux.ibm.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Precedence: bulk

On Thu, Jun 22, 2023 at 12:01 AM +0800, Yu Kuai <yukuai1@huaweicloud.com> w=
rote:
> From: Yu Kuai <yukuai3@huawei.com>
>
> In order to prevent request_queue to be freed before cleaning up
> blktrace debugfs entries, commit db59133e9279 ("scsi: sg: fix blktrace
> debugfs entries leakage") use scsi_device_get(), however,
> scsi_device_get() will also grab scsi module reference and scsi module
> can't be removed.
>
> It's reported that blktests can't unload scsi_debug after block/001:
>
> blktests (master) # ./check block
> block/001 (stress device hotplugging) [failed]
>      +++ /root/blktests/results/nodev/block/001.out.bad 2023-06-19
>       Running block/001
>       Stressing sd
>      +modprobe: FATAL: Module scsi_debug is in use.
>
> Fix this problem by grabbing request_queue reference directly, so that
> scsi host module can still be unloaded while request_queue will be
> pinged by sg device.
>
> Reported-by: Chaitanya Kulkarni <chaitanyak@nvidia.com>
> Link: https://lore.kernel.org/all/1760da91-876d-fc9c-ab51-999a6f66ad50@nv=
idia.com/
> Fixes: db59133e9279 ("scsi: sg: fix blktrace debugfs entries leakage")
> Signed-off-by: Yu Kuai <yukuai3@huawei.com>
> ---
>  drivers/scsi/sg.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/scsi/sg.c b/drivers/scsi/sg.c
> index 2433eeef042a..dcb73787c29d 100644
> --- a/drivers/scsi/sg.c
> +++ b/drivers/scsi/sg.c
> @@ -1497,7 +1497,7 @@ sg_add_device(struct device *cl_dev)
>  	int error;
>  	unsigned long iflags;
>=20=20
> -	error =3D scsi_device_get(scsidp);
> +	error =3D blk_get_queue(scsidp->request_queue);
>  	if (error)
>  		return error;
>=20=20
> @@ -1558,7 +1558,7 @@ sg_add_device(struct device *cl_dev)
>  out:
>  	if (cdev)
>  		cdev_del(cdev);
> -	scsi_device_put(scsidp);
> +	blk_put_queue(scsidp->request_queue);
>  	return error;
>  }
>=20=20
> @@ -1575,7 +1575,7 @@ sg_device_destroy(struct kref *kref)
>  	 */
>=20=20
>  	blk_trace_remove(q);
> -	scsi_device_put(sdp->device);
> +	blk_put_queue(q);
>=20=20
>  	write_lock_irqsave(&sg_index_lock, flags);
>  	idr_remove(&sg_index_idr, sdp->index);
> --=20
> 2.39.2

Hi,

This change (bisected) triggers a regression in our KVM on s390x CI. The
symptom is that a =E2=80=9Cscsi_debug device=E2=80=9D does not bind to the =
scsi_generic
driver. On s390x you can reproduce the problem as follows (I have not
tested on x86):

With this patch applied:

$ sudo modprobe scsi_debug
$ # Get the 'scsi_host,channel,target_number,LUN' tuple for the scsi_debug =
device
$ lsscsi |grep scsi_debug |awk '{ print $1 }'
[0:0:0:0]
$ sudo stat /sys/bus/scsi/devices/0:0:0:0/scsi_generic
stat: cannot statx '/sys/bus/scsi/devices/0:0:0:0/scsi_generic': No such fi=
le or directory


Patch reverted:

$ sudo modprobe scsi_debug
$ lsscsi |grep scsi_debug |awk '{ print $1 }'
[0:0:0:0]
$ sudo stat /sys/bus/scsi/devices/0:0:0:0/scsi_generic
  File: /sys/bus/scsi/devices/0:0:0:0/scsi_generic
  Size: 0         	Blocks: 0          IO Block: 4096   directory
Device: 0,20	Inode: 12155       Links: 3
=E2=80=A6

Any ideas?

 Marc