From:   "Fabio M. De Francesco" <fmdefrancesco@gmail.com>
To:     syzbot <syzbot+355c68b459d1d96c4d06@syzkaller.appspotmail.com>,
        syzkaller-bugs@googlegroups.com, Hillf Danton <hdanton@sina.com>
Cc:     linux-kernel@vger.kernel.org, syzkaller-bugs@googlegroups.com
Subject: Re: [syzbot] WARNING in usb_tx_block/usb_submit_urb
Date:   Thu, 16 Feb 2023 07:54:08 +0100
Message-ID: <2136128.irdbgypaU6@suse>
In-Reply-To: <20230215110515.3833-1-hdanton@sina.com>
References: <20230215110515.3833-1-hdanton@sina.com>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="iso-8859-1"
Precedence: bulk

On mercoled=EC 15 febbraio 2023 12:05:15 CET Hillf Danton wrote:
> On Tue, 14 Feb 2023 23:00:47 -0800
>=20
> > syzbot found the following issue on:
> >=20
> > HEAD commit:    f87b564686ee dt-bindings: usb: amlogic,meson-g12a-usb-
ctrl..
> > git tree:     =20
> > https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git usb-test=
ing
> > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=3D1670f2b3480=
000
> Kill urb in flight after submitting it.
>=20
> #syz test https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git=
=20
> f87b564686ee
>=20
> --- x/drivers/net/wireless/marvell/libertas/if_usb.c
> +++ y/drivers/net/wireless/marvell/libertas/if_usb.c
> @@ -763,9 +763,7 @@ static int if_usb_issue_boot_command(str
>  	memset(bootcmd->pad, 0, sizeof(bootcmd->pad));
>=20
>  	/* Issue command */
> -	usb_tx_block(cardp, cardp->ep_out_buf, sizeof(*bootcmd));
> -
> -	return 0;
> +	return usb_tx_block(cardp, cardp->ep_out_buf, sizeof(*bootcmd));
>  }
>=20
>=20
> @@ -853,10 +851,12 @@ restart:
>  	}

>=20
>  	cardp->bootcmdresp =3D 0;
> +	ret =3D if_usb_issue_boot_command(cardp, BOOT_CMD_FW_BY_USB);
> +	if (ret)
> +		goto done;

I think that you are changing the logic here (please read below)...

>  	do {
>  		int j =3D 0;
>  		i++;
> -		if_usb_issue_boot_command(cardp, BOOT_CMD_FW_BY_USB);

Don't we need to call if_usb_issue_boot_command() in a loop in order to ret=
ry=20
the command?

>  		/* wait for command response */
>  		do {
>  			j++;
> @@ -864,6 +864,8 @@ restart:
>  		} while (cardp->bootcmdresp =3D=3D 0 && j < 10);
>  	} while (cardp->bootcmdresp =3D=3D 0 && i < 5);
>=20
> +	usb_kill_urb(cardp->tx_urb);
> +

I'm not an expert in the USB core, anyway calling usb_kill_urb() looks good=
 to=20
me, but I think we should call it after each call of=20
if_usb_issue_boot_command() in the above outer loop.

>  	if (cardp->bootcmdresp =3D=3D BOOT_CMD_RESP_NOT_SUPPORTED) {
>  		/* Return to normal operation */
>  		ret =3D -EOPNOTSUPP;
> --

Can the following work?

#syz test https://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb.git=20
f87b564686ee

diff --git a/drivers/net/wireless/marvell/libertas/if_usb.c b/drivers/net/
wireless/marvell/libertas/if_usb.c
index 20436a289d5c..626357d0c7b0 100644
=2D-- a/drivers/net/wireless/marvell/libertas/if_usb.c
+++ b/drivers/net/wireless/marvell/libertas/if_usb.c
@@ -859,6 +859,7 @@ static void if_usb_prog_firmware(struct lbs_private *pr=
iv,=20
int ret,
                        j++;
                        msleep_interruptible(100);
                } while (cardp->bootcmdresp =3D=3D 0 && j < 10);
+               usb_kill_urb(cardp->tx_urb):
        } while (cardp->bootcmdresp =3D=3D 0 && i < 5);
=20
        if (cardp->bootcmdresp =3D=3D BOOT_CMD_RESP_NOT_SUPPORTED) {
=2D-

Thanks,

=46abio