Linux USB
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Oliver Neukum" <oneukum@suse.com>, linux-usb@vger.kernel.org
Subject: Re: [PATCH v2 net-next 1/1] net: usb: rtl8150: break cycle between tasklet and URB
Date: Thu, 19 Nov 2026 13:36:09 +0000	[thread overview]
Message-ID: <sashiko-outbox-223185@kernel.org> (raw)
In-Reply-To: <20261118133550.955312-1-oneukum@suse.com>

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [High] net: usb: rtl8150: tasklet livelock on poisoned URB in
  rx_fixup()
- [Medium] net: usb: rtl8150: URBs poisoned too late in
  rtl8150_disconnect()
--

commit 1541c4307c346b500c4b1389da4c563d28fd64b6
Author: Oliver Neukum <oneukum@suse.com>

net: usb: rtl8150: break cycle between tasklet and URB

This commit replaces unlink_all_urbs() with terminate_all_urbs() using
usb_poison_urb() in rtl8150_disconnect(), followed by a second
tasklet_kill(), to break the resubmission cycle between URBs and the
tasklet.

> diff --git a/drivers/net/usb/rtl8150.c b/drivers/net/usb/rtl8150.c
> index d51e43170e033..6d20f52d52efd 100644
> --- a/drivers/net/usb/rtl8150.c
> +++ b/drivers/net/usb/rtl8150.c
> @@ -353,6 +353,13 @@ static void free_all_urbs(rtl8150_t * dev)
>  	usb_free_urb(dev->intr_urb);
>  }
>  
> +static void terminate_all_urbs(rtl8150_t *dev)
> +{
> +	usb_poison_urb(dev->rx_urb);
> +	usb_poison_urb(dev->tx_urb);
> +	usb_poison_urb(dev->intr_urb);
> +}
> +
>  static void unlink_all_urbs(rtl8150_t * dev)
>  {
>  	usb_kill_urb(dev->rx_urb);
> @@ -970,7 +977,8 @@ static void rtl8150_disconnect(struct usb_interface *intf)
>  		set_bit(RTL8150_UNPLUG, &dev->flags);
>  		tasklet_kill(&dev->tl);
>  		unregister_netdev(dev->netdev);
> -		unlink_all_urbs(dev);
> +		terminate_all_urbs(dev);

[Severity: Medium]
Can rx_fixup() still run during unregister_netdev() before
terminate_all_urbs() poisons the URBs?

CPU0                                    CPU1
                                        read_bulk_callback()
                                          test_bit(RTL8150_UNPLUG, ...)
rtl8150_disconnect()
  set_bit(RTL8150_UNPLUG, ...)
  tasklet_kill(&dev->tl)
  unregister_netdev(dev->netdev)
    ...
      rtl8150_close()
        unlink_all_urbs(dev)
          usb_kill_urb(dev->rx_urb)
                                          usb_submit_urb() -> -EPERM
                                          tasklet_schedule(&dev->tl)
          /* urb->reject returns to 0 */
                                        rx_fixup()
                                          usb_submit_urb(dev->rx_urb, ...)
  terminate_all_urbs(dev)

Because rtl8150_close() only uses usb_kill_urb(), dev->rx_urb is unpoisoned
again as soon as usb_kill_urb() returns.  If an in-flight
read_bulk_callback() passed the RTL8150_UNPLUG check just before
rtl8150_disconnect() set the bit, its usb_submit_urb() fails with -EPERM
while usb_kill_urb() is waiting and schedules dev->tl.  rx_fixup() can then
run before terminate_all_urbs() and either resubmit dev->rx_urb after
rtl8150_close() or, if usb_submit_urb() returns -ENODEV, call
netif_device_detach() on an unregistered net_device.

Would calling terminate_all_urbs() before the first tasklet_kill() and
unregister_netdev() avoid this window?

> +		tasklet_kill(&dev->tl);

[Severity: High]
Will tasklet_kill(&dev->tl) livelock if dev->tl is scheduled after
dev->rx_urb is poisoned?

drivers/net/usb/rtl8150.c:rx_fixup() {
	...
try_again:
	status = usb_submit_urb(dev->rx_urb, GFP_ATOMIC);
	if (status == -ENODEV) {
		netif_device_detach(dev->netdev);
	} else if (status) {
		set_bit(RX_URB_FAIL, &dev->flags);
		goto tlsched;
	} else {
		clear_bit(RX_URB_FAIL, &dev->flags);
	}

	return;
tlsched:
	tasklet_schedule(&dev->tl);
}

Submitting a poisoned URB returns -EPERM (or -ENOENT once the endpoint is
disabled), not -ENODEV.  Because rx_fixup() does not check RTL8150_UNPLUG
and treats any error other than -ENODEV as retryable, it sets RX_URB_FAIL
and reschedules dev->tl on every run.  On subsequent runs RX_URB_FAIL makes
rx_fixup() jump straight to try_again, get -EPERM from the poisoned
dev->rx_urb, and call tasklet_schedule(&dev->tl) again, spinning in softirq
and hanging tasklet_kill() in rtl8150_disconnect() during driver unbind or
disconnect.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261118133550.955312-1-oneukum@suse.com?part=1

      reply	other threads:[~2026-11-19 13:36 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18 13:35 [PATCH v2 net-next 1/1] net: usb: rtl8150: break cycle between tasklet and URB Oliver Neukum
2026-11-19 13:36 ` sashiko-bot [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=sashiko-outbox-223185@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=linux-usb@vger.kernel.org \
    --cc=oneukum@suse.com \
    --cc=sashiko-reviews@lists.linux.dev \
    /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