svn commit: r264864 - head/sys/dev/usb/wlan
Rui Paulo
rpaulo at felyko.com
Tue Apr 29 21:07:25 UTC 2014
On Apr 23, 2014, at 20:16, Kevin Lo <kevlo at freebsd.org> wrote:
> Author: kevlo
> Date: Thu Apr 24 03:16:47 2014
> New Revision: 264864
> URL: http://svnweb.freebsd.org/changeset/base/264864
>
> Log:
> Fix panic by adding mtx_assert() to urtwn_init_locked() and
> urtwn_stop_locked().
This was not a panic, but a LOR.
> Modified:
> head/sys/dev/usb/wlan/if_urtwn.c
>
> Modified: head/sys/dev/usb/wlan/if_urtwn.c
> ==============================================================================
> --- head/sys/dev/usb/wlan/if_urtwn.c Thu Apr 24 02:16:23 2014 (r264863)
> +++ head/sys/dev/usb/wlan/if_urtwn.c Thu Apr 24 03:16:47 2014 (r264864)
> @@ -2054,6 +2054,7 @@ urtwn_load_firmware(struct urtwn_softc *
> uint32_t reg;
> int mlen, ntries, page, error;
>
> + URTWN_UNLOCK(sc);
You didn't need to unlock this early. The problem was in firmware_get().
> /* Read firmware image from the filesystem. */
> if ((sc->chip & (URTWN_CHIP_UMC_A_CUT | URTWN_CHIP_92C)) ==
> URTWN_CHIP_UMC_A_CUT)
> @@ -2062,6 +2063,7 @@ urtwn_load_firmware(struct urtwn_softc *
> imagename = "urtwn-rtl8192cfwT";
>
> fw = firmware_get(imagename);
> + URTWN_LOCK(sc);
> if (fw == NULL) {
> device_printf(sc->sc_dev,
> "failed loadfirmware of file %s\n", imagename);
> @@ -2816,6 +2818,8 @@ urtwn_init_locked(void *arg)
> uint32_t reg;
> int error;
>
> + URTWN_ASSERT_LOCKED(sc);
> +
> if (ifp->if_drv_flags & IFF_DRV_RUNNING)
> urtwn_stop_locked(ifp);
>
> @@ -2979,6 +2983,8 @@ urtwn_stop_locked(struct ifnet *ifp)
> {
> struct urtwn_softc *sc = ifp->if_softc;
>
> + URTWN_ASSERT_LOCKED(sc);
> +
> ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
>
> callout_stop(&sc->sc_watchdog_ch);
This is fine, but I wonder if the firmware API should be changed. The problem with this approach, and all USB WiFi drivers, is that by unlocking the mutex, the mutual exclusion protection is lost...
--
Rui Paulo
More information about the svn-src-all
mailing list