Date: Wed, 10 Nov 2010 13:45:14 -0800 From: Pyun YongHyeon <pyunyh@gmail.com> To: M?rcio Luciano Donada <mdonada@auroraalimentos.com.br> Cc: freebsd-stable@freebsd.org Subject: Re: watchdog Message-ID: <20101110214514.GB13340@michelle.cdnetworks.com> In-Reply-To: <4CD9AA69.7070408@auroraalimentos.com.br> References: <4CD92E25.7060304@auroraalimentos.com.br> <20101109192419.GB7766@michelle.cdnetworks.com> <4CD9AA69.7070408@auroraalimentos.com.br>
next in thread | previous in thread | raw e-mail | index | archive | help
--azLHFNyN32YCQGCU
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
On Tue, Nov 09, 2010 at 06:09:13PM -0200, M?rcio Luciano Donada wrote:
> Em 9/11/2010 17:24, Pyun YongHyeon escreveu:
> > On Tue, Nov 09, 2010 at 09:19:01AM -0200, M?rcio Luciano Donada wrote:
> >> Hi list,
> >>
> >> I'm using FreeBSD (uname is below) and these days I'm caught in the log
> >> message below:
> >>
> >> vpn# uname -a
> >> FreeBSD vpn.auroraalimentos.com.br 7.3-STABLE FreeBSD 7.3-STABLE #0: Sat
> >> Jun 19 18:39:18 BRT 2010
> >> root@vpn.xxx.com.br:/usr/obj/usr/src/sys/VPN i386
> >>
> >> Nov 8 11:53:55 vpn kernel: xl0: watchdog timeout (missed Tx interrupts)
> >> -- recovering
> >> Nov 8 12:01:21 vpn kernel: xl0: watchdog timeout (missed Tx interrupts)
> >> -- recovering
> >> Nov 8 12:21:55 vpn kernel: xl0: watchdog timeout (missed Tx interrupts)
> >> -- recovering
> >> Nov 8 16:11:34 vpn kernel: xl0: watchdog timeout (missed Tx interrupts)
> >> -- recovering
> >>
> >> it is possible that the network interface (xl0) is in trouble?
> >>
> >
> > If my memory serve me right there were a couple of reports about
> > xl(4) watchdog timeout in past. The message was added by me to
> > track down possible cause of watchdog timeouts. The message means
> > driver lost Tx completion interrupts but driver noticed that the
> > queued frames were successfully transmitted such that it didn't
> > reset the controller and showed the informational message.
> >
> > To further narrow down the issue, would you show me the following
> > information?
> > o full 'dmesg' output
> > o 'pciconf -lcbv' output
> > o 'devinfo -rv | grep phy' output
>
>
> Hi,
> here's the information you requested:
>
Thanks for the information.
If there is easy way to reproduce it would you let me know?
I've cleaned up some part of xl(4) which looks wrong to me but I
am not sure whether this can address the issue you reported.
Anyway, would you try attached patch?
--azLHFNyN32YCQGCU
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment; filename="xl.watchdog.diff"
Index: sys/dev/xl/if_xl.c
===================================================================
--- sys/dev/xl/if_xl.c (revision 215102)
+++ sys/dev/xl/if_xl.c (working copy)
@@ -1872,6 +1872,8 @@
XL_LOCK_ASSERT(sc);
+ bus_dmamap_sync(sc->xl_ldata.xl_rx_tag, sc->xl_ldata.xl_rx_dmamap,
+ BUS_DMASYNC_POSTREAD);
pos = sc->xl_cdata.xl_rx_head;
for (i = 0; i < XL_RX_LIST_CNT; i++) {
@@ -2227,7 +2229,8 @@
}
#endif
- while ((status = CSR_READ_2(sc, XL_STATUS)) & XL_INTRS &&
+ while (ifp->if_drv_flags & IFF_DRV_RUNNING &&
+ (status = CSR_READ_2(sc, XL_STATUS)) & XL_INTRS &&
status != 0xFFFF) {
CSR_WRITE_2(sc, XL_COMMAND,
XL_CMD_INTR_ACK|(status & XL_INTRS));
@@ -2260,14 +2263,12 @@
xl_init_locked(sc);
}
- if (status & XL_STAT_STATSOFLOW) {
- sc->xl_stats_no_timeout = 1;
+ if (status & XL_STAT_STATSOFLOW)
xl_stats_update_locked(sc);
- sc->xl_stats_no_timeout = 0;
- }
}
- if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
+ if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
+ !IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
if (sc->xl_type == XL_TYPE_905B)
xl_start_90xB_locked(ifp);
else
@@ -2331,11 +2332,8 @@
xl_init_locked(sc);
}
- if (status & XL_STAT_STATSOFLOW) {
- sc->xl_stats_no_timeout = 1;
+ if (status & XL_STAT_STATSOFLOW)
xl_stats_update_locked(sc);
- sc->xl_stats_no_timeout = 0;
- }
}
}
return (rx_npkts);
@@ -2349,13 +2347,19 @@
xl_stats_update(void *xsc)
{
struct xl_softc *sc = xsc;
+ struct mii_data *mii;
XL_LOCK_ASSERT(sc);
- if (xl_watchdog(sc) == EJUSTRETURN)
- return;
+ if (sc->xl_miibus != NULL) {
+ mii = device_get_softc(sc->xl_miibus);
+ mii_tick(mii);
+ }
xl_stats_update_locked(sc);
+
+ xl_watchdog(sc);
+ callout_reset(&sc->xl_stat_callout, hz, xl_stats_update, sc);
}
static void
@@ -2365,15 +2369,11 @@
struct xl_stats xl_stats;
u_int8_t *p;
int i;
- struct mii_data *mii = NULL;
XL_LOCK_ASSERT(sc);
bzero((char *)&xl_stats, sizeof(struct xl_stats));
- if (sc->xl_miibus != NULL)
- mii = device_get_softc(sc->xl_miibus);
-
p = (u_int8_t *)&xl_stats;
/* Read all the stats registers. */
@@ -2395,14 +2395,7 @@
*/
XL_SEL_WIN(4);
CSR_READ_1(sc, XL_W4_BADSSD);
-
- if ((mii != NULL) && (!sc->xl_stats_no_timeout))
- mii_tick(mii);
-
XL_SEL_WIN(7);
-
- if (!sc->xl_stats_no_timeout)
- callout_reset(&sc->xl_stat_callout, hz, xl_stats_update, sc);
}
/*
@@ -2939,9 +2932,7 @@
/* Clear out the stats counters. */
CSR_WRITE_2(sc, XL_COMMAND, XL_CMD_STATS_DISABLE);
- sc->xl_stats_no_timeout = 1;
xl_stats_update_locked(sc);
- sc->xl_stats_no_timeout = 0;
XL_SEL_WIN(4);
CSR_WRITE_2(sc, XL_W4_NET_DIAG, XL_NETDIAG_UPPER_BYTES_ENABLE);
CSR_WRITE_2(sc, XL_COMMAND, XL_CMD_STATS_ENABLE);
Index: sys/dev/xl/if_xlreg.h
===================================================================
--- sys/dev/xl/if_xlreg.h (revision 215102)
+++ sys/dev/xl/if_xlreg.h (working copy)
@@ -614,7 +614,6 @@
u_int32_t xl_xcvr;
u_int16_t xl_media;
u_int16_t xl_caps;
- u_int8_t xl_stats_no_timeout;
u_int16_t xl_tx_thresh;
int xl_pmcap;
int xl_if_flags;
--azLHFNyN32YCQGCU--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20101110214514.GB13340>