Skip to contentSkip to conversation

NETDEV WATCHDOG on U60/SMP

DISCUSSION BERTRAND Joël 3 participants 3 messages Jun 20, 2008

lore
Share & tools
lore
#0 lore raw

Hello,

This mail comes from sparclinux mailing list. I repost it on general
linux kernel mailing list because I'm not sure that this bug is sparc specific. Nevertheless, I can only reproduce it on sparc64/SMP.

My U60 runs linux debian with official 2.6.25 linux kernel (I'm
currently trying 2.6.25.7) and sometimes, when eth2 is stressed, eth2 hangs with NETDEV WATCHDOG :

NETDEV WATCHDOG: eth2: transmit timed out
eth2: transmit timed out, tx_status 00 status 8601.
diagnostics: net 0ccc media 8880 dma 0000003a fifo 0000
eth2: Interrupt posted but not delivered -- IRQ blocked by another device?
Flags; bus-master 1, dirty 2283344(0) current 2283344(0) Transmit list 00000000 vs. fffff800af098200.

  0: @fffff800af098200  length 00000042 status 0c01059a
  1: @fffff800af098260  length 00000042 status 0c01059a
  2: @fffff800af0982c0  length 00000042 status 0c01059a
  3: @fffff800af098320  length 00000042 status 0c01059a
  4: @fffff800af098380  length 00000042 status 0c01059a
  5: @fffff800af0983e0  length 00000042 status 0c01059a
  6: @fffff800af098440  length 00000042 status 0c01059a
  7: @fffff800af0984a0  length 00000042 status 0c01059a
  8: @fffff800af098500  length 8000002a status 0001002a
  9: @fffff800af098560  length 8000002a status 0001002a
  10: @fffff800af0985c0  length 8000002a status 0001002a
  11: @fffff800af098620  length 8000002a status 0001002a
  12: @fffff800af098680  length 8000002a status 0001002a
  13: @fffff800af0986e0  length 8000002a status 0001002a
  14: @fffff800af098740  length 8000002a status 8001002a
  15: @fffff800af0987a0  length 8000002a status 8001002a

eth2: Resetting the Tx ring pointer.
eth2: setting full-duplex.
NETDEV WATCHDOG: eth2: transmit timed out
eth2: transmit timed out, tx_status 00 status 8601.
diagnostics: net 0ccc media 8880 dma 0000003a fifo 0000
eth2: Interrupt posted but not delivered -- IRQ blocked by another device?
Flags; bus-master 1, dirty 16(0) current 16(0)
Transmit list 00000000 vs. fffff800af098200.

  0: @fffff800af098200  length 8000002a status 0001002a
  1: @fffff800af098260  length 8000002a status 0001002a
  2: @fffff800af0982c0  length 8000002a status 0001002a
  3: @fffff800af098320  length 8000002a status 0001002a
  4: @fffff800af098380  length 8000002a status 0001002a
  5: @fffff800af0983e0  length 8000002a status 0001002a
  6: @fffff800af098440  length 8000002a status 0001002a
  7: @fffff800af0984a0  length 8000002a status 0001002a
  8: @fffff800af098500  length 8000002a status 0001002a
  9: @fffff800af098560  length 8000002a status 0001002a
  10: @fffff800af0985c0  length 8000002a status 0001002a
  11: @fffff800af098620  length 8000002a status 0001002a
  12: @fffff800af098680  length 8000002a status 0001002a
  13: @fffff800af0986e0  length 8000002a status 0001002a
  14: @fffff800af098740  length 8000002a status 8001002a
  15: @fffff800af0987a0  length 8000002a status 8001002a

eth2: Resetting the Tx ring pointer.
eth2: setting full-duplex.
...

I have to reboot this server to restore eth2.
This adapter is a 3Com NIC (3C905). I have tried with several different
3Com adapters with the same result. If I change this NIC (for example with a HME or any PCI 2.1 adapter), I cannot reproduce the bug.

It only occurs when ethernet traffic is high on eth2.

I have seen this bug since 2.6.20 even on amd64 (but I'm not sure that
this bug remains in amd64 kernel because I don't have any amd64 workstation to test, and I don't see it on amd64 since 2.6.24. Maybe it is fixed on amd64...).

lspci returns :
0000:00:00.0 Host bridge: Sun Microsystems Computer Corp. Psycho PCI Bus Module
0000:00:01.0 Bridge: Sun Microsystems Computer Corp. EBUS (rev 01) 0000:00:01.1 Ethernet controller: Sun Microsystems Computer Corp. Happy Meal 10/100 Ethernet [hme] (rev 01)
0000:00:02.0 Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 78)
0000:00:03.0 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14)
0000:00:03.1 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14)
0000:00:04.0 SCSI storage controller: Adaptec AIC-7892A U160/m (rev 02) 0000:00:05.0 USB Controller: NEC Corporation USB (rev 43) 0000:00:05.1 USB Controller: NEC Corporation USB (rev 43) 0000:00:05.2 USB Controller: NEC Corporation USB 2.0 (rev 04) 0001:00:00.0 Host bridge: Sun Microsystems Computer Corp. Psycho PCI Bus Module
0001:80:01.0 Bridge: Sun Microsystems Computer Corp. EBUS (rev 01) 0001:80:01.1 Ethernet controller: Sun Microsystems Computer Corp. Happy Meal 10/100 Ethernet [hme] (rev 01)

ifconfig:
eth0 Link encap:Ethernet HWaddr 08:00:20:a1:4b:33

          inet adr:192.168.0.128  Bcast:192.168.0.255  Masque:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:16709366 errors:0 dropped:0 overruns:0 frame:1
          TX packets:21355942 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 lg file transmission:1000
          RX bytes:2391901923 (2.2 GiB)  TX bytes:21605391421 (20.1 GiB)
          Interruption:14 Adresse de base:0x3000

eth1 Link encap:Ethernet HWaddr 08:00:20:a1:4b:33

          inet adr:192.168.254.1  Bcast:192.168.254.255
Masque:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:20207169 errors:0 dropped:0 overruns:0 frame:0
          TX packets:17280402 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 lg file transmission:1000
          RX bytes:19068335140 (17.7 GiB)  TX bytes:8246313479 (7.6 GiB)
          Interruption:24 Adresse de base:0x1800

eth2 Link encap:Ethernet HWaddr 00:04:75:df:1c:6d

          inet adr:192.168.253.1  Bcast:192.168.253.255
Masque:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:1843643 errors:0 dropped:0 overruns:0 frame:0
          TX packets:2416959 errors:13 dropped:0 overruns:0 carrier:0
          collisions:0 lg file transmission:1000
          RX bytes:157416047 (150.1 MiB)  TX bytes:2313298605 (2.1 GiB)
          Interruption:17 Adresse de base:0x8000

lo Link encap:Boucle locale

          inet adr:127.0.0.1  Masque:255.0.0.0
          UP LOOPBACK RUNNING  MTU:16436  Metric:1
          RX packets:7839862 errors:0 dropped:0 overruns:0 frame:0
          TX packets:7839862 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 lg file transmission:0
          RX bytes:3713209874 (3.4 GiB)  TX bytes:3713209874 (3.4 GiB)

Interruptions:
CPU0 CPU2

  0: 1253580857 1253580260     <NULL>  timer
  1:          0          0      sun4u  PSYCHO_PCIERR
  2:          0          0      sun4u  PSYCHO_UE
  3:          0          0      sun4u  PSYCHO_CE
  8:     733411          0      sun4u  su(kbd)
  9:          0    4396224      sun4u  su(mouse)
 10:          0          0      sun4u  parport0
 11:          4          0      sun4u  floppy
 12:          0          0      sun4u  cs4231(capture)
 13:          0          0      sun4u  cs4231(play)
 14:          0   37976886      sun4u  eth0
 15:          0  218660455      sun4u  sym53c8xx
 16:         30          0      sun4u  sym53c8xx
 17:    2042976    2011664      sun4u  eth2
 18:  137883796          0      sun4u  aic7xxx
 19:          0    1208028      sun4u  ohci_hcd:usb2
 20:          0     650947      sun4u  ohci_hcd:usb3
 21:          1          4      sun4u  ehci_hcd:usb1
 22:          0          0      sun4u  PSYCHO_PCIERR
 24:    4957716   33460983      sun4u  eth1

Any idea ?

Regards,

JKB

#1 lore raw
Quoted 32 lines · BERTRAND Jo?l

Hello,

This mail comes from sparclinux mailing list. I repost it on general
linux kernel mailing list because I'm not sure that this bug is sparc specific. Nevertheless, I can only reproduce it on sparc64/SMP.

My U60 runs linux debian with official 2.6.25 linux kernel (I'm
currently trying 2.6.25.7) and sometimes, when eth2 is stressed, eth2 hangs with NETDEV WATCHDOG :

NETDEV WATCHDOG: eth2: transmit timed out
eth2: transmit timed out, tx_status 00 status 8601.
diagnostics: net 0ccc media 8880 dma 0000003a fifo 0000
eth2: Interrupt posted but not delivered -- IRQ blocked by another device?
Flags; bus-master 1, dirty 2283344(0) current 2283344(0) Transmit list 00000000 vs. fffff800af098200.

  0: @fffff800af098200  length 00000042 status 0c01059a
  1: @fffff800af098260  length 00000042 status 0c01059a
  2: @fffff800af0982c0  length 00000042 status 0c01059a
  3: @fffff800af098320  length 00000042 status 0c01059a
  4: @fffff800af098380  length 00000042 status 0c01059a
  5: @fffff800af0983e0  length 00000042 status 0c01059a
  6: @fffff800af098440  length 00000042 status 0c01059a
  7: @fffff800af0984a0  length 00000042 status 0c01059a
  8: @fffff800af098500  length 8000002a status 0001002a
  9: @fffff800af098560  length 8000002a status 0001002a
  10: @fffff800af0985c0  length 8000002a status 0001002a
  11: @fffff800af098620  length 8000002a status 0001002a
  12: @fffff800af098680  length 8000002a status 0001002a
  13: @fffff800af0986e0  length 8000002a status 0001002a
  14: @fffff800af098740  length 8000002a status 8001002a
  15: @fffff800af0987a0  length 8000002a status 8001002a
Quoted 2 lines

eth2: Resetting the Tx ring pointer.
eth2: setting full-duplex.

Some people with similar problems reported, that they can workarround their problems by increasing the rx/tx ring sizes of the 3c59x driver. See http://bugzilla.kernel.org/show_bug.cgi?id=6444 for details.

Would be good to know whether this helps for your problem too.

Steffen

#2 lore raw
Quoted 35 lines · Steffen Klassert
Quoted 31 lines · BERTRAND Jo?l

Hello,

This mail comes from sparclinux mailing list. I repost it on general
linux kernel mailing list because I'm not sure that this bug is sparc specific. Nevertheless, I can only reproduce it on sparc64/SMP.

My U60 runs linux debian with official 2.6.25 linux kernel (I'm
currently trying 2.6.25.7) and sometimes, when eth2 is stressed, eth2 hangs with NETDEV WATCHDOG :

NETDEV WATCHDOG: eth2: transmit timed out
eth2: transmit timed out, tx_status 00 status 8601.
diagnostics: net 0ccc media 8880 dma 0000003a fifo 0000
eth2: Interrupt posted but not delivered -- IRQ blocked by another device?
Flags; bus-master 1, dirty 2283344(0) current 2283344(0) Transmit list 00000000 vs. fffff800af098200.

  0: @fffff800af098200  length 00000042 status 0c01059a
  1: @fffff800af098260  length 00000042 status 0c01059a
  2: @fffff800af0982c0  length 00000042 status 0c01059a
  3: @fffff800af098320  length 00000042 status 0c01059a
  4: @fffff800af098380  length 00000042 status 0c01059a
  5: @fffff800af0983e0  length 00000042 status 0c01059a
  6: @fffff800af098440  length 00000042 status 0c01059a
  7: @fffff800af0984a0  length 00000042 status 0c01059a
  8: @fffff800af098500  length 8000002a status 0001002a
  9: @fffff800af098560  length 8000002a status 0001002a
  10: @fffff800af0985c0  length 8000002a status 0001002a
  11: @fffff800af098620  length 8000002a status 0001002a
  12: @fffff800af098680  length 8000002a status 0001002a
  13: @fffff800af0986e0  length 8000002a status 0001002a
  14: @fffff800af098740  length 8000002a status 8001002a
  15: @fffff800af0987a0  length 8000002a status 8001002a
Quoted 2 lines

eth2: Resetting the Tx ring pointer.
eth2: setting full-duplex.

Quoted 3 lines

Some people with similar problems reported, that they can workarround their problems by increasing the rx/tx ring sizes of the 3c59x driver. See http://bugzilla.kernel.org/show_bug.cgi?id=6444 for details.

	Thanks. I'm trying with RX/TX=256/256 and max_interrupt=1024.

	Regards,

	JKB