Re: ksoftirqd uses 99% CPU triggered by network traffic (maybe RLT-8139 related)

From: Pasi Sjoholm (ptsjohol_at_cc.jyu.fi)
Date: 08/02/04

  • Next message: Dipankar Sarma: "Re: CPU hotplug broken in 2.6.8-rc2 ?" 
    Date:	Mon, 2 Aug 2004 12:57:41 +0300 (EEST)
To: Francois Romieu <romieu@fr.zoreil.com>

    On Sat, 31 Jul 2004, Francois Romieu wrote:

    > Pasi Sjoholm <ptsjohol@cc.jyu.fi> :
    > [interesting report]
    >> The hardest part is to tell where the problem is but I think that
    >> rtl8139_poll-function would be good place to start looking for the bug?
    >> readprofile didn't tell much.. Where to go next? I'm not so good
    >> programmer that I could find the right place to fix..

    > In case it could make a difference: did you check if CONFIG_8139_OLD_RX_RESET
    > changes the behavior or not ?

    Yep, I tried that one also, didn't help a thing.
     
    > If it does not, I'd welcome a test report + log with the two attached patch
    > applied. The first one is just a placebo but the second one could help.

    We are making some sort of a progress but not enough.

    Patch r8139-20.patch didn't help. I made some printk(...)-debug messages
    and that part of a code was never ran.

    With both patch applied and RTL8139DEBUG = 1 I couldn't make the driver
    crash but without DEBUG it did crash. I assume that it has something to
    with fact that syslog did take so much io-bandwidth. (a couple of minutes log
    was ~1GB =))

    But without: 

    --
@@ -2024,17 +2024,17 @@ static int rtl8139_rx(struct net_device
                cur_rx = (cur_rx + rx_size + 4 + 3) & ~3;
                RTL_W16 (RxBufPtr, (u16) (cur_rx - 16));
+       }
-               /* Clear out errors and receive interrupts */
-               status = RTL_R16 (IntrStatus) & RxAckBits;
-               if (likely(status != 0)) {
-                       if (unlikely(status & (RxFIFOOver | RxOverflow))) 
{
-                               tp->stats.rx_errors++;
 -                               if (status & RxFIFOOver)
-                                       tp->stats.rx_fifo_errors++;
-                       }
-                       RTL_W16_F (IntrStatus, RxAckBits);
+       /* Clear out errors and receive interrupts */
+       status = RTL_R16 (IntrStatus) & RxAckBits;
+       if (likely(status != 0)) {
+               if (unlikely(status & (RxFIFOOver | RxOverflow))) {
+                       tp->stats.rx_errors++;
+                       if (status & RxFIFOOver)
+                               tp->stats.rx_fifo_errors++;
                }
+               RTL_W16_F (IntrStatus, RxAckBits);
        }
  done:
--
the driver crashed... even with debug-option was turned on.
Everytime the ksoftirqd started to take cpu-time there were this line in 
the logs:
--
Aug  2 12:10:37 139_interrupt: eth0:.....
--
Notice the 139... it should read rtl8139_interrupt:
Here is a snapshot from the log file when driver is crashing: 
Full logfile is available from: 
http://www.cc.jyu.fi/~ptsjohol/syslog-debug.gz
--
Aug  2 12:10:37 rtl8139_rx: eth0: In rtl8139_rx(), current 03ac BufAddr 036c, free to 039c, Cmd 0c.
Aug  2 12:10:37 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0000.
Aug  2 12:10:37 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0001.
Aug  2 12:10:37 rtl8139_rx: eth0: In rtl8139_rx(), current 0484 BufAddr 0444, free to 0474, Cmd 0c.
Aug  2 12:10:37 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0000.
Aug  2 12:10:37 139_interrupt: eth0: exiting interrupt, intr_status=0x0010.
Aug  2 12:10:37 rtl8139_rx: eth0: In rtl8139_rx(), current 04d0 BufAddr 04cc, free to 04c0, Cmd 0d.
Aug  2 12:10:37 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0010.
Aug  2 12:10:37 rtl8139_rx: eth0: In rtl8139_rx(), current 04d0 BufAddr 04cc, free to 04c0, Cmd 0d.
Aug  2 12:10:37 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0010.
Aug  2 12:10:37 rtl8139_rx: eth0: In rtl8139_rx(), current 04d0 BufAddr 04cc, free to 04c0, Cmd 0d.
......
......
Aug  2 12:12:11 rtl8139_rx: eth0: In rtl8139_rx(), current 04d0 BufAddr 04cc, free to 04c0, Cmd 0d.
Aug  2 12:12:11 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0050.
Aug  2 12:12:11 rtl8139_rx: eth0: In rtl8139_rx(), current 04d0 BufAddr 04cc, free to 04c0, Cmd 0d.
Aug  2 12:12:11 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0050.
Aug  2 12:12:11 rtl8139_rx: eth0: In rtl8139_rx(), current 04d0 BufAddr 04cc, free to 04c0, Cmd 0d.
Aug  2 12:12:11 rtl8139_interrupt: eth0: exiting interrupt, intr_status=0x0050.
--
--
Pasi Sjöholm
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
  • Next message: Dipankar Sarma: "Re: CPU hotplug broken in 2.6.8-rc2 ?"

    Relevant Pages

    • Re: r8169.c
      ... > who tries to work with the 8169 driver on the long run). ... > Btw merging a 20 megaton patch is not the way network drivers changes ... > the suggested format). ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: [PATCH] iSeries virtual disk
      ... > patch for a new driver. ... Can you split them out into a separate patch? ... > better merging it into it's only caller. ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: radeonfb: cannot map FB
      ... I see a big DRM merge in .23 but it's unrelated to the fb driver. ... will be happy to get your patch. ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • RE: 2.6.7-rc3-mm2
      ... The 3w-xxxx.c patch is definately bogus. ... It would have applied against the driver ... > Sony Clie USB driver fix ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Re: [PATCH] race condition in procfs
      ... I got another oops in the very same place with the patch ... If it builds, boots and doesn't crash, I'll post ... send the line "unsubscribe linux-kernel" in ...
      (Linux-Kernel)
    • Quantcast