Re: Network lockup under load

2005-11-23 Thread Jesse Brandeburg
> > > Some e1000 experts need to look into this. > > > > Hmmm, I'll find out who that is and bring them onboard, after > > the T.day break. I'll work with Joe to see where we can get on Monday. Jesse - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [

Re: Network lockup under load

2005-11-23 Thread David S. Miller
From: Joe Korty <[EMAIL PROTECTED]> Date: Wed, 23 Nov 2005 17:37:07 -0500 > On Wed, Nov 23, 2005 at 02:07:04PM -0800, David S. Miller wrote: > > > There is some bug in the e1000 card that causes it to wedge with > > TSO+SG packets, the driver should have a sufficient workaround but > > aparently

Re: Network lockup under load

2005-11-23 Thread Joe Korty
On Wed, Nov 23, 2005 at 02:07:04PM -0800, David S. Miller wrote: > There is some bug in the e1000 card that causes it to wedge with > TSO+SG packets, the driver should have a sufficient workaround but > aparently not. > > Some e1000 experts need to look into this. Hmmm, I'll find out who that is

Re: Network lockup under load

2005-11-23 Thread David S. Miller
From: Joe Korty <[EMAIL PROTECTED]> Date: Wed, 23 Nov 2005 16:32:55 -0500 > On Thu, Nov 24, 2005 at 08:11:05AM +1100, Herbert Xu wrote: > > > Please try turning sg and tso off to see if it helps. > > tso sg result > -- -- -- > on on stallout >

Re: Network lockup under load

2005-11-23 Thread David S. Miller
From: Joe Korty <[EMAIL PROTECTED]> Date: Wed, 23 Nov 2005 16:01:24 -0500 >Then, a following ifdown/ifup with an intervening sleep let another 6 files >through before locking up. This means the e1000 card's transmitter is wedging when using TSO scatter-gather packet buffers. There are so

Re: Network lockup under load

2005-11-23 Thread David S. Miller
From: Herbert Xu <[EMAIL PROTECTED]> Date: Thu, 24 Nov 2005 08:11:05 +1100 > On Wed, Nov 23, 2005 at 04:01:24PM -0500, Joe Korty wrote: > > Will do in a bit, but as everything works fine with 2.6.13-rc1-git7 and > > earlier > > kernels, I don't expect this will be the problem. > > It's probably

Re: Network lockup under load

2005-11-23 Thread Joe Korty
On Thu, Nov 24, 2005 at 08:11:05AM +1100, Herbert Xu wrote: > Please try turning sg and tso off to see if it helps. tso sg result -- -- -- on on stallout on off works off on works off off wo

Re: Network lockup under load

2005-11-23 Thread Herbert Xu
On Wed, Nov 23, 2005 at 04:01:24PM -0500, Joe Korty wrote: > On Thu, Nov 24, 2005 at 07:45:16AM +1100, Herbert Xu wrote: > > > What does ethtool -k say on the e1000 card? > > buttercup# ethtool -k eth0 > Offload parameters for eth0: > rx-checksumming: on > tx-checksumming: on >

Re: Network lockup under load

2005-11-23 Thread Joe Korty
On Thu, Nov 24, 2005 at 07:45:16AM +1100, Herbert Xu wrote: > What does ethtool -k say on the e1000 card? buttercup# ethtool -k eth0 Offload parameters for eth0: rx-checksumming: on tx-checksumming: on scatter-gather: on tcp segmentation offload: on > Looks like buttercu

Re: Network lockup under load

2005-11-23 Thread Herbert Xu
On Wed, Nov 23, 2005 at 02:42:39PM -0500, Joe Korty wrote: > > Intel(R) PRO/1000 Network Driver - version 6.0.60-k2 > Copyright (c) 1999-2005 Intel Corporation. > ACPI: PCI Interrupt :03:01.0[A] -> GSI 28 (level, low) -> IRQ 169 > e1000: eth0: e1000_probe: Intel(R) PRO/1000 Net

Re: Network lockup under load

2005-11-23 Thread Herbert Xu
On Tue, Nov 22, 2005 at 07:39:49PM -0500, Joe Korty wrote: > > 19:32:01.795103 buttercup.ccur.com.34798 > tsunami.ccur.com.ssh: . > 1495105:1496553(1448) ack 28242 win 2372 12059137> (DF) [tos 0x8] > 19:32:08.323773 buttercup.ccur.com.34798 > tsunami.ccur.com.ssh: . > 1495105:1496553(1448) ack

Re: Network lockup under load

2005-11-22 Thread Herbert Xu
On Tue, Nov 22, 2005 at 06:09:30PM -0500, Joe Korty wrote: > > [pid 2479] read(4, "\0", 16384)= 1 > [pid 2479] write(3, > ",}\2169\1777\202T3\326\3572\2037\372\257\37\\\257\235R"..., 6192) = 6192 > [pid 2479] select(7, [3 4], [3], NULL, NULL) = 1 (out [3]) > [pid 2479] write(3, > "R\3

Re: Network lockup under load

2005-11-22 Thread Joe Korty
On Tue, Nov 22, 2005 at 06:09:30PM -0500, Joe Korty wrote: > On Wed, Nov 23, 2005 at 09:07:04AM +1100, Herbert Xu wrote: > > Joe Korty <[EMAIL PROTECTED]> wrote: > >> > >> ssh S 0010 0 2512 2511 (NOTLB) > >> f6d37ea0 0086 f6d37e90 0010 0002 0

Re: Network lockup under load

2005-11-22 Thread Joe Korty
On Wed, Nov 23, 2005 at 09:07:04AM +1100, Herbert Xu wrote: > Joe Korty <[EMAIL PROTECTED]> wrote: >> >> ssh S 0010 0 2512 2511 (NOTLB) >> f6d37ea0 0086 f6d37e90 0010 0002 00d0 0010 >> c218ea40 0001 f6c88530 00

Re: Network lockup under load

2005-11-22 Thread Herbert Xu
Joe Korty <[EMAIL PROTECTED]> wrote: > > ssh S 0010 0 2512 2511 (NOTLB) > f6d37ea0 0086 f6d37e90 0010 0002 00d0 0010 > c218ea40 0001 f6c88530 0010 c064c400 c202b560 0003 > 0120 dd997fff 0d

Re: Network lockup under load

2005-11-22 Thread Joe Korty
[ lkml trimmed from the CC: list ] On Wed, Nov 23, 2005 at 07:40:08AM +1100, Herbert Xu wrote: > On Tue, Nov 22, 2005 at 04:45:02PM +, Joe Korty wrote: >> >> To trigger the lockup, I do a 'scp -rp' of the kernel >> tree from a machine with a defective kernel to any other >> machine. It trigg

Re: Network lockup under load

2005-11-22 Thread Herbert Xu
On Tue, Nov 22, 2005 at 04:45:02PM +, Joe Korty wrote: > > To trigger the lockup, I do a 'scp -rp' of the kernel > tree from a machine with a defective kernel to any other > machine. It triggers anywhere after the first file > transfered, to perhaps 30 files transfered. Please generate a sta