git-net, git-netdev-all and everything else on g5

git-net, git-netdev-all and everything else on g5

[Andrew Morton]

With the full -mm lineup, my tg3-using powerpc g5 spits lots of these:

windfarm: Drive bay control loop started.
audit(1183017094.732:2): audit_pid=2117 old=0 by auid=4294967295
------------[ cut here ]------------
Badness at net/core/dev.c:1303
Call Trace:
[c00000000b45ead0] [c0000000000108c8] .show_stack+0x50/0x1cc (unreliable)
[c00000000b45eb80] [c0000000001b563c] .report_bug+0xa0/0x110
[c00000000b45ec10] [c0000000000250a0] .program_check_exception+0x1fc/0x738
[c00000000b45ecd0] [c000000000004a84] program_check_common+0x104/0x180
--- Exception: 700 at .skb_gso_segment+0xd0/0x298
    LR = .dev_hard_start_xmit+0x23c/0x33c
[c00000000b45efc0] [0000000000000001] 0x1 (unreliable)
[c00000000b45f060] [c00000000037bc2c] .dev_hard_start_xmit+0x23c/0x33c
[c00000000b45f100] [c000000000394934] .__qdisc_run+0x8c/0x410
[c00000000b45f1b0] [c00000000037c0f4] .dev_queue_xmit+0x3c8/0x408
[c00000000b45f240] [c0000000003a7180] .ip_output+0x1d8/0x3d4
[c00000000b45f300] [c0000000003a63f0] .ip_queue_xmit+0x374/0x51c
[c00000000b45f450] [c0000000003bdcc0] .tcp_transmit_skb+0x54c/0x9dc
[c00000000b45f560] [c0000000003bf8f8] .__tcp_push_pending_frames+0x2fc/0xb58
[c00000000b45f6a0] [c0000000003bbccc] .tcp_rcv_established+0x204/0x900
[c00000000b45f750] [c0000000003c6080] .tcp_v4_do_rcv+0x230/0x598
[c00000000b45f830] [c0000000003b1200] .tcp_prequeue_process+0xa0/0xf4
[c00000000b45f8c0] [c0000000003b19c8] .tcp_recvmsg+0x4f0/0x940
[c00000000b45f9b0] [c000000000370fa0] .sock_common_recvmsg+0x68/0x90
[c00000000b45fa40] [c00000000036b2cc] .sock_aio_read+0x120/0x148
[c00000000b45fb50] [c0000000000d029c] .do_sync_read+0xd0/0x160
[c00000000b45fcf0] [c0000000000d04ec] .vfs_read+0x1c0/0x1d8
[c00000000b45fd90] [c0000000000d0888] .sys_read+0x4c/0x90
[c00000000b45fe30] [c00000000000872c] syscall_exit+0x0/0x40
------------[ cut here ]------------
Badness at net/core/dev.c:1303
Call Trace:
[c00000000b45e9c0] [c0000000000108c8] .show_stack+0x50/0x1cc (unreliable)
[c00000000b45ea70] [c0000000001b563c] .report_bug+0xa0/0x110
[c00000000b45eb00] [c0000000000250a0] .program_check_exception+0x1fc/0x738
[c00000000b45ebc0] [c000000000004a84] program_check_common+0x104/0x180
--- Exception: 700 at .skb_gso_segment+0xd0/0x298
    LR = .dev_hard_start_xmit+0x23c/0x33c
[c00000000b45eeb0] [c000000000079a60] .__wake_up_bit+0x4c/0x60 (unreliable)
[c00000000b45ef50] [c00000000037bc2c] .dev_hard_start_xmit+0x23c/0x33c
[c00000000b45eff0] [c000000000394934] .__qdisc_run+0x8c/0x410
[c00000000b45f0a0] [c00000000037c0f4] .dev_queue_xmit+0x3c8/0x408
[c00000000b45f130] [c0000000003a7180] .ip_output+0x1d8/0x3d4
[c00000000b45f1f0] [c0000000003a63f0] .ip_queue_xmit+0x374/0x51c
[c00000000b45f340] [c0000000003bdcc0] .tcp_transmit_skb+0x54c/0x9dc
[c00000000b45f450] [c0000000003bf8f8] .__tcp_push_pending_frames+0x2fc/0xb58
[c00000000b45f590] [c0000000003bbccc] .tcp_rcv_established+0x204/0x900
[c00000000b45f640] [c0000000003c6080] .tcp_v4_do_rcv+0x230/0x598
[c00000000b45f720] [c00000000036fb8c] .release_sock+0xa0/0x16c
[c00000000b45f7c0] [c00000000036fd08] .sk_wait_data+0xb0/0x148
[c00000000b45f8c0] [c0000000003b1a6c] .tcp_recvmsg+0x594/0x940
[c00000000b45f9b0] [c000000000370fa0] .sock_common_recvmsg+0x68/0x90
[c00000000b45fa40] [c00000000036b2cc] .sock_aio_read+0x120/0x148
[c00000000b45fb50] [c0000000000d029c] .do_sync_read+0xd0/0x160
[c00000000b45fcf0] [c0000000000d04ec] .vfs_read+0x1c0/0x1d8
[c00000000b45fd90] [c0000000000d0888] .sys_read+0x4c/0x90
[c00000000b45fe30] [c00000000000872c] syscall_exit+0x0/0x40

That's here, in skb_gso_segment():

        skb_reset_mac_header(skb);
        skb->mac_len = skb->network_header - skb->mac_header;
        __skb_pull(skb, skb->mac_len);

->        if (WARN_ON(skb->ip_summed != CHECKSUM_PARTIAL)) {


config: http://userweb.kernel.org/~akpm/config-g5.txt
dmesg: http://userweb.kernel.org/~akpm/dmesg-g5.txt

Note that that dmesg contains extra stuff at the end which looks like the
box is trying to oops.

Generally ugly.

Re: git-net, git-netdev-all and everything else on g5

[Herbert Xu]

On Thu, Jun 28, 2007 at 09:13:38AM +0000, Andrew Morton wrote:
> 
> With the full -mm lineup, my tg3-using powerpc g5 spits lots of these:
> 
> windfarm: Drive bay control loop started.
> audit(1183017094.732:2): audit_pid=2117 old=0 by auid=4294967295
> ------------[ cut here ]------------
> Badness at net/core/dev.c:1303

Hmm, we really should come up with a better way for checking
checksum support.

[NET]: Fix TX checksum feature check

This patch fixes a boolean error in the new TX checksum check
that causes bogus TSO packets to be generated.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Cheers,
-- 
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
--
diff --git a/net/core/dev.c b/net/core/dev.c
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1509,11 +1509,11 @@ int dev_queue_xmit(struct sk_buff *skb)
 		skb_set_transport_header(skb, skb->csum_start -
 					      skb_headroom(skb));
 
-		if (!(dev->features & NETIF_F_GEN_CSUM)
-		    || ((dev->features & NETIF_F_IP_CSUM)
-			&& skb->protocol == htons(ETH_P_IP))
-		    || ((dev->features & NETIF_F_IPV6_CSUM)
-			&& skb->protocol == htons(ETH_P_IPV6)))
+		if (!(dev->features & NETIF_F_GEN_CSUM) &&
+		    !((dev->features & NETIF_F_IP_CSUM) &&
+		      skb->protocol == htons(ETH_P_IP)) &&
+		    !((dev->features & NETIF_F_IPV6_CSUM) &&
+		      skb->protocol == htons(ETH_P_IPV6)))
 			if (skb_checksum_help(skb))
 				goto out_kfree_skb;
 	}

Re: git-net, git-netdev-all and everything else on g5

[David Miller]

From: Herbert Xu <herbert@gondor.apana.org.au>
Date: Fri, 29 Jun 2007 03:18:31 +0800

> On Thu, Jun 28, 2007 at 09:13:38AM +0000, Andrew Morton wrote:
> > 
> > With the full -mm lineup, my tg3-using powerpc g5 spits lots of these:
> > 
> > windfarm: Drive bay control loop started.
> > audit(1183017094.732:2): audit_pid=2117 old=0 by auid=4294967295
> > ------------[ cut here ]------------
> > Badness at net/core/dev.c:1303
> 
> Hmm, we really should come up with a better way for checking
> checksum support.
> 
> [NET]: Fix TX checksum feature check
> 
> This patch fixes a boolean error in the new TX checksum check
> that causes bogus TSO packets to be generated.
> 
> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Wasn't I just mentioning how much of a mess these tests
are the other day? :-)

Patch applied, thanks Herbert!