Re: 2.6.18-rc1-mm2

Re: 2.6.18-rc1-mm2

[Reuben Farrelly]

On 14/07/2006 5:48 p.m., Andrew Morton wrote:
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.18-rc1/2.6.18-rc1-mm2/
> 
> - Patches were merged, added, dropped and fixed.  Nothing particularly exciting.
> 
> - Added the avr32 architecture.  Review is sought, please.

Ugh.  This on bootup..

Starting HAL daemon: BUG: warning at net/core/dev.c:1171/skb_checksum_help()

Call Trace:
  [<ffffffff8026963e>] show_trace+0xae/0x265
  [<ffffffff8026980a>] dump_stack+0x15/0x1b
  [<ffffffff8043ba7b>] skb_checksum_help+0x61/0x126
  [<ffffffff8802f35f>] :iptable_nat:ip_nat_fn+0x5f/0x1d2
  [<ffffffff8802f71b>] :iptable_nat:ip_nat_local_fn+0x33/0xbc
  [<ffffffff80234c6e>] nf_iterate+0x5a/0x9b
  [<ffffffff802588d0>] nf_hook_slow+0x60/0xcd
  [<ffffffff8023511a>] ip_queue_xmit+0x41a/0x490
  [<ffffffff80221507>] tcp_transmit_skb+0x697/0x6d4
  [<ffffffff80233b37>] __tcp_push_pending_frames+0x867/0x95b
  [<ffffffff80226546>] tcp_sendmsg+0xa76/0xba6
  [<ffffffff80245c49>] inet_sendmsg+0x46/0x53
  [<ffffffff80247ebf>] sock_aio_write+0x13b/0x15c
  [<ffffffff8021758c>] do_sync_write+0xec/0x135
  [<ffffffff802160b5>] vfs_write+0xe5/0x180
  [<ffffffff80216ade>] sys_write+0x47/0x79
  [<ffffffff8025f3a2>] system_call+0x7e/0x83
  [<00002b145e465350>]
BUG: warning at net/core/dev.c:1225/skb_gso_segment()

Call Trace:
  [<ffffffff8026963e>] show_trace+0xae/0x265
  [<ffffffff8026980a>] dump_stack+0x15/0x1b
  [<ffffffff8043b8df>] skb_gso_segment+0xbf/0x1fa
  [<ffffffff8043bf28>] dev_hard_start_xmit+0x168/0x21c
  [<ffffffff80447a35>] __qdisc_run+0x115/0x1e5
  [<ffffffff8022fcef>] dev_queue_xmit+0x11f/0x256
  [<ffffffff802327f7>] ip_output+0x207/0x251
  [<ffffffff80235130>] ip_queue_xmit+0x430/0x490
  [<ffffffff80221507>] tcp_transmit_skb+0x697/0x6d4
  [<ffffffff80233b37>] __tcp_push_pending_frames+0x867/0x95b
  [<ffffffff80226546>] tcp_sendmsg+0xa76/0xba6
  [<ffffffff80245c49>] inet_sendmsg+0x46/0x53
  [<ffffffff80247ebf>] sock_aio_write+0x13b/0x15c
  [<ffffffff8021758c>] do_sync_write+0xec/0x135
  [<ffffffff802160b5>] vfs_write+0xe5/0x180
  [<ffffffff80216ade>] sys_write+0x47/0x79
  [<ffffffff8025f3a2>] system_call+0x7e/0x83
  [<00002b145e465350>]
[  OK  ]
Starting mailman: [  OK  ]
Starting nagios:  Starting network monitor: nagios
[  OK  ]

Fedora Core release 5.90 (Test)
Kernel 2.6.18-rc1-mm2 on an x86_64

tornado.reub.net login:



Reuben

Re: 2.6.18-rc1-mm2

[Andrew Morton]

On Fri, 14 Jul 2006 18:55:42 +1200
Reuben Farrelly <reuben-lkml@reub.net> wrote:

> On 14/07/2006 5:48 p.m., Andrew Morton wrote:
> > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.18-rc1/2.6.18-rc1-mm2/
> > 
> > - Patches were merged, added, dropped and fixed.  Nothing particularly exciting.
> > 
> > - Added the avr32 architecture.  Review is sought, please.
> 
> Ugh.  This on bootup..
> 
> Starting HAL daemon: BUG: warning at net/core/dev.c:1171/skb_checksum_help()
> 
> Call Trace:
>   [<ffffffff8026963e>] show_trace+0xae/0x265
>   [<ffffffff8026980a>] dump_stack+0x15/0x1b
>   [<ffffffff8043ba7b>] skb_checksum_help+0x61/0x126
>   [<ffffffff8802f35f>] :iptable_nat:ip_nat_fn+0x5f/0x1d2
>   [<ffffffff8802f71b>] :iptable_nat:ip_nat_local_fn+0x33/0xbc
>   [<ffffffff80234c6e>] nf_iterate+0x5a/0x9b
>   [<ffffffff802588d0>] nf_hook_slow+0x60/0xcd
>   [<ffffffff8023511a>] ip_queue_xmit+0x41a/0x490
>   [<ffffffff80221507>] tcp_transmit_skb+0x697/0x6d4
>   [<ffffffff80233b37>] __tcp_push_pending_frames+0x867/0x95b
>   [<ffffffff80226546>] tcp_sendmsg+0xa76/0xba6
>   [<ffffffff80245c49>] inet_sendmsg+0x46/0x53
>   [<ffffffff80247ebf>] sock_aio_write+0x13b/0x15c
>   [<ffffffff8021758c>] do_sync_write+0xec/0x135
>   [<ffffffff802160b5>] vfs_write+0xe5/0x180
>   [<ffffffff80216ade>] sys_write+0x47/0x79
>   [<ffffffff8025f3a2>] system_call+0x7e/0x83
>   [<00002b145e465350>]
> BUG: warning at net/core/dev.c:1225/skb_gso_segment()

int skb_checksum_help(struct sk_buff *skb, int inward)
{
	unsigned int csum;
	int ret = 0, offset = skb->h.raw - skb->data;

	if (inward)
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
		static int warned;

		WARN_ON(!warned);

That's a mainline problem, I suspect.

Re: 2.6.18-rc1-mm2

[Herbert Xu]

On Fri, Jul 14, 2006 at 12:05:51AM -0700, Andrew Morton wrote:
>
> > Call Trace:
> >   [<ffffffff8026963e>] show_trace+0xae/0x265
> >   [<ffffffff8026980a>] dump_stack+0x15/0x1b
> >   [<ffffffff8043ba7b>] skb_checksum_help+0x61/0x126
> >   [<ffffffff8802f35f>] :iptable_nat:ip_nat_fn+0x5f/0x1d2

This is tell you that there is a bug in ip_nat_fn in that it completes
the partial checksum even for TSO packets which will cause them to go
out with bogus checksums.

The warning also indicates that the system has detected this and has
worked around it by recomputing the partial checksum after NAT.

The warning is here so someone can fix NAT to not trash the partial
checksum.  It would also tell us if anyone else breaks checksums in
this way.

I've already made the warning appear only once per-boot so I'd really
like to keep it in until

1) NAT is fixed.
2) We're reasonably sure there's nothing else doing this.

Prior to this change your TSO packets would've gone out with corrupted
checksums silently.  Essentially TSO would only slow your machine down
since every transmission it makes has to be retransmitted as non-TSO.

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

Re: 2.6.18-rc1-mm2

[David Miller]

From: Reuben Farrelly <reuben-lkml@reub.net>
Date: Fri, 14 Jul 2006 18:55:42 +1200

> Ugh.  This on bootup..
> 
> Starting HAL daemon: BUG: warning at net/core/dev.c:1171/skb_checksum_help()

It's perfectly normal, and it's there to remind us that
we need to deal with an incremental checksumming issue
wrt. ip_nat_fn for GSO segments at some point.

You can safely ignore it for now.

2.6.18-rc1-mm2: drivers/char/*synclink* compile errors

[Adrian Bunk]

Krzysztof, the following compile errors are caused by your commit 
c2ce920468624d87ec5f91f080ea99681dae6d88 in Linus' tree:

<--  snip  -->

...
  CC      drivers/char/pcmcia/synclink_cs.o
drivers/char/pcmcia/synclink_cs.c: In function ‘dcd_change’:
drivers/char/pcmcia/synclink_cs.c:1171: error: implicit declaration of function ‘hdlc_set_carrier’
...
  CC      drivers/char/synclink.o
drivers/char/synclink.c: In function ‘mgsl_isr_io_pin’:
drivers/char/synclink.c:1348: error: implicit declaration of function ‘hdlc_set_carrier’
...
  CC      drivers/char/synclinkmp.o
drivers/char/synclinkmp.c: In function ‘hdlcdev_open’:
drivers/char/synclinkmp.c:1755: error: implicit declaration of function ‘hdlc_set_carrier’
...
  CC      drivers/char/synclink_gt.o
drivers/char/synclink_gt.c: In function ‘hdlcdev_open’:
drivers/char/synclink_gt.c:1500: error: implicit declaration of function ‘hdlc_set_carrier’

<--   snip  -->

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: 2.6.18-rc1-mm2: drivers/char/*synclink* compile errors

[David Miller]

From: Adrian Bunk <bunk@stusta.de>
Date: Fri, 14 Jul 2006 20:48:25 +0200

> Krzysztof, the following compile errors are caused by your commit 
> c2ce920468624d87ec5f91f080ea99681dae6d88 in Linus' tree:

Yes, we let Krzysztof know about this yesterday.

If we don't get a fix over the weekend I will simply
revert the changeset.

Thanks Adrian.

[-mm patch] remove net/core/skbuff.c:skb_queue_lock_key

[Adrian Bunk]

On Thu, Jul 13, 2006 at 10:48:00PM -0700, Andrew Morton wrote:
>...
> Changes since 2.6.18-rc1-mm1:
>...
> +lockdep-split-the-skb_queue_head_init-lock-class.patch
> 
>  lockdep-versus-net fix.
>...

skb_queue_lock_key is no longer used.

Signed-off-by: Adrian Bunk <bunk@stusta.de>

---

 include/linux/skbuff.h |    2 --
 net/core/skbuff.c      |    7 -------
 2 files changed, 9 deletions(-)

--- linux-2.6.18-rc1-mm2-full/include/linux/skbuff.h.old	2006-07-14 22:15:15.000000000 +0200
+++ linux-2.6.18-rc1-mm2-full/include/linux/skbuff.h	2006-07-14 22:15:23.000000000 +0200
@@ -604,8 +604,6 @@
 	return list_->qlen;
 }
 
-extern struct lock_class_key skb_queue_lock_key;
-
 /*
  * This function creates a split out lock class for each invocation;
  * this is needed for now since a whole lot of users of the skb-queue
--- linux-2.6.18-rc1-mm2-full/net/core/skbuff.c.old	2006-07-14 22:15:30.000000000 +0200
+++ linux-2.6.18-rc1-mm2-full/net/core/skbuff.c	2006-07-14 22:15:38.000000000 +0200
@@ -71,13 +71,6 @@
 static kmem_cache_t *skbuff_fclone_cache __read_mostly;
 
 /*
- * lockdep: lock class key used by skb_queue_head_init():
- */
-struct lock_class_key skb_queue_lock_key;
-
-EXPORT_SYMBOL(skb_queue_lock_key);
-
-/*
  *	Keep out-of-line to prevent kernel bloat.
  *	__builtin_return_address is not used because it is not always
  *	reliable.

Re: [-mm patch] remove net/core/skbuff.c:skb_queue_lock_key

[Arjan van de Ven]

Adrian Bunk wrote:
> On Thu, Jul 13, 2006 at 10:48:00PM -0700, Andrew Morton wrote:
>> ...
>> Changes since 2.6.18-rc1-mm1:
>> ...
>> +lockdep-split-the-skb_queue_head_init-lock-class.patch
>>
>>  lockdep-versus-net fix.
>> ...
> 
> skb_queue_lock_key is no longer used.
> 
> Signed-off-by: Adrian Bunk <bunk@stusta.de>
> 

Acked-by: Arjan van de Ven <arjan@linux.intel.com>