[update] 2.6.2-rc1 - Realtek 8169 patches

[update] 2.6.2-rc1 - Realtek 8169 patches

[Francois Romieu]

An updated version of the r8169 patches against 2.6.2-rc1 is available.

Individual files:
  http://www.fr.zoreil.com/linux/kernel/2.6.x/2.6.2-rc1

Tarball:
  http://www.fr.zoreil.com/linux/kernel/2.6.x/2.6.2-rc1/r8169-blob.tar.bz2

Readme:
  http://www.fr.zoreil.com/linux/kernel/2.6.x/2.6.2-rc1/README.txt

The first patch of the series probably wants to go directly to 2.6.x/2.4.x.

Please Cc: netdev@oss.sgi.com on success/failure.

--
Ueimor

Re: [update] 2.6.2-rc1 - Realtek 8169 patches

[Srihari Vijayaraghavan]

Hello Francois,

On Friday 23 January 2004 09:47, Francois Romieu wrote:
> An updated version of the r8169 patches against 2.6.2-rc1 is available.
>
> Individual files:
>   http://www.fr.zoreil.com/linux/kernel/2.6.x/2.6.2-rc1

I have applied your patches (all of them) and the kernel oops when I ping to 
that machine from a remote machine :-(.

r8169 Gigabit Ethernet driver 1.2 loaded
eth1: RTL8169 at 0xffffff000022e000, 00:0d:61:15:23:e6, IRQ 5
eth1: Auto-negotiation Enabled.
eth1: 100Mbps Full-duplex operation.
skput:over: ffffffffa00e32d8:3068 put:3068 dev:eth1----------- [cut here ] 
-----
Kernel BUG at skbuff:88
invalid operand: 0000 [1]
CPU 0
Pid: 0, comm: swapper Not tainted
RIP: 0010:[<ffffffff802284f2>] <ffffffff802284f2>{skb_over_panic+50}
RSP: 0018:ffffffff80370ff8  EFLAGS: 00010292
RAX: 0000000000000036 RBX: 0000f8650101f865 RCX: 0000000000000000
RDX: 00000000000003f9 RSI: 0000000000000000 RDI: ffffffff802c0260
RBP: 000001003e5c2320 R08: 0000000000000000 R09: 0000000000000003
R10: 00000000ffffffff R11: 0000000000000000 R12: ffffff000022e000
R13: 0000000000000bfc R14: 0000010000091010 R15: 0000000000000001
FS:  00000000005144a0(0000) GS:ffffffff8036d0c0(0000) knlGS:0000000000000000
CS:  0010 DS: 0018 ES: 0018 CR0: 000000008005003b
CR2: 0000002a95d36fe8 CR3: 0000000000101000 CR4: 00000000000006a0
Process swapper (pid: 0, stackpage=ffffffff802b8be0)
Stack: 0000000000000292 ffffffffa00e32e0 0000000000000bfc 000001003e5c2000
       000001003f5d21c0 0000000000008001 0000000000000014 ffffff000022e000
       000001003e5c2320 000001003e5c2000
Call Trace:<IRQ> <ffffffffa00e32e0>{:r8169:rtl8169_rx_interrupt+416}
       <ffffffffa00e3435>{:r8169:rtl8169_interrupt+101} 
<ffffffff801121af>{hand
       <ffffffff80112333>{do_IRQ+147} <ffffffff8010dbe0>{default_idle+0}
       <ffffffff80110089>{ret_from_intr+0}  <EOI> 
<ffffffff8012f3e5>{thread_ret
       <ffffffff8010dc00>{default_idle+32} <ffffffff8010dc8a>{cpu_idle+26}
       <ffffffff803796b2>{start_kernel+322}

Code: 0f 0b c2 c9 29 80 ff ff ff ff 58 00 48 83 c4 08 c3 66 66 66
RIP <ffffffff802284f2>{skb_over_panic+50} RSP <ffffffff80370ff8>
 <0>Kernel panic: Aiee, killing interrupt handler!
In interrupt handler - not syncing
 
I will try to track down which one of your patch introduces the regression.

> The first patch of the series probably wants to go directly to 2.6.x/2.4.x.
>
> Please Cc: netdev@oss.sgi.com on success/failure.

Thanks
Hari

Re: [update] 2.6.2-rc1 - Realtek 8169 patches

[Francois Romieu]

Srihari Vijayaraghavan <harisri@bigpond.com> :
[...]
> I have applied your patches (all of them) and the kernel oops when I ping to 
> that machine from a remote machine :-(.

Ok, can you publish/send me offlist an objdump -S of your r8169 module +
r8169.c source ?

[...]
> I will try to track down which one of your patch introduces the regression.

Thanks, a binary search can be done in 5~6 different rounds.

--
Ueimor

Re: [update] 2.6.2-rc1 - Realtek 8169 patches

[Srihari Vijayaraghavan]

Hello Romieu,

On Saturday 24 January 2004 21:16, you wrote:
> > I have applied your patches (all of them) and the kernel oops when I ping
> > to that machine from a remote machine :-(.
>
> Ok, can you publish/send me offlist an objdump -S of your r8169 module +
> r8169.c source ?

Please read my comments below and tell me if you still want the objdump?

> > I will try to track down which one of your patch introduces the
> > regression.
>
> Thanks, a binary search can be done in 5~6 different rounds.

Oh yes, been there done that :-). The r8169-addr-high.patch introduces the 
regression. The good news is that the previous 13 patches do not introduce 
any issues, and the kernel is stable with them.

Thanks
Hari