e1000e oops

e1000e oops

[maximilian attems]

Linux tau 2.6.23-rc8-mm2-686 #1 SMP Wed Oct 3 23:56:32 CEST 2007 i686
GNU/Linux

eth0 renamed to eth1
sysfs: duplicate filename 'eth1' can not be created
WARNING: at fs/sysfs/dir.c:433 sysfs_add_one()
 [<c0104f9c>] dump_trace+0x68/0x1d5
 [<c0105121>] show_trace_log_lvl+0x18/0x2c
 [<c0105b9b>] show_trace+0xf/0x11
 [<c0105c81>] dump_stack+0x12/0x14
 [<c01ac02e>] sysfs_add_one+0x54/0xb8
 [<c01acc2a>] sysfs_create_link+0xb3/0x125
 [<c023c05f>] device_rename+0x13d/0x196
 [<c025a970>] dev_change_name+0x137/0x205
 [<c025b12c>] dev_ioctl+0x357/0x44f
 [<c017b05f>] do_ioctl+0x1f/0x62
 [<c017b2d9>] vfs_ioctl+0x237/0x249
 [<c017b330>] sys_ioctl+0x45/0x5d
 [<c0103eda>] syscall_call+0x7/0xb
 [<b7edbde4>] 0xb7edbde4
 =======================
net eth1: device_rename: sysfs_create_symlink failed (-17)

00:19.0 Ethernet controller [0200]: Intel Corporation 82801I (ICH9 Family) Gigabit Ethernet Controller [8086:294c] (rev 02)
        Subsystem: Intel Corporation Unknown device [8086:0001]
        Flags: bus master, fast devsel, latency 0, IRQ 218
        Memory at 50380000 (32-bit, non-prefetchable) [size=128K]
        Memory at 503a4000 (32-bit, non-prefetchable) [size=4K]
        I/O ports at 20e0 [size=32]
        Capabilities: [c8] Power Management version 2
        Capabilities: [d0] Message Signalled Interrupts: Mask- 64bit+ Queue=0/0 Enable+
        Capabilities: [e0] Vendor Specific Information


no idea if aboves belongs to netdev or sysfs dep.

-- 
maks

Re: e1000e oops, sysfs related?

[Kok, Auke]

maximilian attems wrote:
> Linux tau 2.6.23-rc8-mm2-686 #1 SMP Wed Oct 3 23:56:32 CEST 2007 i686
> GNU/Linux
> 
> eth0 renamed to eth1
> sysfs: duplicate filename 'eth1' can not be created
> WARNING: at fs/sysfs/dir.c:433 sysfs_add_one()
>  [<c0104f9c>] dump_trace+0x68/0x1d5
>  [<c0105121>] show_trace_log_lvl+0x18/0x2c
>  [<c0105b9b>] show_trace+0xf/0x11
>  [<c0105c81>] dump_stack+0x12/0x14
>  [<c01ac02e>] sysfs_add_one+0x54/0xb8
>  [<c01acc2a>] sysfs_create_link+0xb3/0x125
>  [<c023c05f>] device_rename+0x13d/0x196
>  [<c025a970>] dev_change_name+0x137/0x205
>  [<c025b12c>] dev_ioctl+0x357/0x44f
>  [<c017b05f>] do_ioctl+0x1f/0x62
>  [<c017b2d9>] vfs_ioctl+0x237/0x249
>  [<c017b330>] sys_ioctl+0x45/0x5d
>  [<c0103eda>] syscall_call+0x7/0xb
>  [<b7edbde4>] 0xb7edbde4
>  =======================
> net eth1: device_rename: sysfs_create_symlink failed (-17)
> 
> 00:19.0 Ethernet controller [0200]: Intel Corporation 82801I (ICH9 Family) Gigabit Ethernet Controller [8086:294c] (rev 02)
>         Subsystem: Intel Corporation Unknown device [8086:0001]
>         Flags: bus master, fast devsel, latency 0, IRQ 218
>         Memory at 50380000 (32-bit, non-prefetchable) [size=128K]
>         Memory at 503a4000 (32-bit, non-prefetchable) [size=4K]
>         I/O ports at 20e0 [size=32]
>         Capabilities: [c8] Power Management version 2
>         Capabilities: [d0] Message Signalled Interrupts: Mask- 64bit+ Queue=0/0 Enable+
>         Capabilities: [e0] Vendor Specific Information
> 
> 
> no idea if aboves belongs to netdev or sysfs dep.

I'm seeing the same when removing/inserting network drivers frequently, ixgbe does
the same thing: (git://davem/net-2.6.24)

ixgbe 0000:01:00.0: (PCI Express:2.5Gb/s:Width x8) 00:1a:64:3b:81:c3
ixgbe 0000:01:00.0: MAC: 0, PHY: 0, PBA No: d95854-003
ixgbe 0000:01:00.0: Intel(R) 10 Gigabit Network Connection
net eth2: device_rename: sysfs_create_symlink failed (-17)

but I can't make it throw a stack trace.

perhaps one of the sysfs knowledgeable people can take a look at this. We might be
doing something wrong in e1000e / ixgbe at setup time, but I have no idea what.

Auke

> 

Re: e1000e oops

[David Miller]

From: maximilian attems <max@stro.at>
Date: Fri, 5 Oct 2007 15:11:15 +0200

> net eth1: device_rename: sysfs_create_symlink failed (-17)
 ...
> no idea if aboves belongs to netdev or sysfs dep.

I think the report belongs to udev.

It tries to rename network devices even if they already
have the name it desires.

Re: e1000e oops, sysfs related?

[Eric W. Biederman]

"Kok, Auke" <auke-jan.h.kok@intel.com> writes:

> maximilian attems wrote:
>> Linux tau 2.6.23-rc8-mm2-686 #1 SMP Wed Oct 3 23:56:32 CEST 2007 i686
>> GNU/Linux
>> 
>> eth0 renamed to eth1
>> sysfs: duplicate filename 'eth1' can not be created
>> WARNING: at fs/sysfs/dir.c:433 sysfs_add_one()
>>  [<c0104f9c>] dump_trace+0x68/0x1d5
>>  [<c0105121>] show_trace_log_lvl+0x18/0x2c
>>  [<c0105b9b>] show_trace+0xf/0x11
>>  [<c0105c81>] dump_stack+0x12/0x14
>>  [<c01ac02e>] sysfs_add_one+0x54/0xb8
>>  [<c01acc2a>] sysfs_create_link+0xb3/0x125
>>  [<c023c05f>] device_rename+0x13d/0x196
>>  [<c025a970>] dev_change_name+0x137/0x205
>>  [<c025b12c>] dev_ioctl+0x357/0x44f
>>  [<c017b05f>] do_ioctl+0x1f/0x62
>>  [<c017b2d9>] vfs_ioctl+0x237/0x249
>>  [<c017b330>] sys_ioctl+0x45/0x5d
>>  [<c0103eda>] syscall_call+0x7/0xb
>>  [<b7edbde4>] 0xb7edbde4
>>  =======================
>> net eth1: device_rename: sysfs_create_symlink failed (-17)
>> 

> I'm seeing the same when removing/inserting network drivers frequently, ixgbe
> does
> the same thing: (git://davem/net-2.6.24)
>
> ixgbe 0000:01:00.0: (PCI Express:2.5Gb/s:Width x8) 00:1a:64:3b:81:c3
> ixgbe 0000:01:00.0: MAC: 0, PHY: 0, PBA No: d95854-003
> ixgbe 0000:01:00.0: Intel(R) 10 Gigabit Network Connection
> net eth2: device_rename: sysfs_create_symlink failed (-17)
>
> but I can't make it throw a stack trace.
>
> perhaps one of the sysfs knowledgeable people can take a look at this. We might
> be
> doing something wrong in e1000e / ixgbe at setup time, but I have no idea what.

To reach this path you need to run the equivalent of:
ip link set eth0 name eth1

I have seen this a bit and my memory is that the error is generally
harmless.  It is trying to create a symlink with the same name as the
directory or something like that.  Which makes with one setting of
CONFIG_SYSFS_DEPRECATED but not the other.

So I don't believe this is a driver bug, and I suspect the renames
come from using something like /etc/mactab which forces ethernet
devices with specific mac addresses to specific ethernet device
names.

With whatever is sitting in my development tree I'm pretty certain
this error isn't present.  I don't recall if I did anything specific
to fix this though.  In particular I don't recall if this still
happens with gregkh's queued sysfs patches applied.

Eric