[2.6.29-rc2 regression] e100: netconsole not functional because of missing firmware

[2.6.29-rc2 regression] e100: netconsole not functional because of missing firmware

[Andrey Borzenkov]

[-- Attachment #1: Type: text/plain, Size: 924 bytes --]

[    1.958127] e100 0000:00:0a.0: PCI INT A -> Link[LNKD] -> GSI 11 
(level, low)
 -> IRQ 11
[    1.986079] e100 0000:00:0a.0: PME# disabled
[    1.987930] e100: eth0: e100_probe: addr 0xf7efd000, irq 11, MAC addr 
00:00:39:d7:14:a1
[    1.989358] netconsole: local port 6665
[    1.989848] netconsole: local IP 10.0.0.1
[    1.990427] netconsole: interface eth0
[    1.990895] netconsole: remote port 6666
[    1.991382] netconsole: remote IP 10.0.0.2
[    1.991890] netconsole: remote ethernet address 00:17:42:21:27:e4
[    1.992680] netconsole: device eth0 not up yet, forcing it
[    2.008509] e100: eth0: e100_request_firmware: Failed to load 
firmware "e100/d101m_ucode.bin": -22
[    2.010461] netconsole: failed to open eth0
[    2.010996] netconsole: cleaning up

This has been working all the time without requiring any external 
firmware to be present. Why it suddenly refuses to work now?


[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [2.6.29-rc2 regression] e100: netconsole not functional because of missing firmware

[Jesse Brandeburg]

On Sat, Jan 17, 2009 at 8:08 AM, Andrey Borzenkov <arvidjaar@mail.ru> wrote:
> [    2.008509] e100: eth0: e100_request_firmware: Failed to load
> firmware "e100/d101m_ucode.bin": -22
> [    2.010461] netconsole: failed to open eth0
> [    2.010996] netconsole: cleaning up
>
> This has been working all the time without requiring any external
> firmware to be present. Why it suddenly refuses to work now?

to use netconsole you probably need to set config option
CONFIG_FIRMWARE_IN_KERNEL=y

Hm, I wonder if we couldn't have netconsole=y select
CONFIG_FIRMWARE_IN_KERNEL in Kconfig

the arguments as to why the firmware was split out of the drivers have
already been made on this list.  Not that I agree with them, but the
decision has been made and we're (network devs) going along with it.

Re: [2.6.29-rc2 regression] e100: netconsole not functional because of missing firmware

[Andrey Borzenkov]

[-- Attachment #1: Type: text/plain, Size: 1024 bytes --]

On 19 января 2009 21:30:01 Jesse Brandeburg wrote:
> On Sat, Jan 17, 2009 at 8:08 AM, Andrey Borzenkov <arvidjaar@mail.ru> 
wrote:
> > [    2.008509] e100: eth0: e100_request_firmware: Failed to load
> > firmware "e100/d101m_ucode.bin": -22
> > [    2.010461] netconsole: failed to open eth0
> > [    2.010996] netconsole: cleaning up
> >
> > This has been working all the time without requiring any external
> > firmware to be present. Why it suddenly refuses to work now?
>
> to use netconsole you probably need to set config option
> CONFIG_FIRMWARE_IN_KERNEL=y
>

Yes, I already figured that much :) Unfortunately, there seems to be a 
bug that prevents compiling firmware if src != obj. At least it does not 
work for me.

> Hm, I wonder if we couldn't have netconsole=y select
> CONFIG_FIRMWARE_IN_KERNEL in Kconfig
>

That probably is overkill; unfortunately I am afraid there is no way to 
make kconfig automatically select correct firmware in case e100 is built 
in.  

Thank you!

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 197 bytes --]

Re: [2.6.29-rc2 regression] e100: netconsole not functional because of missing firmware

[Rafael J. Wysocki]

On Monday 19 January 2009, Jesse Brandeburg wrote:
> On Sat, Jan 17, 2009 at 8:08 AM, Andrey Borzenkov <arvidjaar@mail.ru> wrote:
> > [    2.008509] e100: eth0: e100_request_firmware: Failed to load
> > firmware "e100/d101m_ucode.bin": -22
> > [    2.010461] netconsole: failed to open eth0
> > [    2.010996] netconsole: cleaning up
> >
> > This has been working all the time without requiring any external
> > firmware to be present. Why it suddenly refuses to work now?
> 
> to use netconsole you probably need to set config option
> CONFIG_FIRMWARE_IN_KERNEL=y
> 
> Hm, I wonder if we couldn't have netconsole=y select
> CONFIG_FIRMWARE_IN_KERNEL in Kconfig
> 
> the arguments as to why the firmware was split out of the drivers have
> already been made on this list.  Not that I agree with them, but the
> decision has been made and we're (network devs) going along with it.

Still, if that leads to functional regressions, like this one, we need to do
something about it IMO.  David?

Thanks,
Rafael