[elrepo-devel] iwl3945-firmware

[Phil Perry]

Phil Perry wrote:
>  From what I can gather, RHEL5 doesn't currently need the revision -2 
> firmware (RHEL 5.3 uses the rev -1 firmware). I don't know what RHEL 5.4 
> kernels use so it would be good if you could find out/confirm that.
> 
> It seems to me that Red Hat have just used a community package that is 
> designed to be nodist, i.e support all dists including the latest 
> Fedora, hence why the package is updated to the latest rev -2 firmware 
> (this package also updates the rev -1 firmware that *is* required to the 
> latest version which is probably the reason Red Hat have included it in 
> 5.4, rather than that 5.4 needs the rev -2 firmware.
> 
> So what I really need to know is what firmware revision the 5.4 kernel 
> requires. Probably the easiest way to determine this is to remove all 
> firmwares, reboot to a 5.4 series kernel and look in /var/log/messages 
> for errors regarding missing firmwares - you'll normally see an error 
> like unable to load firmware 'iwlwifi-3945-1.ucode' which instantly 
> tells us which revision that kernel driver requires. Then please log 
> that information somewhere (here and/or elrepo.org/bugs) for future 
> reference.
> 
> Likewise, anyone with an Intel iwl5000 series adapter, we also need this 
> information confirmed for the iwl5000-firmware package (I have an 
> iwl4965 so can confirm myself for that package).
> 

Actually this is all easy enough for me to do...

 From the kernel-2.6.18-155 source, in /drivers/net/wireless/iwlwifi/

[buildsys at Quad iwlwifi]$ grep "UCODE_API" * | grep define
iwl-3945.h:#define IWL3945_UCODE_API "-1"
iwl-4965.c:#define IWL4965_UCODE_API "-2"
iwl-5000.c:#define IWL5000_UCODE_API  "-1"

shows us that iwl3945 and iwl5000 both use revision -1 firmwares whereas 
iwl4965 currently uses revision -2 firmware, so now we know. This is the 
same as for RHEL 5.3 so no changes in the 5.4 series kernels.

Our packages in elrepo (and rppmforge) all contain the latest firmware 
for these revisions :)

If anyone wants to repeat that experiment for older 5.0, 5.1 and 5.2 
series kernels, that might be interesting in terms of knowing which 
older revision firmwares we actually need to ship for compatibility.

Thank you Akemi for making me actually *think* about the issue :)