Musings on the PLX kernel sources.

[telzey]

No, I'm not using FileDescriptor's patch in my current 2.6.31.14 patchset.

I've sent you a few different patchsets for 2.6.31.14 over time as I was experimenting with the different manufacturer's source trees, so I'm not sure which version you actually tried.

I've removed all the old versions from my Pogoplug cloud and the latest one is linked in the message viewtopic.php?f=29&t=2472&start=10#p15007

As mentioned in my README, I'm now using the Cloud Engines patches, rebased from kernel 2.6.31.6 to kernel 2.6.31.14. Since you've reported that the basic Cloud Engines patches are stable for SATA/DMA/ethernet, that seemed to be the safest, even though it does make the kernel a bit useless on the Iomega/Silverstone/ZyXEL machines with their extra hardware.

After that I apply the more recent ZyXEL USB patches to fix some of the USB issues that people are seeing. The differences between the Cloud Engines USB patches and ZyXEL USB patches are fairly limited, but all seem to be related to fixing specific problems.

I've run multiple compilations with/without swap memory on both SATA, USB, and SATA-and-USB-at-the-same-time, pegging the CPUs at 190%+ and memory at 90%+ with no problem.

There doesn't seem to be much point in applying FileDescriptor's patch into an already-working system, but it might provide a clue as to what works better or got broken in the changes between the Cloud Engines/Iomega/Silverstone/ZyXEL kernel sources.

I'm not sure if the "hang" that you mention was in reference to FileDescriptor's patch, or to my patchset. If you've had my current patchset hang, then I'd love to know about it, preferably with any details that I could use to try to recreate the problem.

[Philoo]

telzey, I tried that last night ... and failed

1/ there are quite a few patches that were already applied in the 2.6.31.14 I downloaded , I did not reapply
2/ I'll get the exact one later on and but at least one patch failed to apply , it scrolled too fast after successful ones for me to document it. I'll retry, one by one and report.
3/ compilation stops after a couple minutes because the smp.c file is missing.


any pointer for me ?

[telzey]

I've just downloaded the patchset and tried it ... the patches apply cleanly (except for a couple of "Hunk #? Succeeded" messages because of the order of the patches), and it's compiling perfectly for me.

It sounds like you didn't use a clean kernel tree, please make sure that you're applying the patches to a clean unmodified official kernel, i.e. http://www.kernel.org/pub/linux/kernel/ ... 14.tar.bz2

[WarheadsSE]

Key parts there: the vanilla kernel.

I will rebase & re-attempt Telzey, and see if there is less I have to beat over the head.

[FileDescriptor]

$this->bbcode_second_pass_quote('telzey', 'N')o, I'm not using FileDescriptor's patch in my current 2.6.31.14 patchset.

[telzey]

Yeah, I know that you're not actually the author of the patch, but you get the credit here since you brought it to our attention. We've got to call it something unique to refer to it, and Catalin is the author of many ARM patches!

[Philoo]

Woops my bad....

somehow I did download 2.6.32.14 instead of 31.

please do not laugh, sleep deprivation is really bad.

[WarheadsSE]

Can't be worse than being doped from back issues and having people prod you for kernel updates every 12 hours

[telzey]

$this->bbcode_second_pass_quote('WarheadsSE', 'I') will rebase & re-attempt Telzey, and see if there is less I have to beat over the head.

[WarheadsSE]

When considering page size & erase block size, it becomes more limiting, quickly. Safely, you have just under 2.3MiB