> > Any chance this can be applied to stable 2.6.26.x?
> >
> > Half of my agenda was selfish (to be sure future kernels would not
> > hang on my 2 home servers), but the other half was to ensure that
> > Debian would not have hanging kernels in their next stable release.
>
> it's a backport candidate, but i think we should wait a bit with that,
> until after .27-rc5 or -rc6, to make sure there are no side-effects.
ACK
This is my first time here, so I was unsure about standard practices...
and whether I needed to do something myself re 2.6.26.x. The idea
of waiting makes a LOT of sense to me: I don't want to find that other
machines are now broken because of this fix for me!
> The upstream commit is:
>
> commit a2bd7274b47124d2fc4dfdb8c0591f545ba749dd
> Author: Yinghai Lu
> Date: Mon Aug 25 00:56:08 2008 -0700
>
> x86: fix HPET regression in 2.6.26 versus 2.6.25, check hpet against BAR, v3
Thx... will watch for it. :)
> btw., are both of your systems that were hanging fixed by this commit?
Ooohhh...
*sweating profusely...*
I have not been testing "webserver" (since Aug. 5) for two reasons:
1) It has no permanent hard drive at the moment, awaiting one from a
soon-to-disappear older machine
2) It has nearly identical hardware to "fileserver": identical model
motherboard, same CPU family (Athlon 64 X2 3600+ instead of Athlon X2 4850e)
I'm afraid that I was only testing "fileserver" all along, assuming (out of
laziness?) it was the "same".
I am happy to report that -- after copying the '2.6.27-rc4.i386.c-debug'
kernel from fileserver to webserver via NFS -- that it boots fine:
- no hangs (without "hpet=disable")
- no error messages in 'dmesg'
- found these messages in 'dmesg'
=================
[...]
calling pci_subsys_init+0x0/0x120
PCI: Using ACPI for IRQ routing
pci 0000:00:14.0: BAR has HPET at fed00000-fed003ff
initcall pci_subsys_init+0x0/0x120 returned 0 after 0 msecs
[...]
calling hpet_late_init+0x0/0xf7
hpet0: at MMIO 0xfed00000, IRQs 2, 8, 0, 0
hpet0: 4 32-bit timers, 14318180 Hz
initcall hpet_late_init+0x0/0xf7 returned 0 after 0 msecs
[...]
=================
- running 'cat /proc/timer_list' shows HPET is good.
Indeed, while I have no X Windows on fileserver, the
nearly-dead-of-old-age PATA drive I use to test webserver does have
X... and everything in this message below "*sweating profusely...*"
has been typed in the webmail client on webserver!
Thank you for reminding me to test the 2nd machine... it had utterly
slipped my mind, I was so overjoyed to have the fix! :)
Thanks Ingo and Yinghai (and other kernel gurus around the globe),
Dave W.