Bug#330664: Bug still pending - cont give more info ATM ...
I did not forget that bug; and I would like to solve it ASAP.
- it seems to me that my bug is not generic to the Debian-Alpha-SMP kernels, but
rather directly to all
Linux-SMP due to a likely bug in the PCI device Qlogick 4010 (scsi card)
with some module like
qla<something>
I think about that because all Gentoo SMP kernels 2.6 also hang on Alpha, and also
'ANY' SMP kernel in x86 when I insert that PCI card in my computer (either kernel
hand build, Debian, Gentoo) (only SMP kernel, whatever there are 1 or many CPUs, SMP
kernels only - does not hang with non SMP kernels on multi CPU computers )
Thats the actual way I am working ...
but, before confirming any thing, I need to:
come back home where I have access to suffisant hardware (SMP x86 and alpha box, the
faulty QLA, compile SMP and non SMP kernels for all of theose box, greb my serial
cable to dumps Linux logs, and most difficult: boot my Alpha systems on the required
SCSI drives but without having to use the QLA board !!!)
practical infos and relevant test are coming quiet late because it takes ages to
perform enough cross tests ...
(especially when you have only one CPU running on a quadri MB to compile stuff)
* I am not forgetting that I opened a CRITICAL bug. *
Note: qla driver is loaded by busy box in ramdisk on Debian, when I need to do 2
tests on Gentoo to help debing the Debian problem:
- boot scsi with non QLA driver
- boot scsi with non QLA driver, then load the qla driver with hardware NOT present
- boot scsi with non QLA driver, with QLA hardware present but no driver
- boot scsi with non QLA driver, with QLA hardware present, and load driver manually
(disable udev)
- boot scsi with QLA driver build static
(in case some one have time to perform those tests before me)
apart from the fact I could use hand made kernel under Debian, I am not able ATM to
ask the Debian kernels not to use busybox and/or udev at boot time (those fitting
the ramdisk).
Thats why I need to try hand made kernel with Debian, then Gentoo, with both udev
desactivated, and see whats up when loading the QLA driver manually.
A kernel bug /should/ be distribution independant, and should not tell about Gentoo
in Debian BTS ... I know. But some times, it is faster using an alternate distro to
try fast tests than finding out 'who to desactivate that feature' n a given distro.
Logging that is also slown down with 2 other bugs I am having with the Linux kernel
on my machines. Sorry.
--
DEMAINE Benoit-Pierre (aka DoubleHP ) http://www.demaine.info/
\_o< If computing were an exact science, IT engineers would not have work >o_/
Reply to: