[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: RAID trouble with SMP kernels



aphro wrote:

> On Tue, 30 Nov 1999, Adam C Powell IV wrote:
>
> hazels >Is it at all possible this could be a kernel compiler issue?  I built the kernel with
> hazels >the latest gcc, which is 2.95.2-3.
> hazels >
> hazels >If no answer by Friday I'll go to to the vger kernel list.
>
> ouch!!! dont ever ever ever ever ever ever (10000x ever) compile a kernel
> with gcc2.9x, you are asking for trouble there have always been issues
> with most compilers OTHER then gcc 2.7.2.3  read the kernel mailing list
> (http://kt.linuxcare.com)  there are a ton of issues in the kernel(not all
> of them problems with the compiler) that make for BIG problems when
> compiled with something other then 2.7.2.3

Update: the last one I read says Alan Cox uses egcs, I think 1.1.2.  And the 2.2.14 notes have a
number of fixes to let it compile with 2.95 (though not all of it does- see below).  The warning
is in order, but the 10000x never is a bit overblown.

> in short, use gcc2.7.2.3 for the kernel, use whatever you want for
> everything else :)

Okay, after finals and vacation, finally got around to getting gcc272 and building the kernel
(2.2.13) using it.  The compiler was *NOT* the problem- I get exactly the same result as
before.  That is, it seems to do mdadd and mdrun just fine, but when it goes to *mount* the
RAID-5 partition, it freezes.  The only silver lining is that if I don't catch lilo in time,
ckraid running under SMP 2.2.13 (and miraculously, under 2.0.36-scsimod from slink rescue!) is
about six times faster than under kernel-image-2.2.10!!

So, noticed that the 2.2.14 release notes indicate elimination of some IDE hangs, and built
2.2.14 with basically the same config as the default 2.2.13 (from the now-obsolete i386
kernel-patch package) but with SMP, and Ppro/686 instead of 386.  First tried with gcc-2.95
(latest potato), but that failed in drivers/hfmodem, so I did make-kpkg clean and built it with
gcc272.

And it *still* fails, same place.  It mounts a raid0 setup just fine, but not raid5.

To rehash: I'm trying to build an SMP kernel for an Abit BP6 dual-celeron motherboard with
RAID-5 over partitions on four IDE drives (hda-d; yes, I know it would be faster and more robust
over four IDE controllers, but I'm more interested in data integrity than speed, and haven't
tried making the Ultra-66 interfaces work yet).  The stock 2.2.10 kernel package (non-SMP) works
fine, but when I tried building 2.2.10 or 2.2.13 with gcc-2.95, or 2.2.13 or 2.2.14 with
gcc-2.7.2, the machine locks up when mounting the RAID device.

Please help if you can; if no answers within a week, I'll assume it's an unknown problem and go
to linux-kernel@vger.

Zeen,

                  Adam Powell                     http://lyre.mit.edu/~powell/
                  Thomas B. King Assistant Professor of Materials Engineering
                  77 Massachusetts Ave. Rm. 4-117         Phone (617) 452-2086
                  Cambridge, MA 02139 USA                   Fax (617) 253-5418


Reply to: