RE: Software RAID Stopped Working With Aurora Kernel

On Monday May 20, kc130iseo@coastalnet.com wrote:

quoted

    I don't know why, but in one instance it tried to load
      /lib/raid5.o
    and gets :unresolved symbol md_unregister_thread_R4ba824f9

    The other time it correctly loads
      /lib/modules/2.4.18-0.92sparc/kernel/drivers/md/raid5.o

    I suggest removing /lib/raid5.c

Strangely, the only time this error occurred was when I tried to boot with
"md0" devices listed in /etc/fstab. I don't know why it reported this path.
There are no modules directly under /lib. Here's where all the modules are.
I will remove all but the current kernel modules. In fact, I'll be removing
all the 2.2 kernels since I can't boot them any more anyway.

Odd.... you could do an "nm" of each raid5.o and see which one
mentions
   md_unregister_thread_R4ba824f9

Using your source RPM:

...

gcc -Wall -Werror -Wstrict-prototypes -DCONFFILE=\"/etc/mdadm.conf\" -O2 -m3
2 -m
tune=ultrasparc   -c -o mdadm.o mdadm.c
cc1: warnings being treated as errors
In file included from mdadm.h:57,
                 from mdadm.c:30:
md_p.h: In function `md_event':
md_p.h:168: warning: left shift count >= width of type
make: *** [mdadm.o] Error 1
error: Bad exit status from /var/tmp/rpm-tmp.90717 (%build)

Even odder.  It seems to think that __u64 is only 32bits wide.
Sounds like a compiler error, or an include file error.
I suspect __u64 is defined in /usr/include/asm/types.h to be
  typedef unsigned long __u64;

where it must need to be "unsigned long long", but I don't know enough
about the different flavours of sparc or the different compilers to
figure out the most likely problem.
Maybe just put
   #define __64 unsigned long long
in mdadm.h and see if that helps.

NeilBrown