kernel build fails

[michael_tn]

trying to put 2005.1 onto some ES40's. first, booting with the livecd fails when booting to 2.6.x kernel, is that normal? second, working through install guide, install fails at kernel build with:

gcc -Wp,-MD,arch/alpha/lib/.__remqu.o.d -nostdinc -isystem /usr/lib/gcc-lib/alpha-unknown-linux-gnu/3.3.2/include -D__KERNEL__ -Iinclude -D__ASSEMBLY__ -Wall -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing -fno-common -ffreestanding -O2 -fomit-frame-pointer -g -pipe -mno-fp-regs -ffixed-8 -msmall-data -mcpu=ev5 -Wa,-mev6 -Wdeclaration-after-statement -DREM -c -o arch/alpha/lib/__remqu.o arch/alpha/lib/divide.S
AS arch/alpha/lib/callback_srm.o
CC arch/alpha/lib/checksum.o
AS arch/alpha/lib/clear_page.o
....

CC init/version.o
LD init/built-in.o
LD .tmp_vmlinux1
kernel/built-in.o(.spinlock.text+0x5c): In function `migration_init':
kernel/sched.c:4500: undefined reference to `_raw_read_trylock'
kernel/built-in.o(.spinlock.text+0x64):kernel/sched.c:168: undefined reference to `_raw_read_trylock'
make: *** [.tmp_vmlinux1] Error 1


this is after an emerge vanilla-sources, make menu config and make && make modules_install.

any idea where i went wrong?

-- michael

[HPRichard]

AFAIK _raw_read_trylock is a ultra-low-level routine written in assembly language. It looks like it is not being found by the kernel build system, most probably because of not being present for that particular CPU.
I saw a similar issue on a SUN SS20 with 4x90 MHz ROSS HyperSPARC CPUs last week on LinuxTag. At least on this machine this low-level routine is only needed for building a SMP kernel. So you may want to try a UP kernel first and file a bug report/ask an Alpha assembly guru to fix this.
And sorry if I write totally rubbish, I do not know the ES40.