Sorry for the noise - I had missed the mlock patch to hugetlbfs/inode.c .
RChan wrote:
Hi all,
Before taking the plunge and trying Oracle 10g on devel kernels - I have been browsing the Oracle forums.
In upstream kernels there is a shm_hugetlb_group sysctl to allow non-suid Oracle processes to do their thing (otherwise complaints about not being able to get a shared memory segment).
Now devel kernels have the the upstream patch backed out:
# # Ugly hugetlb hack; the real solution (rlimits) is added later on # %patch10 -p1 -R
What do I have to do with rlimit to get Oracle processes to run? Does Oracle
have to fix the binary first?
Cheers richard
