On Sep 01, 2009 15:06 -0400, Brian J. Murrell wrote: > On Tue, 2009-09-01 at 13:47 -0500, Nirmal Seenu wrote: > > In lustre 1.8.0.1 it used to be: > > > > journal->j_commit_timer.expires = round_jiffies(transaction->t_expires); > > > > and it is the following in Lustre 1.8.1: > > > > journal->j_commit_timer.expires = transaction->t_expires; > > Yes, that's one of the differences I listed.
We've since modified this change to use "round_jiffies_up()", which fixes the upstream problem, while avoiding the hangs it caused in Lustre. > > The only other differnece is from the 8kstack-2.6.12.patch which used to > > modify the following line in include/asm-i386/thread_info.h > > > > from > > #define THREAD_SIZE (4096) > > to > > #define THREAD_SIZE (8192) > > > > It seems like, it would be safe for me to rollback the changes to > > fs/jbd/transaction.c and continue using 2.6.22.19 kernel. I think we just ensure that the CONFIG_4K_STACK is not enabled on servers now. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc. _______________________________________________ Lustre-discuss mailing list [email protected] http://lists.lustre.org/mailman/listinfo/lustre-discuss
