On 09/25/2013 04:16 AM, Vyacheslav Dubeyko wrote: > Unfortunately, I can't reproduce the issue, currently. > > Your description of the reproducing patch is slightly complex. So, I can > make something wrong (especially, for the case of rc script modification > because I am using Ubuntu 12.04). Sorry about that. I sent the real scripts so you can see the real settings. Really, I'm just trying to add a little bit of stress before the system daemons (sshd, ntpd, syslogd, klogd) are fully set up, then call shutdown. The case was 100% reproducible that way. > But I suspect that you have some additional workload before the issue > reproducing. Can you reproduce the issue for the case of freshly made > NILFS2 rootfs? Maybe fsstress is not the reason of the issue? Do you > have any NILFS2 related messages in your system log before the issue > reproducing? Yes, it can be done on a freshly made filesystem. In fact, NILFS2 seems more stable after it has filled the drive with checkpoints and cleaned them several times. I'll check my logs again for additional NILFS2 error messages. None have been noted, other than for the recovery messages on the next mount. I could be wrong. The usual load is syslogd, klogd, ntpd, acpid, gpm, and sshd. However, this is a debug kernel with various amounts of debug in it. Tracing, lockdep, netconsole, and kgdb are always in there. A crash kernel is always loaded, taking 9 MB of RAM. [If you wish, I can upload a kernel core, vmlinux, and System.map to my Google Drive, if you are good at working with crash dumps.] Sometimes, the CONFIG_PAGEALLOC and SLUB debugging are added to this mix. Also, in case it makes a difference, SLUB is the slab allocator, and the Deadline I/O scheduler is almost always used. > So, current description of the issue's reproducing path doesn't work for > me. No worries. If I have a problem with sync or umounts, it's either because a) I've compiled a lot of programs before trying a sync or umount, or b) more than one NILFS2/nilfs_cleanerd pair is open at the same time. All I'm hoping is that if such problems happen on your PC, they will be problems that you understand well already. For what it's worth, my problem NILFS2 / partition just withstood two full runs of xfstests (for XFS) and a full run of LTP. For now, the hang on shutdown is completely gone. Good luck! Michael -- To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
