> Hi!
>
> > Statistics only confuse the issue. Disks are only mounted/unmounted
>
> Agreed.
>
> My idea: Last time I checked, linux was not guaranteeing that all data
> are on disk after sync. Sync just means that data are scheduled in que
> somewhere. So you have to sleep 1 after sync, for it to work.
>
> But that is wrong. You should sleep_until_all_data_are written. But
> this is impossible as you don't know it. 1sec was just guess.
>
> I _think_ that there's something similar going on with unmount.
>
> Is it possible that similar issue is here? Could you please try
> recompiling w/o swap _AND_ w/o update? That should produce lots of
> unsaved data. Also watch disk activity few msec before machine is
> physically reset.
>
I will do it.
A long time ago, i.e., about 3 years ago, it was discovered that init
was syncing file-systems every time a child expired. This was removed
from most distributions and also from my SysV init sources. This
"help" by init may have been hiding a long-standing problem.
As I understand it, umount() should flush all buffers before it
marks the file-system R/O and takes it off-line. If it is relying
only on time, this might be the problem.
Cheers,
Dick Johnson
***** FILE SYSTEM MODIFIED *****
Penguin : Linux version 2.1.105 on an i586 machine (66.15 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu