umount and data is lost?

[shimi]

On Sat, Nov 5, 2011 at 11:12 PM, Diego <elcuco at kde.org> wrote:

> Hi list,
>
> I am trying to debug this problem in a system I am developing:
>
> at some point I need to write a number to a file, and then reboot via an
> external device the machine. That device will cut of the power for a
> moment,
> and the machine will power on again (this is to overcome a fault we could
> not
> overcome otherwise). Before the reboot, I "umount" the proper partition and
> then "sync".
>
> For some reason, on one machine we see that the data is not updated and the
> original data is kept in the file. I konw empirically that if I add a
> "sleep(5)" after the umount, and before the "reboot" the data is written to
> the disk.
>
> Any tips how to debug this?
>  * the system is a TI 2.6.32 kernel (OMAP)
>  * the data is saved using fopen()
>  * disk is umounted using system("umount /data"), and then I call from C to
> sync();
>  * Cannot print debug this easily since the reboot is automated at the
> start
> of the APP, and then busybox's "login" breaks my serial terminal (like this
> eats "\n", git fixes if I press enter)
>  * The application is also multi-threaded to add more difficulty ...
>  * FS is JFFS2
>  * again  the workaround is a "sleep(5)"
>
> Any ideas?
>
> man 2 sync says:

BUGS
       According  to  the standard specification (e.g., POSIX.1-2001),
sync() schedules the writes,
       but may return before the actual writing is done.  However, since
version 1.3.20 Linux  does
       actually  wait.   (This  still  does  not  guarantee data integrity:
modern disks have large
       caches.)

maybe that's your culprit?

-- Shimi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.huji.ac.il/pipermail/linux-il/attachments/20111105/6cafe37d/attachment.html>