On Wed, 10 May 2017 17:17:16 -0700
Pádraig Brady <[email protected]> wrote:

> On 10/05/17 15:49, Matias Fonzo wrote:
> > Hi there,
> > 
> > I am writing a shutdown script for the boot process.
> > 
> > I have a little question, which is exactly the order to apply
> > correctly sync(1)?:
> > 
> >   1. Before to unmount swap devices, local/remote filesystems?.
> > 
> >   2. Between the unmount of swap device(s) and local/remote
> >   file systems?.
> > 
> >   3. After to unmount everything?.
> > 
> >   4. ... ?.
> > 
> > Thanks,
> > Matias
> >   
> 
> Note sync(1) doesn't return any errors from sync(2)
> as the later always "succeeds".
> Also syncing is implicit in umount.
> Hence on modern systems calling sync(1) is IMHO redundant
> if doing a umount anyway.
> Consider also data written after the sync and before the umount.
> Therefore the kernel umount is the only place that can
> do that appropriate syncing without races.

Good.  I was thinking in do that -- before and after, I do not consider
the after as "explicit", because the halt implementation that I have,
does the sync, too.

Thanks again,
Matías

Attachment: pgpPnHIdDFu5y.pgp
Description: OpenPGP digital signature

Reply via email to