Re: Reverting "libdiskfs: Shutdown pagers on startup_dosync."

[Justus Winter]

Samuel Thibault <samuel.thibault@ens-lyon.org> writes:

> Hello Justus,
>
> I had to revert your commit:

Yes, I saw that :/

> “
> commit f05b6bae5cbdf3c221eba05987dea7ab14e6524f
> Author: Samuel Thibault <samuel.thibault@ens-lyon.org>
> Date:   Mon Jan 15 22:00:11 2018 +0100
>
>     Revert "libdiskfs: Shutdown pagers on startup_dosync."
>     
>     This reverts commit 89a49ec8b3a847f3b770cff6a4a80b98b94ad7bf.
>     
>     There was a typo in diskfs_shutdown (FSYS_GOAWAY_FORCE || 
> FSYS_GOAWAY_RECURSE)

Damn you C compilers >,<

>     which was making it FSYS_GOAWAY_NOWAIT, and thus not actually syncing if
>     there happens to be clients (it was then returning EBUSY). Fixing this 
> into
>     FSYS_GOAWAY_FORCE | FSYS_GOAWAY_RECURSE then makes it wait for clients 
> (and
>     NOWAIT is not supported), and there are very often some left, thus not
>     syncing either.
> ”
>
> I don't know if the pager actually supports the operation that we want?
> (just crashing clients)

Yes it does.  I had to fix /hurd/startup to cope with that (it was
supposed to wire its pages, but that was broken if memory serves).

I'm still convinced that shutting down the servers is the right thing to
do.  The alternative is that the filesystem goes away with an unclean
disk image.

Justus