Re: [Qemu-devel] [PATCH] Stop VM on ENOSPC error

[Jamie Lokier]

Daniel P. Berrange wrote:
> > A corrupt VM with data loss sounds much worse than a stopped VM to me.
> 
> You're not corrupting data in current code - you're just unable to finish
> new writes, because an IO failure is propagated back to the guest. If the
> guest is properly checking for & handling I/O failures, it should be pretty
> much OK once the host space problem is resolved - perhaps a reboot + journal
> recovery. 

Think about journalling filesystem writes.  If one returns I/O error,
some following queued requests must not proceed, otherwise it puts the
filesystem in an inconsistent state.  That's what I mean by corruption.

What that in mind, please name any OS which properly checks for and
handles write I/O errors, aside from reporting EIO to apps.  I'm
pretty sure Linux is not among them.

(Heck, even ENOSPC isn't handled well in apps.  Why, I remember the
time Firefox corrupted by Bookmarks by storing a zero-length file due
to ENOSPC...  I remember a few times when after Make I had zero-length
.o object files, and we're not talking about my personal crappy
Makefiles but good quality ones, etc.  This seems to be quite common.)

-- Jamie