Re: [RFC 2/5] fs: freeze on suspend and thaw on resume

From: Luis R. Rodriguez
Date: Tue Oct 03 2017 - 16:23:44 EST

On Tue, Oct 03, 2017 at 08:02:22PM +0000, Bart Van Assche wrote:
> On Tue, 2017-10-03 at 11:53 -0700, Luis R. Rodriguez wrote:
> > +static bool super_allows_freeze(struct super_block *sb)
> > +{
> > + return !!(sb->s_type->fs_flags & FS_FREEZE_ON_SUSPEND);
> > +}
> A minor comment: if "!!" would be left out the compiler will perform the
> conversion from int to bool implicitly

For all compilers?

> so I propose to leave out the "!!" and parentheses.


> Anyway, I agree with the approach of this patch and I think
> that freezing filesystems before processes are frozen would be a big step
> forward.

Great! But please note, the current implementation calls fs_suspend_freeze()
*after* try_to_freeze_tasks(), ie: this implementation freezes userspace and
only after then filesystems.

Order will be *critical* here to get right, so we should definitely figure
out if this is definitely the right place (TM) to call fs_suspend_freeze().

Lastly, a final minor implementation note:

I think using a PM notifier would have been much cleaner, in fact it was my the
way I originally implemented this orthogonally to Jiri's work, however to get
this right the semantics of __pm_notifier_call_chain() would need to be
expanded with another state, perhaps PM_USERSPACE_FROZEN, for example. I
decided in the end a new state was not worth it give we would just have one
user: fs freezing. So to be clear using a notifier to wrap this code into
the fs code and not touching kernel/power/process.c is not possible with
today's semantics nor do I think its worth it to expand on these semantics.

This approach is explicit about order and requirements for those that should
care: those that will maintain kernel/power/process.c and friends. Having
this in a notifier would shift this implicitly.