Re: [PATCH v2 2/2] module: Merge same-name module load requests

[Luis Chamberlain <mcgrof@kernel.org>]

On Mon, Nov 14, 2022 at 09:57:56AM +0100, David Hildenbrand wrote:
> On 12.11.22 02:47, Luis Chamberlain wrote:
> > On Wed, Oct 19, 2022 at 02:00:55PM +0200, Petr Pavlu wrote:
> > > On 10/18/22 20:33, Luis Chamberlain wrote:
> > > > On Sat, Oct 15, 2022 at 11:27:10AM +0200, Petr Pavlu wrote:
> > > > > The patch does address a regression observed after commit 6e6de3dee51a
> > > > > ("kernel/module.c: Only return -EEXIST for modules that have finished
> > > > > loading"). I guess it can have a Fixes tag added to the patch.
> > > > > 
> > > > > I think it is hard to split this patch into parts because the implemented
> > > > > "optimization" is the fix.
> > > > 
> > > > git describe --contains 6e6de3dee51a
> > > > v5.3-rc1~38^2~6
> > > > 
> > > > I'm a bit torn about this situation. Reverting 6e6de3dee51a would be the
> > > > right thing to do, but without it, it still leaves the issue reported
> > > > by Prarit Bhargava. We need a way to resolve the issue on stable and
> > > > then your optimizations can be applied on top.
> > > 
> > > Simpler could be to do the following:
> > > 
> > > diff --git a/kernel/module/main.c b/kernel/module/main.c
> > > index d02d39c7174e..0302ac387e93 100644
> > > --- a/kernel/module/main.c
> > > +++ b/kernel/module/main.c
> > > @@ -2386,7 +2386,8 @@ static bool finished_loading(const char *name)
> > >   	sched_annotate_sleep();
> > >   	mutex_lock(&module_mutex);
> > >   	mod = find_module_all(name, strlen(name), true);
> > > -	ret = !mod || mod->state == MODULE_STATE_LIVE;
> > > +	ret = !mod || mod->state == MODULE_STATE_LIVE
> > > +		|| mod->state == MODULE_STATE_GOING;
> > >   	mutex_unlock(&module_mutex);
> > >   	return ret;
> > > @@ -2566,7 +2567,8 @@ static int add_unformed_module(struct module *mod)
> > >   	mutex_lock(&module_mutex);
> > >   	old = find_module_all(mod->name, strlen(mod->name), true);
> > >   	if (old != NULL) {
> > > -		if (old->state != MODULE_STATE_LIVE) {
> > > +		if (old->state == MODULE_STATE_COMING
> > > +		    || old->state == MODULE_STATE_UNFORMED) {
> > >   			/* Wait in case it fails to load. */
> > >   			mutex_unlock(&module_mutex);
> > >   			err = wait_event_interruptible(module_wq,
> > > @@ -2575,7 +2577,7 @@ static int add_unformed_module(struct module *mod)
> > >   				goto out_unlocked;
> > >   			goto again;
> > >   		}
> > > -		err = -EEXIST;
> > > +		err = old->state != MODULE_STATE_LIVE ? -EBUSY : -EEXIST;
> > >   		goto out;
> > >   	}
> > >   	mod_update_bounds(mod);
> > 
> > 
> > Prarit, can you verify this still does not break the issue you reported?
> > David, does this also fix your issue?
> 
> I didn't try, but from a quick glimpse I assume no. Allocating module space
> happens before handling eventual duplicates right now, before a module even
> is "alive" and in the MODULE_STATE_UNFORMED state.

The first two hunks are a revert of commit 6e6de3dee51a and I'm under
the impression that cauased your issues as *more* modules states are
allowed through.

The last hunk tries to fix what 6e6de3dee51a wanted to do.

> But maybe I am missing something important.

Please do test if you can.

  Luis