Date:   Tue, 12 Sep 2017 18:52:42 +0200
From:   "Luis R. Rodriguez" <mcgrof@...nel.org>
To:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc:     "Luis R. Rodriguez" <mcgrof@...nel.org>,
        Arend Van Spriel <arend.vanspriel@...adcom.com>,
        Marcel Holtmann <marcel@...tmann.org>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Gabriel C <nix.or.die@...il.com>, Tejun Heo <tj@...nel.org>,
        "Gustavo F. Padovan" <gustavo@...ovan.org>,
        Sukumar Ghorai <sukumar.ghorai@...el.com>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        "bluez mailin list (linux-bluetooth@...r.kernel.org)" 
        <linux-bluetooth@...r.kernel.org>
Subject: Re: btusb "firmware request while host is not available" at resume

On Mon, Sep 11, 2017 at 05:48:52PM -0700, Greg Kroah-Hartman wrote:
> On Mon, Sep 11, 2017 at 10:06:51PM +0200, Luis R. Rodriguez wrote:
> > On Mon, Sep 11, 2017 at 12:29:55PM -0700, Greg Kroah-Hartman wrote:
> > > On Mon, Sep 11, 2017 at 07:11:38PM +0200, Luis R. Rodriguez wrote:
> > > > On Mon, Sep 11, 2017 at 06:46:47AM -0700, Greg Kroah-Hartman wrote:
> > > > > To confirm, reverting this fixes the problem I was seeing in 4.13.  I've
> > > > > queued it up for the next 4.13-stable release as well.
> > > > 
> > > > Commit 81f95076281f ("firmware: add sanity check on shutdown/suspend") may
> > > > seem kludgy but the reason for it was to cleanup the horrible forced and
> > > > required UMH lock even when the UMH lock was *not* even needed, which was later
> > > > removed via commit 06a45a93e7d34aa ("firmware: move umh try locks into the umh
> > > > code").
> > > 
> > > So what does this mean now that it is reverted?
> > 
> > We discuss what we should do about upkeeping a warning in the future, as
> > I think technically the warning was still valid and it could help avoid
> > racy lookups with the filesystem which otherwise could have gone unnoticed.
> 
> Given that things that used to work before the patch, and then didn't
> work after the patch was added, but then worked again after the patch
> was reverted, I don't think that adding the warning back is good, if it
> still breaks things...

As I noted to Marcel, it would seem this issue was present since before, the
warning either was changed or was more silent before. If reverting commit
81f95076281f "firmware: add sanity check on shutdown/suspend") then one should
consider reverting 06a45a93e7d34aa ("firmware: move umh try locks into the umh
code") as otherwise we end up in a situation where we loose determinism to be
sure of what will happen. From what I gather prior to these two patches we
either got a warning on sync case, or an error on async, after it we get a
warning on either and it should be clearer what the issue should be. I believe
that prior to these two commits the reason why things were failing could have
been pretty obscure. With just commit 81f95076281f reverted we then really
change the logic, and *that* was something I did want to avoid, given I
suspected that kernel was already implicitly depending on the functionality
which the UMH lock had brought upon the non-UMH code path. My suspicions seem
to have been correct.

With just commit 81f95076281f reverted, we may find firmware sometimes, but
we may also just as easily not find firmware some other times. It will depend
on the system.

> > > > Removing the old UMH lock even when the UMH lock was *not* needed was the right
> > > > thing to do but commit 81f95076281f (("firmware: add sanity check on
> > > > shutdown/suspend") was put in place as a safe guard as the lock was also
> > > > placing an implicit sanity check on the API. It ensures the API with the cache
> > > > was used as designed, otherwise you do run the risk of *not getting the
> > > > firmware you may need* -- Marcel seems to acknowledge this possibility.
> > > > 
> > > > It may be possible for us to already have in place safeguards so that upon
> > > > resume we are ensuring the path to the firmware *is* available, so IMHO we
> > > > should remove this *iff* we can provide this guarantee.  Otherwise the check is
> > > > valid. You see, although the UMH lock was bogus, it did implicitly ask the
> > > > question: is it safe for *any* helper to run and make assumptions on the
> > > > filesystem then?
> > > > 
> > > > In lieu of this question being answered the warning is valid given the design
> > > > of the firmware API and the having the cache available as a measure to resolve
> > > > this race.
> > > 
> > > I don't understand what you are trying to say here at all.
> > > 
> > > To be specific, what, if anything, is a problem with the current state
> > > of Linus's tree (and the next 4.13-stable release)?
> > 
> > The warning is issued when drivers issue *new* firmware requests on resume. The
> > firmware API cache was designed to enable drivers to easily be able to request
> > firmware on resume without concern about races against the filesystem, but in order
> > for this to work the drivers must have requested the firmware prior to suspend.
> 
> Ok, then how is this all working today?

Prior to the above *two* mentioned commits a UMH lock was used, and so similar but
different errors should have come up. With just 81f95076281f "firmware: add
sanity check on shutdown/suspend") reverted there is no suspend heuristic
check and we just let the users fend for themselves, it may or may not work
this will very depending on the system.

> I think much more is going on here, sorry, this doesn't make sense.

The UMH lock on the non UMH path was nasty to begin with. We've long depended
on it though, my two commits were an attempt to do us away from UMH lock on
the non-UMH path. It was not trivial to reach a fair middle ground.

> > > If something needs to be fixed, can you make a patch showing that?  Or
> > > do we also need to revert anything else as well to get back to a "better
> > > working" state?
> > 
> > I took a look at the driver and it seems that btusb_setup_intel_new() is
> > not called after the driver is initialized, rather its called only when
> > hci_dev_do_open() is called. Its not clear to me how you can end up calling
> > this on resume but not prior to this on a running system. Feedback from
> > someone more familiar with bt would be useful.
> 
> open is a great place to want to load firmware, that makes sense, you
> get to defer loading stuff until you actually need it.  I see nothing
> wrong here (again, especially as it _does_ work...)

I agree. I'm very well aware of drivers doing "stupid things" on probe
or init. But the issue here seems to have been present since before, I
see one report since 3.13 on the path I expect to be hit in similar
situations but without this commit on the old UMH lock path:

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1356076

> > If it confirmed this information is helping avoid these races we can reconsider
> > re-instating the warn as a firmware dev debugging aid for developers.
> 
> Again, obviously this was not just a "warning" message, you changed the
> logic here.

I'm afraid the logic changes *more* with *just* commit 81f95076281f reverted.
I intentionally tried to avoid that.

> Anyway, all is good now, I guess we don't have to worry about it :)

I'll discuss with Marcel a proper long resolution, as it stands though with
just commit 81f95076281f reverted I'm afraid the situation may be worse off
long term even though it may seem things are peachy for some systems right off
the bat. We've long depended on the UMH lock path, my change was an attempt to
move us away from it and be more explicit about the issue.

  Luis

Powered by blists - more mailing lists