Message-ID: <20200723152054.hrtahg77ycg5ovud@gilmour.lan>
Date:   Thu, 23 Jul 2020 17:20:54 +0200
From:   Maxime Ripard <maxime@...no.tech>
To:     Ondřej Jirman <megous@...ous.com>,
        linux-sunxi@...glegroups.com,
        Vasily Khoruzhick <anarsoul@...il.com>,
        Yangtao Li <tiny.windzz@...il.com>,
        Zhang Rui <rui.zhang@...el.com>,
        Daniel Lezcano <daniel.lezcano@...aro.org>,
        Amit Kucheria <amit.kucheria@...durent.com>,
        Chen-Yu Tsai <wens@...e.org>,
        "open list:ALLWINNER THERMAL DRIVER" <linux-pm@...r.kernel.org>,
        "moderated list:ARM/Allwinner sunXi SoC support" 
        <linux-arm-kernel@...ts.infradead.org>,
        open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] thermal: sun8i: Be loud when probe fails

On Mon, Jul 13, 2020 at 01:29:42AM +0200, Ondřej Jirman wrote:
> Hi Maxime,
> 
> On Wed, Jul 08, 2020 at 03:57:48PM +0200, Maxime Ripard wrote:
> > On Wed, Jul 08, 2020 at 03:44:41PM +0200, Ondřej Jirman wrote:
> > > >
> 
> [...]
> 
> > > > Yeah, but on the other hand, we regularly have people that come up and
> > > > ask if a "legitimate" EPROBE_DEFER error message (as in: the driver
> > > > wasn't there on the first attempt but was there on the second) is a
> > > > cause of concern or not.
> > > 
> > > That's why I also added a success message, to distinguish this case. 
> > 
> > That doesn't really help though. We have plenty of drivers that have
> > some sort of success message and people will still ask about that error
> > message earlier.
> > 
> > > > > And people run several distros for 3-4 months without anyone noticing any
> > > > > issues and that thermal regulation doesn't work. So it seems that lack of a
> > > > > success message is not enough.
> > > > 
> > > > I understand what the issue is, but do you really expect phone users to
> > > > monitor the kernel logs every time they boot their phone to see if the
> > > > thermal throttling is enabled?
> > > 
> > > Not phone users, but people making their own kernels/distributions. Those people
> > > monitor dmesg, and out of 4 distros or more nobody noticed there was an issue
> > > (despite the complaints of overheating by their users).
> > > 
> > > So I thought some warning may be in order, so that distro people more easily
> > > notice they have misconfigured the kernel or sometging.
> > 
> > I mean, then there's nothing we can do to properly address that then.
> > 
> > The configuration system is a gun, we can point at the target, but
> > anyone is definitely free to shot themself in the foot.
> > 
> > You would have exactly the same result if you left the thermal driver
> > disabled, or if you didn't have cpufreq support.
> 
> Right. Though I hope there's some middle ground. I mean all of those dev_err
> in error paths of many drivers are there mostly to help debugging stuff.

Adding all the error messages you have in that patch seems like a good
middle ground to me, and we could definitely use more of them in some
other drivers (like the USB PHY)

> And even though I was part of this driver's development, it took me quite
> some time to figure out it was the missing sunxi-sid driver causing the issue,
> with complete silence from the driver.
> 
> Maybe this can/will be solved at another level entirely, like having a device
> core report devices probes that failed with EPROBE_DEFER some time after
> the boot finished and modules had a chance to load, instead of immediately
> for each probe retry.

The thing is that there's never a point in time where "all the modules
had a chance to load". If you're loading the modules on demand and have
an hotpluggable bus, that might happen after a second or after a year,
you can't say.

The actual fix for this would be to use the on demand probing that seems
to be in the works and avoid EPROBE_DEFER entirely, but that probably
won't happen in a near future.

Maxime

Powered by blists - more mailing lists