[Letux-kernel] [REGRESSION] mmc: Set PROBE_PREFER_ASYNCHRONOUS for drivers that existed in v4.4

[Andreas Kemnade]

On Thu, 3 Dec 2020 14:03:46 -0800
Doug Anderson <dianders at chromium.org> wrote:

> Hi,
> 
> On Thu, Dec 3, 2020 at 1:43 PM Andreas Kemnade <andreas at kemnade.info> wrote:
> >
> > On Tue, 1 Dec 2020 21:00:39 +0100
> > Andreas Kemnade <andreas at kemnade.info> wrote:
> >  
> > > Hi,
> > >
> > > during probing of second mmc host on the GTA04A5 which has a
> > > WL1835 wifi chip attached, I get the following oops. From a first glance
> > > this seems to be a timing sensitive. Maybe it rings some bells for someone...
> > > so I hesitate a bit to start the bisect job.  
> > Did the bisect job.
> > 21b2cec61c04 (mmc: Set PROBE_PREFER_ASYNCHRONOUS for drivers that existed in v4.4)
> >
> > is the evil commit. Reverting it on top of v5.10-rc6 solves the problem.
> >  
> > > Wild guessing: bus width is set while mmc module is not active, so registers
> > > are not accessible.
> > > Kernel configuration is omap2plus_defconfig
> > >
> > > Regards,
> > > Andreas
> > >
> > >  
> 
> ...snip...
> 
> > > [    3.152587]  mmcblk0: p1 p2 p3 p4 < p5 p6 p7 p8 >
> > > [    3.159393] 0x000000880000-0x000020000000 : "File System"
> > > [    3.179534] omap_hsmmc 480b4000.mmc: omap_device_late_idle: enabled but no driver.  Idling  
> 
> I'm guessing that this is the key line.  It seems to come from
> omap_device_late_idle().
> 
> I don't know tons about that function, but a quick guess is that it needs a:
> 
> wait_for_device_probe();
> 
> ...at the start of it.  It seems like it's trying to idle devices that
> never got probed but not actually waiting for all devices to probe.
> It's just assuming that everything is done by late_initcall_sync().
> 
hmm, not sure, we have:
if (od->_driver_status != BUS_NOTIFY_BOUND_DRIVER &&
            od->_driver_status != BUS_NOTIFY_BIND_DRIVER) {
                if (od->_state == OMAP_DEVICE_STATE_ENABLED) {
                        dev_warn(dev, "%s: enabled but no driver.  Idling\n",
                                 __func__);
                        omap_device_idle(pdev);
                }
        }

apparently there is no code to set _driver_status to
BUS_NOTIFY_BOUND_DRIVER. I think that is the key problem. I will try to
create a patch to fix that.

Regards,
Andreas