Re: [PATCH v1 5/5] driver core: Set fw_devlink=on by default

[Geert Uytterhoeven]

Hi Saravana,

On Wed, Jan 13, 2021 at 3:34 AM Saravana Kannan <saravanak@xxxxxxxxxx> wrote:
> On Mon, Jan 11, 2021 at 11:11 PM Marek Szyprowski
> <m.szyprowski@xxxxxxxxxxx> wrote:
> > On 11.01.2021 22:47, Saravana Kannan wrote:
> > > On Mon, Jan 11, 2021 at 6:18 AM Marek Szyprowski
> > > <m.szyprowski@xxxxxxxxxxx> wrote:
> > >> On 11.01.2021 12:12, Marek Szyprowski wrote:
> > >>> On 18.12.2020 04:17, Saravana Kannan wrote:
> > >>>> Cyclic dependencies in some firmware was one of the last remaining
> > >>>> reasons fw_devlink=on couldn't be set by default. Now that cyclic
> > >>>> dependencies don't block probing, set fw_devlink=on by default.
> > >>>>
> > >>>> Setting fw_devlink=on by default brings a bunch of benefits (currently,
> > >>>> only for systems with device tree firmware):
> > >>>> * Significantly cuts down deferred probes.
> > >>>> * Device probe is effectively attempted in graph order.
> > >>>> * Makes it much easier to load drivers as modules without having to
> > >>>>     worry about functional dependencies between modules (depmod is still
> > >>>>     needed for symbol dependencies).
> > >>>>
> > >>>> If this patch prevents some devices from probing, it's very likely due
> > >>>> to the system having one or more device drivers that "probe"/set up a
> > >>>> device (DT node with compatible property) without creating a struct
> > >>>> device for it.  If we hit such cases, the device drivers need to be
> > >>>> fixed so that they populate struct devices and probe them like normal
> > >>>> device drivers so that the driver core is aware of the devices and their
> > >>>> status. See [1] for an example of such a case.
> > >>>>
> > >>>> [1] -
> > >>>> https://lore.kernel.org/lkml/CAGETcx9PiX==mLxB9PO8Myyk6u2vhPVwTMsA5NkD-ywH5xhusw@xxxxxxxxxxxxxx/
> > >>>> Signed-off-by: Saravana Kannan <saravanak@xxxxxxxxxx>
> > >>> This patch landed recently in linux next-20210111 as commit
> > >>> e590474768f1 ("driver core: Set fw_devlink=on by default"). Sadly it
> > >>> breaks Exynos IOMMU operation, what causes lots of devices being
> > >>> deferred and not probed at all. I've briefly checked and noticed that
> > >>> exynos_sysmmu_probe() is never called after this patch. This is really
> > >>> strange for me, as the SYSMMU controllers on Exynos platform are
> > >>> regular platform devices registered by the OF code. The driver code is
> > >>> here: drivers/iommu/exynos-iommu.c, example dts:
> > >>> arch/arm/boot/dts/exynos3250.dtsi (compatible = "samsung,exynos-sysmmu").
> > >> Okay, I found the source of this problem. It is caused by Exynos power
> > >> domain driver, which is not platform driver yet. I will post a patch,
> > >> which converts it to the platform driver.
> > > Thanks Marek! Hopefully the debug logs I added were sufficient to
> > > figure out the reason.
> >
> > Frankly, it took me a while to figure out that device core waits for the
> > power domain devices. Maybe it would be possible to add some more debug
> > messages or hints? Like the reason of the deferred probe in
> > /sys/kernel/debug/devices_deferred ?
>
> There's already a /sys/devices/.../<device>/waiting_for_supplier file
> that tells you if the device is waiting for a supplier device to be
> added. That file goes away once the device probes. If the file has 1,
> then it's waiting for the supplier device to be added (like your
> case). If it's 0, then the device is just waiting on one of the
> existing suppliers to probe. You can find the existing suppliers
> through /sys/devices/.../<device>/supplier:*/supplier. Also, flip
> these dev_dbg() to dev_info() if you need more details about deferred
> probing.

How are we supposed to check the contents of that file, if the system
doesn't even boot into userspace with a ramdisk? All hardware drivers
fail to probe. The only thing that works is "earlycon keep_bootcon",
and kernel output just stops after a while.

Thanks for your suggestions!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds