Re: 2 bug repport

[Ulf Hansson <ulf.hansson@linaro.org>]

On Wed, 22 Apr 2020 at 10:29, Russell King - ARM Linux admin
<linux@armlinux.org.uk> wrote:
>
> On Wed, Apr 22, 2020 at 03:03:57AM -0400, alpha_one_x86 wrote:
> > Hi,
> >
> > On mcbin platform I have uSD problem, repported but no reply on linux kernel
> > bugzilla, https://bugzilla.kernel.org/show_bug.cgi?id=207083
> >
> > Any idea what patch try?
>
> I think that's a question for the MMC people.
>
> If you go back to your working 4.20 kernel, does the problem go away?
> If so, it sounds like a regression in the MMC subsystem.  If not, I
> wonder if it could be the uSD card going bad.
>
> However, I suspect the former.  I've seen one instance here where a
> Clearfog GT8k (Armada 8040 based just like the mcbin) running 5.6 with
> the rootfs on eMMC completely lost the ability to talk to the eMMC to
> the point that the machine had to be power cycled to recover it -
> merely rebooting did not.  I don't know the cause - the initial failure
> had vanished from the kernel logs, and because the eMMC was no longer
> accessible, the rsyslog files did not contain the details either.
> I've since setup remote logging, and I'm currently waiting for it to
> happen again.  I couldn't say if _that_ is a regression because I
> haven't been using the GT8k until very recently, and I tend not to use
> eMMC/uSD on the Macchiatobin that runs 24x7.

In v5.6 the following commits were introduced - and it has turned out
that a few other sdhci controllers had issues with them (which we have
fixed now).

Perhaps you can try a v5.6 kernel and just revert them and see if that
could help - as a start

533a6cfe08f9 ("mmc: core: Default to generic_cmd6_time as timeout in
__mmc_switch()")
ad91619aa9d7 ("mmc: block: Use generic_cmd6_time when modifying
INAND_CMD38_ARG_EXT_CSD")
24ed3bd01d6a ("mmc: core: Specify timeouts for BKOPS and CACHE_FLUSH for eMMC")

Kind regards
Uffe