mmc: tmio: Fixup runtime PM management during probe
authorUlf Hansson <ulf.hansson@linaro.org>
Fri, 13 Sep 2019 09:19:26 +0000 (11:19 +0200)
committerUlf Hansson <ulf.hansson@linaro.org>
Fri, 13 Sep 2019 11:49:04 +0000 (13:49 +0200)
commitaa86f1a3887523d78bfadd1c4e4df8f919336511
treed5cf7ea5e5a9f6b05dbf7ae0499791da730b672c
parent8861474a105c91e812d38cd65d3be795c7f32c4f
mmc: tmio: Fixup runtime PM management during probe

The tmio_mmc_host_probe() calls pm_runtime_set_active() to update the
runtime PM status of the device, as to make it reflect the current status
of the HW. This works fine for most cases, but unfortunate not for all.
Especially, there is a generic problem when the device has a genpd attached
and that genpd have the ->start|stop() callbacks assigned.

More precisely, if the driver calls pm_runtime_set_active() during
->probe(), genpd does not get to invoke the ->start() callback for it,
which means the HW isn't really fully powered on. Furthermore, in the next
phase, when the device becomes runtime suspended, genpd will invoke the
->stop() callback for it, potentially leading to usage count imbalance
problems, depending on what's implemented behind the callbacks of course.

To fix this problem, convert to call pm_runtime_get_sync() from
tmio_mmc_host_probe() rather than pm_runtime_set_active(). Additionally, to
avoid bumping usage counters and unnecessary re-initializing the HW the
first time the tmio driver's ->runtime_resume() callback is called,
introduce a state flag to keeping track of this.

Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
drivers/mmc/host/tmio_mmc.h
drivers/mmc/host/tmio_mmc_core.c