The Amstrad Delta on-board latch2 bit named MODEM_NRESET, now available as a GPIO pin AMS_DELTA_GPIO_PIN_NMODEM_RESET, is used to power up/down (bring into/out of a reset state) two distinct on-board devices simultaneously: the modem, and the voice codec. As a consequence, that bit is, or can be, manipulated concurrently by two drivers, or their platform provided hooks.
Instead of updating those drivers to use the gpiolib API as a new method of controlling the MODEM_NRESET pin state, like it was done to other drivers accessing latch2 pins, and still being vulnerable to potential concurrency conflicts, or trying to solve that sharing issue with a custom piece of code, set up a fixed regulator device on top of that GPIO pin and update both drivers to manipulate that regulator, not the GPIO pin directly.
I'm resubmitting v2 of this series, originally posted on 30 Dec 2011, with Mark's Ack already collected, now rebased on top of current linux-omap/omap1 tip, commit 967809bd7faf71ddc29c8081e0f21db8b201a0f4.
Since patch 2/4, "ASoC: cx20442: add bias control over a platform provided regulator", has already been applied by Liam (thanks!) and is already present in mainline, I've omitted it from this submission.
v2 changes against initial version: * in both the codec and the modem callbacks, don't track the regulator enable/disable state, compare new target bias level (the codec case) or power state (the modem case) with the old value instead; thanks to Mark Brown who suggested this solution, * a few other minor changes, mostly stylistic.
Janusz Krzysztofik (3): ARM: OMAP1: ams-delta: set up regulator over modem reset GPIO pin ARM: OMAP1: ams-delta: update the modem to use regulator API ASoC: OMAP: ams-delta: drop .set_bias_level callback
arch/arm/mach-omap1/Kconfig | 2 + arch/arm/mach-omap1/board-ams-delta.c | 94 +++++++++++++++++++-- arch/arm/plat-omap/include/plat/board-ams-delta.h | 1 - sound/soc/omap/ams-delta.c | 32 ------- 4 files changed, 87 insertions(+), 42 deletions(-)