Le 08/05/2023 à 23:20, Linus Walleij a écrit :
The Nokia 770 is using GPIOs from the global numberspace on the
CBUS node to pass down to the LCD controller. This regresses when we
let the OMAP GPIO driver use dynamic GPIO base.
The Nokia 770 now has dynamic allocation of IRQ numbers, so this
needs to be fixed for it to work.
As this is the only user of LCD MIPID we can easily augment the
driver to use a GPIO descriptor instead and resolve the issue.
The platform data .shutdown() callback wasn't even used in the
code, but we encode a shutdown asserting RESET in the remove()
callback for completeness sake.
The CBUS also has the ADS7846 touchscreen attached.
Populate the devices on the Nokia 770 CBUS I2C using software
nodes instead of platform data quirks. This includes the LCD
and the ADS7846 touchscreen so the conversion just brings the LCD
along with it as software nodes is an all-or-nothing design
pattern.
The ADS7846 has some limited support for using GPIO descriptors,
let's convert it over completely to using device properties and then
fix all remaining boardfile users to provide all platform data using
software nodes.
Dump the of includes and of_match_ptr() in the ADS7846 driver as part
of the job.
Since we have to move ADS7846 over to obtaining the GPIOs it is
using exclusively from descriptors, we provide descriptor tables
for the two remaining in-kernel boardfiles using ADS7846:
- PXA Spitz
- MIPS Alchemy DB1000 development board
It was too hard for me to include software node conversion of
these two remaining users at this time: the spitz is using a
hscync callback in the platform data that would require further
GPIO descriptor conversion of the Spitz, and moving the hsync
callback down into the driver: it will just become too big of
a job, but it can be done separately.
The MIPS Alchemy DB1000 is simply something I cannot test, so take
the easier approach of just providing some GPIO descriptors in
this case as I don't want the patch to grow too intrusive.
As we see that several device trees have incorrect polarity flags
and just expect to bypass the gpiolib polarity handling, fix up
all device trees too, in a separate patch.
Suggested-by: Dmitry Torokhov <dmitry.torokhov@xxxxxxxxx>
Fixes: 92bf78b33b0b ("gpio: omap: use dynamic allocation of base")
Signed-off-by: Linus Walleij <linus.walleij@xxxxxxxxxx>
---
[...]
diff --git a/drivers/video/fbdev/omap/lcd_mipid.c b/drivers/video/fbdev/omap/lcd_mipid.c
index 03cff39d392d..e4a7f0b824ff 100644
--- a/drivers/video/fbdev/omap/lcd_mipid.c
+++ b/drivers/video/fbdev/omap/lcd_mipid.c
@@ -7,6 +7,7 @@
*/
#include <linux/device.h>
#include <linux/delay.h>
+#include <linux/gpio/consumer.h>
#include <linux/slab.h>
#include <linux/workqueue.h>
#include <linux/spi/spi.h>
@@ -41,6 +42,7 @@ struct mipid_device {
when we can issue the
next sleep in/out command */
unsigned long hw_guard_wait; /* max guard time in jiffies */
+ struct gpio_desc *reset;
struct omapfb_device *fbdev;
struct spi_device *spi;
@@ -556,6 +558,12 @@ static int mipid_spi_probe(struct spi_device *spi)
return -ENOMEM;
}
+ /* This will de-assert RESET if active */
+ md->reset = gpiod_get(&spi->dev, "reset", GPIOD_OUT_LOW);
+ if (IS_ERR(md->reset))
+ return dev_err_probe(&spi->dev, PTR_ERR(md->reset),
+ "no reset GPIO line\n");
+
spi->mode = SPI_MODE_0;
md->spi = spi;
dev_set_drvdata(&spi->dev, md);
@@ -574,6 +582,8 @@ static void mipid_spi_remove(struct spi_device *spi)
{
struct mipid_device *md = dev_get_drvdata(&spi->dev);
+ /* Asserts RESET */
+ gpiod_set_value(md->reset, 1);
Hi,
should this also be done in the probe if mipid_detect() fails?
If yes, please also look at [1], that I've just sent, which introduces
an error handling path in the probe.
CJ
[1]:
https://lore.kernel.org/all/8b82e34724755b69f34f15dddb288cd373080390.1620505229.git.christophe.jaillet@xxxxxxxxxx/
mipid_disable(&md->panel);
kfree(md);
}
[...]
