On 06/15/2017 02:19 AM, Andy Shevchenko wrote:
On Thu, Jun 15, 2017 at 2:21 AM,Agree. Setting irq_valid_mask would be the proper approach to skip IRQ for some GPIO pins. But commit 9a752b4c9ab9 added the GPIO index based checks in other IRQ set/unset functions in this driver and missed to add it only in this update_irq_ctrl() function. May be I can submit a separate patch to clean it up and use logic based on setting irq_valid_mask.
From: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@xxxxxxxxxxxxxxx>Shouldn't it have been done using irq_valid_mask flag in the first place?
Commit 9a752b4c9ab9 ("gpio: crystalcove: Do not write regular gpio
registers for virtual GPIOs") added support to skip GPIO register
update for virtual GPIOs, but it missed to add skip logic in
crystalcove_update_irq_ctrl() function. This patch fixes it.
@@ -134,6 +134,9 @@ static void crystalcove_update_irq_ctrl(struct crystalcove_gpio *cg, int gpio)
int reg = to_reg(gpio, CTRL_IN);
+ if (reg < 0)
regmap_update_bits(cg->regmap, reg, CTLI_INTCNT_BE, cg->intcnt_value);
Even if we do the cleanup patch, I would still prefer to have this check. Since to_reg() can return value < 0, its good to check it before passing it to regmap_* functions. Let me know your comments.