On Wed, Jun 11, 2025 at 5:03 PM Bjorn Andersson andersson@kernel.org wrote:
On Wed, Jun 11, 2025 at 04:39:11PM +0200, Bartosz Golaszewski wrote:
From: Bartosz Golaszewski bartosz.golaszewski@linaro.org
When requesting pins whose intr_detection_width setting is not 1 or 2 for interrupts (for example by running `gpiomon -c 0 113` on RB2), we'll hit a BUG() in msm_gpio_irq_set_type(). Potentially crashing the kernel due to an invalid request from user-space is not optimal, so let's go through the pins and mark those that would fail the check as invalid for the irq chip as we should not even register them as available irqs.
I had to go dig into the code to understand why there is a problem with GPIO 113 on RB2 (i.e. UFS_RESET on SM6115)... I think it would have been better to document the actual reason for the problem, which is:
"The UFS_RESET pin doesn't have interrupt logic, but is registered as a GPIO. Requesting the interrupt of this pin hits a BUG() in msm_gpio_irq_set_type() because intr_detection_width is invalid"
The UFS_RESET() case is the one I figured out initially but then realized that the issue will be the same for other non-PINGROUP() macros which set all the interrupt related fields to -1 so I made the message more generic. I will include the above in v2.
Bart