Re: [PATCH] Revert "i2c: designware: defer probe if child GpioInt controllers are not bound"

[Andi Shyti]

Hi Hardik,

On Fri, Aug 14, 2026 at 03:37:19PM +0530, Hardik Prakash wrote:
> This reverts commit 0a4bb2abc3e56d7be6e69b050c88ba52c87e22bf.
> 
> The reverted commit causes a regression on ThinkPad T14s Gen 4 (AMD):
> the touchpad's I2C controller fails with lost arbitration errors,
> because it delays i2c-designware's probe by roughly 500ms, which
> shifts the touchpad's first HID descriptor fetch into a window where
> the platform's embedded controller is still acting as a secondary I2C
> bus master. Debug tracing confirms the GpioInt dependency check itself
> behaves correctly (it defers appropriately and confirms the GPIO
> controller is bound); the arbitration failure happens roughly a
> second after the check passes, when i2c_hid_acpi's own probe attempts
> its first transaction.
> 
> The original fix is still needed for the Lenovo Yoga 7 14AGP11
> touchscreen race the commit addressed, but a corrected version will
> be resubmitted once the EC bus-mastering interaction is understood
> and handled properly, rather than reintroducing a different
> regression on more widely-used ThinkPad hardware in the meantime.
> 
> Reported-by: Thorsten Leemhuis <linux@xxxxxxxxxxxxx>
> Closes: https://lore.kernel.org/all/b4a4eadb-282f-464c-843a-19d415a34d0c@xxxxxxxxxxxxx/
> Signed-off-by: Hardik Prakash <hardikprakash.official@xxxxxxxxx>

merged to i2c/i2c-2.

Thanks,
Andi