increase timeout in ipts_control_wait_flush

[leofricke]

This fixes a rare issue on my Surface Pro 5 (Arch Linux). Happens about every tenth boot.

log: dmesg.txt

As far as i debugged this, the module switches the mode from event to poll (see ipts_eds1_switch_mode) thus calling ipts_control_restart. During shutdown the IPTS_CMD_QUIESCE_IO command sometimes times out in ipts_control_wait_flush returning EAGAIN (-11).
If the module tries to start again, the device is still "flushing" leading to the failure in ipts_control_get_device_info, which obviously makes the touchscreen not working.

Increasing the timeout in ipts_control_wait_flush to 10 sec fixes this on my device, but maybe you have a nicer solution?? Alternatively we could increase the msleep timeout in ipts_control_restart but i think this PR should be the better option.

[DoubleLama]

Same thing here on a Surface Pro 5, kernel 6.19.8-surface-3. Your diagnosis matches exactly what I see: QUIESCE_IO times out with -11 in ipts_control_wait_flush during the event to poll switch, then the restart fails.

More often than on your machine though. Counted over 13 boots in my journal, 7 logged recv failed: -11, so about half rather than one in ten.

When it degenerates it ends up as the kthread_stop use-after-free from issue #1181, with iptsd stuck in D state and suspend and shutdown hanging after that.

For what it is worth, ishland's fork takes a broader approach, bumping the timeout for every ipts_mei_recv call from 1s to 2s in mei.h. I built that branch against 6.19.8 headers and it works here. Your one line is more targeted.

Happy to test this patch if that helps move it along.