Re: [BUG]i2c_hid_acpi broken with 4.17.2 on Framework Laptop 13 AMD

[Sébastien Chaumat <euidzero@xxxxxxxxx>]

Le mar. 2 janv. 2024 à 21:23, Sébastien Chaumat <euidzero@xxxxxxxxx> a écrit :

 output of gpioinfo

kernel alone :

        line   5: unnamed         input active-low consumer=interrupt
        line  84: unnamed         input active-low consumer=interrupt

xen:

        line   5: unnamed         input active-low
        line  84: unnamed         input active-low

xen with skipping IRQ7 double init :

        line   5: unnamed         input active-low consumer=interrupt
        line  84: unnamed         input active-low


So definitely progressing.

Checking /sys/kernel/irq/7

kernel alone :
 actions: pinctrl_amd
 chip_name: IR-IO-APIC
 hwirq: 7
 name: fasteoi
 per_cpu_count: 0,0,0,0,0,20,0,0,0,0,0,0,0,0,0,0
 type: level
 wakeup: enabled

xen skipping IRQ7 double init :

actions: pinctrl_amd
 chip_name: xen-pirq
 hwirq:

 name: ioapic-level
 per_cpu_count: 0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0
 type: edge
 wakeup: disabled

So the skip of IRQ7 in pci_xen_initial_domain() sets the correct handler

 (IIUC xen uses the ioapic-level and handles the eoi separately), but not the correct type (still edge).

I guess this may explains the results above.

 Mario (in CC) patched the pinctrl_amd to flush pending interrupt before starting the driver for the GPIO.

This helped in  the sense of there's no more pending interrupt on IRQ7 (whatever the handler is, level or edge) but then the touchpad is not detected by i2c-hid.

Is there any work in progress related to the incorrect IRQ configuration ?

Thanks,

Sébastien