Le mar. 2 janv. 2024 à 21:23, Sébastien Chaumat <euidz...@gmail.com> 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
