[Bug 238037] [PATCH] Implement ig4 suspend/resume

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=238037

--- Comment #8 from Austin Shafer <ashafer at badland.io> ---
(In reply to J.R. Oldroyd from comment #7)

I remember running into this problem while working on this patch. I think I
know what's causing it, but couldn't find a way to fix it. I also don't think
it's related to this bug's patch, as it would happen to me when first booting.

Basically because of the way that the iichid attaches, I think it is possible
for the iichid device to be initialized before the i2c bus. The iichid will try
to read interrupts before they are available and find no data until things are
reset.

>From D16698:
"I tried a few approaches to identify and attach an iichid device by querying
ACPI directly from the iichid module. That failed on two details: parent
association (in general HID over I2C devices appear below ACPI0 in NewBus but
it is necessary to have them below the iicbus*/iic* nodes, moving these node in
NewBus failed for me with a panic) and also interrupt handling (apparently,
interrupts must be handled by the nodes the IRQs are assigned to (i.e. the node
below ACPI0). Trying to attach to an IRQ from another node (i.e. a node below
iicbus*/i2c*) fails)"

So the iichid is below ACPI0, and I think that gets attached before iicbus. I
have a vague memory of adding printf's to show that this was happening. I'll
try to dig around and see if I have any patches of what I was trying to do to
fix this.

I'm no expert on newbus device ordering, so if I'm wrong I'd love an
explanation.

-- 
You are receiving this mail because:
You are the assignee for the bug.