Re: `ucsi_acpi: probe of USBC000:00 failed with error -12` on Dell XPS 13 9360/9370

Re: `ucsi_acpi: probe of USBC000:00 failed with error -12` on Dell XPS 13 9360/9370

[Greg Kroah-Hartman]

On Fri, May 18, 2018 at 06:08:24PM +0300, Heikki Krogerus wrote:
> 
> Rafael, the problem here with these Dell laptops is that a memory page
> that is used as a mailbox for special communication between EC FW and
> OS (called UCSI mailbox) should be in uncached memory, but the memory
> is marked as cached.
> 
> I'm actually not even sure if the cache mode is something that is set
> based on information from ACPI tables, or is it something that the
> BIOS handles separately?
> 
> If it's information coming from ACPI tables, can we help pin point the
> place the BIOS guys at Dell should fix?

Given that Windows works with this same set of ACPI tables, perhaps the
requirement isn't that this memory location be uncached?

We shouldn't have to fix up ACPI tables just for Linux, we've been down
that path before :(

thanks,

greg k-h

Re: `ucsi_acpi: probe of USBC000:00 failed with error -12` on Dell XPS 13 9360/9370

[Rafael J. Wysocki]

On Fri, May 18, 2018 at 5:15 PM, Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
> On Fri, May 18, 2018 at 06:08:24PM +0300, Heikki Krogerus wrote:
>>
>> Rafael, the problem here with these Dell laptops is that a memory page
>> that is used as a mailbox for special communication between EC FW and
>> OS (called UCSI mailbox) should be in uncached memory, but the memory
>> is marked as cached.

Well, why don't we map it as uncached, then?

>> I'm actually not even sure if the cache mode is something that is set
>> based on information from ACPI tables, or is it something that the
>> BIOS handles separately?

We map the ACPI tables as cached as a rule, so that may be the reason.

>> If it's information coming from ACPI tables, can we help pin point the
>> place the BIOS guys at Dell should fix?

I'm not sure where this particular piece comes from.

> Given that Windows works with this same set of ACPI tables, perhaps the
> requirement isn't that this memory location be uncached?
>
> We shouldn't have to fix up ACPI tables just for Linux, we've been down
> that path before :(

Yes, pretty much.

Thanks,
Rafael

Re: `ucsi_acpi: probe of USBC000:00 failed with error -12` on Dell XPS 13 9360/9370

[Heikki Krogerus]

On Fri, May 18, 2018 at 10:04:02PM +0200, Rafael J. Wysocki wrote:
> On Fri, May 18, 2018 at 5:15 PM, Greg Kroah-Hartman
> > Given that Windows works with this same set of ACPI tables, perhaps the
> > requirement isn't that this memory location be uncached?
> >
> > We shouldn't have to fix up ACPI tables just for Linux, we've been down
> > that path before :(
> 
> Yes, pretty much.

That's not what I'm suggesting. AFAIK these Dell laptops are the only
ones with this problem, so if we accept this, I fear it means we have
to start maintaining DMI quirk table in the driver, and then:

if (dell)
        ioremap_cache()
else
        ioremap_nocache()

I was really hoping we could avoid that. Or is there something else we
can do in the driver?

I have no idea how the windows driver works, so I will double check
and test this next week, but our driver has really worked fine so far.


Thanks,

-- 
heikki