Re: [PATH 1/5 -v2] acpi, IO memory pre-mapping and atomic accessing

[Thomas Renninger <trenn@suse.de>]

On 12/16/2009 02:23 AM, Huang Ying wrote:
> On Wed, 2009-12-16 at 01:47 +0800, Bjorn Helgaas wrote: 
>> On Monday 14 December 2009 06:04:13 pm Huang Ying wrote:
...
>>> The checking here is for bug in firmware not software. I think it is
>>> necessary for the user to know where the bugs may come from, and it is
>>> hard to express the bug in return code.
>>
>> Yes, I understand that this is checking for firmware bugs.  My point
>> is that when a user sees this in his dmesg log:
>>
>> 	Invalid bit width in GAR, firmware bug?
>>
>> we have no context, and even a kernel developer can't figure out what
>> to do.  We could ask for a copy of the FADT and DSDT, but even then,
>> we don't know *which* GAR structure to look at, so we'll probably have
>> to add some instrumentation and ask the user to reproduce the problem.
>>
>> If the check were in the caller, it could at least say something like:
>>
>> 	ACPI: couldn't map generic address [io 0xcf8] for PCI config access
>>
>> which would give us more useful information.
> 
> En... Yes, some information about the invalid GAR is helpful. But the
> GAR information is available in acpi_check_gar too, so we can output
> something as follow in acpi_check_gar:
> 
>     Invalid bit width in GAR [mem 0x8029ff8 24], firmware bug?
> 
> Your message looks like a software issue instead of firmware bug. This
> may confuse the user and developer.

Please use one of these (cmp. with include/linux/kernel.h when to use
which):
#define FW_BUG          "[Firmware Bug]: "
#define FW_WARN         "[Firmware Warn]: "
#define FW_INFO         "[Firmware Info]: "

It's great that someone cares to tell syslog/users that it's a firmware
bug, that should get more common...

      Thomas