Re: [PATCH] pci, add FW_BUG warning to pci= kernel option

[Prarit Bhargava <prarit@redhat.com>]

On 11/17/2014 06:42 PM, Bjorn Helgaas wrote:
> On Mon, Nov 10, 2014 at 8:59 PM, Bjorn Helgaas <bhelgaas@google.com> wrote:
>> [+cc lkml, linux-arch, Linus]
>>
>> On Sat, Nov 01, 2014 at 02:11:19PM -0400, Prarit Bhargava wrote:
>>> The kernel should boot PCI without the use of kernel parameters.  Display
>>> a FW_BUG warning when pci= is used.
>>>
>>> Cc: Myron Stowe <mstowe@redhat.com>
>>> Cc: linux-pci@vger.kernel.org
>>> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
>>> ---
>>>  drivers/pci/pci.c |    2 ++
>>>  1 file changed, 2 insertions(+)
>>>
>>> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>>> index 625a4ac..5172060b 100644
>>> --- a/drivers/pci/pci.c
>>> +++ b/drivers/pci/pci.c
>>> @@ -4515,6 +4515,8 @@ static int __init pci_setup(char *str)
>>>               }
>>>               str = k;
>>>       }
>>> +     add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK);
>>> +     pr_crit(FW_BUG "The PCI configuration has been overridden thorugh the use of pci=.  Please report the issue you are attempting to resolve to your hardware vendor.\n");
>>
>> My goal is to be able to boot without any "pci=" parameters, so from that
>> perspective, I like this.
>>
>> When people have a problem booting Linux, they often try a variety of
>> things like "pci=assign-busses", "pci=nocrs", "pci=nomsi", "pci=nommconf",
>> "pci=noacpi", etc.  If they find something that works, there's a tendency
>> to treat that as a "solution."  I'd rather that they report it, so we can
>> try to fix the bug or add a quirk so the *next* person won't have to figure
>> out the right parameters to use.
>>
>> My worry is that there are a few things where Linux isn't smart enough to
>> do the right thing automatically, and I don't think we'll have a good
>> solution in the near future.  For example:
>>
>>     pci=norom
>>     pci=pcie_bus_perf,pcie_bus_safe,etc.
>>     pci=cbiosize=...
>>     pci=cbmemsize=...
>>     pci=resource_alignment=...
>>     pci=hpiosize=...
>>     pci=hpmemsize=...
>>     pci=realloc
>>
>> I don't like the fact that these options exist, but I suspect there are
>> users that do depend on them and might find this warning too aggressive.
>>
>> So I'm interested in opinions on whether this is a good idea at all and
>> whether we should exclude some options from the warning/tainting.
> 
> Hi Prarit,
> 
> We didn't get any nibbles :)
> 
> I don't think we can realistically taint the kernel for *all* "pci="
> options.  So if you want to pursue this, maybe you could enhance this
> so we can have a whitelist of options that we will support without
> tainting.  Then we can argue about what should be on the list and go
> from there.  Does that seem reasonable?

Sorry for the lower-case typing.  I broke my elbow and am down to one hand.

bjorn, yep sounds reasonable.  i'll do this (slowly unfortunately because of my
arm) ...

P.

> 
> This is a somewhat more aggressive use of tainting than we've had in
> the past, so we might have to iterate on this a bit.
> 
> Bjorn
> --
> To unsubscribe from this list: send the line "unsubscribe linux-pci" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>