[PATCH] Implement retry on linux sysfs pnp id file.

[Drew Moseley]

Kay Sievers wrote:
> On Tue, 2008-09-23 at 09:12 -0700, Drew Moseley wrote:
>> Kay Sievers wrote:
>>>> This is a really dirty hack. 
>>>>
>>>> @Kay: Is this behavior normal? Is there a way to fix sysfs instead of sending 
>>>> HAL into a loop?
>>> Yep, "fix the kernel" is usually the right answer. There is almost
>>> never a valid "the way it has to be" in sysfs. We fixed almost all know
>>> issues so far, so this should be fixed too.
>>>
>>> In:
>>>   drivers/pnp/core.c
>>> we have:
>>>         ret = device_register(&dev->dev);
>>>         if (ret)
>>>                 return ret;
>>>
>>>         pnp_interface_attach_device(dev);
>>>
>>> Which is broken, the attributes must not be created after the device
>>> registration, they should be converted to "default bus attributes" and
>>> all will work fine.
>>>
>> I certainly don't dispute the fact that this patch is less than ideal.
>> I have at lease one more issue that appears to be similar.  I'll take a
>> look at the default bus attributes and see what I find.  Thanks for the
>> pointer.
> 
> I expect pnp can do the same as PCI, and attach the attributes to the
> bus_type struct, which will let the driver core create the attributes
> before the event, not by the pnp code after the event:
> 
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=drivers/pci/pci-sysfs.c;h=9c718583a237f4f6082e7dcc5e84e23e85ee651c;hb=HEAD#l211
> 
> Kay
> 

I posted a note and a patch for discussion at LKML.  Details here:
http://marc.info/?l=linux-kernel&m=122220515804617&w=2

Thanks,
Drew