Re: avoiding pci_disable_device()...

[Jeff Garzik]

Roland Dreier wrote:
>     Jeff> That's an MSI bug.
>
>     Jeff> A current PCI driver -should- be using pci_request_regions().
>
> Hmm... I'm not sure everyone would agree with that.  It does make
> sense that the MSI-X core wants to make sure that it owns the MSI-X
> table without having someone else stomp on it.

The idea is right but the implementation is definitely incorrect, for 
multiple reasons:

* Every PCI driver should call pci_request_regions(), because there 
should only be one pci_driver associated with that set of resources.

* Therefore, any _addition_ to the PCI API that prevents a driver from 
using pci_request_regions() causes multiple problems, by virtue of 
diverging from the other 98% of the kernel's PCI drivers, which use 
pci_request_regions().

IOW, MSI-X's implementation incorrectly constrains PCI driver authors.

* Programmer over-protection.  Who is attempting to stomp on MSI-X 
tables?  If a driver is reading an MSI-X bar and doing something without 
consulting the kernel MSI-X code, that's a bug and most likely a 
violation of the PCI spec.  Fix the driver.

In Linux we don't over-protect programmers with a zillion sanity checks 
for every internal API call; that leads to bloat, and papers over bugs 
that should be fixed.

	Jeff


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/