Re: [RFC PATCH v2 04/11] PCI/P2PDMA: Introduce pci_p2pdma_should_map_bus() and pci_p2pdma_bus_offset()

[Christian König]

Am 24.03.21 um 18:21 schrieb Jason Gunthorpe:
> On Mon, Mar 15, 2021 at 10:27:08AM -0600, Logan Gunthorpe wrote:
>
> > In this case the WARN_ON is just to guard against misuse of the
> > function. It should never happen unless a developer changes the code in
> > a way that is incorrect. So I think that's the correct use of WARN_ON.
> > Though I might change it to WARN and return, that seems safer.
> Right, WARN_ON and return is the right pattern for an assertion that
> must never happen:
>
>    if (WARN_ON(foo))
>        return -1
>
> Linus wants assertions like this to be able to recover. People runing
> the 'panic on warn' mode want the kernel to stop if it detects an
> internal malfunction.

The only justification I can see for a "panic on warn" is to prevent 
further data loss or warn early about a crash.

We only use a BUG_ON() when the alternative would be to corrupt something.

Christian.

> Jason