On Sat, Jan 29, 2022 at 8:59 AM Greg Kroah-Hartman
<gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> It's up to you all, if you think the patch is correct, keep it for now.
In the fixed form (ie with Damien's fix for the wrong test polarity),
it's certainly not wrong, and matches a lot of our standard patterns -
including our documentation in
Documentation/driver-api/driver-model/design-patterns.rst
I did a quick visual grep, and all the cases of
devm_kzalloc(..GFP_KERNEL) I grepped for did indeed have that "if
(!..)" error handling pattern for the return value, including other
cases in the ATA subsystem.
That was very much a "quick visual grep" though, so no guarantees, and
I stopped looking after it was so obvious. IOW, it was just a
git grep -1 devm_kzalloc.*GFP_KERNEL
and then looking at the output and saying "yup, they all seem to do
that allocation failure test".
At the same time it's certainly also true that a small devm_kzalloc()
using GFP_KERNEL should never actually fail, so the patch - while
looking very correct to me - almost certainly makes no difference in
practice.
So I think the only problem with the patch is the original (fixed)
bug, and the source it came from.
Linus
