[REGRESSION] In v5.15.146 an ax88179_178a USB ethernet adapter causes crashes

[REGRESSION] In v5.15.146 an ax88179_178a USB ethernet adapter causes crashes

[Jeffery Miller]

For 5.15 attempting to use an ax88179_178a adapter "0b95:1790 ASIX
Electronics Corp. AX88179 Gigabit Ethernet"
started causing crashes.
This did not reproduce in the 6.6 kernel.

The crashes were narrowed down to the following two commits brought
into v5.15.146:

commit d63fafd6cc28 ("net: usb: ax88179_178a: avoid failed operations
when device is disconnected")
commit f860413aa00c ("net: usb: ax88179_178a: wol optimizations")

Those two use an uninitialized pointer `dev->driver_priv`.

In later kernels this pointer is initialized in commit 2bcbd3d8a7b4
("net: usb: ax88179_178a: move priv to driver_priv").

Picking in the two following commits fixed the issue for me on 5.15:
commit 9718f9ce5b86 ("net: usb: ax88179_178a: remove redundant init code")
commit 2bcbd3d8a7b4 ("net: usb: ax88179_178a: move priv to driver_priv")

The commit 9718f9ce5b86 ("net: usb: ax88179_178a: remove redundant
init code") was required for
the fix to apply cleanly.

Re: [REGRESSION] In v5.15.146 an ax88179_178a USB ethernet adapter causes crashes

[Justin Chen]

On Fri, Jan 12, 2024 at 10:01 AM Jeffery Miller
<jefferymiller@google.com> wrote:
>
> For 5.15 attempting to use an ax88179_178a adapter "0b95:1790 ASIX
> Electronics Corp. AX88179 Gigabit Ethernet"
> started causing crashes.
> This did not reproduce in the 6.6 kernel.
>
Looks like my patch set was not fully backported. The patchset didn't
have a "Fixes" tag, so it looks it was partially pulled for
d63fafd6cc28 ("net: usb: ax88179_178a: avoid failed operations when
device is disconnected") which does have a fixes tag. Just looks like
a bad backport here. Apologies, I should have caught it when I saw the
stable email, I didn't realize it was only for part of the changeset.

Thanks,
Justin

> The crashes were narrowed down to the following two commits brought
> into v5.15.146:
>
> commit d63fafd6cc28 ("net: usb: ax88179_178a: avoid failed operations
> when device is disconnected")
> commit f860413aa00c ("net: usb: ax88179_178a: wol optimizations")
>
> Those two use an uninitialized pointer `dev->driver_priv`.
>
> In later kernels this pointer is initialized in commit 2bcbd3d8a7b4
> ("net: usb: ax88179_178a: move priv to driver_priv").
>
> Picking in the two following commits fixed the issue for me on 5.15:
> commit 9718f9ce5b86 ("net: usb: ax88179_178a: remove redundant init code")
> commit 2bcbd3d8a7b4 ("net: usb: ax88179_178a: move priv to driver_priv")
>
> The commit 9718f9ce5b86 ("net: usb: ax88179_178a: remove redundant
> init code") was required for
> the fix to apply cleanly.

Re: [REGRESSION] In v5.15.146 an ax88179_178a USB ethernet adapter causes crashes

[Greg KH]

On Fri, Jan 12, 2024 at 11:32:13AM -0800, Justin Chen wrote:
> On Fri, Jan 12, 2024 at 10:01 AM Jeffery Miller
> <jefferymiller@google.com> wrote:
> >
> > For 5.15 attempting to use an ax88179_178a adapter "0b95:1790 ASIX
> > Electronics Corp. AX88179 Gigabit Ethernet"
> > started causing crashes.
> > This did not reproduce in the 6.6 kernel.
> >
> Looks like my patch set was not fully backported. The patchset didn't
> have a "Fixes" tag, so it looks it was partially pulled for
> d63fafd6cc28 ("net: usb: ax88179_178a: avoid failed operations when
> device is disconnected") which does have a fixes tag. Just looks like
> a bad backport here. Apologies, I should have caught it when I saw the
> stable email, I didn't realize it was only for part of the changeset.
> 

Now queued up, thanks.

greg k-h