[PATCH] ath10k: clear QMI if failed during init

[PATCH] ath10k: clear QMI if failed during init

[Albert Esteve]

A failed ath10k_qmi_init() left a dangling
pointer (ar_snoc->qmi) in the cleanup path.
Clear the pointer on the way out for hygiene
and consistency with deinit().

Fixes: ba94c753ccb4 ("ath10k: add QMI message handshake for wcn3990 client")
Signed-off-by: Albert Esteve <aesteve@redhat.com>
---
 drivers/net/wireless/ath/ath10k/qmi.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/wireless/ath/ath10k/qmi.c b/drivers/net/wireless/ath/ath10k/qmi.c
index 0d74548a5f34..08a64f737fb4 100644
--- a/drivers/net/wireless/ath/ath10k/qmi.c
+++ b/drivers/net/wireless/ath/ath10k/qmi.c
@@ -1128,6 +1128,7 @@ int ath10k_qmi_init(struct ath10k *ar, u32 msa_size)
 
 err:
 	kfree(qmi);
+	ar_snoc->qmi = NULL;
 	return ret;
 }
 

---
base-commit: bc35965f6940a9bf834d54187b6088b8eb09206d
change-id: 20260914-fix-ath10k-dangling-15676d427cbb

Best regards,
-- 
Albert Esteve <aesteve@redhat.com>

Re: [PATCH] ath10k: clear QMI if failed during init

[Brian Norris]

On Mon, Sep 14, 2026 at 03:05:19PM +0200, Albert Esteve wrote:
> A failed ath10k_qmi_init() left a dangling
> pointer (ar_snoc->qmi) in the cleanup path.
> Clear the pointer on the way out for hygiene
> and consistency with deinit().
> 
> Fixes: ba94c753ccb4 ("ath10k: add QMI message handshake for wcn3990 client")

It's all going away (probe failure / removal) if we get here anyway, so
this hardly qualifies as a "fix". In fact, in a cursory check of the
driver, it seems the NULL-ability of the ar_snoc->qmi field is totally
irrelevant to this driver -- it never is actually checked.

I suppose you can still call it "hygiene" though, and it's probably
reasonable to do the same here as we do in deinit().

> Signed-off-by: Albert Esteve <aesteve@redhat.com>

Reviewed-by: Brian Norris <briannorris@chromium.org>

> ---
>  drivers/net/wireless/ath/ath10k/qmi.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/net/wireless/ath/ath10k/qmi.c b/drivers/net/wireless/ath/ath10k/qmi.c
> index 0d74548a5f34..08a64f737fb4 100644
> --- a/drivers/net/wireless/ath/ath10k/qmi.c
> +++ b/drivers/net/wireless/ath/ath10k/qmi.c
> @@ -1128,6 +1128,7 @@ int ath10k_qmi_init(struct ath10k *ar, u32 msa_size)
>  
>  err:
>  	kfree(qmi);
> +	ar_snoc->qmi = NULL;
>  	return ret;
>  }
>  
> 
> ---
> base-commit: bc35965f6940a9bf834d54187b6088b8eb09206d
> change-id: 20260914-fix-ath10k-dangling-15676d427cbb
> 
> Best regards,
> -- 
> Albert Esteve <aesteve@redhat.com>
> 

Re: [PATCH] ath10k: clear QMI if failed during init

[Albert Esteve]

On Tue, Sep 15, 2026 at 7:26 PM Brian Norris <briannorris@chromium.org> wrote:
>
> On Mon, Sep 14, 2026 at 03:05:19PM +0200, Albert Esteve wrote:
> > A failed ath10k_qmi_init() left a dangling
> > pointer (ar_snoc->qmi) in the cleanup path.
> > Clear the pointer on the way out for hygiene
> > and consistency with deinit().
> >
> > Fixes: ba94c753ccb4 ("ath10k: add QMI message handshake for wcn3990 client")
>
> It's all going away (probe failure / removal) if we get here anyway, so
> this hardly qualifies as a "fix". In fact, in a cursory check of the
> driver, it seems the NULL-ability of the ar_snoc->qmi field is totally
> irrelevant to this driver -- it never is actually checked.

Yes, I agree...

>
> I suppose you can still call it "hygiene" though, and it's probably
> reasonable to do the same here as we do in deinit().

... that's why I phrased the patch body that way. I hesitated about
using the Fixes tag here. I can send a follow-up removing the tag if
you'd prefer.

BR,
Albert

>
> > Signed-off-by: Albert Esteve <aesteve@redhat.com>
>
> Reviewed-by: Brian Norris <briannorris@chromium.org>
>
> > ---
> >  drivers/net/wireless/ath/ath10k/qmi.c | 1 +
> >  1 file changed, 1 insertion(+)
> >
> > diff --git a/drivers/net/wireless/ath/ath10k/qmi.c b/drivers/net/wireless/ath/ath10k/qmi.c
> > index 0d74548a5f34..08a64f737fb4 100644
> > --- a/drivers/net/wireless/ath/ath10k/qmi.c
> > +++ b/drivers/net/wireless/ath/ath10k/qmi.c
> > @@ -1128,6 +1128,7 @@ int ath10k_qmi_init(struct ath10k *ar, u32 msa_size)
> >
> >  err:
> >       kfree(qmi);
> > +     ar_snoc->qmi = NULL;
> >       return ret;
> >  }
> >
> >
> > ---
> > base-commit: bc35965f6940a9bf834d54187b6088b8eb09206d
> > change-id: 20260914-fix-ath10k-dangling-15676d427cbb
> >
> > Best regards,
> > --
> > Albert Esteve <aesteve@redhat.com>
> >
>

Re: [PATCH] ath10k: clear QMI if failed during init

[Jeff Johnson]

On 9/15/2026 11:55 PM, Albert Esteve wrote:
> On Tue, Sep 15, 2026 at 7:26 PM Brian Norris <briannorris@chromium.org> wrote:
>>
>> On Mon, Sep 14, 2026 at 03:05:19PM +0200, Albert Esteve wrote:
>>> A failed ath10k_qmi_init() left a dangling
>>> pointer (ar_snoc->qmi) in the cleanup path.
>>> Clear the pointer on the way out for hygiene
>>> and consistency with deinit().
>>>
>>> Fixes: ba94c753ccb4 ("ath10k: add QMI message handshake for wcn3990 client")
>>
>> It's all going away (probe failure / removal) if we get here anyway, so
>> this hardly qualifies as a "fix". In fact, in a cursory check of the
>> driver, it seems the NULL-ability of the ar_snoc->qmi field is totally
>> irrelevant to this driver -- it never is actually checked.
> 
> Yes, I agree...
> 
>>
>> I suppose you can still call it "hygiene" though, and it's probably
>> reasonable to do the same here as we do in deinit().
> 
> ... that's why I phrased the patch body that way. I hesitated about
> using the Fixes tag here. I can send a follow-up removing the tag if
> you'd prefer.
no need to repost. I'll drop the Fixes tag, update the subject to add the
missing wifi: prefix, and edit/re-flow the commit text to use the 70+ columns
that is preferred when I take it into my pending tree.

/jeff