[PATCH] wifi: brcm80211: handle pmk_op allocation failure

[PATCH] wifi: brcm80211: handle pmk_op allocation failure

[Duoming Zhou]

The kzalloc() in brcmf_pmksa_v3_op() will return null if the
physical memory has run out. As a result, if we dereference
the null value, the null pointer dereference bug will happen.

Return -ENOMEM from brcmf_pmksa_v3_op() if kzalloc() fails
for pmk_op.

Fixes: a96202acaea4 ("wifi: brcmfmac: cfg80211: Add support for PMKID_V3 operations")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
---
 drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
index 28d6a30cc01..3b420b33188 100644
--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
@@ -4322,6 +4322,10 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
 	int ret;
 
 	pmk_op = kzalloc(sizeof(*pmk_op), GFP_KERNEL);
+	if (!pmk_op) {
+		ret = -ENOMEM;
+		goto out;
+	}
 	pmk_op->version = cpu_to_le16(BRCMF_PMKSA_VER_3);
 
 	if (!pmksa) {
@@ -4340,6 +4344,7 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
 	pmk_op->length = cpu_to_le16(length);
 
 	ret = brcmf_fil_iovar_data_set(ifp, "pmkid_info", pmk_op, sizeof(*pmk_op));
+out:
 	kfree(pmk_op);
 	return ret;
 }
-- 
2.17.1

Re: [PATCH] wifi: brcm80211: handle pmk_op allocation failure

[Arend van Spriel]

[-- Attachment #1: Type: text/plain, Size: 1798 bytes --]

On 2/27/2024 9:06 AM, Duoming Zhou wrote:
> The kzalloc() in brcmf_pmksa_v3_op() will return null if the
> physical memory has run out. As a result, if we dereference
> the null value, the null pointer dereference bug will happen.
> 
> Return -ENOMEM from brcmf_pmksa_v3_op() if kzalloc() fails
> for pmk_op.

NAK (see below)

Also this issue was reported earlier by Joe Perches. Not sure if he 
wants to be mentioned as such.

> Fixes: a96202acaea4 ("wifi: brcmfmac: cfg80211: Add support for PMKID_V3 operations")
> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> ---
>   drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c | 5 +++++
>   1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> index 28d6a30cc01..3b420b33188 100644
> --- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> +++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> @@ -4322,6 +4322,10 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
>   	int ret;
>   
>   	pmk_op = kzalloc(sizeof(*pmk_op), GFP_KERNEL);
> +	if (!pmk_op) {
> +		ret = -ENOMEM;
> +		goto out;
> +	}

There is really no need to introduce a new label for this. Although you 
can kfree() a NULL pointer there is no need to do so when you know 
already it is NULL. Just return -ENOMEM and be done with it.

Regards,
Arend

>   	pmk_op->version = cpu_to_le16(BRCMF_PMKSA_VER_3);
>   
>   	if (!pmksa) {
> @@ -4340,6 +4344,7 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
>   	pmk_op->length = cpu_to_le16(length);
>   
>   	ret = brcmf_fil_iovar_data_set(ifp, "pmkid_info", pmk_op, sizeof(*pmk_op));
> +out:
>   	kfree(pmk_op);
>   	return ret;
>   }

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4219 bytes --]

Re: [PATCH] wifi: brcm80211: handle pmk_op allocation failure

[Joe Perches]

On Tue, 2024-02-27 at 12:21 +0100, Arend van Spriel wrote:
> On 2/27/2024 9:06 AM, Duoming Zhou wrote:
> > The kzalloc() in brcmf_pmksa_v3_op() will return null if the
> > physical memory has run out. As a result, if we dereference
> > the null value, the null pointer dereference bug will happen.
> > 
> > Return -ENOMEM from brcmf_pmksa_v3_op() if kzalloc() fails
> > for pmk_op.
> 
> NAK (see below)
> 
> Also this issue was reported earlier by Joe Perches. Not sure if he 
> wants to be mentioned as such.

I think it's unimportant to be mentioned.

I think it's more important that the code be researched
that the simple return of -ENOMEM the appropriate fix
and is handled by all possible callers of the function.

> 
> > Fixes: a96202acaea4 ("wifi: brcmfmac: cfg80211: Add support for PMKID_V3 operations")
> > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > ---
> >   drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c | 5 +++++
> >   1 file changed, 5 insertions(+)
> > 
> > diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.cq b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> > index 28d6a30cc01..3b420b33188 100644
> > --- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> > +++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> > @@ -4322,6 +4322,10 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
> >   	int ret;
> >   
> >   	pmk_op = kzalloc(sizeof(*pmk_op), GFP_KERNEL);
> > +	if (!pmk_op) {
> > +		ret = -ENOMEM;
> > +		goto out;
> > +	}
> 
> There is really no need to introduce a new label for this. Although you 
> can kfree() a NULL pointer there is no need to do so when you know 
> already it is NULL. Just return -ENOMEM and be done with it.
> 
> Regards,
> Arend
> 
> >   	pmk_op->version = cpu_to_le16(BRCMF_PMKSA_VER_3);
> >   
> >   	if (!pmksa) {
> > @@ -4340,6 +4344,7 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
> >   	pmk_op->length = cpu_to_le16(length);
> >   
> >   	ret = brcmf_fil_iovar_data_set(ifp, "pmkid_info", pmk_op, sizeof(*pmk_op));
> > +out:
> >   	kfree(pmk_op);
> >   	return ret;
> >   }

Re: [PATCH] wifi: brcm80211: handle pmk_op allocation failure

[duoming]

On Tue, 27 Feb 2024 03:42:13 -0800 Joe Perches wrote:
> > > The kzalloc() in brcmf_pmksa_v3_op() will return null if the
> > > physical memory has run out. As a result, if we dereference
> > > the null value, the null pointer dereference bug will happen.
> > > 
> > > Return -ENOMEM from brcmf_pmksa_v3_op() if kzalloc() fails
> > > for pmk_op.
> > 
> > NAK (see below)
> > 
> > Also this issue was reported earlier by Joe Perches. Not sure if he 
> > wants to be mentioned as such.
> 
> I think it's unimportant to be mentioned.
> 
> I think it's more important that the code be researched
> that the simple return of -ENOMEM the appropriate fix
> and is handled by all possible callers of the function.

There are three functions that call the brcmf_pmksa_v3_op(),
including brcmf_cfg80211_set_pmksa(), brcmf_cfg80211_del_pmksa()
and brcmf_cfg80211_flush_pmksa(). The return type of the above
three functions is s32. If brcmf_pmksa_v3_op() returns -ENOMEM,
the above three functions will also return -ENOMEM. It could be
handled appropriately by the callers of the function. So I think
the simple return of -ENOMEM is appropriate.

> > > Fixes: a96202acaea4 ("wifi: brcmfmac: cfg80211: Add support for PMKID_V3 operations")
> > > Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
> > > ---
> > >   drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c | 5 +++++
> > >   1 file changed, 5 insertions(+)
> > > 
> > > diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.cq b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> > > index 28d6a30cc01..3b420b33188 100644
> > > --- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> > > +++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
> > > @@ -4322,6 +4322,10 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
> > >   	int ret;
> > >   
> > >   	pmk_op = kzalloc(sizeof(*pmk_op), GFP_KERNEL);
> > > +	if (!pmk_op) {
> > > +		ret = -ENOMEM;
> > > +		goto out;
> > > +	}
> > 
> > There is really no need to introduce a new label for this. Although you 
> > can kfree() a NULL pointer there is no need to do so when you know 
> > already it is NULL. Just return -ENOMEM and be done with it.
> > 
> > Regards,
> > Arend
> > 
> > >   	pmk_op->version = cpu_to_le16(BRCMF_PMKSA_VER_3);
> > >   
> > >   	if (!pmksa) {
> > > @@ -4340,6 +4344,7 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
> > >   	pmk_op->length = cpu_to_le16(length);
> > >   
> > >   	ret = brcmf_fil_iovar_data_set(ifp, "pmkid_info", pmk_op, sizeof(*pmk_op));
> > > +out:
> > >   	kfree(pmk_op);
> > >   	return ret;
> > >   }

Best regards,
Duoming Zhou

Re: [PATCH] wifi: brcm80211: handle pmk_op allocation failure

[Arend van Spriel]

[-- Attachment #1: Type: text/plain, Size: 2416 bytes --]

On 2/27/2024 12:42 PM, Joe Perches wrote:
> On Tue, 2024-02-27 at 12:21 +0100, Arend van Spriel wrote:
>> On 2/27/2024 9:06 AM, Duoming Zhou wrote:
>>> The kzalloc() in brcmf_pmksa_v3_op() will return null if the
>>> physical memory has run out. As a result, if we dereference
>>> the null value, the null pointer dereference bug will happen.
>>>
>>> Return -ENOMEM from brcmf_pmksa_v3_op() if kzalloc() fails
>>> for pmk_op.
>>
>> NAK (see below)
>>
>> Also this issue was reported earlier by Joe Perches. Not sure if he
>> wants to be mentioned as such.
> 
> I think it's unimportant to be mentioned.
> 
> I think it's more important that the code be researched
> that the simple return of -ENOMEM the appropriate fix
> and is handled by all possible callers of the function.

Right. That is what I did after which I replied on Feb 18 to your email.

https://lore.kernel.org/all/3e1f11b3-a689-4c8e-bfda-b43649db7106@broadcom.com/

Regards,
Arend

>>
>>> Fixes: a96202acaea4 ("wifi: brcmfmac: cfg80211: Add support for PMKID_V3 operations")
>>> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
>>> ---
>>>    drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c | 5 +++++
>>>    1 file changed, 5 insertions(+)
>>>
>>> diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.cq b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
>>> index 28d6a30cc01..3b420b33188 100644
>>> --- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
>>> +++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
>>> @@ -4322,6 +4322,10 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
>>>    	int ret;
>>>    
>>>    	pmk_op = kzalloc(sizeof(*pmk_op), GFP_KERNEL);
>>> +	if (!pmk_op) {
>>> +		ret = -ENOMEM;
>>> +		goto out;
>>> +	}
>>
>> There is really no need to introduce a new label for this. Although you
>> can kfree() a NULL pointer there is no need to do so when you know
>> already it is NULL. Just return -ENOMEM and be done with it.
>>
>> Regards,
>> Arend
>>
>>>    	pmk_op->version = cpu_to_le16(BRCMF_PMKSA_VER_3);
>>>    
>>>    	if (!pmksa) {
>>> @@ -4340,6 +4344,7 @@ brcmf_pmksa_v3_op(struct brcmf_if *ifp, struct cfg80211_pmksa *pmksa,
>>>    	pmk_op->length = cpu_to_le16(length);
>>>    
>>>    	ret = brcmf_fil_iovar_data_set(ifp, "pmkid_info", pmk_op, sizeof(*pmk_op));
>>> +out:
>>>    	kfree(pmk_op);
>>>    	return ret;
>>>    }
> 

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4219 bytes --]

Re: [PATCH] wifi: brcm80211: handle pmk_op allocation failure

[Joe Perches]

On Tue, 2024-02-27 at 20:40 +0100, Arend van Spriel wrote:
> On 2/27/2024 12:42 PM, Joe Perches wrote:
> > On Tue, 2024-02-27 at 12:21 +0100, Arend van Spriel wrote:
> > > On 2/27/2024 9:06 AM, Duoming Zhou wrote:
> > > > The kzalloc() in brcmf_pmksa_v3_op() will return null if the
> > > > physical memory has run out. As a result, if we dereference
> > > > the null value, the null pointer dereference bug will happen.
> > > > 
> > > > Return -ENOMEM from brcmf_pmksa_v3_op() if kzalloc() fails
> > > > for pmk_op.
> > > 
> > > NAK (see below)
> > > 
> > > Also this issue was reported earlier by Joe Perches. Not sure if he
> > > wants to be mentioned as such.
> > 
> > I think it's unimportant to be mentioned.
> > 
> > I think it's more important that the code be researched
> > that the simple return of -ENOMEM the appropriate fix
> > and is handled by all possible callers of the function.
> 
> Right. That is what I did after which I replied on Feb 18 to your email.
> 
> https://lore.kernel.org/all/3e1f11b3-a689-4c8e-bfda-b43649db7106@broadcom.com/
> 

Right, I did that as well, but I didn't look up
the entire calling tree.

It likely works.  It's also likely better than
the write through of the null pointer.