Re: [PATCH net-next v2] net: airoha: Validate egress gdm port in airoha_ppe_foe_entry_prepare()

[Lorenzo Bianconi <lorenzo.bianconi@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2262 bytes --]

> On 3/15/25 3:59 PM, Lorenzo Bianconi wrote:
> >>> Fix the issue validating egress gdm port in airoha_ppe_foe_entry_prepare
> >>> routine.
> >>
> >> A more interesting question is, why do you see an invalid port? Is the
> >> hardware broken? Something not correctly configured? Are you just
> >> papering over the crack?
> >>
> >>> -static int airoha_ppe_foe_entry_prepare(struct airoha_foe_entry *hwe,
> >>> +static int airoha_ppe_foe_entry_prepare(struct airoha_eth *eth,
> >>> +					struct airoha_foe_entry *hwe,
> >>>  					struct net_device *dev, int type,
> >>>  					struct airoha_flow_data *data,
> >>>  					int l4proto)
> >>> @@ -224,6 +225,11 @@ static int airoha_ppe_foe_entry_prepare(struct airoha_foe_entry *hwe,
> >>>  	if (dev) {
> >>>  		struct airoha_gdm_port *port = netdev_priv(dev);
> >>
> >> If port is invalid, is dev also invalid? And if dev is invalid, could
> >> dereferencing it to get priv cause an opps?
> > 
> > I do not think this is a hw problem. Running bidirectional high load traffic,
> > I got the sporadic crash reported above. In particular, netfilter runs
> > airoha_ppe_flow_offload_replace() providing the egress net_device pointer used
> > in airoha_ppe_foe_entry_prepare(). Debugging with gdb, I discovered the system
> > crashes dereferencing port pointer in airoha_ppe_foe_entry_prepare() (even if
> > dev pointer is not NULL). Adding this sanity check makes the system stable.
> > Please note a similar check is available even in mtk driver [0].
> 
> I agree with Andrew, you need a better understanding of the root cause.
> This really looks like papering over some deeper issue.
> 
> AFAICS 'dev' is fetched from the airoha driver itself a few lines
> before. Possibly you should double check that code.

Are you referring to airoha_get_dsa_port() routine?
I think dev pointer in airoha_ppe_foe_entry_prepare() is not strictly
necessary a device from a driver itself since it is an egress device
and the flowtable can contain even a wlan or a vlan device. In this
case airoha_get_dsa_port() will just return the original device pointer
and we can't assume priv pointer points to a airoha_gdm_port struct.
Agree?

Regards,
Lorenzo

> 
> Thanks,
> 
> Paolo
> 
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]