[PATCH net v2 2/7] net: dsa: mt7530: fix trapping frames with multiple CPU ports on MT7530

Vladimir Oltean olteanv at gmail.com
Tue Jun 13 13:59:15 PDT 2023


On Tue, Jun 13, 2023 at 11:35:08PM +0300, Arınç ÜNAL wrote:
> On 13.06.2023 23:18, Vladimir Oltean wrote:
> > On Tue, Jun 13, 2023 at 08:58:33PM +0300, Arınç ÜNAL wrote:
> > > On 13.06.2023 20:39, Vladimir Oltean wrote:
> > > > Got it. Then this is really not a problem, and the commit message frames
> > > > it incorrectly.
> > > 
> > > Actually this patch fixes the issue it describes. At the state of this
> > > patch, when multiple CPU ports are defined, port 5 is the active CPU port,
> > > CPU_PORT bits are set to port 6.
> > > 
> > > Once "the patch that prefers port 6, I could easily find the exact name but
> > > your mail snipping makes it hard" is applied, this issue becomes redundant.
> > 
> > Ok. Well, you don't get bonus points for fixing a problem in 2 different
> > ways, once is enough :)
> 
> This is not the case here though.
> 
> This patch fixes an issue that can be stumbled upon in two ways. This is for
> when multiple CPU ports are defined on the devicetree.
> 
> As I explained to Russell, the first is the CPU_PORT field not matching the
> active CPU port.
> 
> The second is when port 5 becomes the only active CPU port. This can only
> happen with the changing the DSA conduit support.
> 
> The "prefer port 6" patch only prevents the first way from happening. The
> latter still can happen. But this feature doesn't exist yet. Hence why I
> think we should apply this series as-is (after some patch log changes) and
> backport it without this patch on kernels older than 5.18.
> 
> Arınç

I was following you until the last phrase. Why should we apply this series
as-is [ to net.git ], if this patch fixes a problem (the *only* problem in
lack of .port_change_master() support, aka for stable kernels) that is
already masked by a different patch targeted to net.git?



More information about the linux-arm-kernel mailing list