Messages in this thread | | | Date | Tue, 19 Jul 2022 02:43:58 +0300 | From | Vladimir Oltean <> | Subject | Re: [net-next RFC PATCH 1/4] net: dsa: qca8k: drop qca8k_read/write/rmw for regmap variant |
| |
On Mon, Jul 18, 2022 at 11:54:44PM +0200, Christian Marangi wrote: > On Mon, Jul 18, 2022 at 11:30:42PM +0300, Vladimir Oltean wrote: > > On Mon, Jul 18, 2022 at 09:30:58PM +0200, Christian Marangi wrote: > > > Tell me if I got this wrong. > > > > > > The suggestion was to move the struct dsa_switch_ops to qca8k.h and add > > > in the specific code probe the needed ops to add to the generic > > > struct... > > > > The declaration yes; the definition to qca8k-common.c. See for example > > where felix_switch_ops is, relative to felix_vsc9959.c, seville_vsc9953.c > > (users), felix.h (declaration), and felix.c (definition). Or how > > mv88e6xxx_switch_ops does things and still supports a gazillion of switches. > > Mhh I checked the example and they doesn't seems to be useful from my > problem. But I think it's better to discuss this to the patch directly > so you can better understand whay I intended with having dsa_switch_ops > set to const.
So you don't modify the common dsa_switch_ops from the switch-specific probe path, but rather, from the common dsa_switch_ops method, you call a second function pointer.
static void felix_phylink_validate(struct dsa_switch *ds, int port, unsigned long *supported, struct phylink_link_state *state) { struct ocelot *ocelot = ds->priv; struct felix *felix = ocelot_to_felix(ocelot);
if (felix->info->phylink_validate) felix->info->phylink_validate(ocelot, port, supported, state); }
| |