RE: [PATCH 07/22] RDMA/irdma: Register an auxiliary driver and implement private channel OPs

Subject: Re: [PATCH 07/22] RDMA/irdma: Register an auxiliary driver and
implement private channel OPs

On Fri, Jan 22, 2021 at 05:48:12PM -0600, Shiraz Saleem wrote:

quoted

+static int irdma_probe(struct auxiliary_device *aux_dev,
+		       const struct auxiliary_device_id *id) {
+	struct irdma_drvdata *drvdata;
+	int ret;
+
+	drvdata = kzalloc(sizeof(*drvdata), GFP_KERNEL);
+	if (!drvdata)
+		return -ENOMEM;
+
+	switch (id->driver_data) {
+	case IRDMA_GEN_2:
+		drvdata->init_dev = irdma_init_dev;
+		drvdata->deinit_dev = irdma_deinit_dev;
+		break;
+	case IRDMA_GEN_1:
+		drvdata->init_dev = i40iw_init_dev;
+		drvdata->deinit_dev = i40iw_deinit_dev;
+		break;
+	default:
+		ret = -ENODEV;
+		goto ver_err;

Also don't do this, if the drivers are so different then give them different aux bus
names and bind two drivers with the different flow.

I suppose we could have a gen1 aux driver and one for gen2 with the flows being a little orthogonal from each other.

The gen2 aux driver auxiliary_device_id table would have entries for iWARP and RoCE
aux dev's.

I suppose the old i40e can keep its weird registration thing, but ice should not
duplicate that, new code must use aux devices properly, as in my other email.

Jason