RE: [PATCH net-next v5 01/10] qed: Add module with basic common support

Yuval Mintz [off-list ref] :
[...]

quoted

quoted

quoted

+struct qed_simd_fp_handler {
+     void    *token;
+     void    (*func)(void *);
+};

Use union * ?

The token is a cookie to be used by a func, so union isn't appropriate.

Lets' reformulate: replace 'void * token' by 'union foobar *token'.

void * silents compiler. union * doesn't.

Not sure I get you here - this hander is used for registering interrupt handlers
in case INTa is used, and multiple modules [both QED and the appropriate
protocol driver] need to use it.
Turning this into an explicit union would require QED to be familiar with the
ISR cookies of all the various protocol drivers. While possible, it'll create
additional .h dependencies between those in a direction that currently doesn't
exist [I.e., QED would have to include protocol driver .h for it].

[...]

quoted

quoted

quoted

+             rc = -ENOMEM;
+             goto ilt_shadow_fail;
+     } else {
+             DP_VERBOSE(p_hwfn, QED_MSG_ILT,
+                        "Allocated 0x%x bytes for ilt shadow\n",
+                        (u32)(size * sizeof(struct qed_dma_mem)));
+     }

The "else" branch after the "goto" isn't idiomatic.

Not that I mind, but is such a prefernce described in any style-guide?

Documentation/CodingStyle gives some hints in its "goto" section but it doesn't
specifically go that far. Is there a reward if I can exhumate some message on
netdev where it would had already been outlined ?

Depends. Do you consider wasted time to be a reward? ;-)