Re: [PATCH v6 02/14] hwtracing: gtrace: Initial implementation of gtrace framework

From: Leo Yu-Chi Liang

Date: Mon Nov 09 2026 - 22:08:30 EST


Hi Mayuresh,

On 10/1/2026 1:22 PM, Mayuresh Chitale wrote:
> +const struct gtrace_component_id *gtrace_match_id(struct gtrace_component *comp,
> + const struct gtrace_component_id *ids)
> +{
> + const struct gtrace_component_id *id;
> +
> + for (id = ids; id->version; id++) {
> + if (comp->id.type != id->type)
> + continue;
> +
> + return id;
> + }

Following up on sashiko's comment: we are adding support for SiFive
pre-ratified trace components on top of this series, and they are
registered with major version 0. Using version 0 as the table terminator
means a driver cannot list a { type, version 0 } entry, which the
eswin/EIC7700 SoC ships.

For now this works because only the type is compared and the
mkversion(1, 0) entry matches any version. But a later version check
would silently drop version 0 components.

Since GTRACE_RVTRACE_ENCODER is 0, the type cannot be the terminator
either. How about reserving type 0:

enum gtrace_component_type {
GTRACE_COMPONENT_TYPE_NONE, /* id table terminator */
GTRACE_RVTRACE_ENCODER,
...
};

for (id = ids; ids && id->type != GTRACE_COMPONENT_TYPE_NONE; id++) {

The NULL ids check would also cover the missing id_table case that
sashiko raised.

Thanks,
Leo