[PATCH 06/27] habanalabs: use memhash_node_export_put() in hl_release_dmabuf()

Stanislaw Gruszka stanislaw.gruszka at linux.intel.com
Thu Feb 16 14:40:34 UTC 2023


On Thu, Feb 16, 2023 at 02:26:50PM +0000, Tomer Tayar wrote:
> > This looks suspicious, I think hnode can be not-NULL here and has
> > hnode->vaddr different than searched addr, in case there is
> > hash collision and we iterate over hlist where there is no
> > searched addr. Not 100% sure about that though.
> > 
> > I think would be better to provide helper like this:
> > 
> > 	hash_for_each_possible(ctx->mem_hash, hnode, node, (unsigned
> > long)addr)
> > 		if (addr == hnode->vaddr)
> > 			return hnode;
> > 	return NULL;
> > 
> > which is basically standard way how hash_for_each_possible() used.
> > 
> > 
> > Regards
> > Stanislaw
> 
> Thanks Stanislaw, we will add such a helper and use it here and in another place with a similar pattern.
> If that is okay, we will do it in another patch, as this one only moves an existing function for code reuse.

Sure, no problem with that.

Regards
Stanislaw



More information about the dri-devel mailing list