On Fri, Feb 14, 2025 at 02:10:27PM +0100, Danilo Krummrich wrote:> On Fri, Feb 14, 2025 at 02:05:36PM +0100, Danilo Krummrich wrote: > > only force FW_CACHE if DRM_NOUVEAU_GSP_DEFAULT? > > Please scratch that, it was a horrible idea.What I recommend is to look into why we disable it by default, I think its sold old obscure reasoning but now suspect it was udev being dumb, in line with why we also try to defensively try to ignore duplicate udev requests for module requests causing a flood. Refer to commit 9b9879fc03275ff ("modules: catch concurrent module loads, treat them as idempotent") for details. Luis
On Tue, Feb 18, 2025 at 06:28:53AM -0800, Luis Chamberlain wrote:> > What I recommend is to look into why we disable it by default, I thinkI think FW_CACHE is enabled by default, no?> its sold old obscure reasoning but now suspect it was udev being dumb, > in line with why we also try to defensively try to ignore duplicate > udev requests for module requests causing a flood. Refer to commit > 9b9879fc03275ff ("modules: catch concurrent module loads, treat them as > idempotent") for details.Thanks, that helps a lot. Given that this commit landed in v6.5 I guess there's not really a concern to force enable FW_CACHE for Nouveau as a fix for commit 176fdcbddfd2 ("drm/nouveau/gsp/r535: add support for booting GSP-RM"), which landed in v6.7.