On Fri, Dec 09, 2016 at 10:35:49AM +0000, Tvrtko Ursulin wrote: > On 09/12/2016 10:18, Chris Wilson wrote: > >Hmm. Could we use the mmio debug. Something like > > > > valid_reg = bitmap_create(0x40000) > > FORCEWAKE_ALL > > for_each_offset() > > read reg > > if (!mmio_debug) > > set_bit(valid_reg); > > > > for_each_bit() > > FORCEWAKE_DISABLE > > read reg using fw_domains > > if (mmio_debug) > > AWOOGA! > >? > > > >Hopefully read reg won't cause system hangs, and less liable to corrupt > >state than write reg. > > I think we need to restate what are we actually trying to test. I at > least am confused now. :) Right, that tests that for all known registers covered by the fw domains (which we force on, including the expected absent domains?), the entry->domains are correct. It may not detect if we have a register that we need that is outside of the set of initialised domains. We still need to catch runtime misuse of mmio (i.e. callers who forget to grab the fw domain). Anyway, I think it does provide some useful validation of the fw getters, so I'll give it a spin and see what happens. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
