On 2019-11-21 at 10:29:09 -0800, Lucas De Marchi wrote: > On Thu, Nov 21, 2019 at 04:42:31PM +0530, Ramalingam C wrote: > > + /* Wa_1604555607:tgl */ > > + val = intel_uncore_read(engine->uncore, FF_MODE2); > > + val &= ~FF_MODE2_TDS_TIMER_MASK; > > + val |= FF_MODE2_TDS_TIMER_128; > > + __wa_write_masked_or(wal, FF_MODE2, FF_MODE2_TDS_TIMER_MASK, val, > > + IS_TGL_REVID(engine->i915, 0, TGL_REVID_A0) ? 0: > > isn't it broken until B0 rather than only on A0? I think we could just > disable the read back for all revs and then selectively enable it > for later.
Thought of excluding the verification when the new stepping comes in. Hence excluded for existing stepping alone. Even your suggestion sounds good to me. I will go with it. With a FIXME Note to indicate that we need to enable for capable steppings. -Ram > > Lucas De Marchi _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
