Hi Wayne, Cooper,

One more difference worth knowing about if v2 gets swapped in:
ae60ec4dcfc4 ("drm/amd/display: Use FreeSync v3 infopacket for sinks
above 255 Hz") is missing the amdgpu_dm_connector.c hunk of patch 2/2.
The posted patch changes four files (92 insertions, 1 deletion) in both
v1 and v2, but the commit only changes three (91 insertions).

The missing hunk is the one that stores the AMD VSDB version for every
DP/eDP sink:

        get_amd_vsdb(amdgpu_dm_connector, &vsdb_info);
+       amdgpu_dm_connector->vsdb_info.amd_vsdb_version = 
vsdb_info.amd_vsdb_version;

        if (vsdb_info.replay_mode) {
                amdgpu_dm_connector->vsdb_info.replay_mode = 
vsdb_info.replay_mode;
-               amdgpu_dm_connector->vsdb_info.amd_vsdb_version = 
vsdb_info.amd_vsdb_version;

Without it, a native DP sink without Replay never records VSDB v3, so it
never gets the v3 packet, and the DP side of the >255 Hz fix has no
effect. The KUnit test added by the same patch catches this on
amd-staging-drm-next 8dd68a79583a:

  # dm_test_fs_caps_dp_vsdb_version: EXPECTATION FAILED at 
drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/tests/amdgpu_dm_connector_test.c:9925
  Expected ctx->aconnector->vsdb_info.amd_vsdb_version == 3, but
      ctx->aconnector->vsdb_info.amd_vsdb_version == 0 (0x0)

The failure is easy to miss right now, because the amdgpu_dm KUnit run
panics earlier in dm_test_i2c_xfer_no_ddc_pin() (I've reported that on
the "Route connector DDC over AUX" thread). With that test fixed and
this hunk restored, all 1931 amdgpu_dm tests pass here. Patch 1/2
(bedad38a7ddf) landed complete.

Thanks,
Eren

Reply via email to