[Intel-gfx] [PATCH v5] drm/i915: pass ELD to HDMI/DP audio driver
Wu Fengguang
fengguang.wu at intel.com
Thu Nov 10 15:12:05 CET 2011
On Thu, Nov 10, 2011 at 09:47:46PM +0800, Wu Fengguang wrote:
> > > Got the delay - it's 72.986623-72.747632 = 239ms.
> > >
> > > [ 72.739944] HDMI hot plug event: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=0
> > > [ 72.742541] HDMI status: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=0
> > > [ 72.745082] HDMI hot plug event: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=1
> > > [ 72.747632] HDMI status: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=1
> > > [ 72.794240] [drm:intel_wait_for_vblank], vblank wait timed out
> > > [ 72.848099] [drm:intel_wait_for_vblank], vblank wait timed out
> > > [ 72.850507] [drm:ironlake_update_wm], FIFO watermarks For pipe A - plane 5, cursor: 6
> > > [ 72.853244] [drm:ironlake_update_wm], FIFO watermarks For pipe B - plane 42, cursor: 6
> > > [ 72.907937] [drm:intel_wait_for_vblank], vblank wait timed out
> > > [ 72.960790] [drm:intel_wait_for_vblank], vblank wait timed out
> > > [ 72.962757] [drm:ironlake_fdi_link_train], FDI_RX_IIR 0x100
> > > [ 72.964880] [drm:ironlake_fdi_link_train], FDI train 1 done.
> > > [ 72.968341] [drm:ironlake_fdi_link_train], FDI_RX_IIR 0x600
> > > [ 72.970274] [drm:ironlake_fdi_link_train], FDI train 2 done.
> > > [ 72.972535] [drm:ironlake_fdi_link_train], FDI train done
> > > [ 72.976751] [drm:intel_update_fbc],
> > > [ 72.977558] [drm:drm_crtc_helper_set_config], Setting connector DPMS state to on
> > > ==> [ 72.981550] [drm:drm_crtc_helper_set_config], [CONNECTOR:12:HDMI-A-2] set DPMS on
> > > [ 72.986623] HDMI: detected monitor RX-V1800 at connection type HDMI
> > > [ 72.988260] HDMI: available speakers: FL/FR LFE FC RL/RR RC RLC/RRC
> > >
> > > I wonder if the DPMS line has anything to do with the delay...
> > >
> > > Thanks,
> > > Fengguang
> >
> > Nice find! It does seem likely since the ELD read worked right after
> > that. Hmm. I am not familiar with the event order for the Intel
> > hardware, and what's calling your edid_to_eld code, etc. Is it possible
> > to move the EDID read call to after the DPMS state is ready?
>
> I added a dump_stack() at the very beginning of drm_crtc_helper_set_config().
>
> Judging from drm_crtc_helper_set_config()'s comment and the below dmesg, it's
> called by user space.
>
> [ 84.470633] Pid: 2893, comm: Xorg Tainted: G W 3.2.0-rc1-eld+ #244
> [ 84.472188] Call Trace:
> [ 84.472795] [<ffffffff8195dfba>] ? printk+0x41/0x43
> [ 84.474011] [<ffffffff8148aa04>] drm_crtc_helper_set_config+0x24/0x852
> [ 84.476342] [<ffffffff81493e52>] ? drm_ut_debug_printk+0x57/0x5e
> [ 84.477743] [<ffffffff8196052d>] ? mutex_unlock+0xe/0x10
> [ 84.479015] [<ffffffff8149a403>] ? drm_mode_object_find+0x61/0x70
> [ 84.481326] [<ffffffff8149b632>] drm_mode_setcrtc+0x35d/0x38e
> [ 84.482679] [<ffffffff8196052d>] ? mutex_unlock+0xe/0x10
> [ 84.483934] [<ffffffff8148efc6>] drm_ioctl+0x2c0/0x38c
> [ 84.486033] [<ffffffff8149b2d5>] ? drm_mode_getencoder+0x9b/0x9b
> [ 84.487463] [<ffffffff811567b0>] do_vfs_ioctl+0x490/0x4d1
> [ 84.489657] [<ffffffff81156838>] sys_ioctl+0x47/0x6b
> [ 84.490780] [<ffffffff8149a3cf>] ? drm_mode_object_find+0x2d/0x70
> [ 84.492235] [<ffffffff81968a82>] system_call_fastpath+0x16/0x1b
> [ 84.495536] [drm:drm_crtc_helper_set_config],
> [ 84.496561] [drm:drm_crtc_helper_set_config], [CRTC:4] [FB:31] #connectors=1 (x y) (0 0)
> [ 84.498323] [drm:drm_crtc_helper_set_config], modes are different, full mode set
> [ 84.499958] [drm:drm_mode_debug_printmodeline], Modeline 27:"720x576" 50 27000 720 732 796 864 576 581 586 625 0x40 0xa
> [ 84.502454] [drm:drm_mode_debug_printmodeline], Modeline 34:"" 0 27000 720 736 798 858 480 489 495 525 0x0 0xa
> [ 84.504672] [drm:drm_crtc_helper_set_config], [CONNECTOR:5:VGA-1] to [CRTC:3]
> [ 84.506205] [drm:drm_crtc_helper_set_config], [CONNECTOR:12:HDMI-A-2] to [CRTC:4]
> [ 84.507869] [drm:drm_crtc_helper_set_config], attempting to set mode from userspace
> [ 84.509699] [drm:drm_mode_debug_printmodeline], Modeline 34:"" 0 27000 720 736 798 858 480 489 495 525 0x0 0xa
> [ 84.511903] [drm:drm_crtc_helper_set_mode], [CRTC:4]
> [ 84.521961] [drm:intel_prepare_page_flip], preparing flip with no unpin work?
> [ 84.564804] [drm:intel_wait_for_vblank], vblank wait timed out
> [ 84.603175] [drm:ironlake_update_wm], FIFO watermarks For pipe A - plane 5, cursor: 6
> [ 84.606681] [drm:ironlake_update_wm], FIFO watermarks For pipe B - plane 42, cursor: 6
> [ 84.610472] [drm:intel_update_fbc],
> [ 84.611501] [drm:intel_choose_pipe_bpp_dither], forcing bpc to 8 for HDMI
> [ 84.613791] [drm:intel_choose_pipe_bpp_dither], setting pipe bpc to 8 (max display bpc 8)
> [ 84.616531] [drm:ironlake_crtc_mode_set], Mode for pipe 1:
> [ 84.617884] [drm:drm_mode_debug_printmodeline], Modeline 34:"" 0 27000 720 736 798 858 480 489 495 525 0x0 0xa
> [ 84.622337] [drm:ironlake_crtc_mode_set], disabling CxSR downclocking
> [ 84.676554] [drm:intel_wait_for_vblank], vblank wait timed out
> [ 84.678981] [drm:ironlake_update_plane], Writing base 00343000 00000000 0 0 4096
> [ 84.681384] [drm:intel_update_fbc],
> [ 84.733441] [drm:intel_wait_for_vblank], vblank wait timed out
> [ 84.735453] [drm:ironlake_update_wm], FIFO watermarks For pipe A - plane 5, cursor: 6
> [ 84.737375] [drm:ironlake_update_wm], FIFO watermarks For pipe B - plane 42, cursor: 6
> [ 84.739922] [drm:drm_crtc_helper_set_mode], [ENCODER:11:TMDS-11] set [MODE:34:]
> [ 84.742634] [drm:intel_hdmi_mode_set], Enabling HDMI audio on pipe B
> [ 84.745045] [drm:intel_write_eld], ELD on [CONNECTOR:12:HDMI-A-2], [ENCODER:11:TMDS-11]
> [ 84.746797] [drm:ironlake_write_eld], ELD on pipe B
> [ 84.747924] [drm:ironlake_write_eld], Audio directed to unknown port
> [ 84.751387] [drm:ironlake_write_eld], ELD size 13
> [ 84.752363] HDMI hot plug event: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=0
> [ 84.753981] HDMI status: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=0
> [ 84.756540] HDMI hot plug event: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=1
> [ 84.759108] HDMI status: Codec=3 Pin=6 Presence_Detect=1 ELD_Valid=1
> [ 84.804189] [drm:intel_wait_for_vblank], vblank wait timed out
> [ 84.857035] [drm:intel_wait_for_vblank], vblank wait timed out
> [ 84.859953] [drm:ironlake_update_wm], FIFO watermarks For pipe A - plane 5, cursor: 6
> [ 84.862637] [drm:ironlake_update_wm], FIFO watermarks For pipe B - plane 42, cursor: 6
> [ 84.916878] [drm:intel_wait_for_vblank], vblank wait timed out
> [ 84.969816] [drm:intel_wait_for_vblank], vblank wait timed out
> [ 84.972300] [drm:ironlake_fdi_link_train], FDI_RX_IIR 0x100
> [ 84.974200] [drm:ironlake_fdi_link_train], FDI train 1 done.
> [ 84.975643] [drm:ironlake_fdi_link_train], FDI_RX_IIR 0x600
> [ 84.977846] [drm:ironlake_fdi_link_train], FDI train 2 done.
> [ 84.980021] [drm:ironlake_fdi_link_train], FDI train done
> [ 84.983699] [drm:intel_update_fbc],
> [ 84.984994] [drm:drm_crtc_helper_set_config], Setting connector DPMS state to on
> [ 84.986531] [drm:drm_crtc_helper_set_config], [CONNECTOR:12:HDMI-A-2] set DPMS on
The above dmesg is associated with very simple assignment code, so
should be irrelevant.
Here, 84.996 should be the last retry time, and ELD is still not
ready.
The below intel_prepare_page_flip() is actually a no-op. So we lose
the obvious clues...
> [ 84.999041] [drm:intel_prepare_page_flip], preparing flip with no unpin work?
> [ 85.006426] HDMI: detected monitor RX-V1800 at connection type HDMI
> [ 85.008394] HDMI: available speakers: FL/FR LFE FC RL/RR RC RLC/RRC
>
> So audio driver got the ELD right after drm_crtc_helper_set_config() returns.
>
> It shows that ironlake_write_eld() is ultimately called by
> drm_crtc_helper_set_config().
More information about the Intel-gfx
mailing list