Message-Id: <201105101237.49686.thomas@fjellstrom.ca>
Date:	Tue, 10 May 2011 12:37:49 -0600
From:	Thomas Fjellstrom <thomas@...llstrom.ca>
To:	Jerome Glisse <j.glisse@...il.com>
Cc:	Alex Deucher <alexdeucher@...il.com>, linux-kernel@...r.kernel.org,
	dri-devel@...ts.freedesktop.org
Subject: Re: long delay when using HMDI output on RS780

On May 10, 2011, you wrote:
> On Mon, May 9, 2011 at 11:22 PM, Thomas Fjellstrom <thomas@...llstrom.ca> 
wrote:
> > On May 9, 2011, Thomas Fjellstrom wrote:
> >> On May 9, 2011, Alex Deucher wrote:
> >> > On Mon, May 9, 2011 at 5:04 PM, Thomas Fjellstrom
> >> > <thomas@...llstrom.ca>
> >> 
> >> wrote:
> >> > > On May 9, 2011, Thomas Fjellstrom wrote:
> >> > >> On May 9, 2011, Jerome Glisse wrote:
> >> > >> > On Mon, May 9, 2011 at 3:50 PM, Thomas Fjellstrom
> >> > >> > <thomas@...llstrom.ca>
> >> > >> 
> >> > >> wrote:
> >> > >> > > On May 7, 2011, Thomas Fjellstrom wrote:
> >> > >> > >> I just switched to using HDMI with my media center, and its
> >> > >> > >> causing a 30+ second delay in the screen turning on, as well
> >> > >> > >> as a 7 second delay in the X startup when it tries to fetch
> >> > >> > >> the EDID
> >> > >> > >> information. Basically I don't get any picture at all once KMS
> >> > >> > >> initializes until after X has been up a few seconds.
> >> > >> > >> 
> >> > >> > >> The odd thing is the monitor seems to think something is going
> >> > >> > >> on, since it doesn't go to sleep or display its "No Signal"
> >> > >> > >> OSD, but just after X starts up, it pops up the "Input
> >> > >> > >> detected/switched" OSD, and the picture appears.
> >> > >> > >> 
> >> > >> > >> The bios, grub2 (in both text mode and graphics mode), and the
> >> > >> > >> initial linux kernel messages all display fine and
> >> > >> > >> immediately. Its only once KMS and radeondrmfb initializes
> >> > >> > >> that there's a problem (at least till X starts up).
> >> > >> > >> 
> >> > >> > >> I've just built with a vanila 2.6.38.4 kernel from the stable
> >> > >> > >> git repo, and have played with some EDID settings, trying to
> >> > >> > >> disable edid where I could thinking thats what caused the
> >> > >> > >> problem. That doesn't seem to be the case though. I also
> >> > >> > >> tried playing with the video= kernel option, trying to force
> >> > >> > >> disable VGA-1, and set a static mode for HDMI-A-1, but if I
> >> > >> > >> try, it seems to force disable HDMI-A-1 instead of force the
> >> > >> > >> mode.
> >> > >> > >> 
> >> > >> > >> With a DVI-D cable instead, the problem goes away.
> >> > >> > >> 
> >> > >> > >> Attached are the dmesg and xorg.log files for the latest boot
> >> > >> > >> with HDMI (no video= parameter, and EDID enabled, most
> >> > >> > >> settings at defaults).
> >> > >> > >> 
> >> > >> > >> What exactly would cause this, and is there a way I can fix
> >> > >> > >> it?
> >> > >> > > 
> >> > >> > > I've been playing around with it more, and got it to not blank
> >> > >> > > the screen after KMS init, /once/. So far no luck repeating
> >> > >> > > that success.
> >> > >> > > 
> >> > >> > > I've tried late, and early kms init, and currently have the
> >> > >> > > radeon module and firmware compiled into the kernel. Boot
> >> > >> > > times at least are fairly decent, about 8-10s till X starts,
> >> > >> > > but about 25-35s till anything shows up.
> >> > >> > > 
> >> > >> > > Some strangeness, I have the kernel set to force the hdmi
> >> > >> > > output to on, with a very specific mode, that X tends to like,
> >> > >> > > the vga port is forced disabled. X is set to ignore EDID, and
> >> > >> > > also set to that specific mode that it tends to auto set
> >> > >> > > itself. Regardless X still wants to pause for 7s 2-3 times
> >> > >> > > while processing EDID info.
> >> > >> > > 
> >> > >> > > I've attached the new dmesg and xorg logs from the latest
> >> > >> > > attempts.
> >> > >> > > 
> >> > >> > > Note, this only happens with KMS, with HDMI. disabling KMS, or
> >> > >> > > using DVI makes the problem go away. Even grub's own graphical
> >> > >> > > mode works fine, its only once KMS inits that things go bad,
> >> > >> > > and its not till after X is up for a few seconds that
> >> > >> > > something displays on my screen.
> >> > >> > > 
> >> > >> > > --
> >> > >> > > Thomas Fjellstrom
> >> > >> > > thomas@...llstrom.ca
> >> > >> > 
> >> > >> > Please boot with drm.debug=4 and attach dmesg it should be more
> >> > >> > verbose. You might also want to try booting with radeon.audio=0
> >> > >> 
> >> > >> Hi, attached both one with just drm.debug=4, and one with that and
> >> > >> radeon.audio=0.
> >> > >> 
> >> > >> Now, I'm guessing its a bug in my monitor claiming it can do HDMI
> >> > >> audio? As setting radeon.audio=0 seems to have fixed the blanking
> >> > >> problem. And the dmesg logs seem to claim that radeon thinks it can
> >> > >> do HDMI audio.
> >> > 
> >> > Most likely the driver is sending the wrong packets for hdmi audio.
> >> > Until that gets fixed up, it's probably best to disable hdmi audio if
> >> > it's not working for you.
> >> 
> >> Ok. I'll do that for now, since I have no need for HDMI audio atm. My
> >> receiver is only capable of Dolby Pro Logic input, so eh.
> >> 
> >> Would be nice to fix the xorg pauses. And whatever is causing the
> >> massive amounts of drm log spam (yes, I could disable drm.debug, but
> >> that would just hide the issue, as whatever it is, will keep doing
> >> whatever it is doing regardless).
> > 
> > Any hints on where to look to fix the stalls when the xorg radeon driver
> > starts up?
> 
> Is the Xserver block or is it just that there is no picture ? ie when
> you see somethings on the screen is it directly the gnome login screen
> or what ever should come first ? or do you then see gnome loading ?
> 
> If it's just black and then you directly got gnome-shell it's likely
> that your tv doesn't live hdmi mode change and take time before
> showing the picture.

I can see X start up, the cursor shows up, but it just pauses for a few 
seconds, then once the wm is up, it pauses for a bit again.

Here's some log info from Xorg that I think might be relevant:

7 second wait:
[    16.161] (II) No input driver/identifier specified (ignoring)
[    23.725] (II) RADEON(0): EDID vendor "BNQ", prod id 30990

1 second:
[    23.726] (II) RADEON(0): Modeline "1920x1080"x0.0   74.25  1920 2558 2602 
2750  1080 1084 1089 1125 +hsync +vsync (27.0 kHz)
[    24.620] (II) RADEON(0): EDID vendor "BNQ", prod id 30990

6 seconds:
[    24.631] (II) RADEON(0): Modeline "1920x1080"x0.0   74.25  1920 2558 2602 
2750  1080 1084 1089 1125 +hsync +vsync (27.0 kHz)
[    30.206] (II) RADEON(0): EDID vendor "BNQ", prod id 30990


It actually prints out the EDID information 5 times. Once at the beginning, 
and theres no discernible pause, then later on at the end, it prints out the 
EDID/DDC information 4 more times, with 3 pauses, once before the second print 
out, once between the second and third, and once between the fourth and the 
fifth.

[    15.849] (II) RADEON(0): EDID for output HDMI-0
[    15.849] (II) RADEON(0): Manufacturer: BNQ  Model: 790e  Serial#: 21573
[    15.849] (II) RADEON(0): Year: 2009  Week: 6
[    15.849] (II) RADEON(0): EDID Version: 1.3

This is what led me to believe it might be related to EDID. Thought maybe 
there was some glitch causing kms+xorg to not like my monitor's EDID info, but 
it seems it likes it just fine, so much it asks for it 4+ times (kms looks, 
then x prints it out 5 times, possibly actually fetching it 3+ times).

Also, I have it setup to auto boot into KDE without compositing enabled, using 
startx in an init script. I'm thinking of switching to icewm or something 
equally light weight, to see if maybe its KDE causing one of the pauses (it 
didn't used to pause like this, before I updated to 2.6.38).

Last bit of info, my "TV" is actually a 24" 1080p LCD monitor.

> Cheers,
> Jerome


-- 
Thomas Fjellstrom
thomas@...llstrom.ca
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists