2014-04-21 20:38:04

by Ken Moffat

[permalink] [raw]
Subject: 3.15.0-rc2 radeon HD 7480D [Aruba] blank display

I've just built 3.15.0-rc2 on this box, and discovered that I get a
blank screen. The boot appears to complete (it sends me information
from SMART which is from my last bootscript), and it responds to
MagicSysRQ to reboot. I tried to login and run startx, but that
didn't seem to make any difference - possibly something went wrong,
it wasn't in my history when I booted a good kernel.

This machine uses a regular old VGA cable to connect it, the
following lines in dmesg make me think it has chosen the wrong
connector:

Apr 21 19:32:45 bluesbreaker kernel: [ 1.273246]
[drm:radeon_dp_link_train_cr] *ERROR* displayport link status failed
Apr 21 19:32:45 bluesbreaker kernel: [ 1.273247]
[drm:radeon_dp_link_train_cr] *ERROR* clock recovery failed
Apr 21 19:32:45 bluesbreaker kernel: [ 1.295116] Console:
switching to colour frame buffer device 85x34

On a boot with 3.14.0 and earlier I get a different console size -
Apr 21 20:42:44 bluesbreaker kernel: [ 1.445992] Console:
switching to colour frame buffer device 133x54

My console font is 12x22, and my screen is 1600x1200 but it looks
as if 3.15-rc2 has fallen back to trying to use 1024x768.

I'm attaching what look like the relevant parts of the messages
from the system logs for 3.15-rc2 and 3.14.0.

Any ideas, please ? This is an AMD A4-5300 APU and not the world's
fastest machine for bisecting :-(

ĸen
--
das eine Mal als Tragödie, dieses Mal als Farce


Attachments:
(No filename) (1.41 kB)
drm-aruba-3.15-rc2 (6.90 kB)
drm-aruba-3.14 (6.67 kB)
Download all attachments

2014-04-21 20:54:17

by Alex Deucher

[permalink] [raw]
Subject: Re: 3.15.0-rc2 radeon HD 7480D [Aruba] blank display

On Mon, Apr 21, 2014 at 4:32 PM, Ken Moffat <[email protected]> wrote:
> I've just built 3.15.0-rc2 on this box, and discovered that I get a
> blank screen. The boot appears to complete (it sends me information
> from SMART which is from my last bootscript), and it responds to
> MagicSysRQ to reboot. I tried to login and run startx, but that
> didn't seem to make any difference - possibly something went wrong,
> it wasn't in my history when I booted a good kernel.
>
> This machine uses a regular old VGA cable to connect it, the
> following lines in dmesg make me think it has chosen the wrong
> connector:

VGA is implemented as a DP to VGA bridge on APUs.

>
> Apr 21 19:32:45 bluesbreaker kernel: [ 1.273246]
> [drm:radeon_dp_link_train_cr] *ERROR* displayport link status failed
> Apr 21 19:32:45 bluesbreaker kernel: [ 1.273247]
> [drm:radeon_dp_link_train_cr] *ERROR* clock recovery failed
> Apr 21 19:32:45 bluesbreaker kernel: [ 1.295116] Console:
> switching to colour frame buffer device 85x34
>
> On a boot with 3.14.0 and earlier I get a different console size -
> Apr 21 20:42:44 bluesbreaker kernel: [ 1.445992] Console:
> switching to colour frame buffer device 133x54
>
> My console font is 12x22, and my screen is 1600x1200 but it looks
> as if 3.15-rc2 has fallen back to trying to use 1024x768.
>
> I'm attaching what look like the relevant parts of the messages
> from the system logs for 3.15-rc2 and 3.14.0.
>
> Any ideas, please ? This is an AMD A4-5300 APU and not the world's
> fastest machine for bisecting :-(

I'd guess it's either the pll rework patches or the switch to the
common dp helper code patches. It would be helpful if you could
bisect.

Alex

2014-04-21 21:05:08

by Ken Moffat

[permalink] [raw]
Subject: Re: 3.15.0-rc2 radeon HD 7480D [Aruba] blank display

On Mon, Apr 21, 2014 at 04:54:15PM -0400, Alex Deucher wrote:
> On Mon, Apr 21, 2014 at 4:32 PM, Ken Moffat <[email protected]> wrote:
> > I've just built 3.15.0-rc2 on this box, and discovered that I get a
> > blank screen. The boot appears to complete (it sends me information
> > from SMART which is from my last bootscript), and it responds to
> > MagicSysRQ to reboot. I tried to login and run startx, but that
> > didn't seem to make any difference - possibly something went wrong,
> > it wasn't in my history when I booted a good kernel.
> >
> > This machine uses a regular old VGA cable to connect it, the
> > following lines in dmesg make me think it has chosen the wrong
> > connector:
>
> VGA is implemented as a DP to VGA bridge on APUs.
>
> >
> > Apr 21 19:32:45 bluesbreaker kernel: [ 1.273246]
> > [drm:radeon_dp_link_train_cr] *ERROR* displayport link status failed
> > Apr 21 19:32:45 bluesbreaker kernel: [ 1.273247]
> > [drm:radeon_dp_link_train_cr] *ERROR* clock recovery failed
> > Apr 21 19:32:45 bluesbreaker kernel: [ 1.295116] Console:
> > switching to colour frame buffer device 85x34
> >
> > On a boot with 3.14.0 and earlier I get a different console size -
> > Apr 21 20:42:44 bluesbreaker kernel: [ 1.445992] Console:
> > switching to colour frame buffer device 133x54
> >
> > My console font is 12x22, and my screen is 1600x1200 but it looks
> > as if 3.15-rc2 has fallen back to trying to use 1024x768.
> >
> > I'm attaching what look like the relevant parts of the messages
> > from the system logs for 3.15-rc2 and 3.14.0.
> >
> > Any ideas, please ? This is an AMD A4-5300 APU and not the world's
> > fastest machine for bisecting :-(
>
> I'd guess it's either the pll rework patches or the switch to the
> common dp helper code patches. It would be helpful if you could
> bisect.
>
> Alex

OK, I guess I can do the builds on a faster machine.

ĸen
--
das eine Mal als Tragödie, dieses Mal als Farce

2014-04-22 02:34:15

by Ken Moffat

[permalink] [raw]
Subject: Re: 3.15.0-rc2 radeon HD 7480D [Aruba] blank display

On Tue, Apr 22, 2014 at 03:31:06AM +0100, Ken Moffat wrote:

[ resending, somehow lkml dropped out of the Cc. ]

> On Tue, Apr 22, 2014 at 12:19:40AM +0100, Ken Moffat wrote:
> > On Mon, Apr 21, 2014 at 05:29:27PM -0400, Ed Tomlinson wrote:
>
> > > Ken,
> > >
> > > You might want to try reverting:
> > >
> > > commit 379dfc25e257ffe10eb53b86d2375f7c0f4f33ef
> > > Author: Alex Deucher <[email protected]>
> > > Date: Mon Apr 7 10:33:46 2014 -0400
> > >
> > > drm/radeon/dp: switch to the common i2c over aux code
> > >
> > > Provides a nice cleanup in radeon.
> > >
> > > Signed-off-by: Alex Deucher <[email protected]>
> > > Signed-off-by: Christian König <[email protected]>
> > >
> > > This fixed a similar problem here (see the drm pull thread for rc1)
> > >
> > > Thanks
> > > Ed Tomlinson
> >
> > Tried reverting that from rc2, but it didn't help. Thanks anyway.
> >
>
> Well, I *believed* I had created a patch for that commit, and
> reverted it from -rc2 with patch -R. But I've now bisected through
> drivers/gpu/drm between v3.14.0 and v3.15-rc2 and the bisection
> pointed to that same commit, so I guess I did something wrong.
>
> There were a couple of other issues along the way (mounting nfs
> hung on one commit, startx hung on several of the commits, but I've
> marked those as "good" because I had a display, even if the system
> was not usable).
>
> I've now gone back to linus' tree that I cloned a few hours ago,
> i.e.
> commit c089b229dfdd09d59a11d8bc2344bf8196d575ce
> Merge: 9ac03675010a 0565103d1adb
> Author: Linus Torvalds <[email protected]>
> Date: Mon Apr 21 10:05:35 2014 -0700
>
> Merge branch 'for-linus' of
> git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml
>
> created a branch, and *definitely* reverted
> 379dfc25e257ffe10eb53b86d2375f7c0f4f33ef (using git revert) in that
> branch. And this version works fine.
>
> So belatedly I see that I seem to have the same problem as Ed. Or
> at least that the same commit is causing both our problems.
>
> Alex - do you need any more information from me to help with this ?
>
> ĸen
> --
> das eine Mal als Tragödie, dieses Mal als Farce

--
das eine Mal als Tragödie, dieses Mal als Farce

2014-04-22 05:54:06

by Alex Deucher

[permalink] [raw]
Subject: Re: 3.15.0-rc2 radeon HD 7480D [Aruba] blank display

On Mon, Apr 21, 2014 at 10:34 PM, Ken Moffat <[email protected]> wrote:
> On Tue, Apr 22, 2014 at 03:31:06AM +0100, Ken Moffat wrote:
>
> [ resending, somehow lkml dropped out of the Cc. ]
>
>> On Tue, Apr 22, 2014 at 12:19:40AM +0100, Ken Moffat wrote:
>> > On Mon, Apr 21, 2014 at 05:29:27PM -0400, Ed Tomlinson wrote:
>>
>> > > Ken,
>> > >
>> > > You might want to try reverting:
>> > >
>> > > commit 379dfc25e257ffe10eb53b86d2375f7c0f4f33ef
>> > > Author: Alex Deucher <[email protected]>
>> > > Date: Mon Apr 7 10:33:46 2014 -0400
>> > >
>> > > drm/radeon/dp: switch to the common i2c over aux code
>> > >
>> > > Provides a nice cleanup in radeon.
>> > >
>> > > Signed-off-by: Alex Deucher <[email protected]>
>> > > Signed-off-by: Christian König <[email protected]>
>> > >
>> > > This fixed a similar problem here (see the drm pull thread for rc1)
>> > >
>> > > Thanks
>> > > Ed Tomlinson
>> >
>> > Tried reverting that from rc2, but it didn't help. Thanks anyway.
>> >
>>
>> Well, I *believed* I had created a patch for that commit, and
>> reverted it from -rc2 with patch -R. But I've now bisected through
>> drivers/gpu/drm between v3.14.0 and v3.15-rc2 and the bisection
>> pointed to that same commit, so I guess I did something wrong.
>>
>> There were a couple of other issues along the way (mounting nfs
>> hung on one commit, startx hung on several of the commits, but I've
>> marked those as "good" because I had a display, even if the system
>> was not usable).
>>
>> I've now gone back to linus' tree that I cloned a few hours ago,
>> i.e.
>> commit c089b229dfdd09d59a11d8bc2344bf8196d575ce
>> Merge: 9ac03675010a 0565103d1adb
>> Author: Linus Torvalds <[email protected]>
>> Date: Mon Apr 21 10:05:35 2014 -0700
>>
>> Merge branch 'for-linus' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml
>>
>> created a branch, and *definitely* reverted
>> 379dfc25e257ffe10eb53b86d2375f7c0f4f33ef (using git revert) in that
>> branch. And this version works fine.
>>
>> So belatedly I see that I seem to have the same problem as Ed. Or
>> at least that the same commit is causing both our problems.
>>
>> Alex - do you need any more information from me to help with this ?

The attached patch should fix it.

Thanks,

Alex


Attachments:
0001-drm-radeon-aux-fix-hpd-assignment-for-aux-bus.patch (1.12 kB)

2014-04-22 15:34:27

by Ed Tomlinson

[permalink] [raw]
Subject: Re: 3.15.0-rc2 radeon HD 7480D [Aruba] blank display

On Tuesday 22 April 2014 01:54:01 Alex Deucher wrote:
> On Mon, Apr 21, 2014 at 10:34 PM, Ken Moffat <[email protected]> wrote:
> > On Tue, Apr 22, 2014 at 03:31:06AM +0100, Ken Moffat wrote:
> >
> > [ resending, somehow lkml dropped out of the Cc. ]
> >
> >> On Tue, Apr 22, 2014 at 12:19:40AM +0100, Ken Moffat wrote:
> >> > On Mon, Apr 21, 2014 at 05:29:27PM -0400, Ed Tomlinson wrote:
> >>
> >> > > Ken,
> >> > >
> >> > > You might want to try reverting:
> >> > >
> >> > > commit 379dfc25e257ffe10eb53b86d2375f7c0f4f33ef
> >> > > Author: Alex Deucher <[email protected]>
> >> > > Date: Mon Apr 7 10:33:46 2014 -0400
> >> > >
> >> > > drm/radeon/dp: switch to the common i2c over aux code
> >> > >
> >> > > Provides a nice cleanup in radeon.
> >> > >
> >> > > Signed-off-by: Alex Deucher <[email protected]>
> >> > > Signed-off-by: Christian König <[email protected]>
> >> > >
> >> > > This fixed a similar problem here (see the drm pull thread for rc1)
> >> > >
> >> > > Thanks
> >> > > Ed Tomlinson
> >> >
> >> > Tried reverting that from rc2, but it didn't help. Thanks anyway.
> >> >
> >>
> >> Well, I *believed* I had created a patch for that commit, and
> >> reverted it from -rc2 with patch -R. But I've now bisected through
> >> drivers/gpu/drm between v3.14.0 and v3.15-rc2 and the bisection
> >> pointed to that same commit, so I guess I did something wrong.
> >>
> >> There were a couple of other issues along the way (mounting nfs
> >> hung on one commit, startx hung on several of the commits, but I've
> >> marked those as "good" because I had a display, even if the system
> >> was not usable).
> >>
> >> I've now gone back to linus' tree that I cloned a few hours ago,
> >> i.e.
> >> commit c089b229dfdd09d59a11d8bc2344bf8196d575ce
> >> Merge: 9ac03675010a 0565103d1adb
> >> Author: Linus Torvalds <[email protected]>
> >> Date: Mon Apr 21 10:05:35 2014 -0700
> >>
> >> Merge branch 'for-linus' of
> >> git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml
> >>
> >> created a branch, and *definitely* reverted
> >> 379dfc25e257ffe10eb53b86d2375f7c0f4f33ef (using git revert) in that
> >> branch. And this version works fine.
> >>
> >> So belatedly I see that I seem to have the same problem as Ed. Or
> >> at least that the same commit is causing both our problems.
> >>
> >> Alex - do you need any more information from me to help with this ?
>
> The attached patch should fix it.

And it does here (on bonaire, r7 260x). You can add my tested by

Tested-by: Ed Tomlinson <[email protected]>

Thanks,
Ed Tomlinson

2014-04-29 00:04:56

by Ken Moffat

[permalink] [raw]
Subject: Re: 3.15.0-rc2 radeon HD 7480D [Aruba] blank display

On Tue, Apr 22, 2014 at 01:54:01AM -0400, Alex Deucher wrote:
>
> The attached patch should fix it.
>
> Thanks,
>
> Alex

> From 9cd764fd57bb2a4e5f618d0f8a64c8154a820688 Mon Sep 17 00:00:00 2001
> From: Alex Deucher <[email protected]>
> Date: Tue, 22 Apr 2014 01:49:28 -0400
> Subject: [PATCH] drm/radeon/aux: fix hpd assignment for aux bus
>
Sorry, my default mailbox (the one where personal mail ends up)
overflowed (the old "procmail thinks it cannot write to it" issue)
and I've been reading mail on my netbook with a short 24-line term,
didn't notice that my overflow mailbox had appeared because it was
on the next screen. So, sorry for not replying.

And then when I did notice, I accidentally sent a personal reply
only to Ed.

The good news is that -rc3 obviously has the fix, and is working
fine. Thanks.

ĸen
--
das eine Mal als Tragödie, dieses Mal als Farce