PROBLEM: i915 Haswell KMS wrong maximum resolution

PROBLEM: i915 Haswell KMS wrong maximum resolution

[Kenneth de Mello]

[-- Attachment #1.1: Type: text/plain, Size: 929 bytes --]

On my system, the maximum resolution of my monitor, 2560x1440, is no longer
automatically detected properly with kernel 3.13.7.   This worked with
3.13.6, and 5 with this hardware and identical kernel configs.    Now the
autodetected resolution is 1920x1080.

The monitor is plugged into the DVI port on my motherboard.

My system:

- i7-4770 (non-K) CPU
- asrock z87 extreme6 motherboard
- Dell U2711 monitor

The integrated graphics adapter is configured as the primary in the BIOS,
and the discreet card is passed through to a Xen guest.   Kernel 3.13.7
mis-detects the resolution regardless if it is being booted with Xen or
without it (and previous versions worked as expected).  More detailed
system information is attached.

I don't notice any unusual messages, other than the wrong resolution.

Steps to reproduce:  Boot with kernel 3.13.7 (with or without Xen)

I'd be happy to provide more detail if required.

Thanks,

[-- Attachment #1.2: Type: text/html, Size: 1214 bytes --]

[-- Attachment #2: 3.13.7_intel_i915_bug.tgz --]
[-- Type: application/x-gzip, Size: 122678 bytes --]

[-- Attachment #3: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: PROBLEM: i915 Haswell KMS wrong maximum resolution

[Bruno Prémont]

Hi Kenneth,

On Fri, 28 March 2014 Kenneth de Mello <kdemello1980@gmail.com> wrote:
> On my system, the maximum resolution of my monitor, 2560x1440, is no longer
> automatically detected properly with kernel 3.13.7.   This worked with
> 3.13.6, and 5 with this hardware and identical kernel configs.    Now the
> autodetected resolution is 1920x1080.
> 
> The monitor is plugged into the DVI port on my motherboard.
> 
> My system:
> 
> - i7-4770 (non-K) CPU
> - asrock z87 extreme6 motherboard
> - Dell U2711 monitor
> 
> The integrated graphics adapter is configured as the primary in the BIOS,
> and the discreet card is passed through to a Xen guest.   Kernel 3.13.7
> mis-detects the resolution regardless if it is being booted with Xen or
> without it (and previous versions worked as expected).  More detailed
> system information is attached.
> 
> I don't notice any unusual messages, other than the wrong resolution.
> 
> Steps to reproduce:  Boot with kernel 3.13.7 (with or without Xen)
> 
> I'd be happy to provide more detail if required.

This smells like the following change in 3.13.6 to 3.13.7 might
be the cause:

commit e3ea8fa6beaf55fee64bf816f3b8a80ad733b2c2
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date:   Mon Mar 3 11:33:36 2014 +0200

    drm/i915: Reject >165MHz modes w/ DVI monitors

    commit 6375b768a9850b6154478993e5fb566fa4614a9c upstream.

    Single-link DVI max dotclock is 165MHz. Filter out modes with higher
    dotclock when the monitor doesn't support HDMI.

    Modes higher than 165 MHz were allowed in

    commit 7d148ef51a657fd04036c3ed7803da600dd0d451
    Author: Daniel Vetter <daniel.vetter@ffwll.ch>
    Date:   Mon Jul 22 18:02:39 2013 +0200

        drm/i915: fix hdmi portclock limits

    Also don't attempt to use 12bpc mode with DVI monitors.

    Cc: Adam Nielsen <a.nielsen@shikadi.net>
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75345
    Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=70331
    Tested-by: Ralf Jung <post+kernel@ralfj.de>
    Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
    Signed-off-by: Jani Nikula <jani.nikula@intel.com>
    Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>


Bruno
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: PROBLEM: i915 Haswell KMS wrong maximum resolution

[Kenneth de Mello]

[-- Attachment #1.1: Type: text/plain, Size: 3309 bytes --]

Bruno,

Thank you for your reply.

What about dual-link DVI?   I though the additional link addressed the
pixel clock limitation.   Has it only been using a single link this entire
time, and it's only worked by ignoring the maximum dotclock, so in other
words, the fact it works at all is the bug?

Also, the commit that rejects mode exceeding 165MHz states:

    "Single-link DVI max dotclock is 165MHz. Filter out modes with higher
     dotclock when the monitor doesn't support HDMI."

Does this mean HDMI in general?   This monitor does support HDMI, but the
maximum resolution when using the physical HDMI ports is 1920x1080.

What is the solution here, to switch to displayport if I want to use kernel
3.13.7 and beyond?   (This is fine, I just need to know so I can buy the
cable).

Kenneth

On Sun, Mar 30, 2014 at 2:36 PM, Bruno Prémont <bonbons@linux-vserver.org>wrote:

> Hi Kenneth,
>
> On Fri, 28 March 2014 Kenneth de Mello <kdemello1980@gmail.com> wrote:
> > On my system, the maximum resolution of my monitor, 2560x1440, is no
> longer
> > automatically detected properly with kernel 3.13.7.   This worked with
> > 3.13.6, and 5 with this hardware and identical kernel configs.    Now the
> > autodetected resolution is 1920x1080.
> >
> > The monitor is plugged into the DVI port on my motherboard.
> >
> > My system:
> >
> > - i7-4770 (non-K) CPU
> > - asrock z87 extreme6 motherboard
> > - Dell U2711 monitor
> >
> > The integrated graphics adapter is configured as the primary in the BIOS,
> > and the discreet card is passed through to a Xen guest.   Kernel 3.13.7
> > mis-detects the resolution regardless if it is being booted with Xen or
> > without it (and previous versions worked as expected).  More detailed
> > system information is attached.
> >
> > I don't notice any unusual messages, other than the wrong resolution.
> >
> > Steps to reproduce:  Boot with kernel 3.13.7 (with or without Xen)
> >
> > I'd be happy to provide more detail if required.
>
> This smells like the following change in 3.13.6 to 3.13.7 might
> be the cause:
>
> commit e3ea8fa6beaf55fee64bf816f3b8a80ad733b2c2
> Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
> Date:   Mon Mar 3 11:33:36 2014 +0200
>
>     drm/i915: Reject >165MHz modes w/ DVI monitors
>
>     commit 6375b768a9850b6154478993e5fb566fa4614a9c upstream.
>
>     Single-link DVI max dotclock is 165MHz. Filter out modes with higher
>     dotclock when the monitor doesn't support HDMI.
>
>     Modes higher than 165 MHz were allowed in
>
>     commit 7d148ef51a657fd04036c3ed7803da600dd0d451
>     Author: Daniel Vetter <daniel.vetter@ffwll.ch>
>     Date:   Mon Jul 22 18:02:39 2013 +0200
>
>         drm/i915: fix hdmi portclock limits
>
>     Also don't attempt to use 12bpc mode with DVI monitors.
>
>     Cc: Adam Nielsen <a.nielsen@shikadi.net>
>     Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75345
>     Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=70331
>     Tested-by: Ralf Jung <post+kernel@ralfj.de>
>     Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
>     Signed-off-by: Jani Nikula <jani.nikula@intel.com>
>     Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
>
>
> Bruno
>

[-- Attachment #1.2: Type: text/html, Size: 4684 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: PROBLEM: i915 Haswell KMS wrong maximum resolution

[Bruno Prémont]

Hi Kenneth,

On Sun, 30 March 2014 Kenneth de Mello wrote:
> What about dual-link DVI?   I though the additional link addressed the
> pixel clock limitation.   Has it only been using a single link this entire
> time, and it's only worked by ignoring the maximum dotclock, so in other
> words, the fact it works at all is the bug?
> 
> Also, the commit that rejects mode exceeding 165MHz states:
> 
>     "Single-link DVI max dotclock is 165MHz. Filter out modes with higher
>      dotclock when the monitor doesn't support HDMI."
> 
> Does this mean HDMI in general?   This monitor does support HDMI, but the
> maximum resolution when using the physical HDMI ports is 1920x1080.
> 
> What is the solution here, to switch to displayport if I want to use kernel
> 3.13.7 and beyond?   (This is fine, I just need to know so I can buy the
> cable).

I don't know the details, just spotting a possible match. Revert that
patch and try, just to make sure it's the cause.
Also try latest 3.14-rc (or release if it happens early enough).

Looking at the patch I don't see anything that would distinguish
single-link and dual-link DVI, but that might be located in the code
around (if so, possibly only in 3.14-rcX).


Note that the debug info you provided has pretty limited extract of
kernel logs.
Try to get more complete dmesg (at least starting from initialization of
i915) as well as Xorg.log.
Possibly you will have to get the same info with non-zero drm.debug
kernel cmdline option (or debug drm module option). But what value is
best is beyond my expertise.

I don't know if it's possible for the GPU to determine if DVI cable is
single or dual-link, but with single-link cable and such a monitor
you get some image to the monitor but with weird "scaling" effects!

Bruno


> Kenneth
> 
> On Sun, Mar 30, 2014 at 2:36 PM, Bruno Prémont wrote:
> > Hi Kenneth,
> >
> > On Fri, 28 March 2014 Kenneth de Mello wrote:
> > > On my system, the maximum resolution of my monitor, 2560x1440, is no
> > longer
> > > automatically detected properly with kernel 3.13.7.   This worked with
> > > 3.13.6, and 5 with this hardware and identical kernel configs.    Now the
> > > autodetected resolution is 1920x1080.
> > >
> > > The monitor is plugged into the DVI port on my motherboard.
> > >
> > > My system:
> > >
> > > - i7-4770 (non-K) CPU
> > > - asrock z87 extreme6 motherboard
> > > - Dell U2711 monitor
> > >
> > > The integrated graphics adapter is configured as the primary in the BIOS,
> > > and the discreet card is passed through to a Xen guest.   Kernel 3.13.7
> > > mis-detects the resolution regardless if it is being booted with Xen or
> > > without it (and previous versions worked as expected).  More detailed
> > > system information is attached.
> > >
> > > I don't notice any unusual messages, other than the wrong resolution.
> > >
> > > Steps to reproduce:  Boot with kernel 3.13.7 (with or without Xen)
> > >
> > > I'd be happy to provide more detail if required.
> >
> > This smells like the following change in 3.13.6 to 3.13.7 might
> > be the cause:
> >
> > commit e3ea8fa6beaf55fee64bf816f3b8a80ad733b2c2
> > Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > Date:   Mon Mar 3 11:33:36 2014 +0200
> >
> >     drm/i915: Reject >165MHz modes w/ DVI monitors
> >
> >     commit 6375b768a9850b6154478993e5fb566fa4614a9c upstream.
> >
> >     Single-link DVI max dotclock is 165MHz. Filter out modes with higher
> >     dotclock when the monitor doesn't support HDMI.
> >
> >     Modes higher than 165 MHz were allowed in
> >
> >     commit 7d148ef51a657fd04036c3ed7803da600dd0d451
> >     Author: Daniel Vetter <daniel.vetter@ffwll.ch>
> >     Date:   Mon Jul 22 18:02:39 2013 +0200
> >
> >         drm/i915: fix hdmi portclock limits
> >
> >     Also don't attempt to use 12bpc mode with DVI monitors.
> >
> >     Cc: Adam Nielsen <a.nielsen@shikadi.net>
> >     Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75345
> >     Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=70331
> >     Tested-by: Ralf Jung <post+kernel@ralfj.de>
> >     Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> >     Signed-off-by: Jani Nikula <jani.nikula@intel.com>
> >     Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> >
> >
> > Bruno
> >
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: PROBLEM: i915 Haswell KMS wrong maximum resolution

[Jani Nikula]

On Sun, 30 Mar 2014, Kenneth de Mello <kdemello1980@gmail.com> wrote:
> What about dual-link DVI?   I though the additional link addressed the
> pixel clock limitation.   Has it only been using a single link this entire
> time, and it's only worked by ignoring the maximum dotclock, so in other
> words, the fact it works at all is the bug?
>
> Also, the commit that rejects mode exceeding 165MHz states:
>
>     "Single-link DVI max dotclock is 165MHz. Filter out modes with higher
>      dotclock when the monitor doesn't support HDMI."
>
> Does this mean HDMI in general?   This monitor does support HDMI, but the
> maximum resolution when using the physical HDMI ports is 1920x1080.
>
> What is the solution here, to switch to displayport if I want to use kernel
> 3.13.7 and beyond?   (This is fine, I just need to know so I can buy the
> cable).

For further details please see the bug report [1].

The reason for the change was that modes with higher than 165 MHz
dotclock are invalid for single-link DVI. We don't support dual-link DVI
natively. Thus this is about HDMI->DVI adapters which are either
dual-link DVI (Ville says highly unlikely) or single-link DVI that allow
higher than 165 MHz dotclock in the monitor end. I'm not sure how we
could distinguish that from a regular single-link DVI that *is* bound by
the maximum dotclock.

Product details for the adapter you're using might be interesting.

BR,
Jani.



[1] https://bugzilla.kernel.org/show_bug.cgi?id=72961


-- 
Jani Nikula, Intel Open Source Technology Center

Re: PROBLEM: i915 Haswell KMS wrong maximum resolution

[Kenneth de Mello]

[-- Attachment #1.1: Type: text/plain, Size: 2097 bytes --]

Jani,

Thank you for explaining further.   After checking the manual for my
motherboard, I am mistaken.   It specifies DVI-I and states the maximum
resolution over DVI is 1920x1200@60Hz, so it appears all is well.   The
fact it worked before is a happy accident.

I have since switched to use displayport, and it is detecting the native
resolution as expected.


Thank you for your time,

BR,
Kenneth

On Mon, Mar 31, 2014 at 3:43 AM, Jani Nikula <jani.nikula@linux.intel.com>wrote:

> On Sun, 30 Mar 2014, Kenneth de Mello <kdemello1980@gmail.com> wrote:
> > What about dual-link DVI?   I though the additional link addressed the
> > pixel clock limitation.   Has it only been using a single link this
> entire
> > time, and it's only worked by ignoring the maximum dotclock, so in other
> > words, the fact it works at all is the bug?
> >
> > Also, the commit that rejects mode exceeding 165MHz states:
> >
> >     "Single-link DVI max dotclock is 165MHz. Filter out modes with higher
> >      dotclock when the monitor doesn't support HDMI."
> >
> > Does this mean HDMI in general?   This monitor does support HDMI, but the
> > maximum resolution when using the physical HDMI ports is 1920x1080.
> >
> > What is the solution here, to switch to displayport if I want to use
> kernel
> > 3.13.7 and beyond?   (This is fine, I just need to know so I can buy the
> > cable).
>
> For further details please see the bug report [1].
>
> The reason for the change was that modes with higher than 165 MHz
> dotclock are invalid for single-link DVI. We don't support dual-link DVI
> natively. Thus this is about HDMI->DVI adapters which are either
> dual-link DVI (Ville says highly unlikely) or single-link DVI that allow
> higher than 165 MHz dotclock in the monitor end. I'm not sure how we
> could distinguish that from a regular single-link DVI that *is* bound by
> the maximum dotclock.
>
> Product details for the adapter you're using might be interesting.
>
> BR,
> Jani.
>
>
>
> [1] https://bugzilla.kernel.org/show_bug.cgi?id=72961
>
>
> --
> Jani Nikula, Intel Open Source Technology Center
>

[-- Attachment #1.2: Type: text/html, Size: 2930 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: PROBLEM: i915 Haswell KMS wrong maximum resolution

[Dave Airlie]

On Mon, Mar 31, 2014 at 5:43 PM, Jani Nikula
<jani.nikula@linux.intel.com> wrote:
> On Sun, 30 Mar 2014, Kenneth de Mello <kdemello1980@gmail.com> wrote:
>> What about dual-link DVI?   I though the additional link addressed the
>> pixel clock limitation.   Has it only been using a single link this entire
>> time, and it's only worked by ignoring the maximum dotclock, so in other
>> words, the fact it works at all is the bug?
>>
>> Also, the commit that rejects mode exceeding 165MHz states:
>>
>>     "Single-link DVI max dotclock is 165MHz. Filter out modes with higher
>>      dotclock when the monitor doesn't support HDMI."
>>
>> Does this mean HDMI in general?   This monitor does support HDMI, but the
>> maximum resolution when using the physical HDMI ports is 1920x1080.
>>
>> What is the solution here, to switch to displayport if I want to use kernel
>> 3.13.7 and beyond?   (This is fine, I just need to know so I can buy the
>> cable).
>
> For further details please see the bug report [1].
>
> The reason for the change was that modes with higher than 165 MHz
> dotclock are invalid for single-link DVI. We don't support dual-link DVI
> natively. Thus this is about HDMI->DVI adapters which are either
> dual-link DVI (Ville says highly unlikely) or single-link DVI that allow
> higher than 165 MHz dotclock in the monitor end. I'm not sure how we
> could distinguish that from a regular single-link DVI that *is* bound by
> the maximum dotclock.
>
> Product details for the adapter you're using might be interesting.

DVI->HDMI adapters are just wires.

If you plug a HDMI monitor via one of those and the hw can do HDMI
speed single links it should work in theory, since the EDID would show
a HDMI monitor.

The DVI limitations are actually the cables rather than the connectors
from what I understand, but I could be misinformed.

Dave.