[gentoo-user] Another X11 problem

[gentoo-user] Another X11 problem

[Ian Zimmerman]

Once or twice per week, when I start X11 (via the startx program), my
display goes all black.  When it happens there is no way back to a
working state: I can see that Xorg is running, I can exit it by hitting
Ctl-Alt-Backspace, but even then display does not recover.  The only
thing I can do is hit Ctl-Alt-Del blindly and reboot, and next time it
usually works.

I did a diff of the Xorg logfile for successful and failed runs, and
they are identical except for some trivial device renumbering.  Clearly
there is no segfault as in the other scenario recently discussed on the
list.  It is also not a monitor problem: turning monitor on & off
doesn't help.

This looks like a problem with the kernel modesetting code, right?  The
log says:

RADEON(0): Chipset: "ATI Radeon HD 4250" (ChipID = 0x9715)

How can I investigate and fix this?  Is there a way without studying the
kernel and Xorg code?

-- 
Please *no* private Cc: on mailing lists and newsgroups
Why does the arrow on Hillary signs point to the right?

Re: [gentoo-user] Another X11 problem

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 564 bytes --]

On Sat, 20 Aug 2016 00:15:15 -0700, Ian Zimmerman wrote:

> Once or twice per week, when I start X11 (via the startx program), my
> display goes all black.  When it happens there is no way back to a
> working state: I can see that Xorg is running, I can exit it by hitting
> Ctl-Alt-Backspace, but even then display does not recover.  The only
> thing I can do is hit Ctl-Alt-Del blindly and reboot, and next time it
> usually works.

Can you get back to a VC with Ctrl-Alt-F1/2?


-- 
Neil Bothwick

There is never enough beer, sex or disk space!

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

[gentoo-user] Re: Another X11 problem

[Ian Zimmerman]

On 2016-08-20 10:58 +0100, Neil Bothwick wrote:

> > Once or twice per week, when I start X11 (via the startx program),
> > my display goes all black.  When it happens there is no way back to
> > a working state: I can see that Xorg is running, I can exit it by
> > hitting Ctl-Alt-Backspace, but even then display does not recover.
> > The only thing I can do is hit Ctl-Alt-Del blindly and reboot, and
> > next time it usually works.
> 
> Can you get back to a VC with Ctrl-Alt-F1/2?

Good question, I am sure I tried that, but I _think_ it didn't work.
(The only way to _know_ would be to hit Ctl-Alt-Del next and see if it
reboots, and I'll make sure I try doing just that next time it happens.)
I do know that nothing appears on the screen even if the switch succeeds.

-- 
Please *no* private Cc: on mailing lists and newsgroups
Why does the arrow on Hillary signs point to the right?

Re: [gentoo-user] Re: Another X11 problem

[wabe]

Ian Zimmerman <itz@buug.org> wrote:

> On 2016-08-20 10:58 +0100, Neil Bothwick wrote:
> 
> > > Once or twice per week, when I start X11 (via the startx program),
> > > my display goes all black.  When it happens there is no way back
> > > to a working state: I can see that Xorg is running, I can exit it
> > > by hitting Ctl-Alt-Backspace, but even then display does not
> > > recover. The only thing I can do is hit Ctl-Alt-Del blindly and
> > > reboot, and next time it usually works.  
> > 
> > Can you get back to a VC with Ctrl-Alt-F1/2?  
> 
> Good question, I am sure I tried that, but I _think_ it didn't work.
> (The only way to _know_ would be to hit Ctl-Alt-Del next and see if it
> reboots, and I'll make sure I try doing just that next time it
> happens.) I do know that nothing appears on the screen even if the
> switch succeeds.

How is your monitor connected to your PC (HDMI, DVI, DP, VGA)? I also had
such a problem some months ago after I connected my monitor via DP. For 
some reason sometimes the video output of my graphics card was only sent 
to its DVI port and not to the DP. 
However after a update of xorg and its components this error was gone.

--
Regards
wabe

Re: [gentoo-user] Another X11 problem

[Adam Carter]

[-- Attachment #1: Type: text/plain, Size: 848 bytes --]

On Sat, Aug 20, 2016 at 5:15 PM, Ian Zimmerman <itz@buug.org> wrote:

> Once or twice per week, when I start X11 (via the startx program), my
> display goes all black.  When it happens there is no way back to a
> working state: I can see that Xorg is running, I can exit it by hitting
> Ctl-Alt-Backspace, but even then display does not recover.  The only
> thing I can do is hit Ctl-Alt-Del blindly and reboot, and next time it
> usually works.
>
> I did a diff of the Xorg logfile for successful and failed runs, and
> they are identical except for some trivial device renumbering.  Clearly
> there is no segfault as in the other scenario recently discussed on the
> list.  It is also not a monitor problem: turning monitor on & off
> doesn't help.
>

Have you checked dmesg and syslogs for messages?

Try a newer kernel and see if it goes away.

[-- Attachment #2: Type: text/html, Size: 1248 bytes --]

[gentoo-user] Re: Another X11 problem

[Ian Zimmerman]

On 2016-08-20, Neil Bothwick wrote:

> > Once or twice per week, when I start X11 (via the startx program),
> > my display goes all black.  When it happens there is no way back to
> > a working state: I can see that Xorg is running, I can exit it by
> > hitting Ctl-Alt-Backspace, but even then display does not recover.
> > The only thing I can do is hit Ctl-Alt-Del blindly and reboot, and
> > next time it usually works.
> 
> Can you get back to a VC with Ctrl-Alt-F1/2?

The answer is yes, as in: the system does the switch.  The screen is not
restored, though.  The only way I know the switch happens is that
Ctl-Alt-Del works after that (it is not bound inside X).

Some system information, probably incomplete (feel free to ask for
more):

 [2+0]~$ uname -a
Linux matica 4.4.6-gentoo #1 SMP Thu Jun 23 23:28:46 PDT 2016 x86_64 AMD Phenom(tm) II X4 955 Processor AuthenticAMD GNU/Linux

(kernel built from gentoo-sources)

 [4+0]~$ fgrep -i chipset /var/log/Xorg.0.log
[    48.269] (II) RADEON: Driver for ATI Radeon chipsets:
[    48.271] (II) VESA: driver for VESA chipsets: vesa
[    48.285] (--) RADEON(0): Chipset: "ATI Radeon HD 4250" (ChipID = 0x9715)

 [6+0]~$ fgrep -i dvi /var/log/Xorg.0.log
[    48.341] (II) RADEON(0): Output DVI-0 has no monitor section
[    48.385] (II) RADEON(0): EDID for output DVI-0
[    48.385] (II) RADEON(0): Printing probed modes for output DVI-0
[    48.385] (II) RADEON(0): Output DVI-0 connected
[    48.386] (II) RADEON(0): Output DVI-0 using initial mode 1920x1200

 [8+0]~$ fgrep -i x.org /var/log/Xorg.0.log
X.Org X Server 1.17.4
[    47.390] 	Before reporting problems, check http://wiki.x.org
[    47.684] 	X.Org ANSI C Emulation: 0.4
[    47.684] 	X.Org Video Driver: 19.0
[    47.684] 	X.Org XInput driver : 21.0
[    47.684] 	X.Org Server Extension : 9.0
[    48.150] (II) Module glx: vendor="X.Org Foundation"
[    48.150] 	ABI class: X.Org Server Extension, version 9.0
[    48.168] (II) Module ati: vendor="X.Org Foundation"
[    48.168] 	Module class: X.Org Video Driver
[    48.168] 	ABI class: X.Org Video Driver, version 19.0
[    48.219] (II) Module radeon: vendor="X.Org Foundation"
[    48.219] 	Module class: X.Org Video Driver
[    48.219] 	ABI class: X.Org Video Driver, version 19.0
[    48.239] (II) Module modesetting: vendor="X.Org Foundation"
[    48.239] 	Module class: X.Org Video Driver
[    48.239] 	ABI class: X.Org Video Driver, version 19.0
[    48.259] (II) Module fbdev: vendor="X.Org Foundation"
[    48.259] 	Module class: X.Org Video Driver
[    48.259] 	ABI class: X.Org Video Driver, version 19.0
[    48.269] (II) Module vesa: vendor="X.Org Foundation"
[    48.269] 	Module class: X.Org Video Driver
[    48.269] 	ABI class: X.Org Video Driver, version 19.0
[    48.285] (II) Module fbdevhw: vendor="X.Org Foundation"
[    48.285] 	ABI class: X.Org Video Driver, version 19.0
[    48.297] (II) Module exa: vendor="X.Org Foundation"
[    48.297] 	ABI class: X.Org Video Driver, version 19.0
[    48.395] (II) Module fb: vendor="X.Org Foundation"
[    48.395] 	ABI class: X.Org ANSI C Emulation, version 0.4
[    50.182] (II) Module evdev: vendor="X.Org Foundation"
[    50.182] 	Module class: X.Org XInput Driver
[    50.182] 	ABI class: X.Org XInput driver, version 21.0


-- 
Please *no* private Cc: on mailing lists and newsgroups
Why does the arrow on Hillary signs point to the right?

[gentoo-user] Re: Another X11 problem

[Ian Zimmerman]

On 2016-08-22, Adam Carter wrote:

> > Once or twice per week, when I start X11 (via the startx program),
> > my display goes all black.  When it happens there is no way back to
> > a working state: I can see that Xorg is running, I can exit it by
> > hitting Ctl-Alt-Backspace, but even then display does not recover.
> > The only thing I can do is hit Ctl-Alt-Del blindly and reboot, and
> > next time it usually works.

> Have you checked dmesg and syslogs for messages?

Yes.  Nothing (significantly) different when this happens.

> Try a newer kernel and see if it goes away.

I'm reluctant to go there before nailing it down a bit more, since
it only happens intermittently and a newer kernel will quite possibly
introduce other problems.

-- 
Please *no* private Cc: on mailing lists and newsgroups
Why does the arrow on Hillary signs point to the right?

Re: [gentoo-user] Re: Another X11 problem

[Corbin Bird]

On 08/25/2016 12:52 PM, Ian Zimmerman wrote:
> On 2016-08-22, Adam Carter wrote:
>
>>> Once or twice per week, when I start X11 (via the startx program),
>>> my display goes all black.  When it happens there is no way back to
>>> a working state: I can see that Xorg is running, I can exit it by
>>> hitting Ctl-Alt-Backspace, but even then display does not recover.
>>> The only thing I can do is hit Ctl-Alt-Del blindly and reboot, and
>>> next time it usually works.
>> Have you checked dmesg and syslogs for messages?
> Yes.  Nothing (significantly) different when this happens.
>
>> Try a newer kernel and see if it goes away.
> I'm reluctant to go there before nailing it down a bit more, since
> it only happens intermittently and a newer kernel will quite possibly
> introduce other problems.
>
I don't remember if you specified your appended command-line kernel 
parameters ( lilo / grub / grub2 ). I have an AMD Phenom II x4 980. The 
parameters have a major effect on the video ( i.e. AGP / IOMMU usage ).

[gentoo-user] Re: Another X11 problem

[Ian Zimmerman]

On 2016-08-25, Corbin Bird wrote:

> 
> On 08/25/2016 12:52 PM, Ian Zimmerman wrote:
> > On 2016-08-22, Adam Carter wrote:
> >
> >>> Once or twice per week, when I start X11 (via the startx program),
> >>> my display goes all black.  When it happens there is no way back to
> >>> a working state: I can see that Xorg is running, I can exit it by
> >>> hitting Ctl-Alt-Backspace, but even then display does not recover.
> >>> The only thing I can do is hit Ctl-Alt-Del blindly and reboot, and
> >>> next time it usually works.
> >> Have you checked dmesg and syslogs for messages?
> > Yes.  Nothing (significantly) different when this happens.
> >
> >> Try a newer kernel and see if it goes away.
> > I'm reluctant to go there before nailing it down a bit more, since
> > it only happens intermittently and a newer kernel will quite possibly
> > introduce other problems.
> >
> I don't remember if you specified your appended command-line kernel 
> parameters ( lilo / grub / grub2 ). I have an AMD Phenom II x4 980. The 
> parameters have a major effect on the video ( i.e. AGP / IOMMU usage ).
> 
> 

I have not posted them yet because I didn't suspect they were
important, but here they are now:

GRUB_CMDLINE_LINUX="rootfstype=ext4"

GRUB_CMDLINE_LINUX_DEFAULT="quiet noresume vt.default_utf8=1 consoleblank=0"

# Uncomment to disable graphical terminal (grub-pc only)
GRUB_TERMINAL=console


I added the consoleblank=0 bit just very recently in the pursuit of this
bug.  It happened without it, too.

The relevant kernel drivers are built as modules.  I load fbcon from
/etc/conf.d/modules, and that seems to be where the switch to graphics
mode occurs.  Note that everything is always okay at this point -
it just switches to a framebuffer enabled console, with a nice small
font.  The bug, if it happens, happens later, when I start xorg.

Also, I pass no module-specific parameters.  The only non-comment line
in /etc/conf.d/modules is:

modules="fbcon it87"

-- 
Please *no* private Cc: on mailing lists and newsgroups
Why does the arrow on Hillary signs point to the right?

Re: [gentoo-user] Re: Another X11 problem

[Corbin Bird]

On 08/25/2016 09:59 PM, Ian Zimmerman wrote:
> On 2016-08-25, Corbin Bird wrote:
>
>> On 08/25/2016 12:52 PM, Ian Zimmerman wrote:
>>> On 2016-08-22, Adam Carter wrote:
>>>
>>>>> Once or twice per week, when I start X11 (via the startx program),
>>>>> my display goes all black.  When it happens there is no way back to
>>>>> a working state: I can see that Xorg is running, I can exit it by
>>>>> hitting Ctl-Alt-Backspace, but even then display does not recover.
>>>>> The only thing I can do is hit Ctl-Alt-Del blindly and reboot, and
>>>>> next time it usually works.
>>>> Have you checked dmesg and syslogs for messages?
>>> Yes.  Nothing (significantly) different when this happens.
>>>
>>>> Try a newer kernel and see if it goes away.
>>> I'm reluctant to go there before nailing it down a bit more, since
>>> it only happens intermittently and a newer kernel will quite possibly
>>> introduce other problems.
>>>
>> I don't remember if you specified your appended command-line kernel
>> parameters ( lilo / grub / grub2 ). I have an AMD Phenom II x4 980. The
>> parameters have a major effect on the video ( i.e. AGP / IOMMU usage ).
>>
>>
> I have not posted them yet because I didn't suspect they were
> important, but here they are now:
>
> GRUB_CMDLINE_LINUX="rootfstype=ext4"
>
> GRUB_CMDLINE_LINUX_DEFAULT="quiet noresume vt.default_utf8=1 consoleblank=0"
>
> # Uncomment to disable graphical terminal (grub-pc only)
> GRUB_TERMINAL=console
>
>
> I added the consoleblank=0 bit just very recently in the pursuit of this
> bug.  It happened without it, too.
>
> The relevant kernel drivers are built as modules.  I load fbcon from
> /etc/conf.d/modules, and that seems to be where the switch to graphics
> mode occurs.  Note that everything is always okay at this point -
> it just switches to a framebuffer enabled console, with a nice small
> font.  The bug, if it happens, happens later, when I start xorg.
>
> Also, I pass no module-specific parameters.  The only non-comment line
> in /etc/conf.d/modules is:
>
> modules="fbcon it87"
>
Try setting up the IOMMU. ( Stabilize memory setup / management )

The IOMMU is really a modified "AGP bridge" built into the CPU.

Typical dmesg output :
[    0.000000] AGP: No AGP bridge found

If you don't create a "reserved for the AGP Aperature" memory block or 
setup the IOMMU, the kernel will setup 64Mg of memory and reserve it for 
"e820 -> AGP Aperature". If the IOMMU is not setup, it will 
automatically use ALL of the "AGP Aperature Reserved" memory.

The added kernel parameter for ( AGP / IOMMU on x86_64 ) : 
iommu=memaper[=1],noaperture

This sets up 64Mg for the IOMMU, and tells it "do not touch 'reserved 
for the AGP Aperature' memory".
( Software IO TLB is now preferred / automatic on 4.1.xx kernels. )

Typical dmesg output :
[    0.636904] AMD-Vi: Found IOMMU at 0000:00:00.2 cap 0x40
[    0.637060] AMD-Vi: Interrupt remapping enabled
[    0.644282] AMD-Vi: Lazy IO/TLB flushing enabled
[    0.717543] PCI-DMA: Using software bounce buffering for IO (SWIOTLB)
[    0.717700] software IO TLB [mem 0xb9ee4000-0xbdee4000] (64MB) mapped 
at [ffff8800b9ee4000-ffff8800bdee3fff]

KMS enabled X drivers need this setup on ( Fam10h / Fam15h ) CPUs.

Also check for any kernel parameters needed for KMS, by your choice of 
video driver.

Hope this helps.

Corbin

[gentoo-user] Re: Another X11 problem

[Ian Zimmerman]

On 2016-08-26 16:09, Corbin Bird wrote:

> >> I don't remember if you specified your appended command-line kernel
> >> parameters ( lilo / grub / grub2 ). I have an AMD Phenom II x4 980. The
> >> parameters have a major effect on the video ( i.e. AGP / IOMMU usage ).

> Try setting up the IOMMU. ( Stabilize memory setup / management )
> 
> The IOMMU is really a modified "AGP bridge" built into the CPU.
> 
> Typical dmesg output :
> [    0.000000] AGP: No AGP bridge found
> 
> If you don't create a "reserved for the AGP Aperature" memory block or 
> setup the IOMMU, the kernel will setup 64Mg of memory and reserve it for 
> "e820 -> AGP Aperature". If the IOMMU is not setup, it will 
> automatically use ALL of the "AGP Aperature Reserved" memory.
> 
> The added kernel parameter for ( AGP / IOMMU on x86_64 ) : 
> iommu=memaper[=1],noaperture
> 
> This sets up 64Mg for the IOMMU, and tells it "do not touch 'reserved 
> for the AGP Aperature' memory".
> ( Software IO TLB is now preferred / automatic on 4.1.xx kernels. )
> 
> Typical dmesg output :
> [    0.636904] AMD-Vi: Found IOMMU at 0000:00:00.2 cap 0x40
> [    0.637060] AMD-Vi: Interrupt remapping enabled
> [    0.644282] AMD-Vi: Lazy IO/TLB flushing enabled
> [    0.717543] PCI-DMA: Using software bounce buffering for IO (SWIOTLB)
> [    0.717700] software IO TLB [mem 0xb9ee4000-0xbdee4000] (64MB) mapped 
> at [ffff8800b9ee4000-ffff8800bdee3fff]
> 
> KMS enabled X drivers need this setup on ( Fam10h / Fam15h ) CPUs.

Some of the dmesg output on my system:

root@matica ~ # dmesg | fgrep -e AGP -e IOMMU -e TLB
[    0.000000] AGP: No AGP bridge found
[    0.000000] AGP: Checking aperture...
[    0.000000] AGP: No AGP bridge found
[    0.000000] AGP: Node 0: aperture [bus addr 0xa4000000-0xa5ffffff] (32MB)
[    0.000000] AGP: Your BIOS doesn't leave an aperture memory hole
[    0.000000] AGP: Please enable the IOMMU option in the BIOS setup
[    0.000000] AGP: This costs you 64MB of RAM
[    0.000000] AGP: Mapping aperture over RAM [mem 0xa4000000-0xa7ffffff] (65536KB)
[    0.005699] Last level iTLB entries: 4KB 512, 2MB 16, 4MB 8
[    0.005700] Last level dTLB entries: 4KB 512, 2MB 128, 4MB 64, 1GB 0
[    0.802036] PCI-DMA: Disabling AGP.
[    0.802111] PCI-DMA: using GART IOMMU.
[    0.802113] PCI-DMA: Reserving 64MB of IOMMU area in the AGP aperture

Do your recommendations still hold?

FWIW my BIOS setup has no option even remotely mentioning IOMMU, so I
don't know what the kernel is trying to tell me there.

This video is integrated with my motherboard (Gigabyte GA-880GM) and it
is not an external card plugged into a slot.

The kernel module is radeon and I'm pretty sure kernel modesetting is
the default.

Lastly, I never had this problem with 3.x kernels and the same hardware
(on Debian, and I think the last I used was 3.12, but I could be wrong
by a couple).

Thanks for trying to help!

-- 
Please *no* private Cc: on mailing lists and newsgroups
Why does the arrow on Hillary signs point to the right?

Re: [gentoo-user] Re: Another X11 problem

[Corbin Bird]

On 08/26/2016 05:23 PM, Ian Zimmerman wrote:
> On 2016-08-26 16:09, Corbin Bird wrote:
>
>>>> I don't remember if you specified your appended command-line kernel
>>>> parameters ( lilo / grub / grub2 ). I have an AMD Phenom II x4 980. The
>>>> parameters have a major effect on the video ( i.e. AGP / IOMMU usage ).
>> Try setting up the IOMMU. ( Stabilize memory setup / management )
>>
>> The IOMMU is really a modified "AGP bridge" built into the CPU.
>>
>> Typical dmesg output :
>> [    0.000000] AGP: No AGP bridge found
>>
>> If you don't create a "reserved for the AGP Aperature" memory block or
>> setup the IOMMU, the kernel will setup 64Mg of memory and reserve it for
>> "e820 -> AGP Aperature". If the IOMMU is not setup, it will
>> automatically use ALL of the "AGP Aperature Reserved" memory.
>>
>> The added kernel parameter for ( AGP / IOMMU on x86_64 ) :
>> iommu=memaper[=1],noaperture
>>
>> This sets up 64Mg for the IOMMU, and tells it "do not touch 'reserved
>> for the AGP Aperature' memory".
>> ( Software IO TLB is now preferred / automatic on 4.1.xx kernels. )
>>
>> Typical dmesg output :
>> [    0.636904] AMD-Vi: Found IOMMU at 0000:00:00.2 cap 0x40
>> [    0.637060] AMD-Vi: Interrupt remapping enabled
>> [    0.644282] AMD-Vi: Lazy IO/TLB flushing enabled
>> [    0.717543] PCI-DMA: Using software bounce buffering for IO (SWIOTLB)
>> [    0.717700] software IO TLB [mem 0xb9ee4000-0xbdee4000] (64MB) mapped
>> at [ffff8800b9ee4000-ffff8800bdee3fff]
>>
>> KMS enabled X drivers need this setup on ( Fam10h / Fam15h ) CPUs.
> Some of the dmesg output on my system:
>
> root@matica ~ # dmesg | fgrep -e AGP -e IOMMU -e TLB
> [    0.000000] AGP: No AGP bridge found
> [    0.000000] AGP: Checking aperture...
> [    0.000000] AGP: No AGP bridge found
> [    0.000000] AGP: Node 0: aperture [bus addr 0xa4000000-0xa5ffffff] (32MB)
> [    0.000000] AGP: Your BIOS doesn't leave an aperture memory hole
> [    0.000000] AGP: Please enable the IOMMU option in the BIOS setup
> [    0.000000] AGP: This costs you 64MB of RAM
> [    0.000000] AGP: Mapping aperture over RAM [mem 0xa4000000-0xa7ffffff] (65536KB)
> [    0.005699] Last level iTLB entries: 4KB 512, 2MB 16, 4MB 8
> [    0.005700] Last level dTLB entries: 4KB 512, 2MB 128, 4MB 64, 1GB 0
> [    0.802036] PCI-DMA: Disabling AGP.
> [    0.802111] PCI-DMA: using GART IOMMU.
> [    0.802113] PCI-DMA: Reserving 64MB of IOMMU area in the AGP aperture
>
> Do your recommendations still hold?
>
> FWIW my BIOS setup has no option even remotely mentioning IOMMU, so I
> don't know what the kernel is trying to tell me there.
>
> This video is integrated with my motherboard (Gigabyte GA-880GM) and it
> is not an external card plugged into a slot.
>
> The kernel module is radeon and I'm pretty sure kernel modesetting is
> the default.
>
> Lastly, I never had this problem with 3.x kernels and the same hardware
> (on Debian, and I think the last I used was 3.12, but I could be wrong
> by a couple).
>
> Thanks for trying to help!
>

The AMD Phenom II x4 980 is on a Gigabyte Motherboard.
They do NOT setup / support  the IOMMU properly for Linux.

Surprised ... you do have a "reserved for the AGP Aperature" memory block.
> [    0.000000] AGP: Node 0: aperture [bus addr 0xa4000000-0xa5ffffff] (32MB)

Use the kernel parameter :

iommu=memaper[=1],noaperture

Documentation is in :
/usr/src/linux/Documentation/kernel-parameters.txt
&&
/usr/src/linux/Documentation/x86/x86_64/boot-options.txt

--------------------

One other thing ... check dmesg for
> [ 0.000000] IOAPIC[0]: apic_id 9, version 33, address 0xfec00000, GSI 
> 0-23 [ 0.000000] IOAPIC[1]: apic_id 10, version 33, address 
> 0xfec20000, GSI 24-55 
Most of the Fam10h CPUs had 2 APIC(s). You need both working. ( IRQ / DMA re-routing ).
Standard Vanilla Kernels usually do not have 2x APIC support enabled.

--------------------

I can't even remember how many times I have gotten this warning on that MB. ( And no way to fix it! )
> [    0.000000] AGP: Your BIOS doesn't leave an aperture memory hole
> [    0.000000] AGP: Please enable the IOMMU option in the BIOS setup
> [    0.000000] AGP: This costs you 64MB of RAM
> [    0.000000] AGP: Mapping aperture over RAM [mem 0xa4000000-0xa7ffffff] (65536KB)

-------------------

Debian has a different way of ( building / configuring ) the graphics. They stick with very old kernels.

-------------------
Interesting reading :

https://www.x.org/wiki/RadeonFeature/ ( more kernel parameters / diagnostics )

-------------------

Hope this helps ...

Corbin

[gentoo-user] Re: Another X11 problem

[Ian Zimmerman]

On 2016-08-26 23:10, Corbin Bird wrote:

> One other thing ... check dmesg for

> > [ 0.000000] IOAPIC[0]: apic_id 9, version 33, address 0xfec00000, GSI 
> > 0-23 [ 0.000000] IOAPIC[1]: apic_id 10, version 33, address 
> > 0xfec20000, GSI 24-55 

> Most of the Fam10h CPUs had 2 APIC(s). You need both working. ( IRQ /
> DMA re-routing ).  Standard Vanilla Kernels usually do not have 2x
> APIC support enabled.

I enabled the option "x2apic" in the "Processor type and features"
section of kernel build, but that didn't change the fact that my kernel
only finds one.

The help text for that build option sounds like it does something very
different, though.  If there is another "2x APIC" option, I don't know
where it is :-(

I should also make clear that I radically underuse the GPU features.  I
don't play any shooting games, I don't watch movies, etc. etc.
Basically plain old 2d acceleration was good enough for me.

I just need to fix the bug :(

-- 
Please *no* private Cc: on mailing lists and newsgroups
Why does the arrow on Hillary signs point to the right?