Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.user > #238380 > unrolled thread

Stuttering Gnome Debian Buster

Back to article view | Back to linux.debian.user

Contents

  Stuttering Gnome Debian Buster solon <lists@13e.eu> - 2021-08-09 11:00 +0200
    Re: Stuttering Gnome Debian Buster Greg Wooledge <greg@wooledge.org> - 2021-08-09 13:40 +0200
      Re: Stuttering Gnome Debian Buster solon <lists@13e.eu> - 2021-08-11 10:20 +0200
        Re: Stuttering Gnome Debian Buster Greg Wooledge <greg@wooledge.org> - 2021-08-11 13:40 +0200
          Re: Stuttering Gnome Debian Buster solon <lists@13e.eu> - 2021-08-12 11:20 +0200

#238380 — Stuttering Gnome Debian Buster

[Multipart message — attachments visible in raw view] — view raw

Hello All,

I have sort of hybrid media server as my living room PC. It's a Dual 
Xeon 2650L-V3 (24 cores total) with 32Gb of DDR4 ram. It functions as a 
storage server for my 40Gb Infiniband SAN, and as a regular fileserver 
providing both samba and nfs. It also runs three virtual machines using 
kvm/qemu, a pfsense load balancer for the network, a pihole in a 
seperate no-gui debian install and a simple webserver debian install. In 
total these VM's have 6 cores and 12Gb of RAM assigned to them. It has a 
Radeon R7 240 2Gb DDR5 video card attached to a 1080p TV. I haven't paid 
any particular attention to optimizing NUMA nodes or the like, but the 
only VM that really has any substantial load is the pfsense VM.

Previously it was running Ubuntu 20.10, but I've recently reinstalled it 
with Debian 10 because Ubuntu development has been taking a development 
direction that I'm finding increasingly annoying. The only thing that 
has been problematic since the reinstall is that I'm noticing extreme 
stuttering on the gnome desktop which wasn't present under Ubuntu, even 
with a positively ancient 512Mb videocard.

I'm using the open source radeon drivers that you get using 
contrib/non-free and I've scoured google trying to figure out the 
problem. The only thing I can come up with is that Gnome 3.30 is 
exceedingly slow, so I installed xfce4, but to a lesser extent this 
still suffers the same problems. Another point that may be related is 
that I have to log in a great many times before I can actually get into 
the desktop. Gdm login screen typically simply comes back 3-5 times 
before the desktop will display. I feel it's probably related, but this 
too is something that I've not been able to find anything substantial 
about through google searches.

I'm wondering if anyone has a suggestion for solving this, or perhaps 
whether this is a known problem dat is likely to be naturally adressed 
by the updates in Debian 11 which is releasing very soon? Does anyone 
have any experience with this issue?

If anyone has any insights to share, I'd love to hear them.

EJ

[toc] | [next] | [standalone]

#238384

On Mon, Aug 09, 2021 at 10:39:24AM +0200, solon wrote:
> I'm using the open source radeon drivers that you get using contrib/non-free
> and I've scoured google trying to figure out the problem. The only thing I
> can come up with is that Gnome 3.30 is exceedingly slow,

The first thing you should do is identify your video chipset.  Use
the command "lspci -nn" and look for "VGA" in the output.  Also look
for any other lines that look like a video device, because sometimes
there's more than one, and sometimes they don't include the word "VGA".

The most important part of the output is the 8-hex-digit PCI identifier
code, usually enclosed in square brackets.  Mine is [8086:5912] for
example.

Once you know what hardware you're dealing with, you can Google for
it and perhaps find solutions.  Failing that, you can post it here
on the mailing list, and maybe someone else will have had experience
with a similar device, or will be better able to find solutions for
you on Google.

Unfortunately, sometimes you need to Google using the model names instead
of the PCI ID, because not all posted solutions include all the hardware
information.

In general, you're looking at three separate issues:

1) Many graphics devices require the installation of non-free firmware
   blobs, in order to function correctly.  In the absence of this
   firmware, the symptom is usually "no 3D acceleration which makes
   GNOME and other desktop environments run really slowly", or even
   "GNOME and other DEs refuse to start, but I installed a display
   manager, so now I'm stuck on a black screen with a cursor".

2) *Some* (not as many) graphics devices require the installation of
   non-free drivers.  This is usually the case with nVidia devices more
   so than Radeons.

3) *Some* of the newest devices require a whole new kernel + X server +
   Mesa stack.

So, knowing which device you're dealing with is critical here.  It will
determine which of these things you need to do.  (And you may need to do
more than one.)

In the best case, you only have to install firmware.  You run a command
like "dmesg | grep -i firmware" to see what firmware is missing, and then
you track down which package contains that firmware, and you install it.
Then reboot, and voila.

If you have to install an nVidia driver, there are wiki pages that
should be helpful.

If you have to install a whole new kernel/X/Mesa stack, you might be
able to use buster-backports.  But that will be a substantial investment
of time and effort that would probably be better spent upgrading to
bullseye (Debian 11) at this time, since bullseye's release is less than
a week away.

[toc] | [prev] | [next] | [standalone]

#238486

[Multipart message — attachments visible in raw view] — view raw

Thanks for the prompt response,

I should probably have specified, but I am certain I am using a 
compatible driver, and backports is enabled.

~ lspci -nn | grep VGA

gives

85:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. 
[AMD/ATI] Oland PRO [Radeon R7 240/340] [1002:6613]

and # radeontop shows no undue load on the device during the stuttering. 
Based on that I feel the problem is unlikely to be video driver related.

The system is running on a backports kernel:

~ uname -a

Linux MyMachine 5.10.0-0.bpo.8-amd64 #1 SMP Debian 5.10.46-2~bpo10+1 
(2021-07-22) x86_64 GNU/Linux

I've since tried a few more things, the most recent of which was 
disabling gnome extensions. This didn't initially appear to make any 
difference, but now after logging out of gnome, starting a xfce, logging 
out of that and back into one of the gnome types, the stuttering appears 
to be gone.

I'll upgrade to bullseye in a few days and will probably only continue 
investigating it if it comes back with enabled gnome extensions after 
that.

Thanks for the help so far! I'll get back if the problem shows up on 
bullseye.

EJ

On 2021-08-09 13:36, Greg Wooledge wrote:

> On Mon, Aug 09, 2021 at 10:39:24AM +0200, solon wrote:
> 
>> I'm using the open source radeon drivers that you get using 
>> contrib/non-free
>> and I've scoured google trying to figure out the problem. The only 
>> thing I
>> can come up with is that Gnome 3.30 is exceedingly slow,
> 
> The first thing you should do is identify your video chipset.  Use
> the command "lspci -nn" and look for "VGA" in the output.  Also look
> for any other lines that look like a video device, because sometimes
> there's more than one, and sometimes they don't include the word "VGA".
> 
> The most important part of the output is the 8-hex-digit PCI identifier
> code, usually enclosed in square brackets.  Mine is [8086:5912] for
> example.
> 
> Once you know what hardware you're dealing with, you can Google for
> it and perhaps find solutions.  Failing that, you can post it here
> on the mailing list, and maybe someone else will have had experience
> with a similar device, or will be better able to find solutions for
> you on Google.
> 
> Unfortunately, sometimes you need to Google using the model names 
> instead
> of the PCI ID, because not all posted solutions include all the 
> hardware
> information.
> 
> In general, you're looking at three separate issues:
> 
> 1) Many graphics devices require the installation of non-free firmware
> blobs, in order to function correctly.  In the absence of this
> firmware, the symptom is usually "no 3D acceleration which makes
> GNOME and other desktop environments run really slowly", or even
> "GNOME and other DEs refuse to start, but I installed a display
> manager, so now I'm stuck on a black screen with a cursor".
> 
> 2) *Some* (not as many) graphics devices require the installation of
> non-free drivers.  This is usually the case with nVidia devices more
> so than Radeons.
> 
> 3) *Some* of the newest devices require a whole new kernel + X server +
> Mesa stack.
> 
> So, knowing which device you're dealing with is critical here.  It will
> determine which of these things you need to do.  (And you may need to 
> do
> more than one.)
> 
> In the best case, you only have to install firmware.  You run a command
> like "dmesg | grep -i firmware" to see what firmware is missing, and 
> then
> you track down which package contains that firmware, and you install 
> it.
> Then reboot, and voila.
> 
> If you have to install an nVidia driver, there are wiki pages that
> should be helpful.
> 
> If you have to install a whole new kernel/X/Mesa stack, you might be
> able to use buster-backports.  But that will be a substantial 
> investment
> of time and effort that would probably be better spent upgrading to
> bullseye (Debian 11) at this time, since bullseye's release is less 
> than
> a week away.

[toc] | [prev] | [next] | [standalone]

#238502

On Wed, Aug 11, 2021 at 10:10:48AM +0200, solon wrote:
> I should probably have specified, but I am certain I am using a compatible
> driver, and backports is enabled.

But what did you *install* from backports?  Simply enabling the repository
doesn't do anything.  Backports are pinned in such a way that they are
never installed by default.  You must explicitly install each package.

> 85:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc.
> [AMD/ATI] Oland PRO [Radeon R7 240/340] [1002:6613]

OK.  First thing I wonder is whether you have installed the
firmware-amd-graphics package (from the non-free repository).  But even
then, I don't know whether it has the correct firmware for your device,
because I couldn't find "R7" or "Oland" in the package description.  I
do see "oland" a few times, but...

  * Radeon HD 8500/8600/8700 series CE microcode (amdgpu/oland_ce.bin)
  * Radeon HD 8500/8600/8700 series K SMC microcode
    (amdgpu/oland_k_smc.bin)
  * Radeon HD 8500/8600/8700 series MC microcode (amdgpu/oland_mc.bin)
  * Radeon HD 8500/8600/8700 series ME microcode (amdgpu/oland_me.bin)
  * Radeon HD 8500/8600/8700 series PFP microcode (amdgpu/oland_pfp.bin)
  * Radeon HD 8500/8600/8700 series RLC microcode (amdgpu/oland_rlc.bin)
  * Radeon HD 8500/8600/8700 series SMC microcode (amdgpu/oland_smc.bin)
  * Radeon HD 8500/8600/8700 series UVD microcode (amdgpu/oland_uvd.bin)

I don't know whether any of those is intended to match your device.

Checking "dmesg | grep -i firmware" would be good.

> Linux MyMachine 5.10.0-0.bpo.8-amd64 #1 SMP Debian 5.10.46-2~bpo10+1
> (2021-07-22) x86_64 GNU/Linux

So you grabbed a kernel image from backports, at least.  That's probably
a good step.  But this doesn't tell us whether you also grabbed the X
server (or Wayland), Mesa, firmware, etc.

> I'll upgrade to bullseye in a few days and will probably only continue
> investigating it if it comes back with enabled gnome extensions after that.

Upgrading to bullseye is probably wise for this new a device.  If nothing
else, it would save you having to figure out all of the backported
packages you need.

Googling gave me a page that's not directly related to Debian:

https://linux-hardware.org/?id=pci:1002-6613-1458-22b0

And an unanswered Debian bug report: https://bugs.debian.org/966420

But the person reporting that bug was using a buster kernel, where you
are using a backported kernel.  So perhaps that doesn't affect you.

[toc] | [prev] | [next] | [standalone]

#238549

[Multipart message — attachments visible in raw view] — view raw

On 2021-08-11 13:30, Greg Wooledge wrote:

> On Wed, Aug 11, 2021 at 10:10:48AM +0200, solon wrote:
> 
>> I should probably have specified, but I am certain I am using a 
>> compatible
>> driver, and backports is enabled.
> 
> But what did you *install* from backports?  Simply enabling the 
> repository
> doesn't do anything.  Backports are pinned in such a way that they are
> never installed by default.  You must explicitly install each package.

These things, I tend to install things from backports as they seem 
useful, not neccesarily just because they exist:
ibverbs-providers/buster-backports,now 24.0-2~bpo10+1 amd64 [installed]
ibverbs-utils/buster-backports,now 24.0-2~bpo10+1 amd64 [installed]
libibumad3/buster-backports,now 24.0-2~bpo10+1 amd64 [installed]
libibverbs1/buster-backports,now 24.0-2~bpo10+1 amd64 [installed]
libnvpair3linux/buster-backports,now 2.0.3-9~bpo10+1 amd64 
[installed,automatic]
librdmacm1/buster-backports,now 24.0-2~bpo10+1 amd64 [installed]
libuutil3linux/buster-backports,now 2.0.3-9~bpo10+1 amd64 
[installed,automatic]
libzfs4linux/buster-backports,now 2.0.3-9~bpo10+1 amd64 
[installed,automatic]
libzpool4linux/buster-backports,now 2.0.3-9~bpo10+1 amd64 
[installed,automatic]
linux-headers-5.10.0-0.bpo.8-amd64/buster-backports,now 
5.10.46-2~bpo10+1 amd64 [installed]
linux-headers-5.10.0-0.bpo.8-common/buster-backports,buster-backports,now 
5.10.46-2~bpo10+1 all [installed,automatic]
linux-image-5.10.0-0.bpo.7-amd64/buster-backports,now 5.10.40-1~bpo10+1 
amd64 [installed]
linux-image-5.10.0-0.bpo.8-amd64/buster-backports,now 5.10.46-2~bpo10+1 
amd64 [installed]
linux-kbuild-5.10/buster-backports,now 5.10.46-2~bpo10+1 amd64 
[installed,automatic]
rdma-core/buster-backports,now 24.0-2~bpo10+1 amd64 [installed]
rdmacm-utils/buster-backports,now 24.0-2~bpo10+1 amd64 [installed]
spl/buster-backports,buster-backports,now 2.0.3-9~bpo10+1 all 
[installed]
zfs-dkms/buster-backports,buster-backports,now 2.0.3-9~bpo10+1 all 
[installed]
zfsutils-linux/buster-backports,now 2.0.3-9~bpo10+1 amd64 [installed]

>> 85:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc.
>> [AMD/ATI] Oland PRO [Radeon R7 240/340] [1002:6613]
> 
> OK.  First thing I wonder is whether you have installed the
> firmware-amd-graphics package (from the non-free repository).  But even
> then, I don't know whether it has the correct firmware for your device,
> because I couldn't find "R7" or "Oland" in the package description.  I
> do see "oland" a few times, but...

I didn't have backports graphics firmware. I've now installed it. Good 
suggestion. I'll see if that makes any difference:
firmware-amd-graphics/buster-backports,buster-backports,now 
20210315-2~bpo10+1 all [installed]

> * Radeon HD 8500/8600/8700 series CE microcode (amdgpu/oland_ce.bin)
> * Radeon HD 8500/8600/8700 series K SMC microcode
> (amdgpu/oland_k_smc.bin)
> * Radeon HD 8500/8600/8700 series MC microcode (amdgpu/oland_mc.bin)
> * Radeon HD 8500/8600/8700 series ME microcode (amdgpu/oland_me.bin)
> * Radeon HD 8500/8600/8700 series PFP microcode (amdgpu/oland_pfp.bin)
> * Radeon HD 8500/8600/8700 series RLC microcode (amdgpu/oland_rlc.bin)
> * Radeon HD 8500/8600/8700 series SMC microcode (amdgpu/oland_smc.bin)
> * Radeon HD 8500/8600/8700 series UVD microcode (amdgpu/oland_uvd.bin)
> 
> I don't know whether any of those is intended to match your device.
> 
> Checking "dmesg | grep -i firmware" would be good.

Reporting as (after backports firmware install):
[ 2.967504] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/oland_pfp.bin
[ 2.967525] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/oland_me.bin
[ 2.967545] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/oland_ce.bin
[ 2.967564] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/oland_rlc.bin
[ 2.967594] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/si58_mc.bin
[ 2.967636] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/oland_smc.bin
[ 2.976662] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/TAHITI_uvd.bin
[ 2.976711] radeon 0000:85:00.0: firmware: direct-loading firmware 
radeon/TAHITI_vce.bin

>> Linux MyMachine 5.10.0-0.bpo.8-amd64 #1 SMP Debian 5.10.46-2~bpo10+1
>> (2021-07-22) x86_64 GNU/Linux
> 
> So you grabbed a kernel image from backports, at least.  That's 
> probably
> a good step.  But this doesn't tell us whether you also grabbed the X
> server (or Wayland), Mesa, firmware, etc.

To my knowledge there is, by design, nothing X related in backports. 
I've looked for xserver and gnome packages and discovered that nothing 
GUI is included in backports, though I can't remember exactly where I 
found that.

>> I'll upgrade to bullseye in a few days and will probably only continue
>> investigating it if it comes back with enabled gnome extensions after 
>> that.
> 
> Upgrading to bullseye is probably wise for this new a device.  If 
> nothing
> else, it would save you having to figure out all of the backported
> packages you need.
> 
> Googling gave me a page that's not directly related to Debian:
> 
> https://linux-hardware.org/?id=pci:1002-6613-1458-22b0
> 
> And an unanswered Debian bug report: https://bugs.debian.org/966420
> 
> But the person reporting that bug was using a buster kernel, where you
> are using a backported kernel.  So perhaps that doesn't affect you.

It's a possibility, but this problem has been present over two different 
radeon videocards. The previous one a ASUS ATI/Radeon 512Mb HD5450, 
which exhibited the same behaviour. I upgraded to this 2Gb R7 240, which 
didn't change anything.

I'll wait for the official bullseye release and then upgrade. This 
system runs a VM that does the load balancing for my home network, so I 
can't take it offline while the better half is busy with endless Teams 
meetings, it requires a little planning, so I'd prefer just doing it 
once when bullsseye goes to stable.

So far I'm suspecting that the problem is related to gnome-extensions, 
with those disabled it certainly behaves alot better. That wasn't a 
problem under Ubuntu though, so I'm guessing there is an issue with 
gnome versions somewhere there.

Thanks for the responses so far,

EJ

[toc] | [prev] | [standalone]

Back to top | Article view | linux.debian.user

csiph-web