AMD fglrx not working under centos 7.3

Issues related to hardware problems
s1m3mu3
Posts: 3
Joined: 2016/12/29 11:59:16

AMD fglrx not working under centos 7.3

Postby s1m3mu3 » 2016/12/29 12:30:44

Hi,
I have a Bonaire XT Radeon HD 7790, AMD graphics card.
I recently updated my system to centos 7.3. Now the AMD fglrx graphics driver stopped working. I need it because of the openCL support.
The package from ELRepo did not work even with 7.2. I used the install script provided on the amd website.
Under centos 7.2 I had to reinstall for every new kernel, because it creates a kernel module. But after that it worked fine.
On the AMD website, RHEL 7.2 is mentioned as an supported distribution but not 7.3.

With some changes in 7.3 fglrx broke. Installing works and the kernel module is created. But after restarting instead of the login screen I get only a black screen with a mouse pointer. Changing to a virtual console (tty2) works fine. So this might be related a new xserver version.

Is there a way to downgrade to a previous xserver until this issue is resolved?

I tried the following:

Code: Select all

yum downgrade xorg\*
...
No Match for available package: xorg-sgml-doctools-1.10-5.el7.noarch
No Match for available package: xorg-x11-apps-7.7-6.el7.x86_64
No Match for available package: xorg-x11-docs-1.6-7.el7.noarch
...


It seems the old xorg is no longer found in the repository. Even yum history undo to revert to 7.2 is not possible due to unavailable packages.

Any help would be greatly appreciated,
s1m3mu3

nugai
Posts: 13
Joined: 2014/08/01 23:03:19

Re: AMD fglrx not working under centos 7.3

Postby nugai » 2017/01/15 03:37:12

FYI, I just wanted to acknowledge that you're not the only one with such a problem. A couple of days ago, after running "yum upgrade", which upgraded the kernel from 3.10.0-327.36.3 to 3.10.0-514.2.2 (and "rpm --query centos-release" showed centos-release-7-3.1611.el7.centos.x86_64), I encountered blank screens on a Xfce 4-monitor system with dual HD 6450 cards. When I selected the previous kernel version in the boot menu, everything ran fine.

Running "Xorg -version" on both kernels showed that they use the exact same version of X11 server (namely xorg-x11-server 1.17.2-22.el7). Therefore, it appeared the X11 server was not the culprit, and downgrading it probably will not solve your problem.

As I documented further below, looking at the xserver log files "/var/log/Xorg.0.log" and looking for entries in the boot log with "dmesg | grep fglrx" showed that the fglrx driver did not load in the latest kernel version (failing to create "/proc/ati" and its contents). This hint led me to an entry in the Elrepo bug tracking system, where developer "wolfy" suggested to try the files from http://wdl.lug.ro/fglrx. I downloaded and installed these files (date stamped 30-Dec-2016 09:33 at the time of this post), which immediately solved the problem upon reboot for me. So, I'd say my problem (and hopefully yours, too) seems to be just a "driver out of sync" problem and a matter of patience until the drivers percolate through development and early testing into elrepo-testing (which at the time of this posting contained fglrx-x11-drv-15.12-4.el7.elrepo.x86_64.rpm dated Aug-2016) and then the main distribution repos that yum catches up with.

Here's how I installed the files manually (via ssh console from another system, logged in as user, then using su or sudo for executing the commands with root privileges):

Code: Select all

cd Downloads
wget http://wdl.lug.ro/fglrx/fglrx-x11-drv-15.12-5.el7.elrepo.x86_64.rpm
wget http://wdl.lug.ro/fglrx/kmod-fglrx-15.12-5.el7.elrepo.x86_64.rpm
rpm -Uvh kmod-fglrx-15.12-5.el7.elrepo.x86_64.rpm
rpm -Uvh fglrx-x11-drv-15.12-5.el7.elrepo.x86_64.rpm
reboot


Note added 19-Jan-2107: I just noticed today that the driver issue has been reported last week in the Elrepo bug tracker here. While the driver fixes problems with existing setups, please note that aticonfig may not be working yet for new or first installs. wolfy explained the problem was caused by a "library that was removed from the distro" and that aticonfig was the reason why the drivers have not yet been updated in elrepo and elrepo-testing. So, I'm guessing it will be just a matter of time until all problems will be completely resolved.

I hope this helps.

A big 'Thank You!' to all the developers and volunteers whose work helps keeping our systems alive!!!

=================================================================

And now, as an afterthought, for the sake of my own laziness and not having to remember where to find my notes, and for the benefit of others who may need to troubleshoot their own systems, here's a description of how a tracked down the problem (which required booting into the current and previous kernel several times to find differences).

1. Check Xorg version

On kernel 3.10.0-327.36.3 (which runs OK):

Code: Select all

[root@mysystem Downloads]# Xorg -version

X.Org X Server 1.17.2
Release Date: 2015-06-16
X Protocol Version 11, Revision 0
Build Operating System:  2.6.32-573.18.1.el6.x86_64
Current Operating System: Linux mysystem.mydomain.com 3.10.0-327.36.3.el7.x86_64 #1 SMP Mon Oct 24 16:09:20 UTC 2016 x86_64
Kernel command line: BOOT_IMAGE=/vmlinuz-3.10.0-327.36.3.el7.x86_64 root=/dev/mapper/centos-root ro rd.lvm.lv=centos/swap vconsole.font=latarcyrheb-sun16 rd.lvm.lv=centos/root crashkernel=auto vconsole.keymap=us rhgb quiet LANG=en_US.UTF-8 radeon.modeset=0 rd.driver.blacklist=radeon
Build Date: 06 November 2016  12:43:39AM
Build ID: xorg-x11-server 1.17.2-22.el7
Current version of pixman: 0.34.0
        Before reporting problems, check http://wiki.x.org
        to make sure that you have the latest version.
[root@mysystem Downloads]#


On kernel 3.10.0-514.2.2 (which fails):

Code: Select all

[root@mysystem Downloads]# Xorg -version

X.Org X Server 1.17.2
Release Date: 2015-06-16
X Protocol Version 11, Revision 0
Build Operating System:  2.6.32-573.18.1.el6.x86_64
Current Operating System: Linux mysystem.mydomain.com 3.10.0-514.2.2.el7.x86_64 #1 SMP Tue Dec 6 23:06:41 UTC 2016 x86_64
Kernel command line: BOOT_IMAGE=/vmlinuz-3.10.0-514.2.2.el7.x86_64 root=/dev/mapper/centos-root ro rd.lvm.lv=centos/swap vconsole.font=latarcyrheb-sun16 rd.lvm.lv=centos/root crashkernel=auto vconsole.keymap=us rhgb quiet LANG=en_US.UTF-8 radeon.modeset=0 rd.driver.blacklist=radeon
Build Date: 06 November 2016  12:43:39AM
Build ID: xorg-x11-server 1.17.2-22.el7
Current version of pixman: 0.34.0
        Before reporting problems, check http://wiki.x.org
        to make sure that you have the latest version.
[root@mysystem Downloads]#

Result: Exact same version. Hence X11 is not the culprit.

2. Check and and see if server starts

Looking the end of the log file xserver log file "/var/log/Xorg.0.log" shows that X11 fails under 3.10.0-514.2.2 :

Code: Select all

[root@mysystem Downloads]# tail -14 /var/log/Xorg.0.log
[    24.118] (EE)
[    24.118] (EE) Segmentation fault at address 0x22b0
[    24.118] (EE)
Fatal server error:
[    24.118] (EE) Caught signal 11 (Segmentation fault). Server aborting
[    24.118] (EE)
[    24.118] (EE)
Please consult the The X.Org Foundation support
         at http://wiki.x.org
 for help.
[    24.118] (EE) Please also check the log file at "/var/log/Xorg.0.log" for additional information.
[    24.118] (EE)
[    24.119] (EE) fglrx(0): firegl_SetSuspendResumeState FAILED -9.
[    24.414] (EE) fglrx(1): firegl_SetSuspendResumeState FAILED -9.

Result: Ah, not good.

3. Check where differences start

Scrolling back in the log files shows, by comparison, where and how the same version of the X11 server fails in one instance but not the other, after loading /usr/lib64/xorg/modules/linux/libfglrxdrm.so. Specifically, note the difference between “ukiDynamicMajor: found major device number 248” and “ukiDynamicMajor: failed to open /proc/ati/major” about 11 lines from the top.

On kernel 3.10.0-327.36.3 (which runs OK):

Code: Select all

(II) Loading /usr/lib64/xorg/modules/linux/libfglrxdrm.so
[    16.329] (II) Module fglrxdrm: vendor="FireGL - AMD Technologies Inc."
[    16.329]    compiled for 1.4.99.906, module version = 15.30.3
[    16.329] (II) AMD Proprietary Linux Driver Version Identifier:15.30.3
[    16.329] (II) AMD Proprietary Linux Driver Release Identifier: UNSUPPORTED-15.302                   
[    16.329] (II) AMD Proprietary Linux Driver Build Date: Dec 17 2015 02:43:16
[    16.329] (++) using VT number 1

[    16.331] (WW) Falling back to old probe method for fglrx
[    16.368] (II) Loading PCS database from /etc/ati/amdpcsdb /etc/ati/amdpcsdb.default
[    16.377] ukiDynamicMajor: found major device number 248
[    16.377] ukiDynamicMajor: found major device number 248
[    16.377] ukiOpenByBusid: Searching for BusID PCI:1:0:0
[    16.377] ukiOpenDevice: node name is /dev/ati/card0
[    16.377] ukiOpenDevice: open result is 9, (OK)
[    16.549] ukiOpenByBusid: ukiOpenMinor returns 9
[    16.549] ukiOpenByBusid: ukiGetBusid reports PCI:1:0:0
[    16.551] ukiDynamicMajor: found major device number 248
[    16.552] ukiDynamicMajor: found major device number 248
[    16.552] ukiOpenByBusid: Searching for BusID PCI:1:0:0
[    16.552] ukiOpenDevice: node name is /dev/ati/card0
[    16.552] ukiOpenDevice: open result is 9, (OK)
[    16.552] ukiOpenByBusid: ukiOpenMinor returns 9
[    16.552] ukiOpenByBusid: ukiGetBusid reports PCI:1:0:0
[    16.552] ukiDynamicMajor: found major device number 248
[    16.552] ukiDynamicMajor: found major device number 248
[    16.552] ukiOpenByBusid: Searching for BusID PCI:4:0:0
[    16.552] ukiOpenDevice: node name is /dev/ati/card0
[    16.552] ukiOpenDevice: open result is 9, (OK)
[    16.552] ukiOpenByBusid: ukiOpenMinor returns 9
[    16.552] ukiOpenByBusid: ukiGetBusid reports PCI:1:0:0
[    16.552] ukiOpenDevice: node name is /dev/ati/card1
[    16.885] ukiOpenDevice: open result is 9, (OK)
[    16.885] ukiOpenByBusid: ukiOpenMinor returns 9
[    16.885] ukiOpenByBusid: ukiGetBusid reports PCI:4:0:0
[    16.886] ukiDynamicMajor: found major device number 248
[    16.886] ukiDynamicMajor: found major device number 248
[    16.886] ukiOpenByBusid: Searching for BusID PCI:4:0:0
[    16.886] ukiOpenDevice: node name is /dev/ati/card0
[    16.886] ukiOpenDevice: open result is 9, (OK)
[    16.886] ukiOpenByBusid: ukiOpenMinor returns 9
[    16.886] ukiOpenByBusid: ukiGetBusid reports PCI:1:0:0
[    16.886] ukiOpenDevice: node name is /dev/ati/card1
[    16.886] ukiOpenDevice: open result is 9, (OK)
[    16.886] ukiOpenByBusid: ukiOpenMinor returns 9
[    16.886] ukiOpenByBusid: ukiGetBusid reports PCI:4:0:0
[    16.891] (--) Chipset Supported AMD Graphics Processor (0x6779) found
[    16.891] (--) Chipset Supported AMD Graphics Processor (0x6779) found
[    16.891] (--) Chipset Supported AMD Graphics Processor (0x6779) found
[    16.891] (--) Chipset Supported AMD Graphics Processor (0x6779) found


On kernel 3.10.0-514.2.2 (which fails):

Code: Select all

[    23.267] (II) Loading /usr/lib64/xorg/modules/linux/libfglrxdrm.so
[    23.267] (II) Module fglrxdrm: vendor="FireGL - AMD Technologies Inc."
[    23.267]    compiled for 1.4.99.906, module version = 15.30.3
[    23.267] (II) AMD Proprietary Linux Driver Version Identifier:15.30.3
[    23.267] (II) AMD Proprietary Linux Driver Release Identifier: UNSUPPORTED-15.302                   
[    23.267] (II) AMD Proprietary Linux Driver Build Date: Dec 17 2015 02:43:16
[    23.267] (++) using VT number 1

[    23.267] (WW) Falling back to old probe method for fglrx
[    23.298] (II) Loading PCS database from /etc/ati/amdpcsdb /etc/ati/amdpcsdb.default
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.303] ukiDynamicMajor: failed to open /proc/ati/major
[    23.307] (--) Chipset Supported AMD Graphics Processor (0x6779) found
[    23.307] (--) Chipset Supported AMD Graphics Processor (0x6779) found
[    23.307] (--) Chipset Supported AMD Graphics Processor (0x6779) found
[    23.307] (--) Chipset Supported AMD Graphics Processor (0x6779) found

Result: "failed to open /proc/ati/major" gives a hint.

4. Check for presence of /proc/ati

Doing an "ls -l /proc" shows that no /proc/ati (and therefore also no /proc/ati/major) existed at all under 3.10.0-514.2.2:

Code: Select all

[root@mysystem Downloads]# ls -l /proc
– snip –
dr-xr-xr-x.  9 root           root                         0 Jan 14 13:52 99
dr-xr-xr-x.  2 root           root                         0 Jan 14 17:24 acpi
dr-xr-xr-x.  7 root           root                         0 Jan 14 17:24 asound
-r--r--r--.  1 root           root                         0 Jan 14 17:24 buddyinfo
– snip –


...while under 3.10.0-327.36.3, the following /proc/ati/major structure exists:

Code: Select all

[root@mysystem Downloads]# ls -l /proc/ati/
total 0
dr-xr-xr-x. 2 root root 0 Jan 14 18:08 0
dr-xr-xr-x. 2 root root 0 Jan 14 18:08 1
-r--r--r--. 1 root root 0 Jan 14 18:08 debug
-r--r--r--. 1 root root 0 Jan 14 18:08 major
[root@mysystem Downloads]#


5. Check boot log entries for "fglrx"

The command “dmesg | grep fglrx” produces a number of entries on 3.10.0-327.36.3, but there are no (zero) entries on 3.10.0-514.2.2. This is further evidence that the driver does not load at boot time under 3.10.0-514.2.2.

Hence, most likely the driver is the culprit.

Since AMD stopped official support for HD 6xxx+ cards in Nov-2015 (see here), I briefly considered compiling AMD drivers based on what is described for FC23 on this page here, but luckily found somebody else already did that in the solution described above.

Undoubtedly, users with AMD HD cards will be in a world of hurt going forward, and therefore I'm considering upgrading my graphics card to a card that takes advantage of using AMD's latest supported drivers (currently 16.50, see list of supported cards about halfway down the page here -- but there's no guarantee AMD will release updated drivers in time, either). However, I'm lazy and cheap, and it would break my heart having to rip apart a system and ditch hardware that works perfectly well.

s1m3mu3
Posts: 3
Joined: 2016/12/29 11:59:16

Re: AMD fglrx not working under centos 7.3

Postby s1m3mu3 » 2017/01/31 11:14:08

thank you for your detailed report. Installing the packagages individually is not working for me:

Code: Select all

# rpm -Uvh kmod-fglrx-15.12-5.el7.elrepo.x86_64.rpm
error: Failed dependencies:
   fglrx-x11-drv = 15.12 is needed by kmod-fglrx-15.12-5.el7.elrepo.x86_64
   
# rpm -Uvh fglrx-x11-drv-15.12-5.el7.elrepo.x86_64.rpm
error: Failed dependencies:
   fglrx-kmod = 15.12-5.el7.elrepo is needed by fglrx-x11-drv-15.12-5.el7.elrepo.x86_64

When I try to install both (with rpm or yum, the following happens):

Code: Select all

rpm -ivh kmod-fglrx-15.12-5.el7.elrepo.x86_64.rpm fglrx-x11-drv-15.12-5.el7.elrepo.x86_64.rpm
Preparing...                          ################################# [100%]
Updating / installing...
   1:kmod-fglrx-15.12-5.el7.elrepo    ################################# [ 50%]
Working. This may take some time ...
Done.
   2:fglrx-x11-drv-15.12-5.el7.elrepo ################################# [100%]
Failed to install module fglrx

Broadcast message from systemd-journald@xxx (Tue 2017-01-31 12:05:42 CET):

dracut[22659]: Failed to install module fglrx

So I can not test your approach.
This already happened before (centos 7.2, old kernel), which is why I installed the driver with the install script from the amd website. Unfortunately this is no longer possible with centos 7.3.

pjwelsh
Posts: 2580
Joined: 2007/01/07 02:18:02
Location: Central IL USA

Re: AMD fglrx not working under centos 7.3

Postby pjwelsh » 2017/02/01 17:03:01

@s1m3mu3
Try to use yum instead to install those. The rpm command does not do dependency resolution:
yum install fglrx-x11-drv-15.12-5.el7.elrepo.x86_64.rpm kmod-fglrx-15.12-5.el7.elrepo.x86_64.rpm

@nugai
Sadly, after install the above 2 packages, I'm left with a black window that only shows a small cursor (no login screen). The "EE" line in /var/log/Xorg.0.log had error on /usr/X11R6/lib64/modules/dri/fglrx_dri.so no such file. I created the directory and 'ln -s /usr/lib64/dri/fglrx_dri.so /usr/X11R6/lib64/modules/dri/' and the blank screen persists. I did run 'amdconfig --initial' from the local console of that laptop.

s1m3mu3
Posts: 3
Joined: 2016/12/29 11:59:16

Re: AMD fglrx not working under centos 7.3

Postby s1m3mu3 » 2017/02/02 08:58:02

@ pjwelsh
as I mentioned above, using yum results in the same error

Code: Select all

yum install kmod-fglrx-15.12-5.el7.elrepo.x86_64.rpm fglrx-x11-drv-15.12-5.el7.elrepo.x86_64.rpm


Update 06 Feb 2017:
The new AMDGPU-Pro Driver Version 16.60 now also supports my HD 7790 so I am using this driver instead of the fglrx.

supertight
Posts: 92
Joined: 2017/02/07 21:47:51

Re: AMD fglrx not working under centos 7.3

Postby supertight » 2017/02/08 22:17:23

I'm having the same issues. AMD Radion HD 8450g / Lenovo Thinkpad e545 laptop.
Main display working / HDMI output not working. from gui no video size settings available in menu for main or hdmi output.
I cant find the x11 .conf file to change via command line.

"yum -y install kmod-fglrx"

Code: Select all

-> Running transaction check
---> Package kmod-fglrx.x86_64 0:15.12-3.el7.elrepo will be installed
--> Processing Dependency: kernel(kill_fasync) = 0xdd9d7a7d for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(fasync_helper) = 0x124106ba for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_unbind_pasid) = 0xc3022564 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_set_invalidate_ctx_cb) = 0x904f86f0 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_set_invalid_ppr_cb) = 0xe693e19b for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_init_device) = 0xbf4e6777 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_free_device) = 0x90b26027 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_enable_device_erratum) = 0x4ea602e8 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_device_info) = 0x0eab57ba for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_bind_pasid) = 0x8bd98934 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: fglrx-x11-drv = 15.12 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Running transaction check
---> Package fglrx-x11-drv.x86_64 0:15.12-3.el7.elrepo will be installed
---> Package kmod-fglrx.x86_64 0:15.12-3.el7.elrepo will be installed
--> Processing Dependency: kernel(kill_fasync) = 0xdd9d7a7d for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(fasync_helper) = 0x124106ba for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_unbind_pasid) = 0xc3022564 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_set_invalidate_ctx_cb) = 0x904f86f0 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_set_invalid_ppr_cb) = 0xe693e19b for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_init_device) = 0xbf4e6777 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_free_device) = 0x90b26027 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_enable_device_erratum) = 0x4ea602e8 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_device_info) = 0x0eab57ba for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Processing Dependency: kernel(amd_iommu_bind_pasid) = 0x8bd98934 for package: kmod-fglrx-15.12-3.el7.elrepo.x86_64
--> Finished Dependency Resolution
 You could try using --skip-broken to work around the problem
** Found 3 pre-existing rpmdb problem(s), 'yum check' output follows:
ipa-client-4.4.0-14.el7.centos.4.x86_64 has installed conflicts freeipa-client: ipa-client-4.4.0-14.el7.centos.4.x86_64
ipa-client-common-4.4.0-14.el7.centos.4.noarch has installed conflicts freeipa-client-common: ipa-client-common-4.4.0-14.el7.centos.4.noarch
ipa-common-4.4.0-14.el7.centos.4.noarch has installed conflicts freeipa-common: ipa-common-4.4.0-14.el7.centos.4.noarch


This hangs the video. I can ssh into the system and run.
"yum -y install fglrx-x11-drv-15.12-5.el7.elrepo.x86_64.rpm kmod-fglrx-15.12-5.el7.elrepo.x86_64.rpm"

Both install clean without error. However, I still have no video. I have to pass yum remove command via ssh to restore video in the main monitor and the hdmi output is still broken.

nugai
Posts: 13
Joined: 2014/08/01 23:03:19

Re: AMD fglrx not working under centos 7.3

Postby nugai » 2017/03/02 04:33:41

FYI, developer wolfy today posted a message on the elrepo bug tracker (see here) that the fglrx-15.12-5 drivers are now in his stable repository and that newer fglrx-15.12-6 drivers (dated 26-Feb-2017) are now in his testing repository at http://wdl.lug.ro/fglrx/testing/ , but he thinks the newer drivers still need further improvement (although I encountered no problems on my system):
They do work but with a caveat: for a reason that I was not yet able to identify, X can be started as startx but not automatic as graphical login from multi-user.target. I am still debugging this issue which puzzles me, as X is clearly happy, only the DE is not.

cwatson
Posts: 127
Joined: 2007/10/02 18:55:02

Re: AMD fglrx not working under centos 7.3

Postby cwatson » 2017/03/02 19:30:26

nugai wrote:FYI, developer wolfy today posted a message on the elrepo bug tracker (see here) that the fglrx-15.12-5 drivers are now in his stable repository and that newer fglrx-15.12-6 drivers (dated 26-Feb-2017) are now in his testing repository at http://wdl.lug.ro/fglrx/testing/ , but he thinks the newer drivers still need further improvement (although I encountered no problems on my system):
They do work but with a caveat: for a reason that I was not yet able to identify, X can be started as startx but not automatic as graphical login from multi-user.target. I am still debugging this issue which puzzles me, as X is clearly happy, only the DE is not.


nugai, would you mind listing the steps you took to install wolfy's latest driver? I attempted with the fglrx-15.12-5 a couple weeks ago (per the advice on the ElRepo bug tracker page), but as reported by other users in this thread there was a dependency issue and I couldn't install either. Which kernel version are you using? Do you have to run aticonfig --initial, then reboot? Thanks.

nugai
Posts: 13
Joined: 2014/08/01 23:03:19

Re: AMD fglrx not working under centos 7.3

Postby nugai » 2017/03/03 20:53:13

cwatson wrote:nugai, would you mind listing the steps you took to install wolfy's latest driver? I attempted with the fglrx-15.12-5 a couple weeks ago (per the advice on the ElRepo bug tracker page), but as reported by other users in this thread there was a dependency issue and I couldn't install either. Which kernel version are you using? Do you have to run aticonfig --initial, then reboot? Thanks.

Sure. I did the driver install on kernel 3.10.0-514.6.2.el7.x86_64 but "yum update" (which I always run before and after fglrx driver updates) since then installed the 3.10.0-514.10.2.el7.x86_64 kernel.

I used a quick and ugly hack (using "rpm" rather than "yum") that I'm not particularly proud of -- and yum now occasionally complains about "rpms were updated outside of database" or something like that. I have a hunch there are probably better ways to do it (such as adding wolfy's testing repo to the repo list), but here's what I did:

1. Logged in remotely via SSH (using Putty from a windows machine), then switched to root via "su" (I don't allow ssh to root, I hate "sudo" and I don't like my user account in the wheel group because for testing purposes, I want to experience the same constraints unprivileged users do), then ran "yum update":

Code: Select all

su
yum update

2. If you have a running installation with a complex configuration (like I do on my 4-monitor system), backup the /etc/X11/xorg.conf file, as the new rpms will overwrite it:

Code: Select all

cp /etc/X11/xorg.conf /etc/X11/xorg.conf-backup

3. Switched to Download directory with:

Code: Select all

cd Download

4. Got the 15.12-6 drivers from wolfy's testing repository:

Code: Select all

 wget http://wdl.lug.ro/fglrx/testing/kmod-fglrx-15.12-6.el7.elrepo.x86_64.rpm
 wget http://wdl.lug.ro/fglrx/testing/fglrx-x11-drv-15.12-6.el7.elrepo.x86_64.rpm

5. Forced the removal of the 15.12-5 drivers (the "-e" option is for "erase" and the "--nodeps" option ignores dependencies):

Code: Select all

 rpm -e --nodeps kmod-fglrx-15.12-5.el7.elrepo.x86_64 fglrx-x11-drv-15.12-5.el7.elrepo.x86_64
In case you don't know what fglrx drivers you have installed, you can list them with:

Code: Select all

yum list installed | grep fglrx

6. Installed the downloaded 15.12-6 drivers with:

Code: Select all

rpm -Uvh kmod-fglrx-15.12-6.el7.elrepo.x86_64.rpm  fglrx-x11-drv-15.12-6.el7.elrepo.x86_64.rpm

7. If you did step #2, copy the config file back:

Code: Select all

cp /etc/X11/xorg.conf-backup /etc/X11/xorg.conf
-- else, run:

Code: Select all

aticonfig --initial

8. Rebooted with:

Code: Select all

reboot

That is what worked for me. Your mileage may differ.

cwatson
Posts: 127
Joined: 2007/10/02 18:55:02

Re: AMD fglrx not working under centos 7.3

Postby cwatson » 2017/03/04 17:34:26

Dear nugai, thank you for your detailed instructions. Before I install them, do you know if I have to install the RPM's while I am logged into the latest kernel? I suspect that is not required (as I am currently logged in running 3.10.0-327.36.3, the last one to work) but want to clarify before I install them.

Thanks!