$Id$
Symptom: Had set the RTC clock in far future, such as
2038/02/01 (Y2038 overflowed date).
When rebooted, the system time is 2022-04-07.
timedatectl shows that RTC is in correct future date,
but system time is not.
Cause: Not kernel, but systemd, resets the clock to "sane" value when it detects the system clock is far from the systemd build date.
Apr 07 00:00:00 systemd[1]: System time is further ahead than 15y after build time, resetting clock to build time.
#define CLOCK_VALID_RANGE_USEC_MAX 473364000000000,
15 years.
src/shared/clock-util.c:clock_apply_epoch() does
/usr/lib/clock-epoch file exists, use the
mtime of the file for the epoch.
Usually you do not have this file.
TIME_EPOCH, which is a value dynamically generated when
systemd was built.
meson.build tries
$SOURCE_DATE_EPOCH environment on build.
RHEL 9 rpmbuild picks this up from %changelog of
RPM *.spec file.
For systemd, currently it sets this to
1649289600 (2022-04-07 00:00:00 UTC)
Workaround: None.
Placing a /usr/lib/clock-epoch file, and
timestamping it near the desired system date, makes
the system boot time to that date (doesn't honour RTC),
when experimenting with future system time.
Workaround: Place a new systemd unit file
/etc/systemd/system/hwclock-hctosys.service as following:
[Unit] Description Set date from RTC After=basic.target Wants=basic.target [Service] Type=oneshot ExecStart=/usr/sbin/hwclock --hctosys RemainAfterExit=yes [Install] WantedBy=basic.targetThen, invoke
systemctl enable hwclock-hctosys
to enable it.
Synptom: rpm and/or dnf command refuses to open /var/lib/rpm/rpmdb.sqlite file, which is clear in sight, no weird ACLs, no SELinux problems.
[kabe@i686 ix86.repo]$ sudo dnf --installroot=/mnt --disablerepo=\* --enablerepo=ix86\* reinstall kbd error: Unable to open sqlite database /mnt/var/lib/rpm/rpmdb.sqlite: unable to open database file error: cannot open Packages index using sqlite - Operation not permitted (1) error: cannot open Packages database in /mnt/var/lib/rpm Error: Error: rpmdb open failed [kabe@i686 ix86.repo]$ ls -l /mnt/var/lib/rpm/ total 16056 -rw-r--r--. 1 root root 16408576 Sep 22 20:45 rpmdb.sqlite -rw-r--r--. 1 root root 32768 Sep 23 06:08 rpmdb.sqlite-shm -rw-r--r--. 1 root root 0 Sep 22 20:45 rpmdb.sqlite-wal
Cause: plain stat(2) on the file was failing, due to rpmdb.sqlite file having Y2038-overflow timestamp.
[kabe@i686 ix86.repo]$ stat /mnt/var/lib/rpm/rpmdb.sqlite File: /mnt/var/lib/rpm/rpmdb.sqlite Size: 16408576 Blocks: 32048 IO Block: 4096 regular file Device: 812h/2066d Inode: 12583042 Links: 1 Access: (0644/-rw-r--r--) Uid: ( 0/ root) Gid: ( 0/ root) Access: 2038-02-02 13:28:25.030000000 +0900 Modify: 2022-09-22 20:45:55.593398092 +0900 Change: 2022-09-22 20:45:55.593398092 +0900 Birth: 2022-09-22 17:00:03.843346104 +0900
Rocky Linux supplied sqlite-libs-3.34.1-5.el9.i686 is
not compiled with -D_TIME_BITS=64, so
it ceases to work when any of the file it touches has
Y2038-overflow time attributes.
Workaround: Use sqlite-libs package supplied by this site.
Symptom: anaconda installer dies at text mode stage:
File "/usr/lib/python3.9/site-packages/pyanaconda/core/configuration/storage_constraints.py", line 22, in <module>
from pyanaconda.core.storage import DEVICE_TYPE_LVM, DEVICE_TYPE_MD, DEVICE_TYPE_PARTITION, \
File "/usr/lib/python3.9/site-packages/pyanaconda/core/storage.py", line 22, in <module>
from blivet import udev
File "/usr/lib/python3.9/site-packages/blivet/udev.py", line 39, in <module>
global_udev = pyudev.Context()
File "/usr/lib/python3.9/site-packages/pyudev/core.py", line 61, in __init__
self._libudev = load_ctypes_library('udev', SIGNATURES, ERROR_CHECKERS)
File "/usr/lib/python3.9/site-packages/pyudev/_ctypeslib/utils.py", line 55, in load_ctypes_library
raise ImportError('No library named %s' % name)
ImportError: No library named udev
Exception ignored in: <function Context.__del__ at 0xb4edb418>
Traceback (most recent call last):
File "/usr/lib/python3.9/site-packages/pyudev/core.py", line 65, in __del__
self._libudev.udev_unref(self)
AttributeError: 'Context' object has no attribute '_libudev'
Pane is dead (status 1, Sun Oct 16 12:35:14 2022)
Cause: For unknown reason, python3.i586 tries to load
/usr/lib/libudev.so, not
/usr/lib/libudev.so.1, and fails.
Workaround: Add /usr/lib/libudev.so symlink into
the installer medium. Adding lorax --installpkgs=systemd-devel
will do it.
Symptom: When installing the system with time set beyond Y2038, anaconda installer halts during package install:
The following error occured while installing the payload. This is fatal error and installation will be aborted.DNF error: Error in POSTTRANS scriptlet in rpm package kernel-core
/tmp/packaging.log has log:
20:09:41,788 INF packaging: Configuring (running scriptlet for): kernel-core-5.14.0-70.26.1.el9.v1.i686 1664514571 71c15d5731de2e5c4a824e989abb8981e08e8b8148dc9c08a919218fd7d3a50c 20:12:23,810 INF dnf.rpm: grub2-probe: error: cannot find a GRUB drive for /dev/sda2. Check your device.map. sed: couldn't edit /var/tmp/dracut.eQcVl8/initramfs/usr/lib/systemd/system/dbus.socket: not a regular file sed: couldn't edit /var/tmp/dracut.eQcVl8/initramfs/usr/lib/systemd/system/systemd-tmpfiles-setup.service: not a regular file sed: couldn't edit /var/tmp/dracut.1VlkEL/initramfs/usr/lib/systemd/system/dbus.socket: not a regular file sed: couldn't edit /var/tmp/dracut.1VlkEL/initramfs/usr/lib/systemd/system/systemd-tmpfiles-setup.service: not a regular file sed: couldn't edit /boot/loader/entries/ac6ed28bcbe9456b977b68236ce04be4-0-rescue.conf: not a regular file grep: /boot/grub2/grubenv: No such file or directory grep: /boot/grub2/grubenv: No such file or directory grep: /boot/grub2/grubenv: No such file or directory grep: /boot/grub2/grubenv: No such file or directory grep: /boot/grub2/grubenv: No such file or directory grep: /boot/grub2/grubenv: No such file or directory warning: %posttrans(kernel-core-5.14.0-70.26.1.el9.v1.i686) scriptlet failed, exit status 4 20:12:23,864 ERR dnf.rpm: Error in POSTTRANS scriptlet in rpm package kernel-core
Cause 1: grub-probe command is not compiled with
-D_TIME_BITS=64. Without it, it fails to probe
/dev/sda2 and fails with message
grub2-probe: error: cannot find a GRUB drive for /dev/sda2. Check your device.map.
Cause 2: sed is not compiled with
-D_TIME_BITS=64.
Without it, it fails to sed -i a file and returns error.
exit status 4 is a sum of errors during the
posttrans scriptlet. It may vary across installations.
Resolution: Use the grub and sed
package provided from this site.
Symptom: postgresql.src compile fails with:
+ /usr/bin/make -O -j4 V=1 VERBOSE=1 world
...
In file included from ../../src/include/postgres.h:47,
from hashfn.c:24:
../../src/include/utils/elog.h:71:10: fatal error: utils/errcodes.h: No such file or directory
71 | #include "utils/errcodes.h"
| ^~~~~~~~~~~~~~~~~~
compilation terminated.
or:
PATH="/home/kabe/r9builds/postgresql-r9/BUILD/postgresql-13.7/tmp_install/usr/bin:$PATH" LD_LIBRARY_PATH="/home/kabe/r9builds/postgresql-r9/BUILD/postgresql-13.7/tmp_install/usr/lib" ../../src/test/regress/pg_regress --temp-instance=./tmp_check --inputdir=. --bindir= --dbname=contrib_regression --load-extension=ltree python3/ltree_plpython ============== creating temporary instance ============== ============== initializing database system ============== pg_regress: initdb failed Examine /home/kabe/r9builds/postgresql-r9/BUILD/postgresql-13.7/contrib/ltree_plpython/log/initdb.log for the reason. Command was: "initdb" -D "/home/kabe/r9builds/postgresql-r9/BUILD/postgresql-13.7/contrib/ltree_plpython/./tmp_check/data" --no-clean --no-sync > "/home/kabe/r9builds/postgresql-r9/BUILD/postgresql-13.7/contrib/ltree_plpython/log/initdb.log" 2>&1 initdb.log shows: sh: line 1: initdb: command not found
Cause: postgresql Makefile is touchy on MAKELEVEL environment variable.
If ./rpmbuild is invoked via make,
MAKELEVEL is set to MAKELEVEL=1 for make
invoked inside the *.spec, and confuses postgresql Makefiles.
Resolution: Set to MAKELEVEL=0 in *.spec, if the variable is set.
Symptom: Booting into text console. The KMS console tries to use full-screen resolution, but I want it to be 800x600.
Resolution: add video=800x600 to kernel command line.
RHEL 9 derivatives will have Hyper-V console as a KMS driver, sovideo=kernel option will work. (Info: for RHEL 8,video=hyperv_fb:800x600was needed)
Edit /boot/loader/entries/*.conf to be
... options root=UUID=244150bf-638d-423a-8ba6-a7e424fcad37 ro crashkernel=1G-4G:192M,4G-64G:256M,64G-:512M resume=UUID=c4d7fb5b-125b-4b2e-8bdb-eeebb7c5d8de video=800x600 ...
Symptom: When I start Xorg by xinit, the
Xorg modesettings driver ignores the current KMS resolution,
and uses ex. 1024x768.
Workaround: Drop-in the following file as
/etc/X11/xorg.conf.d/40-monitor800.conf:
Section "Monitor"
Identifier "Monitor800x600"
Modeline "m800x600" 40.00 800 840 968 1056 600 601 605 628 +hsync +vsync
Option "PreferredMode" "m800x600"
EndSection
Section "Screen"
Identifier "Screen800x600"
Monitor "Monitor800x600"
EndSection
Option "PreferredMode" "m800x600" is the important part.
Symptom: I am satisfied with Xorg resolution by xinit,
but when with KDE 5 installed, startx resets the
screen resolution to 1024x768.
I do not want KDE to change resolution.
Workaround: Stop the "KScreen 2" background service. Invoke "System Setting", navigate to "Startup and Shutdown" -> "Background Services", disable and stop "KScreen 2" service.
make check fails:
FAIL: tests/check-all ===================== Running suite(s): main Abstract addresses Netlink attributes Clone ematch tree netns 37%: Checks: 29, Failures: 18, Errors: 0 tests/nl-test-util.c:127:S:Core:cache_and_clone:0: assert(r == 0) failed (errno=1, Operation not permitted) tests/nl-test-util.c:127:S:Core:test_create_iface:0: assert(r == 0) failed (errno=1, Operation not permitted) tests/nl-test-util.c:127:S:Core:test_create_iface:1: assert(r == 0) failed (errno=1, Operation not permitted) ...
Cause: unshare(CLONE_NEWUSER) seems to be not
available under setarch i686 chroot /chroot/i686 environment.
unshare -U ls also fails on commandline.
Symptom: Compiling ruby. %check fails with:
1) Failure: TestGem#test_default_path [/home/kabe/r9builds/ruby-r9/BUILD/ruby-3.0.4/test/rubygems/test_gem.rb:616]: <["/usr/share/gems"]> expected but was <["/usr/share/gems", "/usr/local/share/gems"]>. ...
Workaround: You must temporary disable ruby on the compilation environment.
sudo sh -c ' \
if [ -d /usr/share/rubygems ]; then mv /usr/share/rubygems{,-}; fi; \
if [ -d /usr/share/ruby ]; then mv /usr/share/ruby{,-}; fi; \
if [ -d /usr/lib/ruby ]; then mv /usr/lib/ruby{,-}; fi; \
if [ -d /usr/share/gems ]; then mv /usr/share/gems{,-}; fi; \
if [ -e /usr/bin/ruby ]; then mv /usr/bin/ruby{,-}; fi; \
'
Do the opposite when you finished compiling.
Symptom: Compile of librttopo (EPEL) fails:
Processing files: mingw32-librttopo-debuginfo-1.1.0-9.el9.noarch error: Could not open %files file /home/kabe/r9builds/librttopo-epel-r9/BUILD/librttopo/mingw32-debugfiles.list: No such file or directory
Workaround: Invoke rpmbuild with following options:
./rpmbuild --target=i586 -v -bb \ --without mingw \ -D 'mingw_build_win32 0' \ -D 'mingw_build_win64 0' \ -D 'mingw_build_ucrt64 0' \ ...
Symptom: clang compile fails in %check:
+ : 'RUN: at line 1'
+ %clang_dxc -Tlib_6_7 -fcgl -Fo - /home/kabe/r9builds/clang-r9/BUILD/clang-16.0.6.src/test/Driver/hlsl_no_stdinc.hlsl -###
+ /usr/bin/FileCheck /home/kabe/r9builds/clang-r9/BUILD/clang-16.0.6.src/test/Driver/hlsl_no_stdinc.hlsl --check-prefix=STDINC
/home/kabe/r9builds/clang-r9/BUILD/clang-16.0.6.src/test/Driver/hlsl_no_stdinc.hlsl:9:11: error: STDINC: expected string not found in input
// STDINC:"-finclude-default-header"
^
<stdin>:1:1: note: scanning from here
/home/kabe/r9builds/clang-r9/BUILD/clang-16.0.6.src/redhat-linux-build/test/Driver/Output/hlsl_no_stdinc.hlsl.script: line 1: %clang_dxc: command not found
^
<stdin>:1:58: note: possible intended match here
/home/kabe/r9builds/clang-r9/BUILD/clang-16.0.6.src/redhat-linux-build/test/Driver/Output/hlsl_no_stdinc.hlsl.script: line 1: %clang_dxc: command not found
Cause: python3-lit-15.0.7-1.el9.noarch.rpm package is
too old.
Newer python3-lit-16.0.6-2.el9.noarch.rpm has
definitions for ToolSubst('%clang_dxc', command=self.config.clang,
.
Symptom: lorax fails with:
2023-02-23 18:30:02,225: libavahi-common.so.3, needed by /usr/bin/ostree, not found libavahi-common.so.3, needed by /usr/bin/ostree, not found 2023-02-23 18:30:02,226: libavahi-client.so.3, needed by /usr/bin/ostree, not found libavahi-client.so.3, needed by /usr/bin/ostree, not found
Cause: lorax erases avahi-glib package during run
to save space:
pylorax.log:template line 257: removepkg opus libtheora libvisual flac-libs gsm avahi-glib avahi-libs ModemManager-glib ... pylorax.log:removepkg avahi-glib: 58kb pylorax.log:removed /var/tmp/lorax/lorax.o1jrqzmp/installtree/usr/lib/libavahi-glib.so.1.0.2 pylorax.log:removed /var/tmp/lorax/lorax.o1jrqzmp/installtree/usr/lib/libavahi-glib.so.1 pylorax.log:removepkg avahi-libs: 489kb
Resolution:
Compile ostree package without avahi dependency:
with_avahi=no CPPFLAGS="-D_TIME_BITS=64 -D_FILE_OFFSET_BITS=64" ./rpmbuild
Symptom: Spinned DVD1.iso exceeds 4.7GB limit, like 5.2GB.
Workaround: Exclude texlive, mingw and gcc-toolset-12 from the media. You are not likely to need it immediately.
xorrisofs -o ./DVD1.iso \ -b isolinux/isolinux.bin -c isolinux/boot.cat \ -no-emul-boot -boot-load-size 4 -boot-info-table \ -R -T -J \ -v \ -V "Rocky-Linux-9-1-i386" \ -m upgrade.img -m boot.iso \ -m 'texlive-*' -m 'mingw*' -m 'gcc-toolset-12-*' \ -no-pad -iso-level 3 -D --hardlinks -joliet-long \ ./img9
Symptom: Under heavy load, login on vt console is rejected by
Login timed out after 60 seconds
Resolution: Add login(1) LOGIN_TIMEOUT in /etc/login.defs:
## default 60secs is short under heavy load LOGIN_TIMEOUT 300
You may have to boot into single user mode to make this change, since multi-user mode keeps CPU load > 5 for several minutes.
Symptom: Using the "Breeze dark" scheme, which is default in RHEL 9.1 era EPEL. When I add multiple input methods, the tray icon disappears. It is still there; hovering the mouse cursor will show the icon in thin color.
Workaround: Use "Breeze" scheme, from Settings - Appearance - Plasma Style. The tray icon coloring seems to be scheme dependent.
Symptom: Starting KDE by startkde (startx)
does not work.
journalctl shows plasma related services repeatedly timeout:
Mar 19 17:32:21 cygnus.five.ten systemd[598]: plasma-polkit-agent.service: Failed with result 'timeout'. ?? Subject: Unit failed ?? Defined-By: systemd ?? Support: https://access.redhat.com/support ?? ?? The unit UNIT has entered the 'failed' state with result 'timeout'. Mar 19 17:32:21 cygnus.five.ten systemd[598]: Failed to start KDE PolicyKit Authentication Agent. Mar 19 17:32:16 cygnus.five.ten systemd[598]: plasma-powerdevil.service: start operation timed out. Terminating. Mar 19 17:32:16 cygnus.five.ten systemd[598]: plasma-powerdevil.service: Failed with result 'timeout'. ?? Subject: Unit failed ?? Defined-By: systemd ?? Support: https://access.redhat.com/support ?? ?? The unit UNIT has entered the 'failed' state with result 'timeout'. Mar 19 17:32:16 cygnus.five.ten systemd[598]: Failed to start Powerdevil. ?? Subject: A start job for unit UNIT has failed ?? Defined-By: systemd ?? Support: https://access.redhat.com/support ?? ?? A start job for unit UNIT has finished with a failure. ?? ?? The job identifier is 13121 and the job result is failed. Mar 19 17:32:17 cygnus.five.ten systemd[598]: plasma-powerdevil.service: Scheduled restart job, restart counter is at 372.
Workaround: Enlenghen timeout for systemctl --user services. Use a systemd drop-in facility. Create the following directories:
mkdir /etc/systemd/user/gmenudbusmenuproxy.service.d mkdir /etc/systemd/user/plasma-baloorunner.service.d mkdir /etc/systemd/user/plasma-kactivitymanagerd.service.d mkdir /etc/systemd/user/plasma-ksmserver.service.d mkdir /etc/systemd/user/plasma-kwin_x11.service.d mkdir /etc/systemd/user/plasma-plasmashell.service.d mkdir /etc/systemd/user/plasma-polkit-agent.service.d mkdir /etc/systemd/user/plasma-powerdevil.service.d mkdir /etc/systemd/user/plasma-xembedsniproxy.service.dand drop-in the following as
timeout.conf :
[Service] TimeoutSec=600sec
Scripted as:
for i in \ gmenudbusmenuproxy.service \ plasma-baloorunner.service \ plasma-kactivitymanagerd.service \ plasma-ksmserver.service \ plasma-kwin_x11.service \ plasma-plasmashell.service \ plasma-polkit-agent.service \ plasma-powerdevil.service \ plasma-xembedsniproxy.service \ ; do mkdir /etc/systemd/user/$i.d; cat << 'EOF' > /etc/systemd/user/$i.d/timeout.conf; [Service] TimeoutSec=600sec EOF done
Do systemctl --user daemon-reload, and
check that setting is valid by
systemctl --user show plasma-plasmashell and search for
TimeoutStartUSec=10min .
By default, plasma-powerdevil.service and plasma-polkit-agent.service has TimeoutSec=5sec, which is way too short for slow machines.
Symptom: Wi-Fi adapter was recognised and could connect during anaconda installation, but not after HDD reboot.
Cause: Probably NetworkManager-wifi package
is missing. It is not included in "minimum" install.
Resolution: Install NetworkManager-wifi package,
then systemctl restart NetworkManager .
Resolution: Recompile the named binary in "setarch i686" environment with glibc.i586 .
The restriction is checked by glibc.
$ objdump -j .note.gnu.property -s /usr/bin/dconf
/usr/bin/dconf: file format elf32-i386
Contents of section .note.gnu.property:
01ec 04000000 18000000 05000000 474e5500 ............GNU.
01fc 020000c0 04000000 03000000 028000c0 ................
020c 04000000 01000000 |
| | |
| flag(0x01) [swab]
4 bytes size follows |
|
GNU_PROPERTY_X86_ISA_1_NEEDED 0xc0008002
Symptom: Selecting "Properties" in the keyboard icon in the task tray does not invoke anything. From command line, it dies with error:
$ python3 /usr/share/ibus/setup/main.py ibus-setup (ibus-setup:1720): Gtk-WARNING **: 20:53:09.622: Could not load a pixbuf from /org/gtk/libgtk/theme/Adwaita/assets/check-symbolic.svg. This may indicate that pixbuf loaders or the mime database could not be found. ** Gtk:ERROR:../gtk/gtkiconhelper.c:494:ensure_surface_for_gicon: assertion failed (error == NULL): Failed to load /usr/share/icons/Adwaita/16x16/status/image-missing.png: Unrecognized image file format (gdk-pixbuf-error-quark, 3) Bail out! Gtk:ERROR:../gtk/gtkiconhelper.c:494:ensure_surface_for_gicon: assertion failed (error == NULL): Failed to load /usr/share/icons/Adwaita/16x16/status/image-missing.png: Unrecognized image file format (gdk-pixbuf-error-quark, 3)
Points to check:
/usr/lib/gdk-pixbuf-2.0/2.10.0/loaders/libpixbufloader-svg.so exists.
/usr/lib/gdk-pixbuf-2.0/2.10.0/loaders.cache mentions
libpixbufloader-svg.so .
If missing, reinstall gdk-pixbuf2 package, or
regenerate with gdk-pixbuf-query-loaders-32 --update-cache .
/usr/share/mime/ is populated with *.xml files.
Invoke rpm -q -V -v shared-mime-info to check.
update-mime-database -V manually.
(it is invoked via rpm --filetriggers shared-mime-info)
XDG_DATA_DIRS=/usr/share/kde-settings/kde-profile/default/share:/usr/local/share:/usr/share
Symptom: Using Wi-Fi adapter using ath5k kernel driver. SSID scan works, but cannot connect to access points with WPA disabled (only WPA2-PSK).
Cause: Some ath5k hardware seems to not capable of WPA2 hardware offloading.
Workaround: use nohwcrypt=1 ath5k option.
Add ath5k.nohwcrypt=1 to kernel boot options, or place
file below as /etc/modprobe.d/ath5k.conf (see modprobe.d(5))
options ath5k nohwcrypt=1
Symptom:Trying to switch ibus input method. Click-selecting the keyboard icon, or using [Super]-Space, does show the alternate input method (after terrible delay), but keyboard icon doesn't change and input method doesn't change.
Cause: Default timeout of 5 seconds may be too short for slow machines
Resolution: set a longer timeout on ibus-daemon.
% pkill ibus-daemon % IBUS_TIMEOUT=120000 /usr/bin/ibus-daemon -d -r --xim --timeout=120000
There are two configurable timeouts for ibus-daemon; ibus timeout (IBUS_TIMEOUT), which defaults to 6 seconds, and gdbus timeout (--timeout) which defaults to 5 seconds. These may be too short for slow machines. Above example sets them to 120 seconds. (--timeout seems to be enough)
Edit /etc/X11/xinit/xinput.d/ibus.conf as:
XIM=ibus XIM_PROGRAM="/usr/bin/ibus-daemon" ICON="ibus" XIM_ARGS="-r --xim --timeout=120000" PREFERENCE_PROGRAM=/usr/bin/ibus-setup ...
Symptom: It is very annoying when bash goes to find into noexistent repository for mistyped command:
$ aa bash: aa: command not found... Long pause Failed to search for file: cannot update repo 'extras': Cannot prepare internal mirrorlist: Status code: 404 for https://mirrors.rockylinux.org/mirrorlist?arch=i386&repo=extras-9 (IP: 151.101.42.132); Last error: Status code: 404 for https://mirrors.rockylinux.org/mirrorlist?arch=i386&repo=extras-9 (IP: 146.75.94.132)
Resolution: Uninstall PackageKit-command-not-found package. The find hook is defined in /etc/profile.d/PackageKit.sh .
rpm -ev PackageKit-command-not-found
xbench result of nouveau.ko modeset and xorg-x11-drv-nv does not much differ. xorg-x11-drv-nv has about 20% overall performance gain on X11 primitive operations, but since nouveau.ko kernel module is very stable comapred to Rocky Linux 8 counterpart, just sticking to nouveau.ko is not unwise.
Symptom: Blacklisted radeon.ko KMS driver by kernel option
module_blacklist=radeon, nomodeset or
/etc/modprobe.d/radeon.conf with blacklist radeon,
since KMS radeon.ko driver has horrible performance when scrolling
the console.
Installed xorg-x11-drv-ati driver.
When I try starting Xorg, it fails:
[ 5552.173] (II) [KMS] drm report modesetting isn't supported. [ 5552.175] (EE) Screen 0 deleted because of no matching config section.
Cause: xorg-x11-drv-ati needs KMS driver to work. Do not blacklist radeon.ko kernel driver.
Symptom: Blacklisted mgag200.ko kernel module, since KMS (DRM) console has horrible scrollong performance. Xorg does not start, with /var/log/Xorg.0.log saying:
[ 560.593] (II) LoadModule: "mga" [ 560.621] (II) Loading /usr/lib/xorg/modules/drivers/mga_drv.so [ 560.640] (II) Module mga: vendor="X.Org Foundation" [ 560.640] compiled for 1.20.11, module version = 2.0.1 [ 560.641] Module class: X.Org Video Driver [ 560.642] ABI class: X.Org Video Driver, version 24.1 [ 560.656] (II) LoadModule: "modesetting" [ 560.677] (II) Loading /usr/lib/xorg/modules/drivers/modesetting_drv.so [ 560.809] (II) Module modesetting: vendor="X.Org Foundation" [ 560.810] compiled for 1.20.11, module version = 1.20.11 [ 560.811] Module class: X.Org Video Driver [ 560.811] ABI class: X.Org Video Driver, version 24.1 [ 560.812] (II) LoadModule: "fbdev" [ 560.829] (II) Loading /usr/lib/xorg/modules/drivers/fbdev_drv.so [ 560.850] (II) Module fbdev: vendor="X.Org Foundation" [ 560.851] compiled for 1.20.11, module version = 0.5.0 [ 560.852] Module class: X.Org Video Driver [ 560.852] ABI class: X.Org Video Driver, version 24.1 [ 560.853] (II) LoadModule: "vesa" [ 560.887] (WW) Warning, couldn't open module vesa [ 560.888] (EE) Failed to load module "vesa" (module does not exist, 0) [ 560.889] (II) MGA: driver for Matrox chipsets: mga2064w, mga1064sg, mga2164w, mga2164w AGP, mgag100, mgag100 PCI, mgag200, mgag200 PCI, mgag200 SE A PCI, mgag200 SE B PCI, mgag200 EV Maxim, mgag200 ER SH7757, mgag200 eW Nuvoton, mgag200 eW3 Nuvoton, mgag200eH, mgag200eH3, mgag400, mgag550 [ 560.914] (II) modesetting: Driver for Modesetting Kernel Drivers: kms [ 560.915] (II) FBDEV: driver for framebuffer: fbdev [ 560.963] (WW) Falling back to old probe method for modesetting [ 560.965] (EE) open /dev/dri/card0: No such file or directory [ 560.965] (WW) Falling back to old probe method for fbdev [ 560.986] (II) Loading sub module "fbdevhw" [ 560.987] (II) LoadModule: "fbdevhw" [ 560.993] (II) Loading /usr/lib/xorg/modules/libfbdevhw.so [ 561.023] (II) Module fbdevhw: vendor="X.Org Foundation" [ 561.023] compiled for 1.20.11, module version = 0.0.2 [ 561.024] ABI class: X.Org Video Driver, version 24.1 [ 561.045] (EE) open /dev/fb0: No such file or directory [ 561.099] (EE) No devices detected. [ 561.100] (EE) Fatal server error: [ 561.102] (EE) no screens found(EE) [ 561.104] (EE)
Cause: RHEL 9 does not support video cards without DRM drivers, as mentioned in patch of xorg-x11-server package, 0001-mustard-xfree86-Disable-the-PCI-probe-path.patch . Un-applying this patch makes Xorg to not work on Hyper-V.
Workaround: Do not disable mgag200.ko kernel module.
Symptom: xinit -- /usr/bin/Xorg -retro causes weird color for root window stipple, when using xorg-x11-drv-ati Xorg driver. Does not occur without xorg-x11-drv-ati and using modeset driver.
Resolution: Add Option "AccelMethod" "glamor"
(The newer AccelMethod) to the Xorg setting.
Add following as /etc/X11/xorg.conf.d/30-radeon.conf:
Section "Device" Identifier "radeon-dev" Driver "radeon" Option "AccelMethod" "glamor" EndSection
Symptom: Using Matrox MGA G200 video card. On boot, mgag200.ko KMS module is loaded, and video is limited to 1024x768, although the monitor is 1360x768.
Workaround:
/usr/X11/xorg.conf.d/40-mgamonitor.conf as below:
Section "Monitor" Identifier "Monitor0" #Option "ReducedBlanking" VertRefresh 55-65 #Option "MaxClock" "67000" ModeLine "m1360x768" 85.5 1360 1424 1536 1792 768 771 777 795 +hsync +vsync #Option "PreferredMode" "m1360x768" #Option "DefaultMode" "m1360x768" EndSection Section "Screen" Identifier "Screen0" Monitor "Monitor0" SubSection "Display" Virtual 1360 768 Viewport 1360 768 #Modes "m1360x768" EndSubSection EndSection"Virtual 1360 768" "Viewport 1360 768" is the important part. xorg-x11-drv-mga seems to use 1368x768 for unknown reason.
You still need this 40-mgamonitor.conf with
#Modes "m1360x768" uncommented, when using
modesetting driver (without xorg-x11-drv-mga) and setting
console to 1360x768 by kernel option video=1360x768 .
Symptom: Scrolling in DRM KMS fbdev console is slow.
Workaround: There is Scroll speed impovement patch, which is merged into kernel-6.x series (commit 6f29e049,0d030118,61bfcb6a,7438f52b ) but not backported for kernel-5.x line, which is the base for RHEL 9.
The kernel provided by this site has this acceleration patch backported, and additional acceleration by assembly. It makes scrolling faster by x2 for modern hardware, +30% gain in i586.
+ SOURCE_DATA_EPOCH=1629181597
+ LD_LIBRARY_PATH=/home/kabe/r9builds/clang-r9/BUILDROOT/clang-15.0.7-2.el9_1.v1.i386//usr/lib
+ /usr/bin/ninja check-all -C redhat-linux-build
ninja: Entering directory `redhat-linux-build'
[0/1] Running all regression tests
lit: /usr/lib/python3.9/site-packages/lit/TestingConfig.py:102: fatal: unable to parse config file '/home/kabe/r9builds/clang-r9/BUILD/clang-15.0.7.src/redhat-linux-build/tools/extra/include-cleaner/test/lit.site.cfg.py', traceback: Traceback (most recent call last):
File "/usr/lib/python3.9/site-packages/lit/TestingConfig.py", line 91, in load_from_path
exec(compile(data, path, 'exec'), cfg_globals, None)
File "/home/kabe/r9builds/clang-r9/BUILD/clang-15.0.7.src/redhat-linux-build/tools/extra/include-cleaner/test/lit.site.cfg.py", line 14, in
config.llvm_tools_dir = path(lit_config.substitute("/usr/bin"))
AttributeError: 'LitConfig' object has no attribute 'substitute'
FAILED: CMakeFiles/check-all
Resolution:python3-lit.noarch package is old.
Replace with Rocky Linux 9.2 version,
python3-lit-15.0.7-1.el9.noarch.rpm .
This version has substitute method defined.
Symptom: When compiling the kernel, it fails at very last stage with following errors:
error: Illegal sequence ".." in: 0x........ error: Dependency tokens must begin with alpha-numeric, '_' or '/': . error: Illegal char '<' (0x3c) in: 0x......<.
Cause: /usr/lib/rpm/redhat/find-provides.ksyms assumes
output of readelf -R .rodata module.ko
is always 16-byte aligned.
When .rodata is not aliged at 16 bytes, e.g
$ readelf -R .rodata /tmp/module.ko Hex dump of section '.rodata': 0x00000000 00000000 12001000 13000000 13080800 ................ 0x00000010 00000000 13000800 29000000 14080800 ........)....... 0x00000020 3c000000 14000800 <......./usr/lib/rpm/redhat/find-provides.ksyms assembles an hex string like:
$ readelf -R .rodata /tmp/module.ko | awk '/0x/{printf $2$3$4$5}'
00000000120010001300000013080800000000001300080029000000140808003c00000014000800<.......
which has ASCII dump part appended as a garbage.
Resolution: Patch /usr/lib/rpm/redhat/find-provides.ksyms .
--- /usr/lib/rpm/redhat/find-provides.ksyms.dist 2022-05-16 21:07:11.000000000 +0900
+++ /usr/lib/rpm/redhat/find-provides.ksyms 2023-06-29 18:20:52.545271482 +0900
@@ -39,7 +39,7 @@ for module in $(grep -E '/lib/modules/.+
| awk --non-decimal-data '{printf("'"${dep_pfx}"'(%s) = 0x%08x\n", $2, $1)}' \
| LC_ALL=C sort -u
else
- ELFRODATA=$(readelf -R .rodata $module | awk '/0x/{printf $2$3$4$5}')
+ ELFRODATA=$(readelf -R .rodata $module | cut -b 1-48 | awk '/0x/{printf $2$3$4$5}')
if [[ -n $(readelf -h $module | grep "little endian") ]]; then
RODATA=$(echo $ELFRODATA | sed 's/\(..\)\(..\)\(..\)\(..\)/\4\3\2\1/g')
else
Symptom: libssh2 (from EPEL) fails in %check stage.
FAIL: ssh2.sh ============================================================================ Testsuite summary for libssh2 - ============================================================================ # TOTAL: 3 # PASS: 2 # SKIP: 0 # XFAIL: 0 # FAIL: 1 # XPASS: 0 # ERROR: 0 ============================================================================ See tests/test-suite.log Please report to libssh2-devel@cool.haxx.se ============================================================================
tests/test-suite.log:
command-line line 0: Deprecated option UsePrivilegeSeparation main: sshd: ssh-rsa algorithm is disabled command-line line 0: Deprecated option UsePrivilegeSeparation Failure establishing SSH session FAIL ssh2.sh (exit status: 1)
Cause: ssh-rsa cipher used in tests/etc/{host,user}
is not supported by DEFAULT policy of
update-crypto-policies --show.
Workaround: do one of following:
ssh-keygen -t ecdsa -f tests/etc/host ssh-keygen -t ecdsa -f tests/etc/user
sudo update-crypto-policies --set DEFAULT:SHA1 make sudo update-crypto-policies --set DEFAULT
Symptom: During qt5-qtbase package build, at the end, orphaned files are reported and build fails:
Installed (but unpackaged) file(s) found:
/usr/lib/cmake/Qt5Gui/Qt5Gui_QTsLibPlugin.cmake
/usr/lib/debug/usr/lib/qt5/plugins/generic/libqtslibplugin.so-5.15.3-1.el9.v1.i386.debug
/usr/lib/qt5/plugins/generic/libqtslibplugin.so
Cause: tslib functionality is automatically turned on
when tslib-devel is installed, but qt5-qtbase.spec
does not count it.
Resolution: in qt5-qtbase.spec, do
./configure ... -no-tslib .
Symptom: When compiling rust-* packages in EPEL, %install fails to find /usr/share/cargo/registry/*:
Processing files: rust-cargo-c-devel-0.9.12-4.el9.noarch error: File not found: /home/kabe/r9builds/rust-cargo-c-epel-r9/BUILDROOT/rust-cargo-c-0.9.12-4.el9.i386/usr/share/cargo/registry/cargo-c-0.9.12/LICENSE error: File not found: /home/kabe/r9builds/rust-cargo-c-epel-r9/BUILDROOT/rust-cargo-c-0.9.12-4.el9.i386/usr/share/cargo/registry/cargo-c-0.9.12/README.md error: Directory not found: /home/kabe/r9builds/rust-cargo-c-epel-r9/BUILDROOT/rust-cargo-c-0.9.12-4.el9.i386/usr/share/cargo/registry/cargo-c-0.9.12
Cause: %__cargo_install rpmbuild macro is redefined to bogus one
by rust-toolset package.
Workaround: Uninstall rust-toolset
(and python3-setuptools-rust dependency)
before compiling.
Symptom: lorax bombs out by:
Problem 1: package perl-interpreter-4:5.32.1-480.el9.v1.i586 requires libperl.so.5.32, but none of the providers can be installed - package perl-interpreter-4:5.32.1-480.el9.v1.i586 requires perl(:MODULE_COMPAT_5.32.1), but none of the providers can be installed - package perl-interpreter-4:5.32.1-480.el9.v1.i586 requires perl-libs, but none of the providers can be installed - package perl-interpreter-4:5.32.1-480.el9.v1.i586 requires perl-libs(x86-32) = 4:5.32.1-480.el9.v1, but none of the providers can be installed - conflicting requests - nothing provides perl(Exporter) needed by perl-libs-4:5.32.1-480.el9.v1.i586 - nothing provides perl(Carp) needed by perl-libs-4:5.32.1-480.el9.v1.i586 - nothing provides perl(Encode) needed by perl-libs-4:5.32.1-480.el9.v1.i586 - nothing provides perl(File::Spec) needed by perl-libs-4:5.32.1-480.el9.v1.i586
Cause: perl-Exporter package was not built during perl.src build.
Resolution: Add flag to rpmbuild:
./rpmbuild -D 'perl_bootstrap 1' ... SPECS/perl.spec
This will build perl-Exporter package.
Symptom: Conflict error reported during
dnf group install kde-desktop-environment:
Error: Problem: problem with installed package jack-audio-connection-kit-1.9.21-1.el9.v1.i586 - package pipewire-jack-audio-connection-kit-0.3.47-3.el9_2.v1.i586 conflicts with jack-audio-connection-kit provided by jack-audio-connection-kit-1.9.21-1.el9.v1.i586 - conflicting requests
Resolution: You should not have been compiling
jack-audio-connection-kit (EPEL) package in first place.
pipewire-jack-audio-connection-kit-libs package compiled from
pipewire.src package will provide libjack.so.0 dependency.
Check rpmbuild --with jack is in effect during
pipewire.src build.
Wrong Resolution: Rebuild pipewire.src package with
./rpmbuild --without jack.
This will make it not build the conflicting pipewire-jack-audio-connection-kit
package.
Symptom: aplay -L shows
pipewire device,
$ aplay -L
null
Discard all samples (playback) or generate zero samples (capture)
sysdefault
Default Audio Device
pipewire
PipeWire Sound Server
default
Default ALSA Output (currently PipeWire Media Server)
sysdefault:CARD=I82801DBICH4
Intel 82801DB-ICH4, Intel 82801DB-ICH4
Default Audio Device
...
but
aplay /usr/share/sounds/alsa/Front_Center.wavplayes no sound and does not return. (after checking volume is up by
alsamixer -D hw:0)
Cause: RHEL 9.2 supplied
pipewire-0.3.47-3.el9_2 and
wireplumber-0.4.8-1.el9
seems to have a
bug.
Resolution:
Compile and install pipewire and wireplumber package with
Fedora 39 supplied ones, that is,
pipewire-0.3.80-1.fc39 and
wireplumber-0.4.14-2.fc39 .
Symptom: pw-play says:
$ pw-play /usr/share/sounds/alsa/Front_Center.wav stream node 32 error: no node available remote error: id=2 seq:7 res:-2 (No such file or directory): no node available $ _
Workaround: The error is intermittent,
if you had already replaced pipewire and wireplumber
packages to Fedora 39 version.
The pipewire.service did not woke up fast enough
in response to pipewire.socket access.
$ pw-play /usr/share/sounds/alsa/Front_Center.wav stream node 32 error: no node available remote error: id=2 seq:7 res:-2 (No such file or directory): no node available $ pw-play /usr/share/sounds/alsa/Front_Center.wav the sound plays $ _
You could reset the state (first pw-play fails) by
systemctl --user stop pipewire.service .
Symptom: Using Creative Sound Blaster AWE32 sound card. This card is an ISA-PNP card, so manual module load is required. When the following sequence is executed, entire system freezes.
$ sudo modprobe snd-sbawe seq_ports=0 $ systemctl --user start pipewire wireplumber
Cause: Unknown.
Workaround: Invoke pipewire and wireplumber one by one:
$ systemctl --user mask wireplumber $ sudo modprobe snd-sbawe seq_ports=0 $ systemctl --user start pipewire $ systemctl --user unmask wireplumber $ systemctl --user start wireplumber(This is not a bulletproof solution; some kernels still freeze by above sequence)
Above sequence will keep CPU usage in "top" around "50.0 sy", so Invoke pw-play command to revive CPU usage back to sane value:
$ pw-play /usr/share/sounds/alsa/Front_Center.wavWait for couple of minutes. Sound plays, and CPU usage goes down to normal. (Reason that pw-play fixes CPU usage is unknown)
Symptom: Starting Xorg by adding args to Xorg:
$ xinit -- /usr/bin/Xorg -retro $ startx /usr/bin/xterm -- /usr/bin/Xorg -retroInside the X session,
aplay aborts:
$ aplay /usr/share/sounds/alsa/Front_Center.wav Playing WAVE '/usr/share/sounds/alsa/Front_Center.wav' : Signed 16 bit Little Endian, Rate 48000 Hz, Mono aplay: set_params:1416: Unable to install hw params: ACCESS: RW_INTERLEAVED FORMAT: S16_LE SUBFORMAT: STD SAMPLE_BITS: 16 FRAME_BITS: 16 CHANNELS: 1 RATE: 48000 PERIOD_TIME: 125000 PERIOD_SIZE: 6000 PERIOD_BYTES: 12000 PERIODS: 4 BUFFER_TIME: 500000 BUFFER_SIZE: 24000 BUFFER_BYTES: 48000 TICK_TIME: [0 0] $ _
Cause: When Xorg is invoked without vt1 parameter,
it tries to run on the next vt, thus vt2,
but systemd-logind does not recognise that vt2 is owned by the user
and does not add write permission to /dev/snd/controlC0.
Proper permission of the sound device file should be:
$ getfacl /dev/snd/controlC0 getfacl: Removing leading '/' from absolute path names # file: dev/snd/controlC0 # owner: root # group: audio user::rw- user:kabe:rw- group::rw- mask::rw- other::---When audio is not working, the emphasised part is not present. You can switch back to tty1 (Ctrl-Alt-F1) and test
aplay,
which works.
Workaround:
-- /usr/bin/Xorg to startx.
startx automatically adds vt1 argument only when
no Xserver argument is specified.
vt1 for Xorg argument, as
xinit -- /usr/bin/Xorg -retro :0 vt1 .
Symptom: Started KDE by startx on a slow machine.
Since it's slow, it invokes the default screenlocker during startup,
and the machine is slow enough to be
screenlocker reinvoked after disarming the screenlocker.
I can't access the desktop.
systemsettings can't be invoked to change the setting.
Workaround: Manually disable the spash screen and screenlocker by creating the following files:
; ~/.config/ksplashrc [KSplash] Engine=none Theme=None
; ~/.config/kscreenlockerrc [Daemon] Autolock=false
Symptom: Initial user, started KDE desktop.
The wallpaper is black.
journalctl -xe reveals:
Oct 22 15:51:18 rocky9i686.five.ten plasmashell[1757]: qml: The backend got an unknown wallpaper provider type. The wallpaper will now fall back to the default. Please check your wallpaper configuration!
Cause: The plugin for reading *.webp background/wallpaper is missing.
Resolution (Compile and) install qt5-qtimageformats
package. This will install
/usr/lib/qt5/plugins/imageformats/libqwebp.so, the plugin
for reading images written in *.webp format.
Symptom: Started KDE desktop, but plasmashell
seems to not working. Black background/wallpaper, no task bar.
journalctl is recording:
Oct 30 20:01:03 cygnus.five.ten kded5[3121]: QProcess: Destroyed while process ("/usr/libexec/kf5/kconf_update") is still running.
...
Oct 31 19:43:42 cygnus.five.ten kconf_update[4849]: QProcess: Destroyed while process ("/usr/lib/kconf_update_bin/spectacle-migrate-shortcuts") is still running
Cause: Hardcoded default timeout in /usr/include/qt5/QtCore/qprocess.h:
bool waitForFinished(int msecs = 30000);
(30 secs) is too short for slow machines.
Workaround: Patch that line in qt5-qtbase.src to be:
bool waitForFinished(int msecs = 300000); //5 minutes
and, install it on the compile machine and target machine, then
recompile kf5-kconfig and plasma-workspace,
and install them on target machine.
Resolution: Install plasma-pa (EPEL) package.
Be sure you have pipewire-pulseaudio
(not pulseaudio) package
to satisfy pulseaudio-daemon dependency.
Symptom: On insertion of Broadcom BCM43xx based (Cardbus)
WiFi adapter, dmesg or /var/log/messages says as below and
wlan0 interface does not appear under ip link:
[ 2835.272754] pcmcia_socket pcmcia_socket0: pccard: CardBus card inserted into slot 0 [ 2835.272803] pci 0000:03:00.0: [14e4:4320] type 00 class 0x028000 [ 2835.272828] pci 0000:03:00.0: reg 0x10: [mem 0x00000000-0x00001fff] [ 2835.273081] pci 0000:03:00.0: BAR 0: assigned [mem 0xd4000000-0xd4001fff] [ 2836.492280] b43-pci-bridge 0000:03:00.0: enabling device (0000 -> 0002) [ 2836.492418] ssb: Found chip with id 0x4306, rev 0x03 and package 0x00 [ 2836.497229] b43-pci-bridge 0000:03:00.0: Sonics Silicon Backplane found on PCI device 0000:03:00.0 [ 2836.742010] b43-phy1: Broadcom 4306 WLAN found (core revision 5) [ 2836.767037] b43-phy1: Found PHY: Analog 2, Type 2 (G), Revision 2 [ 2836.767064] b43-phy1: Found Radio: Manuf 0x17F, ID 0x2050, Revision 2, Version 0 [ 2836.797769] Broadcom 43xx driver loaded [ Features: PNLS ] [ 2836.902466] b43 ssb0:0: Direct firmware load for b43/ucode5.fw failed with error -2 [ 2836.902543] b43 ssb0:0: Direct firmware load for b43/ucode5.fw failed with error -2 [ 2836.902636] b43 ssb0:0: Direct firmware load for b43-open/ucode5.fw failed with error -2 [ 2836.903994] b43 ssb0:0: Direct firmware load for b43-open/ucode5.fw failed with error -2 [ 2836.904004] b43-phy1 ERROR: Firmware file "b43/ucode5.fw" not found [ 2836.904009] b43-phy1 ERROR: Firmware file "b43-open/ucode5.fw" not found [ 2836.904013] b43-phy1 ERROR: You must go to https://wireless.wiki.kernel.org/en/users/Drivers/b43#devicefirmware and download the correct firmware for this driver version. Please carefully read all instructions on this website.
Resolution: Install a firmware package. b43-openfwwf-5.2-10.el6.noarch.rpm supplied with CentOS 6 works. If it doesn't, you need to use b43-fwcutter by yourself to extract a firmware.
Note: b43.ko driver is not supplied in stock Rocky Linux 9 kernel.
To make the driver work in installation/rescue disc, you must add
b43-openfwwf-5.2-10.el6.noarch.rpm
in ix86.repo/BaseOS/Packages/, createrepo, and
pass --installpkgs=b43-openfwwf to lorax.