Created attachment 258137 [details] part of journal log OS is openSUSE tumbleweed with radeon 4350. After Installation of kernel 4.13.rc6-2.1.g43edc52 (obs://build.opensuse.org/Kernel) booting problem with radeon 4350. Graphic Card could not be installed, booting only with default(?) graphic mode. See attachment. Building my own kernel from mainline: 4.13-rc7 2017-08-28 [tarball] has the same error. RV710_pfp.bin was always present in: /lib/firmware/4.13.0-rc7..../radeon/ and /lib/firmware/radeon/ Why kernel fails to load it? kernel 4.13.0-rc2-1.gb545b87 (obs://build.opensuse.org/Kernel) was booting (Jul 30.) without this error. Actual kernel 4.12.8-1.2 is running without problems.
Did you rebuild initrd after updating kernel-firmware containing /lib/firmware/radeon/RV710_pfb.bin?
(In reply to Takashi Iwai from comment #1) > Did you rebuild initrd after updating kernel-firmware containing > /lib/firmware/radeon/RV710_pfb.bin? kernel-firmware was installed at Di 15 Aug 2017 16:16:40 CEST. In the meantime there were a lot of updates of dracut, udev etc. Today, installed 4.13.rc6-2.1.g43edc52-vanilla in order to get a new entry of dracut in zypper/log I will add a new attachment taken var/log/zypper/log with 3 entries of dracut in var/log/zypper/log. Maybe this could answer your question properly.
Created attachment 258139 [details] dracut logs during kernel installation
Correction: Taken from: var/log/zypper/history
It doesn't matter what zypper says. Simply check initrd content via lsinitrd. If the target firmware file is found there, it's OK. If not, that's the problem.
Created attachment 258143 [details] lsinitrd kernel 4.13.0-rc6 parts the firmware is there
(In reply to Takashi Iwai from comment #5) > It doesn't matter what zypper says. Simply check initrd content via > lsinitrd. > If the target firmware file is found there, it's OK. If not, that's the > problem. As I said in my description kernel 4.13.0-rc2-1.gb545b87 was booting with out any problems. So my conclusion is, that radeon.ko has become broken in the meantime, and that is. what dracut printed out for me.
The only reason the firmware would fail to load is if it's not available. From the log, it shows direct fw loading failed. Did you build radeon into your kernel rather than as a module? If so, you need to build the firmware into your kernel as well.
I guess it's module since Walther is using SUSE kernels. Could you give the full dmesg output from 4.13.0-rc6-vanilla you've installed recently? That has the firmware file in initrd as in comment 6, we can double-check it.
(In reply to Alex Deucher from comment #8) > The only reason the firmware would fail to load is if it's not available. > From the log, it shows direct fw loading failed. Did you build radeon into > your kernel rather than as a module? If so, you need to build the firmware > into your kernel as well. This is an argument, which belongs to my own kernel. I learned from https://wiki.gentoo.org/wiki/Radeon that these 3 firmware files are needed to be build-in in my own localyesconfig kernel. So far, I did not do so. But i filed this bug, because I found an openSUSE kernel with the described problems.
(In reply to Takashi Iwai from comment #9) > I guess it's module since Walther is using SUSE kernels. > > Could you give the full dmesg output from 4.13.0-rc6-vanilla you've > installed recently? That has the firmware file in initrd as in comment 6, > we can double-check it. There were a lot of updates the last few days for tumblewwed, the latest came from http://download.opensuse.org/update/tumbleweed/ with a updates for Mesa and libvdpau. Maybe this is the reason why: I have to write, that this bug will be closed as fixed. Today kernel 4.13.0-rc6-vanilla boots without problems for the first time! Thank you for your time, but I could not know this development before. P.S. "full dmesg output" seems to be obsolete now.