You are not logged in.
Pages: 1
Upgraded from 5.7.5.arch1-1 -> 5.7.6.arch1-1 today (I update every day) and now my audio now emits nothing but a screech. Downgrading to 5.7.5.arch1-1 fixes it. I switched versions back and forth a few times to confirm this. Could not see anything in the journal hinting about the problem.
Offline
There are audio related changes listed in https://cdn.kernel.org/pub/linux/kernel … eLog-5.7.6
Offline
Found a bug just raised on the kernel bug tracker about this issue: https://bugzilla.kernel.org/show_bug.cgi?id=208317.
Offline
I can also confirm this with a thunderbolt dock for a thinkpad. It definitively comes from upgrading 5.7.5 -> 5.7.6 as reverting to 5.7.5 fixes the issue.
Offline
The only commit that mentions thunderbolt is https://git.kernel.org/pub/scm/linux/ke … 1c91b54cd1 what if you revert it?
Edit:
Change the PKGBUILD prepare function to
prepare() {
cd $_srcname
echo "Setting version..."
scripts/setlocalversion --save-scmversion
echo "-$pkgrel" > localversion.10-pkgrel
echo "${pkgbase#linux}" > localversion.20-pkgname
git revert -n afaff825e3a436f9d1e3986530133b1c91b54cd1
local src
for src in "${source[@]}"; do
src="${src%%::*}"
src="${src##*/}"
[[ $src = *.patch ]] || continue
echo "Applying patch $src..."
patch -Np1 < "../$src"
done
echo "Setting config..."
cp ../config .config
make olddefconfig
make -s kernelrelease > version
echo "Prepared $pkgbase version $(<version)"
}
Last edited by loqs (2020-06-27 17:02:50)
Offline
@loqs, thanks for looking at this issue but unfortunately reverting that commit made no difference.
Last edited by bulletmark (2020-06-28 10:52:30)
Offline
Please place this at the end of the PKGBUILD
pkgver() {
cd $_srcname
git describe --long | sed -E 's/^v//;s/([^-]*-g)/r\1/;s/-/./g;s/\.rc/rc/'
}
Then
cd src/archlinux-linux
git reset --hard
git checkout v5.7.5
cd ../..
makepkg -rsi #Check 5.7.5 does work
Next
cd src/archlinux-linux
$ git checkout v5.7.6^1 #the parent of 5.7.6 to avoid changing the version in the makefile which might trigger a full rebuild
$ cd ../..
$ makepkg -ersi #This is to confirm 5.7.6 as built on your system does have the issue
Start bisection
cd src/archlinux-linux
$ git bisect start
$ git bisect good v5.7.5
$ git bisect bad v5.7.6
$ cd ../..
$ makepkg -ersif
Bisection loop
cd src/archlinux-linux
$ git bisect $result #Substitue good or bad here
$ cd ../..
$ makepkg -ersif #Repeat these four lines and test the generated kernel until git has found the bad commit
Offline
@loqs, this is where I got to (after at least 4 full kernel builds at 1 hour each):
1. It seems to me that I don't want the pkgver function because it forces a version change each update which forces a full rebuild. I think I should just set pkgver=5.7.5.arch1 for the point of these tests?
2. The git tags should be v5.7.5-arch1 not v5.7.5, etc.
3. Using the current 5.7.6 source tree, I did a build at git tag v5.7.6-arch1 and at v5.7.5-arch1 but the problem still existed at both. So I looked at the Arch changes for 5.7.6 and saw that the config file changed as well. Using the 5.7.5 config file I rebuilt 5.7.6 but the problem still existed. I then built 5.7.5 with the 5.7.5 config file and the problem disappeared. Even though this seems odd (and I don't know where the new config file should be introduced), I started to bisect upwards but the first bisect run terminated with:
INSTALL sound/usb/line6/snd-usb-podhd.ko
INSTALL sound/usb/line6/snd-usb-toneport.ko
INSTALL sound/usb/line6/snd-usb-variax.ko
INSTALL sound/usb/misc/snd-ua101.ko
INSTALL sound/usb/snd-usb-audio.ko
INSTALL sound/usb/snd-usbmidi-lib.ko
INSTALL sound/usb/usx2y/snd-usb-us122l.ko
INSTALL sound/usb/usx2y/snd-usb-usx2y.ko
INSTALL sound/x86/snd-hdmi-lpe-audio.ko
INSTALL sound/xen/snd_xen_front.ko
INSTALL virt/lib/irqbypass.ko
DEPMOD 5.7.5-1-custom
rm: cannot remove '/home/mark/build/linux/pkg/linux-custom/usr/lib/modules/5.7.5-arch1-1-custom/source': No such file or directory
rm: cannot remove '/home/mark/build/linux/pkg/linux-custom/usr/lib/modules/5.7.5-arch1-1-custom/build': No such file or directory
==> ERROR: A failure occurred in package_linux-custom().
Aborting...
Offline
$ git bisect start
$git bisect bad v5.7.6-arch1
$ git bisect good v5.7.5-arch1
Bisecting: a merge base must be tested
[66dfe45221605e11f38a0bf5eb2ee808cea7cfe7] Linux 5.7.5
$ git bisect good
Bisecting: 238 revisions left to test after this (roughly 8 steps)
[6dc1d4511e0ff6db45fafa76c74012ad05c30c34] firmware: imx: scu: Fix possible memory leak in imx_scu_probe()
I forgot the arch kernel repository does not have the upstream tags
git bisect start
git bisect bad a06eb423367eea436351ef2a92f61a4e625cf558 #v5.7.6
git bisect good 66dfe45221605e11f38a0bf5eb2ee808cea7cfe7 #v5.7.5
Bisecting: 237 revisions left to test after this (roughly 8 steps)
[74c5fd5dedde91e58acc52eb5948788496a017a2] scsi: core: Fix incorrect usage of shost_for_each_device
Which would have saved testing the merge base.
The pkgver function does not change the working tree it changes the pkgver recorded in the PKGBUILD so each bisect step would have its own set of built packages.
makepkg -Codd # this ill clean the src directory so do not do this while bisecting
cd src/archlinux-linux/
git checkout v5.7.5-arch1
make olddefconfig
diff .config .config.old
3c3
< # Linux/x86 5.7.5-arch1 Kernel Configuration
---
> # Linux/x86 5.7.6-arch1 Kernel Configuration
The only diference in the config should be the version.
The build failure looks to be from the end of _package()
# remove build and source links
rm "$modulesdir"/{source,build}
}
I do not know why that would fail.
Offline
I assume that audio codec inside a thunderbolt dock is a USB-attached audio codec, and I have a similar issue with another USB DAC(Shanling M0).
After 5.7.6 update, the DAC has severe glitches with or without pulseaudio tsched settings, reverting to 5.7.5 solved this issue.
I checked the changes of the 5.7.6 release and found the following patchset, after reverting this patchset the problem solved.
https://patchwork.kernel.org/patch/11506943/
Has anyone submitted a bug for that issue? Or could anyone check if this works with other audio codecs(e.g. USB codec intergrated in the tb dock)?
Last edited by imi415 (2020-06-28 15:45:56)
Offline
@imi415 https://git.kernel.org/pub/scm/linux/ke … 8627a49bf3 matches https://patchwork.kernel.org/patch/11506943/
https://patchwork.kernel.org/patch/11506941/ has not been applied to 5.7.6, has it?
Possible upstream bug reports
https://bugzilla.kernel.org/show_bug.cgi?id=208317
https://bugzilla.kernel.org/show_bug.cgi?id=208311
See also the reply instructions in
https://lore.kernel.org/alsa-devel/2020 … r@tsoy.me/
https://lore.kernel.org/stable/20200623 … ation.org/
Last edited by loqs (2020-06-28 14:28:22)
Offline
@loqs Yes you are right, I noticed that those patches are in the same series, so I ended up with an unreverted warning. The second patch wasn't in 5.7.6.
Offline
The pkgver function does not change the working tree it changes the pkgver recorded in the PKGBUILD so each bisect step would have its own set of built packages.
I know, but the bisect changes the git describe hash so that means the pkgver changes every build and that seems to cause a full rebuild. So I was asking that, for the point of the bisection builds at least, shouldn't I just hard code pkgver?
The only diference in the config should be the version.
I was referring to the Arch config file change between 5.7.5 and 5.7.6. That is not in the Linux source tree so is not bisected.
Offline
loqs wrote:The pkgver function does not change the working tree it changes the pkgver recorded in the PKGBUILD so each bisect step would have its own set of built packages.
I know, but the bisect changes the git describe hash so that means the pkgver changes every build and that seems to cause a full rebuild. So I was asking that, for the point of the bisection builds at least, shouldn't I just hard code pkgver?
If it works without the pkgver function remove it but if https://git.kernel.org/pub/scm/linux/ke … 8627a49bf3 is the cause it might be quicker to tes just reverting that.
The only diference in the config should be the version.
I was referring to the Arch config file change between 5.7.5 and 5.7.6. That is not in the Linux source tree so is not bisected.
That config change was introduced by commits added with 5.7.4 but the config was not updated until 5.7.6.arch1-1 if those changes are not present they would be added by `make olddefconfig` in prepare() or if prepare was skipped by `make all` in build.
Offline
I built the kernel 5.7.6 source cleanly with that usb-audio commit reverted and can confirm that my thunderbolt connected audio problem goes away.
Offline
The author of the offending commit has produced a single line fix which I confirmed on 5.7.6. See https://bugzilla.kernel.org/show_bug.cgi?id=208353.
Offline
Pages: 1