I have been working to update the pathset for 5.15.61 but am finding lots of patches are seemingly upstreamed. Many center on the bcm27xx vc4 subsystem. There currently 146 patches for this platform containing vc4 in the name. I have personally reviewed 14 of them but am finding the process tedious.
The process is running
update_kernel.sh -v -u 5.15 5.15.61 and manually reviewing the patches that the script fails to apply. I do this by inspecting each target file and the patch's code. Is there an automated method to check these patches against 5.15.61?
Nothing I have ever found. If your intent is only a target for yourself (i.e. not to submit a PR for a kernel upgrade), you can easily short circuit the upgrade script, worked yesterday for the mvebu target without issue.
Yes, I was able to do this to skip bcm27xx but I am running a Raspberry Pi 4 which is bcm27xx so I need to work on this
I began to see if I could hack the make target to delete the offending patch if it can be reverse applied and continue. From what I gather, include/quilt.mk seems to be doing all the work when the update script runs:
make target/linux/refresh V=ws KERNEL_PATCHVER=5.15 LINUX_VERSION=5.15.61 LINUX_KERNEL_HASH=skip
Can it be modified to run
quilt delete -n and continue if it detects a reverse applied patch?
Applying patch platform/950-0004-drm-vc4-hdmi-Remove-the-DDC-probing-for-status-detec.patch
patching file drivers/gpu/drm/vc4/vc4_hdmi.c
Hunk #1 FAILED at 176.
1 out of 1 hunk FAILED -- rejects in file drivers/gpu/drm/vc4/vc4_hdmi.c
Patch platform/950-0004-drm-vc4-hdmi-Remove-the-DDC-probing-for-status-detec.patch can be reverse-applied
I was hoping quilt would end in a non-1 exit code for these patches, but it does not seem to.
Before getting too deep in this:
- Is modifying the make behavior to remove a patch that can be reverse applied sane?
- If so, how would one accomplish this?
@Noltari - I see you added all the bcm27xx specific 5.15 patches. Any thoughts?
There is now a pending PR.