Closed liu-song-6 closed 3 years ago
Probably a bug. I failed to reproduce it however. Can you invoke kpatch-build
with -d
option and then upload ${HOME}/.kpatch/tmp/orig/drivers/md/raid1.o
and ${HOME}/.kpatch/tmp/patched/drivers/md/raid1.o
somewhere?
Does this fix it? jpoimboe@066c4ba
It doesn't seem to work.
Probably a bug. I failed to reproduce it however. Can you invoke
kpatch-build
with-d
option and then upload${HOME}/.kpatch/tmp/orig/drivers/md/raid1.o
and${HOME}/.kpatch/tmp/patched/drivers/md/raid1.o
somewhere?
I added these files and build.log to https://github.com/liu-song-6/kpatch/tree/debug_1160 .
Thanks, Song
Probably a bug. I failed to reproduce it however. Can you invoke
kpatch-build
with-d
option and then upload${HOME}/.kpatch/tmp/orig/drivers/md/raid1.o
and${HOME}/.kpatch/tmp/patched/drivers/md/raid1.o
somewhere?I added these files and build.log to https://github.com/liu-song-6/kpatch/tree/debug_1160 .
@sm00th does this help reproducing the issue? Thanks!
@sm00th does this help reproducing the issue? Thanks!
Yes, I didn't have much time to look at it today, but I can reprdoruce it with these files.
I wasn't able to recreate, but it may be a bug in the ".cold" subfunction handling, i.e. kpatch_detect_child_functions().
I wasn't able to recreate, but it may be a bug in the ".cold" subfunction handling, i.e. kpatch_detect_child_functions().
Right, we expect '.cold' functions to have an id at the end and the ones in provided objectfiles don't have that. So it could be as easy as this https://github.com/sm00th/kpatch/commit/c1f4f77a77e54b6159fbf888c2c5ce64cb160be0? Or are there any pitfalls here?
Right, we expect '.cold' functions to have an id at the end and the ones in provided objectfiles don't have that. So it could be as easy as this sm00th@c1f4f77? Or are there any pitfalls here?
Oh, good catch. That should be the fix. I made the same fix to objtool (and forgot to fix it for here, oops).
I wasn't able to recreate, but it may be a bug in the ".cold" subfunction handling, i.e. kpatch_detect_child_functions().
Right, we expect '.cold' functions to have an id at the end and the ones in provided objectfiles don't have that. So it could be as easy as this sm00th@c1f4f77? Or are there any pitfalls here?
Yes, this fixes kpatch-build for my test. Thanks a lot!
Hi folks,
I am testing kpatch on a very simple patch:
which hit errors like:
Is this expected or a bug? I am using gcc 9.1.1, and I have
CONFIG_MD_RAID1=m
Thanks in advance.