2.4.21-pre5 kksymoops for i386/ia64
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 3 lines · Keith Owens
Automatic decoding of oops on 2.5 has been very useful, so this patch adds kksymoops support to 2.4.21-pre5. Currently only for i386 and ia64, other architectures are easy to add.
This 2.4 kallsyms patch doesn't work for cross compilation because the modutils are terminally broken in this regard.
Rather than using this patch I would rather backport the 2.5 code which works fine even for cross and compresses the symbol tables too.
-Andi
Quoted 11 lines · Andi Kleen
Quoted 3 lines · Keith Owens
Automatic decoding of oops on 2.5 has been very useful, so this patch adds kksymoops support to 2.4.21-pre5. Currently only for i386 and ia64, other architectures are easy to add.
This 2.4 kallsyms patch doesn't work for cross compilation because the modutils are terminally broken in this regard.
Rather than using this patch I would rather backport the 2.5 code which works fine even for cross and compresses the symbol tables too.
The 2.5 kallsyms does not have section data which is required for kdb (and probably kgdb). That data used to be there until Rusty deleted it, the lack of section data is one of the reasons that I no longer do kdb for 2.5. If you want to kill kdb for 2.4 kernels as well, go ahead and backport the incomplete 2.5 code.
Quoted 10 lines · Keith Owens
Quoted 2 lines · Andi Kleen
Rather than using this patch I would rather backport the 2.5 code which works fine even for cross and compresses the symbol tables too.
The 2.5 kallsyms does not have section data which is required for kdb (and probably kgdb). That data used to be there until Rusty deleted it, the lack of section data is one of the reasons that I no longer do kdb for 2.5. If you want to kill kdb for 2.4 kernels as well, go ahead and backport the incomplete 2.5 code.
ps. The 2.5 kallsyms code is incompatible with modutils 2.4, backporting the incomplete 2.5 kallsyms would only get debugging symbols for the kernel, not for modules. Changing modutils 2.4 is not an option, I will not introduce an incompatible change in the middle of a stable kernel series unless there is no choice (e.g. to fix a critical bug).
Quoted 6 lines · Keith Owens
ps. The 2.5 kallsyms code is incompatible with modutils 2.4, backporting the incomplete 2.5 kallsyms would only get debugging symbols for the kernel, not for modules. Changing modutils 2.4 is not an option, I will not introduce an incompatible change in the middle of a stable kernel series unless there is no choice (e.g. to fix a critical bug).
At least for me not working in cross compile setups is a critical bug. YMMV.
-Andi
Quoted 10 lines · Andi Kleen
Quoted 6 lines · Keith Owens
ps. The 2.5 kallsyms code is incompatible with modutils 2.4, backporting the incomplete 2.5 kallsyms would only get debugging symbols for the kernel, not for modules. Changing modutils 2.4 is not an option, I will not introduce an incompatible change in the middle of a stable kernel series unless there is no choice (e.g. to fix a critical bug).
At least for me not working in cross compile setups is a critical bug. YMMV.
How the hell can it be a critical bug when 2.4 kernels do not currently have _any_ kksymoops support?
Quoted 3 lines · Keith Owens
How the hell can it be a critical bug when 2.4 kernels do not currently have _any_ kksymoops support?
A critical bug of the 2.4 kksymoops patches, making them mostly unusable.
-Andi
Quoted 2 lines · Andi Kleen
At least for me not working in cross compile setups is a critical bug. YMMV.
I can make modutils 2.4 (including kallsyms and depmod) work in cross compile mode and still maintain backwards compatibility. Conceptually the change is simple, but it affects a lot of code and needs a lot of testing. I am not going to spend any time on adding cross compile ability to modutils 2.4 until I see a defined need for it, people saying "would be nice to have does not count". If I cannot get kallsyms into the kernel for native compiles, why should I waste my time making kallsyms work for the even smaller group of people who do cross compiles?
Bottom line - when, and only when, the kksymoops patch is in the 2.4 kernel, then I will spend the time to make modutils 2.4 work in cross compile mode. If you insist that kallsyms work in cross compile mode before the patch goes in, then it is not going to happen and nobody gets automatic oops decoding in 2.4.
Quoted 5 lines · Keith Owens
Bottom line - when, and only when, the kksymoops patch is in the 2.4 kernel, then I will spend the time to make modutils 2.4 work in cross compile mode. If you insist that kallsyms work in cross compile mode before the patch goes in, then it is not going to happen and nobody gets automatic oops decoding in 2.4.
I can understand your frustration, but you know perfectly well that features which don't work for non-i386, don't work on SMP, or don't work when cross-compiled just _don't_ get merged. (In general, at least).
Signature · 1 line
dwmw2