On Tue, Jun 20, 2023 at 10:51:25AM +0530, Naveen N Rao wrote:
Christophe Leroy wrote:
quoted
This reverts commit 1e688dd2a3d6759d416616ff07afc4bb836c4213.
That commit aimed at optimising the code around generation of
WARN_ON/BUG_ON but this leads to a lot of dead code erroneously
generated by GCC.
text data bss dec hex filename
9551585 3627834 224376 13403795 cc8693 vmlinux.before
9535281 3628358 224376 13388015 cc48ef vmlinux.after
Once this change is reverted, in a standard configuration (pmac32 +
function tracer) the text is reduced by 16k which is around 1.7%
Aneesh recently reported a build failure due to the use of 'asm goto' in
WARN_ON(). We were able to root-cause it to the use of 'asm goto' with two
config options: CONFIG_CC_OPTIMIZE_FOR_SIZE and
CONFIG_DEBUG_SECTION_MISMATCH.
FWIW;
I recently had clang-powerpc report a very dodgy build error that was
due to a combination of these asm-goto and the usage of __cleanup__.
For some reason the label of the asm-goto crossed over the __cleanup__
variable declaration -- which is not valid, but also was completely
insane for that's not what the code called for.
https://lkml.kernel.org/r/20230610082005.GB1370249@hirez.programming.kicks-ass.net (local)
But in my book that's a compiler issue, not a kernel issue and I'd be
hesitant to pull the asm-goto use just for that.