https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97445
--- Comment #28 from Christophe Leroy <christophe.leroy at csgroup dot eu> ---
Find attached the temporary files for net/core/skbuff.c, as it is indeed what
initially triggered my focus on this issue. I don't expect such a function at
all:
00000cc8 <csum_partial>:
cc8: 48 00 00 00 b cc8 <csum_partial>
cc8: R_PPC_REL24 __csum_partial
Every function should call __csum_partial() directly.
I guess that's the same problem, powerpc csum_partial() does sums 'manually'
for some particular values of constant 'len', then fallbacks on
__csum_partial().
When 'len' is not a constant, csum_partial() is reduced to a call to
__csum_partial(), and I'd expect GCC 10 to call __csum_partial() directly as
all previous GCC have always done.
Though, getting the following warning when building:
warning: inlining failed in call to 'csum_partial': --param
max-inline-insns-single limit reached [-Winline]
