On Thu, Aug 26, 2021 at 11:21 AM Sean Christopherson <seanjc@xxxxxxxxxx> wrote: > > On Wed, Aug 25, 2021, Bill Wendling wrote: > > Clang may decide to inline some functions that have inline asm with labels. > > For all changlogs, it's probably worth clarifying that they have _global_ labels, > e.g. local labels within an asm block are perfectly ok for inlining, as are local > labels in the function (but outside of the block) used by asm goto. > > And maybe add a "#define noinline ..." to match the kernel and convert the two > existing uses in pmu_lbr.c as a prep patch? > I sent out a patch for this. Thanks! As for the changelogs, I'll send an updated patch series once the noinline patch is resolved. -bw > > Doing this duplicates the labels, causing the assembler to be complain. These > > patches add the "noinline" attribute to the functions to prevent this. > > > > Bill Wendling (4): > > x86: realmode: mark exec_in_big_real_mode as noinline > > x86: svm: mark test_run as noinline > > x86: umip: mark do_ring3 as noinline > > x86: vmx: mark some test_* functions as noinline > > > > x86/realmode.c | 2 +- > > x86/svm.c | 2 +- > > x86/umip.c | 2 +- > > x86/vmx.c | 6 +++--- > > 4 files changed, 6 insertions(+), 6 deletions(-) > > > > -- > > 2.33.0.rc2.250.ged5fa647cd-goog > >
