@@ -257,13 +257,18 @@ hardware_interrupt_hv:
BEGIN_FTR_SECTION
_MASKABLE_EXCEPTION_PSERIES(0x502, hardware_interrupt,
EXC_HV, SOFTEN_TEST_HV)
- KVM_HANDLER(PACA_EXGEN, EXC_HV, 0x502)
FTR_SECTION_ELSE
_MASKABLE_EXCEPTION_PSERIES(0x500, hardware_interrupt,
EXC_STD, SOFTEN_TEST_PR)
- KVM_HANDLER(PACA_EXGEN, EXC_STD, 0x500)
ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206)
+ /*
+ * Relon code jumps to these KVM handlers too so can't put them
+ * in the feature sections.
+ */
+ KVM_HANDLER(PACA_EXGEN, EXC_HV, 0x502)
+ KVM_HANDLER(PACA_EXGEN, EXC_STD, 0x500)
+
STD_EXCEPTION_PSERIES(0x600, alignment)
KVM_HANDLER(PACA_EXGEN, EXC_STD, 0x600)
In preparation for reworking the alternate feature patching code, remove a case of cross-fixup branching in the exception code. That is, a branch from one alt feature section targeting another alt feature section. We have the following: BEGIN_FTR_SECTION b do_kvm_0x502 ... do_kvm_0x502: ... FTR_SECTION_ELSE b do_kvm_0x500 ... do_kvm_0x500: ... ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE | CPU_FTR_ARCH_206) BEGIN_FTR_SECTION b do_kvm_0x502 FTR_SECTION_ELSE b do_kvm_0x500 ALT_FTR_SECTION_END_IFSET(CPU_FTR_HVMODE) This somehow manages to work today. The first branch to do_kvm_0x500 works as expected and branches to the do_kvm_0x500 code that is moved to the destination location. The second branch to it branches to the linked location of the do_kvm_0x500 code. Two copies of the code are being executed. A subsequent patch moves the patchcode far away and discards it after boot, so this type of branch doesn't work. Signed-off-by: Nick Piggin <npiggin@gmail.com> --- arch/powerpc/kernel/exceptions-64s.S | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)