On 20/2/19 3:26 pm, Daniel Axtens wrote:
I needed the following diff to get this booting on a T4240RDB:
diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile
index 961f44eabb65..fbe9894d6305 100644
--- a/arch/powerpc/kernel/Makefile
+++ b/arch/powerpc/kernel/Makefile
@@ -3,6 +3,10 @@
# Makefile for the linux kernel.
#
+KCOV_INSTRUMENT_cputable.o := n
+KCOV_INSTRUMENT_setup_64.o := n
+KCOV_INSTRUMENT_paca.o := n
+
CFLAGS_ptrace.o += -DUTS_MACHINE='"$(UTS_MACHINE)"'
# Disable clang warning for using setjmp without setjmp.h header
diff --git a/arch/powerpc/mm/Makefile b/arch/powerpc/mm/Makefile
index f965fc33a8b7..0140e7e12c29 100644
--- a/arch/powerpc/mm/Makefile
+++ b/arch/powerpc/mm/Makefile
@@ -5,6 +5,9 @@
ccflags-$(CONFIG_PPC64) := $(NO_MINIMAL_TOC)
+KCOV_INSTRUMENT_tlb_nohash.o := n
+KCOV_INSTRUMENT_fsl_booke_mmu.o := n
+
CFLAGS_REMOVE_slb.o = $(CC_FLAGS_FTRACE)
obj-y := fault.o mem.o pgtable.o mmap.o \
The change to kernel/ is required to get the kernel to even begin to
boot, and the change to mm/ is required to successfully set up SMP. I'm
not sure preciesly why they cause issues.
Thanks for testing this - I'll roll this into v3.
I was then able to run kcovtrace and the results seem to make sense. >
Perhaps in the future some further stuff should be trimmed down to make
the coverage results less noisy (restore_math is probably not telling us
anything interesting, for example), but certainly this is a great start.
I think syzkaller (as the main kcov consumer) can probably cope...
With those changes,
Tested-by: Daniel Axtens <d...@axtens.net> # e6500
--
Andrew Donnellan OzLabs, ADL Canberra
andrew.donnel...@au1.ibm.com IBM Australia Limited