powerpc/book3s64/kexec: Clear CIABR on kexec
authorJordan Niethe <jniethe5@gmail.com>
Mon, 7 Dec 2020 01:05:18 +0000 (12:05 +1100)
committerMichael Ellerman <mpe@ellerman.id.au>
Mon, 7 Dec 2020 12:26:01 +0000 (23:26 +1100)
The value in CIABR persists across kexec which can lead to unintended
results when the new kernel hits the old kernel's breakpoint. For
example:

0:mon> bi $loadavg_proc_show
0:mon> b
   type            address
1 inst   c000000000519060  loadavg_proc_show+0x0/0x130
0:mon> x

$ kexec -l /mnt/vmlinux --initrd=/mnt/rootfs.cpio.gz --append='xmon=off'
$ kexec -e

$ cat /proc/loadavg
Trace/breakpoint trap

Make sure CIABR is cleared so this does not happen.

Signed-off-by: Jordan Niethe <jniethe5@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20201207010519.15597-1-jniethe5@gmail.com
arch/powerpc/include/asm/book3s/64/kexec.h

index 6b5c3a2..d4b9d47 100644 (file)
@@ -3,6 +3,7 @@
 #ifndef _ASM_POWERPC_BOOK3S_64_KEXEC_H_
 #define _ASM_POWERPC_BOOK3S_64_KEXEC_H_
 
+#include <asm/plpar_wrappers.h>
 
 #define reset_sprs reset_sprs
 static inline void reset_sprs(void)
@@ -14,6 +15,10 @@ static inline void reset_sprs(void)
 
        if (cpu_has_feature(CPU_FTR_ARCH_207S)) {
                mtspr(SPRN_IAMR, 0);
+               if (cpu_has_feature(CPU_FTR_HVMODE))
+                       mtspr(SPRN_CIABR, 0);
+               else
+                       plpar_set_ciabr(0);
        }
 
        /*  Do we need isync()? We are going via a kexec reset */