Current we store the userspace r1 to PACATMSCRATCH before finally
saving it to the thread struct.
In theory an exception could be taken here (like a machine check or
SLB miss) that could write PACATMSCRATCH and hence corrupt the
userspace r1. The SLB fault currently doesn't touch PACATMSCRATCH, but
others do.
We've never actually seen this happen but it's theoretically
possible. Either way, the code is fragile as it is.
This patch saves r1 to the kernel stack (which can't fault) before we
turn MSR[RI] back on. PACATMSCRATCH is still used but only with
MSR[RI] off. We then copy r1 from the kernel stack to the thread
struct once we have MSR[RI] back on.
Suggested-by: Breno Leitao <lei...@debian.org>
Signed-off-by: Michael Neuling <mi...@neuling.org>
---
arch/powerpc/kernel/tm.S | 8 +++++++-
1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/arch/powerpc/kernel/tm.S b/arch/powerpc/kernel/tm.S
index 701b0f5b09..8207816a1e 100644
--- a/arch/powerpc/kernel/tm.S
+++ b/arch/powerpc/kernel/tm.S
@@ -178,6 +178,12 @@ _GLOBAL(tm_reclaim)
std r11, GPR11(r1) /* Temporary stash */
+ /* Move r1 to kernel stack in case PACATMSCRATCH is used once
+ * we turn on RI
+ */
+ ld r11, PACATMSCRATCH(r13)
+ std r11, GPR1(r1)
+
/* Store r13 away so we can free up the scratch SPR for the
* SLB fault handler (needed once we start access the
* thread_struct)
@@ -214,7 +220,7 @@ _GLOBAL(tm_reclaim)
SAVE_GPR(8, r7) /* user r8 */
SAVE_GPR(9, r7) /* user r9 */
SAVE_GPR(10, r7) /* user r10 */
- ld r3, PACATMSCRATCH(r13) /* user r1 */
+ ld r3, GPR1(r1) /* user r1 */
ld r4, GPR7(r1) /* user r7 */
ld r5, GPR11(r1) /* user r11 */
ld r6, GPR12(r1) /* user r12 */
--
2.17.1
