[PATCH RFC] s390: Fix nospec table alignments

From: Josh Poimboeuf
Date: Fri Aug 26 2022 - 19:56:00 EST


Add proper alignment for .nospec_call_table and .nospec_return_table in
vmlinux.

Fixes: f19fbd5ed642 ("s390: introduce execute-trampolines for branches")
Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
---
This is RFC because I don't know anything about s390 behavior for
unaligned data accesses, but this seemed to fix an issue for me.

While working on another s390 issue, I was getting intermittent boot
failures in __nospec_revert() when it tried to access 'instr[0]'. I
noticed the __nospec_call_start address ended in 'ff'. This patch
seemed to fix it. I have no idea why it was (only sometimes) failing in
the first place.

The intermittent part of it is probably at least partially explained by
CONFIG_RANDOMIZE_BASE. Except now I can no longer recreate it :-/

Regardless, this patch seems correct. I just can't explain what I saw.
Any ideas?

arch/s390/kernel/vmlinux.lds.S | 1 +
1 file changed, 1 insertion(+)

diff --git a/arch/s390/kernel/vmlinux.lds.S b/arch/s390/kernel/vmlinux.lds.S
index 2e526f11b91e..5ea3830af0cc 100644
--- a/arch/s390/kernel/vmlinux.lds.S
+++ b/arch/s390/kernel/vmlinux.lds.S
@@ -131,6 +131,7 @@ SECTIONS
/*
* Table with the patch locations to undo expolines
*/
+ . = ALIGN(4);
.nospec_call_table : {
__nospec_call_start = . ;
*(.s390_indirect*)
--
2.37.2