On Fri, 3 May 2019 12:22:47 +0200 Peter Zijlstra peterz@infradead.org wrote:
Given that the entry_*.S changes for this functionality are somewhat tricky, make sure the paths are tested every boot, instead of on the rare occasion when we trip an INT3 while rewriting text.
Getting the INT3 frame setup even slightly wrong will make this come unstuck something spectacular.
Requested-by: Andy Lutomirski luto@kernel.org Signed-off-by: Peter Zijlstra (Intel) peterz@infradead.org
arch/x86/kernel/alternative.c | 81 ++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 77 insertions(+), 4 deletions(-)
I'll add this to my queue, as I tested this on x86_32 and 64. I also put in a printk("HERE in int3_magic\n") to make sure it really did get there (although, the BUG_ON should be good enough).
The printk will be removed from the official commit. It was just for my peace of mind.
-- Steve