Re: [PATCH 3/7] openrisc: restore call-saved regs on sigreturn

From: Jonas Bonn
Date: Mon Sep 19 2016 - 10:47:12 EST

On 09/16/2016 04:43 PM, Stafford Horne wrote:
From: Jonas Bonn <jonas@xxxxxxxxxxxx>

Return to userspace via _resume_userspace instead of via syscall return
path for the rt_sigreturn syscall.

I'll rework this comment more later, but this patch needs testing.

This whole patch was reworked later... what happened to that work? I recall it being posted for testing without feedback...

In any case, the patch comment needs cleaning up.


Old comment from previous patch:

The sigreturn syscall is more like a context switch than a function call;
it entails a return from one context (the signal handler) to another
(the process in question). For a context switch like this there are
effectively no call-saved regs that remain constant across the transition.

This patch restores the call-saved regs from pt_regs before returning from
the syscall, effectively restoring the context that the process had before
being interrupted by the signal handler. Restoring the call-saved regs
in this way allows us to return to userspace via the usual syscall fast

Reported-by: Sebastian Macke <sebastian@xxxxxxxx>
Signed-off-by: Jonas Bonn <jonas@xxxxxxxxxxxx>
Signed-off-by: Stafford Horne <shorne@xxxxxxxxx>
arch/openrisc/kernel/entry.S | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/arch/openrisc/kernel/entry.S b/arch/openrisc/kernel/entry.S
index fec8bf9..572d223 100644
--- a/arch/openrisc/kernel/entry.S
+++ b/arch/openrisc/kernel/entry.S
@@ -1101,8 +1101,16 @@ ENTRY(__sys_fork)
l.addi r3,r1,0
- l.j _sys_rt_sigreturn
+ l.jal _sys_rt_sigreturn
l.addi r3,r1,0
+ l.sfne r30,r0
+ l.bnf _no_syscall_trace
+ l.nop
+ l.jal do_syscall_trace_leave
+ l.addi r3,r1,0
+ l.j _resume_userspace
+ l.nop
/* This is a catch-all syscall for atomic instructions for the OpenRISC 1000.
* The functions takes a variable number of parameters depending on which