When I added entry_SYSCALL_64_after_hwframe, I left TRACE_IRQS_OFF
before it. This means that users of entry_SYSCALL_64_after_hwframe
were responsible for invoking TRACE_IRQS_OFF, and the one and only
user (added in the same commit) got it wrong.
I think this would manifest as a warning if a Xen PV guest with
CONFIG_DEBUG_LOCKDEP=y were used with context tracking. (The
context tracking bit is to cause lockdep to get invoked before we
turn IRQs back on.) I haven't tested that for real yet because I
can't get a kernel configured like that to boot at all on Xen PV.
I've reported it upstream. The problem seems to be that Xen PV is
missing early #UD handling, is hitting some WARN, and we rely on
Move TRACE_IRQS_OFF below the label.
Cc: stable@xxxxxxxxxxxxxxx
Cc: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
Cc: Juergen Gross <jgross@xxxxxxxx>
Fixes: 8a9949bc71a7 ("x86/xen/64: Rearrange the SYSCALL entries")
Signed-off-by: Andy Lutomirski <luto@xxxxxxxxxx>
---
arch/x86/entry/entry_64.S | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
index a2b30ec69497..5063ed1214dd 100644
--- a/arch/x86/entry/entry_64.S
+++ b/arch/x86/entry/entry_64.S
@@ -148,8 +148,6 @@ ENTRY(entry_SYSCALL_64)
movq %rsp, PER_CPU_VAR(rsp_scratch)
movq PER_CPU_VAR(cpu_current_top_of_stack), %rsp
- TRACE_IRQS_OFF
-
/* Construct struct pt_regs on stack */
pushq $__USER_DS /* pt_regs->ss */
pushq PER_CPU_VAR(rsp_scratch) /* pt_regs->sp */
@@ -170,6 +168,8 @@ GLOBAL(entry_SYSCALL_64_after_hwframe)
sub $(6*8), %rsp /* pt_regs->bp, bx, r12-15 not saved */
UNWIND_HINT_REGS extra=0
+ TRACE_IRQS_OFF
+
/*
* If we need to do entry work or if we guess we'll need to do
* exit work, go straight to the slow path.
--
2.13.6