Dear Josh,
On 03/27/17 17:01, Paul Menzel wrote:
On 03/27/17 16:54, Josh Poimboeuf wrote:
On x86-32, with CONFIG_FIRMWARE and multiple CPUs, if you enable
function graph tracing and then suspend to RAM, it will triple fault and
reboot when it resumes.
The first fault happens when booting a secondary CPU:
startup_32_smp()
load_ucode_ap()
prepare_ftrace_return()
ftrace_graph_is_dead()
(accesses 'kill_ftrace_graph')
The early head_32.S code calls into load_ucode_ap(), which has an an
ftrace hook, so it calls prepare_ftrace_return(), which calls
ftrace_graph_is_dead(), which tries to access the global
'kill_ftrace_graph' variable with a virtual address, causing a fault
because the CPU is still in real mode.
The fix is to add a check in prepare_ftrace_return() to make sure it's
running in protected mode before continuing. The check makes sure the
stack pointer is a virtual kernel address. It's a bit of a hack, but
it's not very intrusive and it works well enough.
For reference, here are a few other ways this could have potentially
been fixed:
- Move startup_32_smp()'s call to load_ucode_ap() down to *after* paging
is enabled. (No idea what that would break.)
- Track down load_ucode_ap()'s entire callee tree and mark all the
functions 'notrace'. (Probably not realistic.)
- Pause graph tracing in ftrace_suspend_notifier_call() or bringup_cpu()
or __cpu_up(), and ensure that the pause facility can be queried from
real mode.
Reported-by: Paul Menzel <pmenzel@xxxxxxxxxxxxx>
Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
---
arch/x86/kernel/ftrace.c | 11 +++++++++++
1 file changed, 11 insertions(+)
Thank you for debugging this. It’s great that you were able to reproduce
this in QEMU. Hopefully, that’ll make for an easy test case. ;-)
diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index 8f3d9cf..1c5c4e2 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -983,6 +983,17 @@ void prepare_ftrace_return(unsigned long
self_addr, unsigned long *parent,
unsigned long return_hooker = (unsigned long)
&return_to_handler;
+ /*
+ * When resuming from suspend-to-ram, this function can be
indirectly
+ * called from early CPU startup code while the CPU is in real mode,
+ * which would fail miserably. Make sure the stack pointer is a
+ * virtual address.
+ *
+ * This check isn't as accurate as virt_addr_valid(), but it
should be
+ * good enough for this purpose, and it's fast.
+ */
+ if (unlikely((long)__builtin_frame_address(0) >= 0)) return;
The coding style requires the `return;` to be on a separate line.
+
if (unlikely(ftrace_graph_is_dead()))
return;
I’ll test your change this evening.
With both patches applied `./analyze_suspend.py -config
suspend-callgraph.cfg -filter i915` succeeds on a Lenovo X60t, so
suspend and resume work perfectly, when tracing is enabled.
Tested-by: Paul Menzel <pmenzel@xxxxxxxxxxxxx>
It’d be awesome, if you could tag both patches for inclusion into the
stable Linux Kernel series.
Kind regards,
Paul
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
