Re: [BUG] spinlock recursion when enabling function tracer on 32-bit

[Steven Rostedt]

On Fri, 20 Dec 2024 14:55:27 +0100
Ludwig Rydberg <ludwig.rydb...@gaisler.com> wrote:

> On 2024-12-18 22:35, Steven Rostedt wrote:
> 
> > 
> > Can you test this patch?
> >   
> 
> I have tested the patch on both riscv32 (qemu) & sparc32 (leon) and it seems
> to be working fine.
> 
> But I can mention that I ran into a deadlock situation on sparc32.
> I had to remove CC_FLAGS_FTRACE for the code that is responsible for
> flushing TLBs (SMP). It uses an NMI to signal other CPUs which triggered
> the deadlock if tracing was enabled on that path.

Can you try this patch on top of the last one?

-- Steve

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index 7e257e855dd1..a9fe54b79ce5 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -4398,8 +4398,13 @@ rb_reserve_next_event(struct trace_buffer *buffer,
        int nr_loops = 0;
        int add_ts_default;
 
-       /* ring buffer does cmpxchg, make sure it is safe in NMI context */
-       if (!IS_ENABLED(CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG) &&
+       /*
+        * ring buffer does cmpxchg as well as atomic64 operations
+        * (which some archs use locking for atomic64), make sure this
+        * is safe in NMI context
+        */
+       if ((!IS_ENABLED(CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG) ||
+            IS_ENABLED(CONFIG_GENERIC_ATOMIC64)) &&
            (unlikely(in_nmi()))) {
                return NULL;
        }