x86/alternatives: Sync bp_patching update for avoiding NULL pointer exception

ftracetest multiple_kprobes.tc testcase hits the following NULL pointer
exception:

 BUG: kernel NULL pointer dereference, address: 0000000000000000
 PGD 800000007bf60067 P4D 800000007bf60067 PUD 7bf5f067 PMD 0
 Oops: 0000 [#1] PREEMPT SMP PTI
 RIP: 0010:poke_int3_handler+0x39/0x100
 Call Trace:
  <IRQ>
  do_int3+0xd/0xf0
  int3+0x42/0x50
  RIP: 0010:sched_clock+0x6/0x10

poke_int3_handler+0x39 was alternatives:958:

  static inline void *text_poke_addr(struct text_poke_loc *tp)
  {
          return _stext + tp->rel_addr; <------ Here is line #958
  }

This seems to be caused by tp (bp_patching.vec) being NULL but
bp_patching.nr_entries != 0. There is a small chance for this
to happen, because we have no synchronization between the zeroing
of bp_patching.nr_entries and before clearing bp_patching.vec.

Steve suggested we could fix this by adding sync_core(), because int3
is done with interrupts disabled, and the on_each_cpu() requires
all CPUs to have had their interrupts enabled.

 [ mingo: Edited the comments and the changelog. ]

Suggested-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Tested-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: bristot@redhat.com
Fixes: c0213b0 ("x86/alternative: Batch of patch operations")
Link: https://lkml.kernel.org/r/157483421229.25881.15314414408559963162.stgit@devnote2
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Author: mhiramat

arch/x86/kernel/alternative.c

@@ -1134,8 +1134,14 @@ static void text_poke_bp_batch(struct text_poke_loc *tp, unsigned int nr_entries
 	 * sync_core() implies an smp_mb() and orders this store against
 	 * the writing of the new instruction.
 	 */
-	bp_patching.vec = NULL;
 	bp_patching.nr_entries = 0;
+	/*
+	 * This sync_core () call ensures that all INT3 handlers in progress
+	 * have finished. This allows poke_int3_handler() after this to
+	 * avoid touching bp_paching.vec by checking nr_entries == 0.
+	 */
+	text_poke_sync();
+	bp_patching.vec = NULL;
 }
 
 void text_poke_loc_init(struct text_poke_loc *tp, void *addr,