Re: [tip:perf/core] x86: Introduce text_poke_smp_batch() for batch-code modifying

[Peter Zijlstra <peterz@infradead.org>]

On Mon, 2010-12-06 at 18:17 +0000, tip-bot for Masami Hiramatsu wrote:

> @@ -631,10 +639,13 @@ static int __kprobes stop_machine_text_poke(void *data)
>  void *__kprobes text_poke_smp(void *addr, const void *opcode, size_t len)
>  {
>  	struct text_poke_params tpp;
> +	struct text_poke_param p;
>  
> -	tpp.addr = addr;
> -	tpp.opcode = opcode;
> -	tpp.len = len;
> +	p.addr = addr;
> +	p.opcode = opcode;
> +	p.len = len;
> +	tpp.params = &p;
> +	tpp.nparams = 1;
>  	atomic_set(&stop_machine_first, 1);
>  	wrote_text = 0;
>  	/* Use __stop_machine() because the caller already got online_cpus. */

 ^^^^^^^^^

> @@ -642,6 +653,26 @@ void *__kprobes text_poke_smp(void *addr, const void *opcode, size_t len)
>  	return addr;
>  }
>  
> +/**
> + * text_poke_smp_batch - Update instructions on a live kernel on SMP
> + * @params: an array of text_poke parameters
> + * @n: the number of elements in params.
> + *
> + * Modify multi-byte instruction by using stop_machine() on SMP. Since the
> + * stop_machine() is heavy task, it is better to aggregate text_poke requests
> + * and do it once if possible.
> + *
> + * Note: Must be called under get_online_cpus() and text_mutex.
> + */
> +void __kprobes text_poke_smp_batch(struct text_poke_param *params, int n)
> +{
> +	struct text_poke_params tpp = {.params = params, .nparams = n};
> +
> +	atomic_set(&stop_machine_first, 1);
> +	wrote_text = 0;
> +	stop_machine(stop_machine_text_poke, (void *)&tpp, NULL);
> +}

 ^^^^^^^^^^^^^^


---
Subject: x86: Fix text_poke_smp_batch() deadlock

=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.38-rc4-test+ #1
-------------------------------------------------------
bash/1850 is trying to acquire lock:
 (text_mutex){+.+.+.}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f

but task is already holding lock:
 (smp_alt){+.+...}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f

which lock already depends on the new lock.


the existing dependency chain (in reverse order) is:

-> #2 (smp_alt){+.+...}:
       [<ffffffff81082d02>] lock_acquire+0xcd/0xf8
       [<ffffffff8192e119>] __mutex_lock_common+0x4c/0x339
       [<ffffffff8192e4ca>] mutex_lock_nested+0x3e/0x43
       [<ffffffff8101050f>] alternatives_smp_switch+0x77/0x1d8
       [<ffffffff81926a6f>] do_boot_cpu+0xd7/0x762
       [<ffffffff819277dd>] native_cpu_up+0xe6/0x16a
       [<ffffffff81928e28>] _cpu_up+0x9d/0xee
       [<ffffffff81928f4c>] cpu_up+0xd3/0xe7
       [<ffffffff82268d4b>] kernel_init+0xe8/0x20a
       [<ffffffff8100ba24>] kernel_thread_helper+0x4/0x10

-> #1 (cpu_hotplug.lock){+.+.+.}:
       [<ffffffff81082d02>] lock_acquire+0xcd/0xf8
       [<ffffffff8192e119>] __mutex_lock_common+0x4c/0x339
       [<ffffffff8192e4ca>] mutex_lock_nested+0x3e/0x43
       [<ffffffff810568cc>] get_online_cpus+0x41/0x55
       [<ffffffff810a1348>] stop_machine+0x1e/0x3e
       [<ffffffff819314c1>] text_poke_smp_batch+0x3a/0x3c
       [<ffffffff81932b6c>] arch_optimize_kprobes+0x10d/0x11c
       [<ffffffff81933a51>] kprobe_optimizer+0x152/0x222
       [<ffffffff8106bb71>] process_one_work+0x1d3/0x335
       [<ffffffff8106cfae>] worker_thread+0x104/0x1a4
       [<ffffffff810707c4>] kthread+0x9d/0xa5
       [<ffffffff8100ba24>] kernel_thread_helper+0x4/0x10

-> #0 (text_mutex){+.+.+.}:


other info that might help us debug this:

6 locks held by bash/1850:
 #0:  (&buffer->mutex){+.+.+.}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f
 #1:  (s_active#75){.+.+.+}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f
 #2:  (x86_cpu_hotplug_driver_mutex){+.+.+.}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f
 #3:  (cpu_add_remove_lock){+.+.+.}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f
 #4:  (cpu_hotplug.lock){+.+.+.}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f
 #5:  (smp_alt){+.+...}, at: [<ffffffff8100a9c1>] return_to_handler+0x0/0x2f

stack backtrace:
Pid: 1850, comm: bash Not tainted 2.6.38-rc4-test+ #1
Call Trace:

 [<ffffffff81080eb2>] print_circular_bug+0xa8/0xb7
 [<ffffffff8192e4ca>] mutex_lock_nested+0x3e/0x43
 [<ffffffff81010302>] alternatives_smp_unlock+0x3d/0x93
 [<ffffffff81010630>] alternatives_smp_switch+0x198/0x1d8
 [<ffffffff8102568a>] native_cpu_die+0x65/0x95
 [<ffffffff818cc4ec>] _cpu_down+0x13e/0x202
 [<ffffffff8117a619>] sysfs_write_file+0x108/0x144
 [<ffffffff8111f5a2>] vfs_write+0xac/0xff
 [<ffffffff8111f7a9>] sys_write+0x4a/0x6e

Reported-by: Steven Rostedt <rostedt@goodmis.org>
Tested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 arch/x86/kernel/alternative.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
index 1236085..7038b95 100644
--- a/arch/x86/kernel/alternative.c
+++ b/arch/x86/kernel/alternative.c
@@ -671,7 +671,7 @@ void __kprobes text_poke_smp_batch(struct text_poke_param *params, int n)
 
 	atomic_set(&stop_machine_first, 1);
 	wrote_text = 0;
-	stop_machine(stop_machine_text_poke, (void *)&tpp, NULL);
+	__stop_machine(stop_machine_text_poke, (void *)&tpp, NULL);
 }
 
 #if defined(CONFIG_DYNAMIC_FTRACE) || defined(HAVE_JUMP_LABEL)