perf: recursive locking of ctx->mutex

[Sasha Levin <sasha.levin@oracle.com>]

Hi all,

While fuzzing with trinity inside a KVM tools guest running the latest -next
kernel I've stumbled on the following spew:

[  460.391146] =============================================
[  460.391146] [ INFO: possible recursive locking detected ]
[  460.391146] 4.1.0-rc1-next-20150429-sasha-00037-g3e011d3-dirty #2192 Not tainted
[  460.391146] ---------------------------------------------
[  460.391146] trinity-main/9652 is trying to acquire lock:
[  460.391146] (&ctx->mutex){+.+.+.}, at: perf_event_ctx_lock_nested (kernel/events/core.c:965)
[  460.391146] Mutex: counter: 1 owner: None
[  460.391146]
[  460.391146] but task is already holding lock:
[  460.391146] (&ctx->mutex){+.+.+.}, at: perf_event_init_task (kernel/events/core.c:8732 kernel/events/core.c:8800)
[  460.391146] Mutex: counter: 0 owner: trinity-main
[  460.391146]
[  460.391146] other info that might help us debug this:
[  460.391146]  Possible unsafe locking scenario:
[  460.391146]
[  460.391146]        CPU0
[  460.391146]        ----
[  460.391146]   lock(&ctx->mutex);
[  460.391146]   lock(&ctx->mutex);
[  460.391146]
[  460.391146]  *** DEADLOCK ***
[  460.391146]
[  460.391146]  May be due to missing lock nesting notation
[  460.391146]
[  460.391146] 2 locks held by trinity-main/9652:
[  460.391146] #0: (&ctx->mutex){+.+.+.}, at: perf_event_init_task (kernel/events/core.c:8732 kernel/events/core.c:8800)
[  460.438470] Mutex: counter: 0 owner: trinity-main
[  460.438470] #1: (&pmus_srcu){......}, at: perf_init_event (kernel/events/core.c:7385)
[  460.438470]
[  460.438470] stack backtrace:
[  460.438470] CPU: 3 PID: 9652 Comm: trinity-main Not tainted 4.1.0-rc1-next-20150429-sasha-00037-g3e011d3-dirty #2192
[  460.438470]  ffffffffaef7d2c0 00000000f948113a ffff8802f4e4f528 ffffffffa8dddba3
[  460.438470]  0000000000000000 ffffffffaef7d2c0 ffff8802f4e4f6c8 ffffffff9f302223
[  460.438470]  ffff880a70f240c8 ffff880a0000bb5a ffffed0000000547 ffff8802f61d0dc8
[  460.438470] Call Trace:
[  460.438470] dump_stack (lib/dump_stack.c:52)
[  460.438470] __lock_acquire (kernel/locking/lockdep.c:1776 kernel/locking/lockdep.c:1820 kernel/locking/lockdep.c:2152 kernel/locking/lockdep.c:3238)
[  460.438470] ? lockdep_init_map (kernel/locking/lockdep.c:3105)
[  460.438470] ? sched_clock_cpu (kernel/sched/clock.c:311)
[  460.438470] ? __lock_is_held (kernel/locking/lockdep.c:3572)
[  460.438470] lock_acquire (kernel/locking/lockdep.c:3658)
[  460.438470] ? perf_event_ctx_lock_nested (kernel/events/core.c:965)
[  460.438470] mutex_lock_nested (kernel/locking/mutex.c:526 kernel/locking/mutex.c:617)
[  460.438470] ? perf_event_ctx_lock_nested (kernel/events/core.c:965)
[  460.438470] ? get_parent_ip (kernel/sched/core.c:2556)
[  460.438470] ? get_lock_stats (kernel/locking/lockdep.c:249)
[  460.438470] ? perf_event_ctx_lock_nested (kernel/events/core.c:965)
[  460.438470] ? _mutex_lock_nest_lock (kernel/locking/mutex.c:615)
[  460.438470] ? perf_event_ctx_lock_nested (include/linux/rcupdate.h:969 kernel/events/core.c:962)
[  460.438470] perf_event_ctx_lock_nested (kernel/events/core.c:965)
[  460.438470] ? perf_event_ctx_lock_nested (include/linux/rcupdate.h:912 kernel/events/core.c:956)
[  460.438470] ? perf_init_event (include/linux/rcupdate.h:969 kernel/events/core.c:7394)
[  460.438470] perf_try_init_event (kernel/events/core.c:977 kernel/events/core.c:7368)
[  460.438470] perf_init_event (kernel/events/core.c:7404)
[  460.438470] ? perf_bp_event (kernel/events/core.c:7385)
[  460.438470] perf_event_alloc (kernel/events/core.c:7564)
[  460.438470] inherit_event.isra.57 (kernel/events/core.c:8564)
[  460.438470] inherit_task_group.isra.59 (kernel/events/core.c:8646 kernel/events/core.c:8682)
[  460.438470] perf_event_init_task (kernel/events/core.c:8735 kernel/events/core.c:8800)
[  460.438470] copy_process (kernel/fork.c:1418)
[  460.438470] do_fork (kernel/fork.c:1705)
[  460.438470] SyS_clone (kernel/fork.c:1794)
[  460.438470] system_call_fastpath (arch/x86/kernel/entry_64.S:261)


Thanks,
Sasha