Re: [PATCH v4 7/8] lockdep: Change hardirq{s_enabled,_context} to per-cpu variables

[Peter Zijlstra <peterz@infradead.org>]

On Tue, Jun 23, 2020 at 09:13:35PM +0200, Marco Elver wrote:
> [   10.182354] ------------[ cut here ]------------
> [   10.183058] WARNING: CPU: 7 PID: 136 at kernel/locking/lockdep.c:398 lockdep_hardirqs_on_prepare+0x1c6/0x270
> [   10.184347] Modules linked in:
> [   10.184771] CPU: 7 PID: 136 Comm: systemd-journal Not tainted 5.8.0-rc1+ #3
> [   10.185706] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-1 04/01/2014
> [   10.186821] RIP: 0010:lockdep_hardirqs_on_prepare+0x1c6/0x270
> [   10.187594] Code: 75 28 65 48 8b 04 25 28 00 00 00 48 3b 44 24 08 0f 85 b9 00 00 00 48 83 c4 10 5b 41 5e 41 5f c3 65 48 ff 05 d4 24 4e 75 eb d8 <0f> 0b 90 41 c7 86 c4 08 00 00 00 00 00 00 eb c8 e8 65 09 71 01 85
> [   10.190203] RSP: 0018:ffffa7ee802b7848 EFLAGS: 00010017
> [   10.190989] RAX: 0000000000000001 RBX: ffff955e92a34ab0 RCX: 0000000000000001
> [   10.192053] RDX: 0000000000000006 RSI: ffff955e92a34a88 RDI: ffff955e92a341c0
> [   10.193117] RBP: ffffa7ee802b7be8 R08: 0000000000000000 R09: 0000ffffffffffff
> [   10.194186] R10: 0000ffffffffffff R11: 0000ffff8d07e268 R12: 0000000000000001
> [   10.195249] R13: ffffffff8e41bb10 R14: ffff955e92a341c0 R15: 0000000000000001
> [   10.196312] FS:  00007fd6862aa8c0(0000) GS:ffff955e9fd80000(0000) knlGS:0000000000000000
> [   10.197513] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [   10.198373] CR2: 00007fd6837dd000 CR3: 0000000812acc001 CR4: 0000000000760ee0
> [   10.199436] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> [   10.200494] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> [   10.201554] PKRU: 55555554
> [   10.201967] Call Trace:
> [   10.202348]  ? _raw_spin_unlock_irqrestore+0x40/0x70
> [   10.203093]  trace_hardirqs_on+0x56/0x60			<----- enter IRQ flags tracing code?
> [   10.203686]  _raw_spin_unlock_irqrestore+0x40/0x70  		<----- take report_lock
> [   10.204406]  prepare_report+0x11f/0x150
> [   10.204986]  kcsan_report+0xca/0x6c0 			<----- generating a KCSAN report

Oh, duh.. that's because lockdep_off() ;-)

diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index ab7571c1a1f5..c9ea05edce25 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -395,7 +395,7 @@ void lockdep_init_task(struct task_struct *task)
 
 static __always_inline void lockdep_recursion_finish(void)
 {
-	if (WARN_ON_ONCE(--current->lockdep_recursion))
+	if (WARN_ON_ONCE((--current->lockdep_recursion) & LOCKDEP_RECURSION_MASK))
 		current->lockdep_recursion = 0;
 }