irqchip/gic-v3: Change pr_debug message to pr_devel
authorMark Salter <msalter@redhat.com>
Fri, 2 Feb 2018 14:20:29 +0000 (09:20 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 26 Apr 2018 09:02:21 +0000 (11:02 +0200)
commitde16dfcc510d8ec2981e4ed97e51d17c043cdaec
tree429d2f52ea4409b9dfef34e2cd06a06b7ef234a6
parent4032cd4fd3aec5d76d5ad9451f6d31ff6003c7cf
irqchip/gic-v3: Change pr_debug message to pr_devel

[ Upstream commit b6dd4d83dc2f78cebc9a7e6e7e4bc2be4d29b94d ]

The pr_debug() in gic-v3 gic_send_sgi() can trigger a circular locking
warning:

 GICv3: CPU10: ICC_SGI1R_EL1 5000400
 ======================================================
 WARNING: possible circular locking dependency detected
 4.15.0+ #1 Tainted: G        W
 ------------------------------------------------------
 dynamic_debug01/1873 is trying to acquire lock:
  ((console_sem).lock){-...}, at: [<0000000099c891ec>] down_trylock+0x20/0x4c

 but task is already holding lock:
  (&rq->lock){-.-.}, at: [<00000000842e1587>] __task_rq_lock+0x54/0xdc

 which lock already depends on the new lock.

 the existing dependency chain (in reverse order) is:

 -> #2 (&rq->lock){-.-.}:
        __lock_acquire+0x3b4/0x6e0
        lock_acquire+0xf4/0x2a8
        _raw_spin_lock+0x4c/0x60
        task_fork_fair+0x3c/0x148
        sched_fork+0x10c/0x214
        copy_process.isra.32.part.33+0x4e8/0x14f0
        _do_fork+0xe8/0x78c
        kernel_thread+0x48/0x54
        rest_init+0x34/0x2a4
        start_kernel+0x45c/0x488

 -> #1 (&p->pi_lock){-.-.}:
        __lock_acquire+0x3b4/0x6e0
        lock_acquire+0xf4/0x2a8
        _raw_spin_lock_irqsave+0x58/0x70
        try_to_wake_up+0x48/0x600
        wake_up_process+0x28/0x34
        __up.isra.0+0x60/0x6c
        up+0x60/0x68
        __up_console_sem+0x4c/0x7c
        console_unlock+0x328/0x634
        vprintk_emit+0x25c/0x390
        dev_vprintk_emit+0xc4/0x1fc
        dev_printk_emit+0x88/0xa8
        __dev_printk+0x58/0x9c
        _dev_info+0x84/0xa8
        usb_new_device+0x100/0x474
        hub_port_connect+0x280/0x92c
        hub_event+0x740/0xa84
        process_one_work+0x240/0x70c
        worker_thread+0x60/0x400
        kthread+0x110/0x13c
        ret_from_fork+0x10/0x18

 -> #0 ((console_sem).lock){-...}:
        validate_chain.isra.34+0x6e4/0xa20
        __lock_acquire+0x3b4/0x6e0
        lock_acquire+0xf4/0x2a8
        _raw_spin_lock_irqsave+0x58/0x70
        down_trylock+0x20/0x4c
        __down_trylock_console_sem+0x3c/0x9c
        console_trylock+0x20/0xb0
        vprintk_emit+0x254/0x390
        vprintk_default+0x58/0x90
        vprintk_func+0xbc/0x164
        printk+0x80/0xa0
        __dynamic_pr_debug+0x84/0xac
        gic_raise_softirq+0x184/0x18c
        smp_cross_call+0xac/0x218
        smp_send_reschedule+0x3c/0x48
        resched_curr+0x60/0x9c
        check_preempt_curr+0x70/0xdc
        wake_up_new_task+0x310/0x470
        _do_fork+0x188/0x78c
        SyS_clone+0x44/0x50
        __sys_trace_return+0x0/0x4

 other info that might help us debug this:

 Chain exists of:
   (console_sem).lock --> &p->pi_lock --> &rq->lock

  Possible unsafe locking scenario:

        CPU0                    CPU1
        ----                    ----
   lock(&rq->lock);
                                lock(&p->pi_lock);
                                lock(&rq->lock);
   lock((console_sem).lock);

  *** DEADLOCK ***

 2 locks held by dynamic_debug01/1873:
  #0:  (&p->pi_lock){-.-.}, at: [<000000001366df53>] wake_up_new_task+0x40/0x470
  #1:  (&rq->lock){-.-.}, at: [<00000000842e1587>] __task_rq_lock+0x54/0xdc

 stack backtrace:
 CPU: 10 PID: 1873 Comm: dynamic_debug01 Tainted: G        W        4.15.0+ #1
 Hardware name: GIGABYTE R120-T34-00/MT30-GS2-00, BIOS T48 10/02/2017
 Call trace:
  dump_backtrace+0x0/0x188
  show_stack+0x24/0x2c
  dump_stack+0xa4/0xe0
  print_circular_bug.isra.31+0x29c/0x2b8
  check_prev_add.constprop.39+0x6c8/0x6dc
  validate_chain.isra.34+0x6e4/0xa20
  __lock_acquire+0x3b4/0x6e0
  lock_acquire+0xf4/0x2a8
  _raw_spin_lock_irqsave+0x58/0x70
  down_trylock+0x20/0x4c
  __down_trylock_console_sem+0x3c/0x9c
  console_trylock+0x20/0xb0
  vprintk_emit+0x254/0x390
  vprintk_default+0x58/0x90
  vprintk_func+0xbc/0x164
  printk+0x80/0xa0
  __dynamic_pr_debug+0x84/0xac
  gic_raise_softirq+0x184/0x18c
  smp_cross_call+0xac/0x218
  smp_send_reschedule+0x3c/0x48
  resched_curr+0x60/0x9c
  check_preempt_curr+0x70/0xdc
  wake_up_new_task+0x310/0x470
  _do_fork+0x188/0x78c
  SyS_clone+0x44/0x50
  __sys_trace_return+0x0/0x4
 GICv3: CPU0: ICC_SGI1R_EL1 12000

This could be fixed with printk_deferred() but that might lessen its
usefulness for debugging. So change it to pr_devel to keep it out of
production kernels. Developers working on gic-v3 can enable it as
needed in their kernels.

Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Sasha Levin <alexander.levin@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/irqchip/irq-gic-v3.c