git.ipfire.org Git - thirdparty/kernel/stable-queue.git/blob

   1 From 66ff0fe9e7bda8aec99985b24daad03652f7304e Mon Sep 17 00:00:00 2001
   2 From: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
   3 Date: Tue, 16 Apr 2013 14:08:50 -0400
   4 Subject: xen/smp/spinlock: Fix leakage of the spinlock interrupt line for every CPU online/offline
   5
   6 From: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
   7
   8 commit 66ff0fe9e7bda8aec99985b24daad03652f7304e upstream.
   9
  10 While we don't use the spinlock interrupt line (see for details
  11 commit f10cd522c5fbfec9ae3cc01967868c9c2401ed23 -
  12 xen: disable PV spinlocks on HVM) - we should still do the proper
  13 init / deinit sequence. We did not do that correctly and for the
  14 CPU init for PVHVM guest we would allocate an interrupt line - but
  15 failed to deallocate the old interrupt line.
  16
  17 This resulted in leakage of an irq_desc but more importantly this splat
  18 as we online an offlined CPU:
  19
  20 genirq: Flags mismatch irq 71. 0002cc20 (spinlock1) vs. 0002cc20 (spinlock1)
  21 Pid: 2542, comm: init.late Not tainted 3.9.0-rc6upstream #1
  22 Call Trace:
  23  [<ffffffff811156de>] __setup_irq+0x23e/0x4a0
  24  [<ffffffff81194191>] ? kmem_cache_alloc_trace+0x221/0x250
  25  [<ffffffff811161bb>] request_threaded_irq+0xfb/0x160
  26  [<ffffffff8104c6f0>] ? xen_spin_trylock+0x20/0x20
  27  [<ffffffff813a8423>] bind_ipi_to_irqhandler+0xa3/0x160
  28  [<ffffffff81303758>] ? kasprintf+0x38/0x40
  29  [<ffffffff8104c6f0>] ? xen_spin_trylock+0x20/0x20
  30  [<ffffffff810cad35>] ? update_max_interval+0x15/0x40
  31  [<ffffffff816605db>] xen_init_lock_cpu+0x3c/0x78
  32  [<ffffffff81660029>] xen_hvm_cpu_notify+0x29/0x33
  33  [<ffffffff81676bdd>] notifier_call_chain+0x4d/0x70
  34  [<ffffffff810bb2a9>] __raw_notifier_call_chain+0x9/0x10
  35  [<ffffffff8109402b>] __cpu_notify+0x1b/0x30
  36  [<ffffffff8166834a>] _cpu_up+0xa0/0x14b
  37  [<ffffffff816684ce>] cpu_up+0xd9/0xec
  38  [<ffffffff8165f754>] store_online+0x94/0xd0
  39  [<ffffffff8141d15b>] dev_attr_store+0x1b/0x20
  40  [<ffffffff81218f44>] sysfs_write_file+0xf4/0x170
  41  [<ffffffff811a2864>] vfs_write+0xb4/0x130
  42  [<ffffffff811a302a>] sys_write+0x5a/0xa0
  43  [<ffffffff8167ada9>] system_call_fastpath+0x16/0x1b
  44 cpu 1 spinlock event irq -16
  45 smpboot: Booting Node 0 Processor 1 APIC 0x2
  46
  47 And if one looks at the /proc/interrupts right after
  48 offlining (CPU1):
  49
  50   70:          0          0  xen-percpu-ipi       spinlock0
  51   71:          0          0  xen-percpu-ipi       spinlock1
  52   77:          0          0  xen-percpu-ipi       spinlock2
  53
  54 There is the oddity of the 'spinlock1' still being present.
  55
  56 Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
  57 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  58
  59 ---
  60  arch/x86/xen/smp.c |    1 +
  61  1 file changed, 1 insertion(+)
  62
  63 --- a/arch/x86/xen/smp.c
  64 +++ b/arch/x86/xen/smp.c
  65 @@ -662,6 +662,7 @@ static void xen_hvm_cpu_die(unsigned int
  66         unbind_from_irqhandler(per_cpu(xen_debug_irq, cpu), NULL);
  67         unbind_from_irqhandler(per_cpu(xen_callfuncsingle_irq, cpu), NULL);
  68         unbind_from_irqhandler(per_cpu(xen_irq_work, cpu), NULL);
  69 +       xen_uninit_lock_cpu(cpu);
  70         xen_teardown_timer(cpu);
  71         native_cpu_die(cpu);
  72  }