releases/3.4.28/pci-pciehp-use-per-slot-workqueues-to-avoid-deadlock.patch

   1 From c2be6f93b383c873a4f9d521afa49b1b67d06085 Mon Sep 17 00:00:00 2001
   2 From: Yijing Wang <wangyijing@huawei.com>
   3 Date: Fri, 11 Jan 2013 10:15:54 +0800
   4 Subject: PCI: pciehp: Use per-slot workqueues to avoid deadlock
   5
   6 From: Yijing Wang <wangyijing@huawei.com>
   7
   8 commit c2be6f93b383c873a4f9d521afa49b1b67d06085 upstream.
   9
  10 When we have a hotplug-capable PCIe port with a second hotplug-capable
  11 PCIe port below it, removing the device below the upstream port causes
  12 a deadlock.
  13
  14 The deadlock happens because we use the pciehp_wq workqueue to run
  15 pciehp_power_thread(), which uses pciehp_disable_slot() to remove devices
  16 below the upstream port.  When we remove the downstream PCIe port, we call
  17 pciehp_remove(), the pciehp driver's .remove() method.  That calls
  18 flush_workqueue(pciehp_wq), which deadlocks because the
  19 pciehp_power_thread() work item is still running.
  20
  21 This patch avoids the deadlock by creating a workqueue for every PCIe port
  22 and removing the single shared workqueue.
  23
  24 Here's the call path that leads to the deadlock:
  25
  26   pciehp_queue_pushbutton_work
  27     queue_work(pciehp_wq)                   # queue pciehp_power_thread
  28     ...
  29
  30   pciehp_power_thread
  31     pciehp_disable_slot
  32       remove_board
  33         pciehp_unconfigure_device
  34           pci_stop_and_remove_bus_device
  35             ...
  36               pciehp_remove                 # pciehp driver .remove method
  37                 pciehp_release_ctrl
  38                   pcie_cleanup_slot
  39                     flush_workqueue(pciehp_wq)
  40
  41 This is fairly urgent because it can be caused by simply unplugging a
  42 Thunderbolt adapter, as reported by Daniel below.
  43
  44 [bhelgaas: changelog]
  45 Reference: http://lkml.kernel.org/r/CAMVG2ssiRgcTD1bej2tkUUfsWmpL5eNtPcNif9va2-Gzb2u8nQ@mail.gmail.com
  46 Reported-and-tested-by: Daniel J Blueman <daniel@quora.org>
  47 Reviewed-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
  48 Signed-off-by: Yijing Wang <wangyijing@huawei.com>
  49 Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
  50 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  51
  52 ---
  53  drivers/pci/hotplug/pciehp.h      |    2 +-
  54  drivers/pci/hotplug/pciehp_core.c |   11 ++---------
  55  drivers/pci/hotplug/pciehp_ctrl.c |    8 ++++----
  56  drivers/pci/hotplug/pciehp_hpc.c  |   11 ++++++++++-
  57  4 files changed, 17 insertions(+), 15 deletions(-)
  58
  59 --- a/drivers/pci/hotplug/pciehp.h
  60 +++ b/drivers/pci/hotplug/pciehp.h
  61 @@ -44,7 +44,6 @@ extern bool pciehp_poll_mode;
  62  extern int pciehp_poll_time;
  63  extern bool pciehp_debug;
  64  extern bool pciehp_force;
  65 -extern struct workqueue_struct *pciehp_wq;
  66
  67  #define dbg(format, arg...)                                            \
  68  do {                                                                   \
  69 @@ -78,6 +77,7 @@ struct slot {
  70         struct hotplug_slot *hotplug_slot;
  71         struct delayed_work work;       /* work for button event */
  72         struct mutex lock;
  73 +       struct workqueue_struct *wq;
  74  };
  75
  76  struct event_info {
  77 --- a/drivers/pci/hotplug/pciehp_core.c
  78 +++ b/drivers/pci/hotplug/pciehp_core.c
  79 @@ -42,7 +42,6 @@ bool pciehp_debug;
  80  bool pciehp_poll_mode;
  81  int pciehp_poll_time;
  82  bool pciehp_force;
  83 -struct workqueue_struct *pciehp_wq;
  84
  85  #define DRIVER_VERSION "0.4"
  86  #define DRIVER_AUTHOR  "Dan Zink <dan.zink@compaq.com>, Greg Kroah-Hartman <greg@kroah.com>, Dely Sy <dely.l.sy@intel.com>"
  87 @@ -340,18 +339,13 @@ static int __init pcied_init(void)
  88  {
  89         int retval = 0;
  90
  91 -       pciehp_wq = alloc_workqueue("pciehp", 0, 0);
  92 -       if (!pciehp_wq)
  93 -               return -ENOMEM;
  94 -
  95         pciehp_firmware_init();
  96         retval = pcie_port_service_register(&hpdriver_portdrv);
  97         dbg("pcie_port_service_register = %d\n", retval);
  98         info(DRIVER_DESC " version: " DRIVER_VERSION "\n");
  99 -       if (retval) {
 100 -               destroy_workqueue(pciehp_wq);
 101 +       if (retval)
 102                 dbg("Failure to register service\n");
 103 -       }
 104 +
 105         return retval;
 106  }
 107
 108 @@ -359,7 +353,6 @@ static void __exit pcied_cleanup(void)
 109  {
 110         dbg("unload_pciehpd()\n");
 111         pcie_port_service_unregister(&hpdriver_portdrv);
 112 -       destroy_workqueue(pciehp_wq);
 113         info(DRIVER_DESC " version: " DRIVER_VERSION " unloaded\n");
 114  }
 115
 116 --- a/drivers/pci/hotplug/pciehp_ctrl.c
 117 +++ b/drivers/pci/hotplug/pciehp_ctrl.c
 118 @@ -49,7 +49,7 @@ static int queue_interrupt_event(struct
 119         info->p_slot = p_slot;
 120         INIT_WORK(&info->work, interrupt_event_handler);
 121
 122 -       queue_work(pciehp_wq, &info->work);
 123 +       queue_work(p_slot->wq, &info->work);
 124
 125         return 0;
 126  }
 127 @@ -344,7 +344,7 @@ void pciehp_queue_pushbutton_work(struct
 128                 kfree(info);
 129                 goto out;
 130         }
 131 -       queue_work(pciehp_wq, &info->work);
 132 +       queue_work(p_slot->wq, &info->work);
 133   out:
 134         mutex_unlock(&p_slot->lock);
 135  }
 136 @@ -377,7 +377,7 @@ static void handle_button_press_event(st
 137                 if (ATTN_LED(ctrl))
 138                         pciehp_set_attention_status(p_slot, 0);
 139
 140 -               queue_delayed_work(pciehp_wq, &p_slot->work, 5*HZ);
 141 +               queue_delayed_work(p_slot->wq, &p_slot->work, 5*HZ);
 142                 break;
 143         case BLINKINGOFF_STATE:
 144         case BLINKINGON_STATE:
 145 @@ -439,7 +439,7 @@ static void handle_surprise_event(struct
 146         else
 147                 p_slot->state = POWERON_STATE;
 148
 149 -       queue_work(pciehp_wq, &info->work);
 150 +       queue_work(p_slot->wq, &info->work);
 151  }
 152
 153  static void interrupt_event_handler(struct work_struct *work)
 154 --- a/drivers/pci/hotplug/pciehp_hpc.c
 155 +++ b/drivers/pci/hotplug/pciehp_hpc.c
 156 @@ -874,23 +874,32 @@ static void pcie_shutdown_notification(s
 157  static int pcie_init_slot(struct controller *ctrl)
 158  {
 159         struct slot *slot;
 160 +       char name[32];
 161
 162         slot = kzalloc(sizeof(*slot), GFP_KERNEL);
 163         if (!slot)
 164                 return -ENOMEM;
 165
 166 +       snprintf(name, sizeof(name), "pciehp-%u", PSN(ctrl));
 167 +       slot->wq = alloc_workqueue(name, 0, 0);
 168 +       if (!slot->wq)
 169 +               goto abort;
 170 +
 171         slot->ctrl = ctrl;
 172         mutex_init(&slot->lock);
 173         INIT_DELAYED_WORK(&slot->work, pciehp_queue_pushbutton_work);
 174         ctrl->slot = slot;
 175         return 0;
 176 +abort:
 177 +       kfree(slot);
 178 +       return -ENOMEM;
 179  }
 180
 181  static void pcie_cleanup_slot(struct controller *ctrl)
 182  {
 183         struct slot *slot = ctrl->slot;
 184         cancel_delayed_work(&slot->work);
 185 -       flush_workqueue(pciehp_wq);
 186 +       destroy_workqueue(slot->wq);
 187         kfree(slot);
 188  }
 189