IB/umad: Fix kernel crash while unloading ib_umad
authorYonatan Cohen <yonatanc@mellanox.com>
Wed, 12 Feb 2020 07:26:34 +0000 (09:26 +0200)
committerJason Gunthorpe <jgg@mellanox.com>
Thu, 13 Feb 2020 14:00:50 +0000 (10:00 -0400)
When disassociating a device from umad we must ensure that the sysfs
access is prevented before blocking the fops, otherwise assumptions in
syfs don't hold:

    CPU0                     CPU1
 ib_umad_kill_port()        ibdev_show()
    port->ib_dev = NULL
                                      dev_name(port->ib_dev)

The prior patch made an error in moving the device_destroy(), it should
have been split into device_del() (above) and put_device() (below). At
this point we already have the split, so move the device_del() back to its
original place.

  kernel stack
  PF: error_code(0x0000) - not-present page
  Oops: 0000 [#1] SMP DEBUG_PAGEALLOC PTI
  RIP: 0010:ibdev_show+0x18/0x50 [ib_umad]
  RSP: 0018:ffffc9000097fe40 EFLAGS: 00010282
  RAX: 0000000000000000 RBX: ffffffffa0441120 RCX: ffff8881df514000
  RDX: ffff8881df514000 RSI: ffffffffa0441120 RDI: ffff8881df1e8870
  RBP: ffffffff81caf000 R08: ffff8881df1e8870 R09: 0000000000000000
  R10: 0000000000001000 R11: 0000000000000003 R12: ffff88822f550b40
  R13: 0000000000000001 R14: ffffc9000097ff08 R15: ffff8882238bad58
  FS:  00007f1437ff3740(0000) GS:ffff888236940000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00000000000004e8 CR3: 00000001e0dfc001 CR4: 00000000001606e0
  Call Trace:
   dev_attr_show+0x15/0x50
   sysfs_kf_seq_show+0xb8/0x1a0
   seq_read+0x12d/0x350
   vfs_read+0x89/0x140
   ksys_read+0x55/0xd0
   do_syscall_64+0x55/0x1b0
   entry_SYSCALL_64_after_hwframe+0x44/0xa9:

Fixes: cf7ad3030271 ("IB/umad: Avoid destroying device while it is accessed")
Link: https://lore.kernel.org/r/20200212072635.682689-9-leon@kernel.org
Signed-off-by: Yonatan Cohen <yonatanc@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Reviewed-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
drivers/infiniband/core/user_mad.c

index d1407fa378e832bdfe45ba4c4b35645bd59b3a89..1235ffb2389b1c00feed521311fb54d019337c4a 100644 (file)
@@ -1312,6 +1312,9 @@ static void ib_umad_kill_port(struct ib_umad_port *port)
        struct ib_umad_file *file;
        int id;
 
+       cdev_device_del(&port->sm_cdev, &port->sm_dev);
+       cdev_device_del(&port->cdev, &port->dev);
+
        mutex_lock(&port->file_mutex);
 
        /* Mark ib_dev NULL and block ioctl or other file ops to progress
@@ -1331,8 +1334,6 @@ static void ib_umad_kill_port(struct ib_umad_port *port)
 
        mutex_unlock(&port->file_mutex);
 
-       cdev_device_del(&port->sm_cdev, &port->sm_dev);
-       cdev_device_del(&port->cdev, &port->dev);
        ida_free(&umad_ida, port->dev_num);
 
        /* balances device_initialize() */