net/mlx5: Check that driver was probed prior attaching the device
authorLeon Romanovsky <leonro@nvidia.com>
Sun, 21 Mar 2021 17:57:14 +0000 (19:57 +0200)
committerSaeed Mahameed <saeedm@nvidia.com>
Wed, 16 Jun 2021 22:36:35 +0000 (15:36 -0700)
The device can be requested to be attached despite being not probed.
This situation is possible if devlink reload races with module removal,
and the following kernel panic is an outcome of such race.

 mlx5_core 0000:00:09.0: firmware version: 4.7.9999
 mlx5_core 0000:00:09.0: 0.000 Gb/s available PCIe bandwidth (8.0 GT/s PCIe x255 link)
 BUG: unable to handle page fault for address: fffffffffffffff0
 #PF: supervisor read access in kernel mode
 #PF: error_code(0x0000) - not-present page
 PGD 3218067 P4D 3218067 PUD 321a067 PMD 0
 Oops: 0000 [#1] SMP KASAN NOPTI
 CPU: 7 PID: 250 Comm: devlink Not tainted 5.12.0-rc2+ #2836
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
 RIP: 0010:mlx5_attach_device+0x80/0x280 [mlx5_core]
 Code: f8 48 c1 e8 03 42 80 3c 38 00 0f 85 80 01 00 00 48 8b 45 68 48 8d 78 f0 48 89 fe 48 c1 ee 03 42 80 3c 3e 00 0f 85 70 01 00 00 <48> 8b 40 f0 48 85 c0 74 0d 48 89 ef ff d0 85 c0 0f 85 84 05 0e 00
 RSP: 0018:ffff8880129675f0 EFLAGS: 00010246
 RAX: 0000000000000000 RBX: 0000000000000001 RCX: ffffffff827407f1
 RDX: 1ffff110011336cf RSI: 1ffffffffffffffe RDI: fffffffffffffff0
 RBP: ffff888008e0c000 R08: 0000000000000008 R09: ffffffffa0662ee7
 R10: fffffbfff40cc5dc R11: 0000000000000000 R12: ffff88800ea002e0
 R13: ffffed1001d459f7 R14: ffffffffa05ef4f8 R15: dffffc0000000000
 FS:  00007f51dfeaf740(0000) GS:ffff88806d5c0000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: fffffffffffffff0 CR3: 000000000bc82006 CR4: 0000000000370ea0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  mlx5_load_one+0x117/0x1d0 [mlx5_core]
  devlink_reload+0x2d5/0x520
  ? devlink_remote_reload_actions_performed+0x30/0x30
  ? mutex_trylock+0x24b/0x2d0
  ? devlink_nl_cmd_reload+0x62b/0x1070
  devlink_nl_cmd_reload+0x66d/0x1070
  ? devlink_reload+0x520/0x520
  ? devlink_nl_pre_doit+0x64/0x4d0
  genl_family_rcv_msg_doit+0x1e9/0x2f0
  ? mutex_lock_io_nested+0x1130/0x1130
  ? genl_family_rcv_msg_attrs_parse.constprop.0+0x240/0x240
  ? security_capable+0x51/0x90
  genl_rcv_msg+0x27f/0x4a0
  ? genl_get_cmd+0x3c0/0x3c0
  ? lock_acquire+0x1a9/0x6d0
  ? devlink_reload+0x520/0x520
  ? lock_release+0x6c0/0x6c0
  netlink_rcv_skb+0x11d/0x340
  ? genl_get_cmd+0x3c0/0x3c0
  ? netlink_ack+0x9f0/0x9f0
  ? lock_release+0x1f9/0x6c0
  genl_rcv+0x24/0x40
  netlink_unicast+0x433/0x700
  ? netlink_attachskb+0x730/0x730
  ? _copy_from_iter_full+0x178/0x650
  ? __alloc_skb+0x113/0x2b0
  netlink_sendmsg+0x6f1/0xbd0
  ? netlink_unicast+0x700/0x700
  ? netlink_unicast+0x700/0x700
  sock_sendmsg+0xb0/0xe0
  __sys_sendto+0x193/0x240
  ? __x64_sys_getpeername+0xb0/0xb0
  ? copy_page_range+0x2300/0x2300
  ? __up_read+0x1a1/0x7b0
  ? do_user_addr_fault+0x219/0xdc0
  __x64_sys_sendto+0xdd/0x1b0
  ? syscall_enter_from_user_mode+0x1d/0x50
  do_syscall_64+0x2d/0x40
  entry_SYSCALL_64_after_hwframe+0x44/0xae
 RIP: 0033:0x7f51dffb514a
 Code: d8 64 89 02 48 c7 c0 ff ff ff ff eb b8 0f 1f 00 f3 0f 1e fa 41 89 ca 64 8b 04 25 18 00 00 00 85 c0 75 15 b8 2c 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 76 c3 0f 1f 44 00 00 55 48 83 ec 30 44 89 4c
 RSP: 002b:00007ffcaef22e78 EFLAGS: 00000246 ORIG_RAX: 000000000000002c
 RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 00007f51dffb514a
 RDX: 0000000000000030 RSI: 000055750daf2440 RDI: 0000000000000003
 RBP: 000055750daf2410 R08: 00007f51e0081200 R09: 000000000000000c
 R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
 Modules linked in: mlx5_core(-) ptp pps_core ib_ipoib rdma_ucm rdma_cm iw_cm ib_cm ib_umad ib_uverbs ib_core [last unloaded: mlx5_ib]
 CR2: fffffffffffffff0
 ---[ end trace 7789831bfe74fa42 ]---

Fixes: a925b5e309c9 ("net/mlx5: Register mlx5 devices to auxiliary virtual bus")
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
drivers/net/ethernet/mellanox/mlx5/core/dev.c

index a9166cd..8de118a 100644 (file)
@@ -320,6 +320,16 @@ int mlx5_attach_device(struct mlx5_core_dev *dev)
                        }
                } else {
                        adev = &priv->adev[i]->adev;
+
+                       /* Pay attention that this is not PCI driver that
+                        * mlx5_core_dev is connected, but auxiliary driver.
+                        *
+                        * Here we can race of module unload with devlink
+                        * reload, but we don't need to take extra lock because
+                        * we are holding global mlx5_intf_mutex.
+                        */
+                       if (!adev->dev.driver)
+                               continue;
                        adrv = to_auxiliary_drv(adev->dev.driver);
 
                        if (adrv->resume)
@@ -350,6 +360,10 @@ void mlx5_detach_device(struct mlx5_core_dev *dev)
                        continue;
 
                adev = &priv->adev[i]->adev;
+               /* Auxiliary driver was unbind manually through sysfs */
+               if (!adev->dev.driver)
+                       goto skip_suspend;
+
                adrv = to_auxiliary_drv(adev->dev.driver);
 
                if (adrv->suspend) {
@@ -357,6 +371,7 @@ void mlx5_detach_device(struct mlx5_core_dev *dev)
                        continue;
                }
 
+skip_suspend:
                del_adev(&priv->adev[i]->adev);
                priv->adev[i] = NULL;
        }