Skip to contentSkip to conversation

ehea: Fix memory hotplug oops

PATCH Applied Anton Blanchard 3 participants 3 messages May 11, 2011 – May 12, 2011

Subsystems (2)
Networking Network drivers
lore
Share & tools
lore

Anton Blanchard

author
#0 lore raw

PATCH 1/1 ehea: Fix memory hotplug oops

The ehea driver oopses during memory hotplug if the ports are not up. A simple testcase:

# ifconfig ethX down
# echo offline > /sys/devices/system/memory/memory32/state
Oops: Kernel access of bad area, sig: 11 [#1]

last sysfs file: /sys/devices/system/memory/memory32/state
REGS: c000000709393110 TRAP: 0300 Not tainted (2.6.39-rc2-01385-g7ef73bc-dirty)
DAR: 0000000000000000, DSISR: 40000000
...
NIP [c000000000067c98] .__wake_up_common+0x48/0xf0 LR [c00000000006d034] .__wake_up+0x54/0x90

Call Trace:
[c00000000006d034] .__wake_up+0x54/0x90
[d000000006bb6270] .ehea_rereg_mrs+0x140/0x730 [ehea]
[d000000006bb69c4] .ehea_mem_notifier+0x164/0x170 [ehea]
[c0000000006fc8a8] .notifier_call_chain+0x78/0xf0
[c0000000000b3d70] .__blocking_notifier_call_chain+0x70/0xb0
[c000000000458d78] .memory_notify+0x28/0x40
[c0000000001871d8] .remove_memory+0x208/0x6d0
[c000000000458264] .memory_section_action+0x94/0x140
[c0000000004583ec] .memory_block_change_state+0xdc/0x1d0
[c0000000004585cc] .store_mem_state+0xec/0x160
[c00000000044768c] .sysdev_store+0x3c/0x50
[c00000000020b48c] .sysfs_write_file+0xec/0x1f0
[c00000000018f86c] .vfs_write+0xec/0x1e0
[c00000000018fa88] .SyS_write+0x58/0xd0

To fix this, initialise the waitqueues during port probe instead of port open.

Signed-off-by Anton Blanchard <[email protected]>
Cc [email protected]

---

drivers/net/ehea/ehea_main.c +3 -3
Index: linux-net/drivers/net/ehea/ehea_main.c
===================================================================
@@ -2688,9 +2688,6 @@ static int ehea_open(struct net_device *
 		netif_start_queue(dev);
 	}
 
-	init_waitqueue_head(&port->swqe_avail_wq);
-	init_waitqueue_head(&port->restart_wq);
-
 	mutex_unlock(&port->port_lock);
 
 	return ret;

@@ -3278,6 +3275,9 @@ struct ehea_port *ehea_setup_single_port
 
 	INIT_WORK(&port->reset_task, ehea_reset_port);
 
+	init_waitqueue_head(&port->swqe_avail_wq);
+	init_waitqueue_head(&port->restart_wq);
+
 	ret = register_netdev(dev);
 	if (ret) {
 		pr_err("register_netdev failed. ret=%d\n", ret);
Acked-by on v1 1/1
#1 lore raw
Quoted 10 lines · Anton Blanchard

The ehea driver oopses during memory hotplug if the ports are not up. A simple testcase:

 # ifconfig ethX down
 # echo offline > /sys/devices/system/memory/memory32/state
 Oops: Kernel access of bad area, sig: 11 [#1]
 last sysfs file: /sys/devices/system/memory/memory32/state
 REGS: c000000709393110 TRAP: 0300 Not tainted

(2.6.39-rc2-01385-g7ef73bc-dirty)

Quoted 20 lines

DAR: 0000000000000000, DSISR: 40000000
...
NIP [c000000000067c98] .__wake_up_common+0x48/0xf0 LR [c00000000006d034] .__wake_up+0x54/0x90

 Call Trace:
 [c00000000006d034] .__wake_up+0x54/0x90
 [d000000006bb6270] .ehea_rereg_mrs+0x140/0x730 [ehea]
 [d000000006bb69c4] .ehea_mem_notifier+0x164/0x170 [ehea]
 [c0000000006fc8a8] .notifier_call_chain+0x78/0xf0
 [c0000000000b3d70] .__blocking_notifier_call_chain+0x70/0xb0
 [c000000000458d78] .memory_notify+0x28/0x40
 [c0000000001871d8] .remove_memory+0x208/0x6d0
 [c000000000458264] .memory_section_action+0x94/0x140
 [c0000000004583ec] .memory_block_change_state+0xdc/0x1d0
 [c0000000004585cc] .store_mem_state+0xec/0x160
 [c00000000044768c] .sysdev_store+0x3c/0x50
 [c00000000020b48c] .sysfs_write_file+0xec/0x1f0
 [c00000000018f86c] .vfs_write+0xec/0x1e0
 [c00000000018fa88] .SyS_write+0x58/0xd0
Quoted 4 lines

To fix this, initialise the waitqueues during port probe instead of port open.

Signed-off-by Anton Blanchard <[email protected]>
Acked-by Breno Leitao <[email protected]>
maintainer Applied
#2 lore raw

From: Breno Leitao <[email protected]>
Date: Wed, 11 May 2011 09:46:51 -0300

Quoted 32 lines · Anton Blanchard
Quoted 10 lines

The ehea driver oopses during memory hotplug if the ports are not up. A simple testcase:

 # ifconfig ethX down
 # echo offline > /sys/devices/system/memory/memory32/state
 Oops: Kernel access of bad area, sig: 11 [#1]
 last sysfs file: /sys/devices/system/memory/memory32/state
 REGS: c000000709393110 TRAP: 0300 Not tainted

(2.6.39-rc2-01385-g7ef73bc-dirty)

Quoted 20 lines

DAR: 0000000000000000, DSISR: 40000000
...
NIP [c000000000067c98] .__wake_up_common+0x48/0xf0 LR [c00000000006d034] .__wake_up+0x54/0x90

 Call Trace:
 [c00000000006d034] .__wake_up+0x54/0x90
 [d000000006bb6270] .ehea_rereg_mrs+0x140/0x730 [ehea]
 [d000000006bb69c4] .ehea_mem_notifier+0x164/0x170 [ehea]
 [c0000000006fc8a8] .notifier_call_chain+0x78/0xf0
 [c0000000000b3d70] .__blocking_notifier_call_chain+0x70/0xb0
 [c000000000458d78] .memory_notify+0x28/0x40
 [c0000000001871d8] .remove_memory+0x208/0x6d0
 [c000000000458264] .memory_section_action+0x94/0x140
 [c0000000004583ec] .memory_block_change_state+0xdc/0x1d0
 [c0000000004585cc] .store_mem_state+0xec/0x160
 [c00000000044768c] .sysdev_store+0x3c/0x50
 [c00000000020b48c] .sysfs_write_file+0xec/0x1f0
 [c00000000018f86c] .vfs_write+0xec/0x1e0
 [c00000000018fa88] .SyS_write+0x58/0xd0
Quoted 5 lines
Quoted 4 lines

To fix this, initialise the waitqueues during port probe instead of port open.

Signed-off-by Anton Blanchard <[email protected]>
Acked-by Breno Leitao <[email protected]>

Applied to net-2.6, thanks.