2010-06-15 15:35:53

by Jan-Bernd Themann

Subject: [PATCH 2/2] ehea: Fix kernel deadlock in DLPAR-mem processing

Port reset operations and memory add/remove operations need to
be serialized to avoid a kernel deadlock. The deadlock is caused
by calling the napi_disable() function twice.
Therefore we have to employ the dlpar_mem_lock in the ehea_reset_port
function as well


Signed-off-by: Jan-Bernd Themann <[email protected]>

---
created against kernel-2.6.35-rc3

drivers/net/ehea/ehea.h | 2 +-
drivers/net/ehea/ehea_main.c | 8 +++-----
2 files changed, 4 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ehea/ehea.h b/drivers/net/ehea/ehea.h
index 0630980..0060e42 100644
--- a/drivers/net/ehea/ehea.h
+++ b/drivers/net/ehea/ehea.h
@@ -40,7 +40,7 @@
#include <asm/io.h>

#define DRV_NAME "ehea"
-#define DRV_VERSION "EHEA_0103"
+#define DRV_VERSION "EHEA_0105"

/* eHEA capability flags */
#define DLPAR_PORT_ADD_REM 1
diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c
index fd890fa..8b92acb 100644
--- a/drivers/net/ehea/ehea_main.c
+++ b/drivers/net/ehea/ehea_main.c
@@ -2860,6 +2860,7 @@ static void ehea_reset_port(struct work_struct *work)
container_of(work, struct ehea_port, reset_task);
struct net_device *dev = port->netdev;

+ mutex_lock(&dlpar_mem_lock);
port->resets++;
mutex_lock(&port->port_lock);
netif_stop_queue(dev);
@@ -2882,6 +2883,7 @@ static void ehea_reset_port(struct work_struct *work)
netif_wake_queue(dev);
out:
mutex_unlock(&port->port_lock);
+ mutex_unlock(&dlpar_mem_lock);
}

static void ehea_rereg_mrs(struct work_struct *work)
@@ -3543,10 +3545,7 @@ static int ehea_mem_notifier(struct notifier_block *nb,
int ret = NOTIFY_BAD;
struct memory_notify *arg = data;

- if (!mutex_trylock(&dlpar_mem_lock)) {
- ehea_info("ehea_mem_notifier must not be called parallelized");
- goto out;
- }
+ mutex_lock(&dlpar_mem_lock);

switch (action) {
case MEM_CANCEL_OFFLINE:
@@ -3575,7 +3574,6 @@ static int ehea_mem_notifier(struct notifier_block *nb,

out_unlock:
mutex_unlock(&dlpar_mem_lock);
-out:
return ret;
}

--
1.7.0

2010-06-17 01:05:38

by David Miller

Subject: Re: [PATCH 2/2] ehea: Fix kernel deadlock in DLPAR-mem processing

From: "Jan-Bernd Themann" <[email protected]>
Date: Tue, 15 Jun 2010 17:35:42 +0200

> Port reset operations and memory add/remove operations need to
> be serialized to avoid a kernel deadlock. The deadlock is caused
> by calling the napi_disable() function twice.
> Therefore we have to employ the dlpar_mem_lock in the ehea_reset_port
> function as well
>
>
> Signed-off-by: Jan-Bernd Themann <[email protected]>

Applied.