Skip to contentSkip to conversation

omap: resource: Make resource_refresh() thread safe.

PATCH Applied Mike Chan 3 participants 5 messages Sep 3, 2009 – Sep 11, 2009

Subsystems (2)
lore
Share & tools
lore

Mike Chan

author
#0 lore raw

PATCH 1/1 omap: resource: Make resource_refresh() thread safe.

Need to lock the res_mutex when traversing the res_list.

---
 arch/arm/plat-omap/resource.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
arch/arm/plat-omap/resource.c +2 -0
index 25072cd..4631912 100644
@@ -234,11 +234,13 @@ int resource_refresh(void)
 	struct shared_resource *resp = NULL;
 	int ret = 0;
 
+	down(&res_mutex);
 	list_for_each_entry(resp, &res_list, node) {
 		ret = update_resource_level(resp);
 		if (ret)
 			break;
 	}
+	up(&res_mutex);
 	return ret;
 }
 
Signature · 1 line
1.5.4.5
Applied
#1 lore raw
Quoted 3 lines · Mike Chan

Need to lock the res_mutex when traversing the res_list.

Looks good, thanks.

Pushed to PM branch and pm-2.6.29.

Kevin

Quoted 24 lines
---
 arch/arm/plat-omap/resource.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
arch/arm/plat-omap/resource.c +2 -0
index 25072cd..4631912 100644
@@ -234,11 +234,13 @@ int resource_refresh(void)
 	struct shared_resource *resp = NULL;
 	int ret = 0;
 
+	down(&res_mutex);
 	list_for_each_entry(resp, &res_list, node) {
 		ret = update_resource_level(resp);
 		if (ret)
 			break;
 	}
+	up(&res_mutex);
 	return ret;
 }
 
Signature · 1 line
1.5.4.5
#2 lore raw
Quoted 7 lines · Kevin Hilman
Quoted 3 lines · Mike Chan

Need to lock the res_mutex when traversing the res_list.

Looks good, thanks.

This patch causes a hang for me when transitioning to OFF mode. This was tested on the Android 2.6.29 tree and is 100% reproducible. The moment a user runs 'echo 1 > /sys/power/enable_off_mode' the board hangs without any further output.

Reverting the patch allows me to hit OFF mode again. I haven't yet tested this on vanilla 2.6.29 or latest L-O.

Mike

Quoted 33 lines

Pushed to PM branch and pm-2.6.29.

Kevin

Quoted 24 lines
---
 arch/arm/plat-omap/resource.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
arch/arm/plat-omap/resource.c +2 -0
index 25072cd..4631912 100644
@@ -234,11 +234,13 @@ int resource_refresh(void)
 	struct shared_resource *resp = NULL;
 	int ret = 0;
 
+	down(&res_mutex);
 	list_for_each_entry(resp, &res_list, node) {
 		ret = update_resource_level(resp);
 		if (ret)
 			break;
 	}
+	up(&res_mutex);
 	return ret;
 }
 
Signature · 1 line
1.5.4.5

--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#3 lore raw
Quoted 9 lines · Mike Turquette
Quoted 7 lines · Kevin Hilman
Quoted 3 lines · Mike Chan

Need to lock the res_mutex when traversing the res_list.

Looks good, thanks.

Quoted 8 lines

This patch causes a hang for me when transitioning to OFF mode. This was tested on the Android 2.6.29 tree and is 100% reproducible. The moment a user runs 'echo 1 > /sys/power/enable_off_mode' the board hangs without any further output.

Reverting the patch allows me to hit OFF mode again. I haven't yet tested this on vanilla 2.6.29 or latest L-O.

OK, reverting this in both PM branches. Looks like a deadlock to me, as update_resource_level() must cause a call to something else that takes the mutex.

Kevin

Quoted 32 lines
Quoted 32 lines

Pushed to PM branch and pm-2.6.29.

Kevin

Quoted 21 lines
---
 arch/arm/plat-omap/resource.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
arch/arm/plat-omap/resource.c +1 -0
index 25072cd..4631912 100644
@@ -234,11 +234,13 @@ int resource_refresh(void)
 	struct shared_resource *resp = NULL;
 	int ret = 0;
 +	down(&res_mutex);
 	list_for_each_entry(resp, &res_list, node) {
 		ret = update_resource_level(resp);
 		if (ret)
 			break;
 	}
+	up(&res_mutex);
 	return ret;
 }
 -- 
Quoted 1 line

1.5.4.5

--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

author
#4 lore raw
Quoted 16 lines · Kevin Hilman
Quoted 14 lines · Mike Turquette
Quoted 7 lines · Kevin Hilman
Quoted 3 lines · Mike Chan

Need to lock the res_mutex when traversing the res_list.

Looks good, thanks.

This patch causes a hang for me when transitioning to OFF mode.  This was tested on the Android 2.6.29 tree and is 100% reproducible.  The moment a user runs 'echo 1 > /sys/power/enable_off_mode' the board hangs without any further output.

Quoted 8 lines
Quoted 3 lines

Reverting the patch allows me to hit OFF mode again.  I haven't yet tested this on vanilla 2.6.29 or latest L-O.

OK, reverting this in both PM branches.  Looks like a deadlock to me, as update_resource_level() must cause a call to something else that takes the mutex.

Good catch, I see the deadlock are in the change_level function pointers. Where set_xxx will call resource_request() or resource_release() => resource_lookup() which grabs the list mutex we grabbed in resource_refresh().

Thoughts on changing the mutex to a RW semaphore?

-- Mike

Quoted 35 lines

Kevin

Quoted 32 lines
Quoted 32 lines

Pushed to PM branch and pm-2.6.29.

Kevin

Quoted 20 lines

---
 arch/arm/plat-omap/resource.c |    2 ++

 1 files changed, 2 insertions(+), 0 deletions(-)
arch/arm/plat-omap/resource.c +1 -0
index 25072cd..4631912 100644
@@ -234,11 +234,13 @@ int resource_refresh(void)
     struct shared_resource *resp = NULL;
     int ret = 0;
  +  down(&res_mutex);
     list_for_each_entry(resp, &res_list, node) {
             ret = update_resource_level(resp);
             if (ret)
                     break;
     }
+   up(&res_mutex);
     return ret;
  }
Quoted 2 lines

 --
1.5.4.5

--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html