Skip to contentSkip to conversation

drm: Move two seq_printf's outside of locked mutex

PATCH In review Jonas Lundqvist 2 participants 4 messages Dec 30, 2014 – Dec 31, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm: Move two seq_printf's outside of locked mutex

In drm_info.c: drm_bufs_info() and drm_vm_info() two seq_printf() was done unnecessarily while locking a mutex. This patch flips the order of the print and lock/unlock where applicable.

---
 drivers/gpu/drm/drm_info.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/gpu/drm/drm_info.c +3 -4
index 51efebd..981afe7 100644
@@ -81,11 +81,10 @@ int drm_vm_info(struct seq_file *m, void *data)
 	   _DRM_SCATTER_GATHER and _DRM_CONSISTENT */
 	const char *types[] = { "FB", "REG", "SHM", "AGP", "SG", "PCI" };
 	const char *type;
-	int i;
+	int i = 0;
 
-	mutex_lock(&dev->struct_mutex);
 	seq_printf(m, "slot	 offset	      size type flags	 address mtrr\n\n");
-	i = 0;
+	mutex_lock(&dev->struct_mutex);
 	list_for_each_entry(r_list, &dev->maplist, head) {
 		map = r_list->map;
 		if (!map)
@@ -147,8 +146,8 @@ int drm_bufs_info(struct seq_file *m, void *data)
 			seq_printf(m, "\n");
 		seq_printf(m, " %d", dma->buflist[i]->list);
 	}
-	seq_printf(m, "\n");
 	mutex_unlock(&dev->struct_mutex);
+	seq_printf(m, "\n");
 	return 0;
 }
 
Signature · 1 line
2.1.4
#1 lore raw

Jonas,

Quoted 28 lines · Jonas Lundqvist

In drm_info.c: drm_bufs_info() and drm_vm_info() two seq_printf() was done unnecessarily while locking a mutex. This patch flips the order of the print and lock/unlock where applicable.

---
 drivers/gpu/drm/drm_info.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/gpu/drm/drm_info.c +2 -3
index 51efebd..981afe7 100644
@@ -81,11 +81,10 @@ int drm_vm_info(struct seq_file *m, void *data)
 	   _DRM_SCATTER_GATHER and _DRM_CONSISTENT */
 	const char *types[] = { "FB", "REG", "SHM", "AGP", "SG", "PCI" };
 	const char *type;
-	int i;
+	int i = 0;
 
-	mutex_lock(&dev->struct_mutex);
 	seq_printf(m, "slot	 offset	      size type flags	 address mtrr\n\n");
-	i = 0;
+	mutex_lock(&dev->struct_mutex);
 	list_for_each_entry(r_list, &dev->maplist, head) {
 		map = r_list->map;
 		if (!map)

drivers/gpu/drm/drm_info.c int drm_bufs_info(struct seq_file *m, void *data) · line 153

Quoted 9 lines
drivers/gpu/drm/drm_info.c +1 -1
@@ -147,8 +146,8 @@ int drm_bufs_info(struct seq_file *m, void *data)
 			seq_printf(m, "\n");
 		seq_printf(m, " %d", dma->buflist[i]->list);
 	}
-	seq_printf(m, "\n");
 	mutex_unlock(&dev->struct_mutex);
+	seq_printf(m, "\n");
 	return 0;
 }
 

You changed 'i' but you didn't explain in your log message why you did this.

Does this change really improve anything? It may work the same with the locks moved around. But if you look at the function as a whole, the locks encapsulate the body of this function nicely. I like the original design better.

Quoted 8 lines
Signature · 7 lines
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Signature · 1 line
- Jeremiah Mahler
#2 lore raw

Hi Jeremiah,

Quoted 1 line · Jeremiah Mahler

You changed 'i' but you didn't explain in your log message why you did this.

I can change the commit message to something more generic. "Move code outside of locked mutex" or similar.

Quoted 4 lines

Does this change really improve anything? It may work the same with the locks moved around. But if you look at the function as a whole, the locks encapsulate the body of this function nicely. I like the original design better.

The locking was already done this way, ie after the seq_printf, in the functions drm_clients_info() and drm_gem_name_info() in thr same file. So this change is really more of an alignment.

Best regards
Jonas

#3 lore raw

Jonas,

Quoted 8 lines · Jonas Lundqvist

Hi Jeremiah,

Quoted 1 line · Jeremiah Mahler

You changed 'i' but you didn't explain in your log message why you did this.

I can change the commit message to something more generic. "Move code outside of locked mutex" or similar.

That still doesn't explain why you changed the 'i' variable.

Quoted 9 lines
Quoted 4 lines

Does this change really improve anything? It may work the same with the locks moved around. But if you look at the function as a whole, the locks encapsulate the body of this function nicely. I like the original design better.

The locking was already done this way, ie after the seq_printf, in the functions drm_clients_info() and drm_gem_name_info() in thr same file. So this change is really more of an alignment.

Your right, those two have have the lock after the seq_printf. But the drm_bufs_info() function has its lock before the seq_printf. So before your change about half are one way and half are the other.

I am still not convinced that either of these ways is better or makes any difference whatsoever.

Quoted 3 lines

Best regards
Jonas

Signature · 1 line
- Jeremiah Mahler