Skip to contentSkip to conversation

pid: tighten pidmap spinlock section and clean up

PATCH In review André Goddard Rosa 4 participants 7 messages Nov 22, 2009 – Dec 4, 2009

lore
Share & tools
lore
#0 lore raw

cover pid: tighten pidmap spinlock section and clean up

Avoid calling kfree() from under pidmap spinlock and reduces code size by 16 bytes on gcc 4.4.1 on Core 2.

André Goddard Rosa (2):
pid: tighten pidmap spinlock critical section by removing kfree()
pid: reduce code size by using a pointer to iterate over array

 kernel/pid.c |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)
#1 lore raw

PATCH 1/2 pid: tighten pidmap spinlock critical section by removing kfree()

Avoid calling kfree() under pidmap spinlock, calling it afterwards.

Normally kfree() is very fast, but sometimes it can be slow, so avoid calling it under the spinlock if we can.

---
 kernel/pid.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
kernel/pid.c +4 -3
index d3f722d..55fd590 100644
@@ -141,11 +141,12 @@ static int alloc_pidmap(struct pid_namespace *pid_ns)
 			 * installing it:
 			 */
 			spin_lock_irq(&pidmap_lock);
-			if (map->page)
-				kfree(page);
-			else
+			if (!map->page) {
 				map->page = page;
+				page = NULL;
+			}
 			spin_unlock_irq(&pidmap_lock);
+			kfree(page);
 			if (unlikely(!map->page))
 				break;
 		}
Signature · 1 line
1.6.5.3.298.g39add
#6 lore raw

PATCH 2/2 pid: reduce code size by using a pointer to iterate over array

In reply to André Goddard Rosa

It decreases code size by 16 bytes on my gcc 4.4.1 on Core 2:

  text    data     bss     dec     hex filename
  4314    2216       8    6538    198a kernel/pid.o-BEFORE
  4298    2216       8    6522    197a kernel/pid.o-AFTER
---
 kernel/pid.c |    5 ++---
 1 files changed, 2 insertions(+), 3 deletions(-)
kernel/pid.c +2 -3
index 55fd590..2e17c9c 100644
@@ -269,12 +269,11 @@ struct pid *alloc_pid(struct pid_namespace *ns)
 	for (type = 0; type < PIDTYPE_MAX; ++type)
 		INIT_HLIST_HEAD(&pid->tasks[type]);
 
+	upid = pid->numbers + ns->level;
 	spin_lock_irq(&pidmap_lock);
-	for (i = ns->level; i >= 0; i--) {
-		upid = &pid->numbers[i];
+	for ( ; upid >= pid->numbers; --upid)
 		hlist_add_head_rcu(&upid->pid_chain,
 				&pid_hash[pid_hashfn(upid->nr, upid->ns)]);
-	}
 	spin_unlock_irq(&pidmap_lock);
 
 out:
Signature · 1 line
1.6.5.3.298.g39add
#2 lore raw

Re: [PATCH 1/2] pid: tighten pidmap spinlock critical section by removing kfree()

In reply to André Goddard Rosa

(Adding some CC's.)

Quoted 40 lines · André Goddard Rosa

Avoid calling kfree() under pidmap spinlock, calling it afterwards.

Normally kfree() is very fast, but sometimes it can be slow, so avoid calling it under the spinlock if we can.

---
 kernel/pid.c |    7 ++++---

 1 files changed, 4 insertions(+), 3 deletions(-)
kernel/pid.c +4 -4
index d3f722d..55fd590 100644
@@ -141,11 +141,12 @@ static int alloc_pidmap(struct pid_namespace *pid_ns)
                         * installing it:
                         */
                        spin_lock_irq(&pidmap_lock);
-                       if (map->page)
-                               kfree(page);
-                       else
+                       if (!map->page) {
                                map->page = page;
+                               page = NULL;
+                       }
                        spin_unlock_irq(&pidmap_lock);
+                       kfree(page);
                        if (unlikely(!map->page))
                                break;
                }
--

1.6.5.3.298.g39add

--
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/

#3 lore raw

Re: [PATCH 1/2] pid: tighten pidmap spinlock critical section by removing kfree()

Quoted 8 lines · Pekka Enberg

(Adding some CC's.)

Quoted 4 lines · André Goddard Rosa

Avoid calling kfree() under pidmap spinlock, calling it afterwards.

Normally kfree() is very fast, but sometimes it can be slow, so avoid calling it under the spinlock if we can.

kfree() is called when we race with another process which also finds map->page == NULL, allocs the new page and takes pidmap_lock before us. This is extremely unlikely case, right?

Quoted 5 lines
Quoted 5 lines
kernel/pid.c +0 -1
@@ -141,11 +141,12 @@ static int alloc_pidmap(struct pid_namespace *pid_ns)
                         * installing it:
                         */
                        spin_lock_irq(&pidmap_lock);
-                       if (map->page)
Quoted 8 lines

kernel/pid.c static int alloc_pidmap(struct pid_namespace *pid_ns) · line 149

Quoted 8 lines
kernel/pid.c +4 -2
-                               kfree(page);
-                       else
+                       if (!map->page) {
                                map->page = page;
+                               page = NULL;
+                       }
                        spin_unlock_irq(&pidmap_lock);
+                       kfree(page);

And this change pessimizes (a little bit) the likely case, when the race doesn't happen. And imho this change doesn't make the code more readable.

But this is subjective, and technically the patch is correct afaics.

Quoted 2 lines
Quoted 2 lines
kernel/pid.c +0 -0
                        if (unlikely(!map->page))

                        �

Hmm. Off-topic, but why alloc_pidmap() does not do this right after kzalloc() ?

Oleg.

#4 lore raw

Re: [PATCH 1/2] pid: tighten pidmap spinlock critical section by removing kfree()

Hi, Oleg!

Quoted 5 lines · Oleg Nesterov
Quoted 4 lines · Pekka Enberg

(Adding some CC's.)

On Sat, Nov 21, 2009 at 2:16 PM, André Goddard Rosa <[email protected]> wrote:

Quoted 8 lines
Quoted 4 lines
Quoted 4 lines

Avoid calling kfree() under pidmap spinlock, calling it afterwards.

Normally kfree() is very fast, but sometimes it can be slow, so avoid calling it under the spinlock if we can.

kfree() is called when we race with another process which also finds map->page == NULL, allocs the new page and takes pidmap_lock before us. This is extremely unlikely case, right?

Right, somehow.

Quoted 18 lines
Quoted 5 lines
Quoted 5 lines
kernel/pid.c +0 -1
@@ -141,11 +141,12 @@ static int alloc_pidmap(struct pid_namespace *pid_ns)
                         * installing it:
                         */
                        spin_lock_irq(&pidmap_lock);
-                       if (map->page)
Quoted 8 lines

kernel/pid.c static int alloc_pidmap(struct pid_namespace *pid_ns) · line 149

Quoted 8 lines
kernel/pid.c +4 -2
-                               kfree(page);
-                       else
+                       if (!map->page) {
                                map->page = page;
+                               page = NULL;
+                       }
                        spin_unlock_irq(&pidmap_lock);
+                       kfree(page);

And this change pessimizes (a little bit) the likely case, when the race doesn't happen. And imho this change doesn't make the code more readable.

Quoted 2 lines

But this is subjective, and technically the patch is correct afaics.

It does not affect the likely case which happens when the pidmap is already allocated.

In the unlikely case where the pidmap must be allocated, if we think that we could have
let's say 8 processes contending for that spinlock, while one process got it first and allocated
the page, having the kfree() out of the spinlock would make those other 7 processes doing
useful work (performing the release of the page) before, because it would avoid all of them
spinning around waiting until the all the others also free their allocated pages.

Quoted 5 lines
Quoted 2 lines
Quoted 2 lines
kernel/pid.c +0 -0
                        if (unlikely(!map->page))

                        �

Hmm. Off-topic, but why alloc_pidmap() does not do this right after kzalloc() ?

Hmm... I would say that it's an optimistic best effort. We avoid failing right away
hoping that another process (racing) had success allocating the page. That is unlikely! :)

Thank you,
André

#5 lore raw

Re: [PATCH 1/2][resend] pid: tighten pidmap spinlock critical section by removing kfree()

In reply to André Goddard Rosa
Quoted 21 lines · On Thu, 3 Dec 2009 12:53:37 -0200 Andr__ Goddard Rosa…

Avoid calling kfree() under pidmap spinlock, calling it afterwards.

Normally kfree() is fast, but sometimes it can be slow, so avoid calling it under the spinlock if we can do it.

---
 kernel/pid.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
kernel/pid.c +0 -2
index d3f722d..55fd590 100644
@@ -141,11 +141,12 @@ static int alloc_pidmap(struct pid_namespace *pid_ns)
 			 * installing it:
 			 */
 			spin_lock_irq(&pidmap_lock);
-			if (map->page)
-				kfree(page);

kernel/pid.c static int alloc_pidmap(struct pid_namespace *pid_ns) · line 152

Quoted 10 lines
kernel/pid.c +4 -1
-			else
+			if (!map->page) {
 				map->page = page;
+				page = NULL;
+			}
 			spin_unlock_irq(&pidmap_lock);
+			kfree(page);
 			if (unlikely(!map->page))
 				break;
 		}

um, OK, but the chances of that kfree() actually being executed are very small.