Mempolicy: Use MPOL_PREFERRED for system-wide default policy

PATCH Randy Dunlap 2 participants 3 messages Apr 14, 2008 3 messages not in the archive Memory management

Share & tools

A message that is not in the archive (sent off-list or not yet received): 20080404145944.5442.2684.sendpatchset@localhost

A message that is not in the archive (sent off-list or not yet received): 20080404150040.5442.49132.sendpatchset@localhost

Randy Dunlap

author

2008-04-14 17:23 UTC #2 lore raw

Quoted 17 lines · On Fri, 04 Apr 2008 11:00:40 -0400 Lee Schermerhorn wrote:

PATCH 09/15 Mempolicy: Use MPOL_PREFERRED for system-wide default policy
Against: 2.6.25-rc8-mm1
Signed-off-by Lee Schermerhorn <[email protected]>
 Documentation/vm/numa_memory_policy.txt |   52 ++++++++----------------
 mm/mempolicy.c                          |   68 +++++++++++++++++++-------------
 2 files changed, 59 insertions(+), 61 deletions(-)
Index: linux-2.6.25-rc8-mm1/Documentation/vm/numa_memory_policy.txt
===================================================================
Documentation/vm/numa_memory_policy.txt +0 -0
@@ -187,19 +170,18 @@ Components of Memory Policies
 

Documentation/vm/numa_memory_policy.txt Components of Memory Policies · line 176

Quoted 10 lines

Documentation/vm/numa_memory_policy.txt +3 -4

 	MPOL_PREFERRED:  This mode specifies that the allocation should be
 	attempted from the single node specified in the policy.  If that
-	allocation fails, the kernel will search other nodes, exactly as
-	it would for a local allocation that started at the preferred node
-	in increasing distance from the preferred node.  "Local" allocation
-	policy can be viewed as a Preferred policy that starts at the node
+	allocation fails, the kernel will search other nodes, in order of
+	increasing distance from the preferred node based on information
+	provided by the platform firmware.
 	containing the cpu where the allocation takes place.

Something in the patch lines above seems to be foobarred. I.e., the sentences/lines don't flow correctly.

Documentation/vm/numa_memory_policy.txt Components of Memory Policies · line 177

Quoted 1 line

Documentation/vm/numa_memory_policy.txt +0 -0
 	    Internally, the Preferred policy uses a single node--the

---
~Randy

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

Lee Schermerhorn

2008-04-14 20:05 UTC #3 lore raw

Quoted 19 lines · Randy Dunlap

Quoted 17 lines · On Fri, 04 Apr 2008 11:00:40 -0400 Lee Schermerhorn wrote:
PATCH 09/15 Mempolicy: Use MPOL_PREFERRED for system-wide default policy
Against: 2.6.25-rc8-mm1
Signed-off-by Lee Schermerhorn <[email protected]>
 Documentation/vm/numa_memory_policy.txt |   52 ++++++++----------------
 mm/mempolicy.c                          |   68 +++++++++++++++++++-------------
 2 files changed, 59 insertions(+), 61 deletions(-)
Index: linux-2.6.25-rc8-mm1/Documentation/vm/numa_memory_policy.txt
===================================================================
Documentation/vm/numa_memory_policy.txt +0 -0
@@ -187,19 +170,18 @@ Components of Memory Policies
 

Quoted 13 lines

Documentation/vm/numa_memory_policy.txt Components of Memory Policies · line 176

Quoted 10 lines

Documentation/vm/numa_memory_policy.txt +3 -4

 	MPOL_PREFERRED:  This mode specifies that the allocation should be
 	attempted from the single node specified in the policy.  If that
-	allocation fails, the kernel will search other nodes, exactly as
-	it would for a local allocation that started at the preferred node
-	in increasing distance from the preferred node.  "Local" allocation
-	policy can be viewed as a Preferred policy that starts at the node
+	allocation fails, the kernel will search other nodes, in order of
+	increasing distance from the preferred node based on information
+	provided by the platform firmware.
 	containing the cpu where the allocation takes place.

Something in the patch lines above seems to be foobarred. I.e., the sentences/lines don't flow correctly.

Looks like I forgot to delete the line starting with 'containing ...'. I'll fix it, along with the 'expecially' in patch 8. [That's not a word, huh? :)]

Lee

A message that is not in the archive (sent off-list or not yet received): 20080404150034.5442.92020.sendpatchset@localhost

Randy Dunlap

author

2008-04-14 17:32 UTC #5 lore raw

Re: [PATCH 8/15] Mempolicy: Rework mempolicy Reference Counting [yet again]

Quoted 13 lines · On Fri, 04 Apr 2008 11:00:34 -0400 Lee Schermerhorn wrote:

PATCH 08/15 Mem Policy: rework mempolicy reference counting [yet again]
Against: 2.6.25-rc8-mm1
 Documentation/vm/numa_memory_policy.txt |   68 ++++++++++++++
Index: linux-2.6.25-rc8-mm1/Documentation/vm/numa_memory_policy.txt
===================================================================
Documentation/vm/numa_memory_policy.txt +0 -0
@@ -311,6 +311,74 @@ Components of Memory Policies
 	    MPOL_PREFERRED policies that were created with an empty nodemask
 	    (local allocation).

...

Documentation/vm/numa_memory_policy.txt

Quoted 3 lines

Documentation/vm/numa_memory_policy.txt +3 -0

+   Because of this extra reference counting, and because we must lookup
+   shared policies in a tree structure under spinlock, shared policies are
+   more expensive to use in the page allocation path.  This is expecially

especially

Documentation/vm/numa_memory_policy.txt

Quoted 9 lines

Documentation/vm/numa_memory_policy.txt +6 -0

+   true for shared policies on shared memory regions shared by tasks running
+   on different NUMA nodes.  This extra overhead can be avoided by always
+   falling back to task or system default policy for shared memory regions,
+   or by prefaulting the entire shared memory region into memory and locking
+   it down.  However, this might not be appropriate for all applications.
+
 MEMORY POLICY APIs
 
 Linux supports 3 system calls for controlling memory policy.  These APIS

---
~Randy