summaryrefslogtreecommitdiffstats
path: root/Documentation
diff options
context:
space:
mode:
authorLee Schermerhorn <lee.schermerhorn@hp.com>2010-05-26 14:45:06 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2010-05-27 09:12:57 -0700
commitb9498bfe86530fd54fb855906383c0c905a52c80 (patch)
tree6171b659409b2a4f3e404878f0cd1010edb6a273 /Documentation
parent3dd6b5fb436083fdd8af7ad2baa07db19deb73b2 (diff)
numa: update Documentation/vm/numa, add memoryless node info
Kamezawa Hiroyuki requested documentation for the numa_mem_id() and slab related changes. He suggested Documentation/vm/numa for this documentation. Looking at this file, it seems to me to be hopelessly out of date relative to current Linux NUMA support. At the risk of going down a rathole, I have made an attempt to rewrite the doc at a slightly higher level [I think] and provide pointers to other in-tree documents and out-of-tree man pages that cover the details. Let the games begin. Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Tejun Heo <tj@kernel.org> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Christoph Lameter <cl@linux-foundation.org> Cc: Nick Piggin <npiggin@suse.de> Cc: David Rientjes <rientjes@google.com> Cc: Eric Whitney <eric.whitney@hp.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Pekka Enberg <penberg@cs.helsinki.fi> Cc: Randy Dunlap <randy.dunlap@oracle.com> Cc: <linux-arch@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'Documentation')
-rw-r--r--Documentation/vm/numa186
1 files changed, 147 insertions, 39 deletions
diff --git a/Documentation/vm/numa b/Documentation/vm/numa
index e93ad9425e2..a200a386429 100644
--- a/Documentation/vm/numa
+++ b/Documentation/vm/numa
@@ -1,41 +1,149 @@
Started Nov 1999 by Kanoj Sarcar <kanoj@sgi.com>
-The intent of this file is to have an uptodate, running commentary
-from different people about NUMA specific code in the Linux vm.
-
-What is NUMA? It is an architecture where the memory access times
-for different regions of memory from a given processor varies
-according to the "distance" of the memory region from the processor.
-Each region of memory to which access times are the same from any
-cpu, is called a node. On such architectures, it is beneficial if
-the kernel tries to minimize inter node communications. Schemes
-for this range from kernel text and read-only data replication
-across nodes, and trying to house all the data structures that
-key components of the kernel need on memory on that node.
-
-Currently, all the numa support is to provide efficient handling
-of widely discontiguous physical memory, so architectures which
-are not NUMA but can have huge holes in the physical address space
-can use the same code. All this code is bracketed by CONFIG_DISCONTIGMEM.
-
-The initial port includes NUMAizing the bootmem allocator code by
-encapsulating all the pieces of information into a bootmem_data_t
-structure. Node specific calls have been added to the allocator.
-In theory, any platform which uses the bootmem allocator should
-be able to put the bootmem and mem_map data structures anywhere
-it deems best.
-
-Each node's page allocation data structures have also been encapsulated
-into a pg_data_t. The bootmem_data_t is just one part of this. To
-make the code look uniform between NUMA and regular UMA platforms,
-UMA platforms have a statically allocated pg_data_t too (contig_page_data).
-For the sake of uniformity, the function num_online_nodes() is also defined
-for all platforms. As we run benchmarks, we might decide to NUMAize
-more variables like low_on_memory, nr_free_pages etc into the pg_data_t.
-
-The NUMA aware page allocation code currently tries to allocate pages
-from different nodes in a round robin manner. This will be changed to
-do concentratic circle search, starting from current node, once the
-NUMA port achieves more maturity. The call alloc_pages_node has been
-added, so that drivers can make the call and not worry about whether
-it is running on a NUMA or UMA platform.
+What is NUMA?
+
+This question can be answered from a couple of perspectives: the
+hardware view and the Linux software view.
+
+From the hardware perspective, a NUMA system is a computer platform that
+comprises multiple components or assemblies each of which may contain 0
+or more CPUs, local memory, and/or IO buses. For brevity and to
+disambiguate the hardware view of these physical components/assemblies
+from the software abstraction thereof, we'll call the components/assemblies
+'cells' in this document.
+
+Each of the 'cells' may be viewed as an SMP [symmetric multi-processor] subset
+of the system--although some components necessary for a stand-alone SMP system
+may not be populated on any given cell. The cells of the NUMA system are
+connected together with some sort of system interconnect--e.g., a crossbar or
+point-to-point link are common types of NUMA system interconnects. Both of
+these types of interconnects can be aggregated to create NUMA platforms with
+cells at multiple distances from other cells.
+
+For Linux, the NUMA platforms of interest are primarily what is known as Cache
+Coherent NUMA or ccNUMA systems. With ccNUMA systems, all memory is visible
+to and accessible from any CPU attached to any cell and cache coherency
+is handled in hardware by the processor caches and/or the system interconnect.
+
+Memory access time and effective memory bandwidth varies depending on how far
+away the cell containing the CPU or IO bus making the memory access is from the
+cell containing the target memory. For example, access to memory by CPUs
+attached to the same cell will experience faster access times and higher
+bandwidths than accesses to memory on other, remote cells. NUMA platforms
+can have cells at multiple remote distances from any given cell.
+
+Platform vendors don't build NUMA systems just to make software developers'
+lives interesting. Rather, this architecture is a means to provide scalable
+memory bandwidth. However, to achieve scalable memory bandwidth, system and
+application software must arrange for a large majority of the memory references
+[cache misses] to be to "local" memory--memory on the same cell, if any--or
+to the closest cell with memory.
+
+This leads to the Linux software view of a NUMA system:
+
+Linux divides the system's hardware resources into multiple software
+abstractions called "nodes". Linux maps the nodes onto the physical cells
+of the hardware platform, abstracting away some of the details for some
+architectures. As with physical cells, software nodes may contain 0 or more
+CPUs, memory and/or IO buses. And, again, memory accesses to memory on
+"closer" nodes--nodes that map to closer cells--will generally experience
+faster access times and higher effective bandwidth than accesses to more
+remote cells.
+
+For some architectures, such as x86, Linux will "hide" any node representing a
+physical cell that has no memory attached, and reassign any CPUs attached to
+that cell to a node representing a cell that does have memory. Thus, on
+these architectures, one cannot assume that all CPUs that Linux associates with
+a given node will see the same local memory access times and bandwidth.
+
+In addition, for some architectures, again x86 is an example, Linux supports
+the emulation of additional nodes. For NUMA emulation, linux will carve up
+the existing nodes--or the system memory for non-NUMA platforms--into multiple
+nodes. Each emulated node will manage a fraction of the underlying cells'
+physical memory. NUMA emluation is useful for testing NUMA kernel and
+application features on non-NUMA platforms, and as a sort of memory resource
+management mechanism when used together with cpusets.
+[see Documentation/cgroups/cpusets.txt]
+
+For each node with memory, Linux constructs an independent memory management
+subsystem, complete with its own free page lists, in-use page lists, usage
+statistics and locks to mediate access. In addition, Linux constructs for
+each memory zone [one or more of DMA, DMA32, NORMAL, HIGH_MEMORY, MOVABLE],
+an ordered "zonelist". A zonelist specifies the zones/nodes to visit when a
+selected zone/node cannot satisfy the allocation request. This situation,
+when a zone has no available memory to satisfy a request, is called
+"overflow" or "fallback".
+
+Because some nodes contain multiple zones containing different types of
+memory, Linux must decide whether to order the zonelists such that allocations
+fall back to the same zone type on a different node, or to a different zone
+type on the same node. This is an important consideration because some zones,
+such as DMA or DMA32, represent relatively scarce resources. Linux chooses
+a default zonelist order based on the sizes of the various zone types relative
+to the total memory of the node and the total memory of the system. The
+default zonelist order may be overridden using the numa_zonelist_order kernel
+boot parameter or sysctl. [see Documentation/kernel-parameters.txt and
+Documentation/sysctl/vm.txt]
+
+By default, Linux will attempt to satisfy memory allocation requests from the
+node to which the CPU that executes the request is assigned. Specifically,
+Linux will attempt to allocate from the first node in the appropriate zonelist
+for the node where the request originates. This is called "local allocation."
+If the "local" node cannot satisfy the request, the kernel will examine other
+nodes' zones in the selected zonelist looking for the first zone in the list
+that can satisfy the request.
+
+Local allocation will tend to keep subsequent access to the allocated memory
+"local" to the underlying physical resources and off the system interconnect--
+as long as the task on whose behalf the kernel allocated some memory does not
+later migrate away from that memory. The Linux scheduler is aware of the
+NUMA topology of the platform--embodied in the "scheduling domains" data
+structures [see Documentation/scheduler/sched-domains.txt]--and the scheduler
+attempts to minimize task migration to distant scheduling domains. However,
+the scheduler does not take a task's NUMA footprint into account directly.
+Thus, under sufficient imbalance, tasks can migrate between nodes, remote
+from their initial node and kernel data structures.
+
+System administrators and application designers can restrict a task's migration
+to improve NUMA locality using various CPU affinity command line interfaces,
+such as taskset(1) and numactl(1), and program interfaces such as
+sched_setaffinity(2). Further, one can modify the kernel's default local
+allocation behavior using Linux NUMA memory policy.
+[see Documentation/vm/numa_memory_policy.]
+
+System administrators can restrict the CPUs and nodes' memories that a non-
+privileged user can specify in the scheduling or NUMA commands and functions
+using control groups and CPUsets. [see Documentation/cgroups/CPUsets.txt]
+
+On architectures that do not hide memoryless nodes, Linux will include only
+zones [nodes] with memory in the zonelists. This means that for a memoryless
+node the "local memory node"--the node of the first zone in CPU's node's
+zonelist--will not be the node itself. Rather, it will be the node that the
+kernel selected as the nearest node with memory when it built the zonelists.
+So, default, local allocations will succeed with the kernel supplying the
+closest available memory. This is a consequence of the same mechanism that
+allows such allocations to fallback to other nearby nodes when a node that
+does contain memory overflows.
+
+Some kernel allocations do not want or cannot tolerate this allocation fallback
+behavior. Rather they want to be sure they get memory from the specified node
+or get notified that the node has no free memory. This is usually the case when
+a subsystem allocates per CPU memory resources, for example.
+
+A typical model for making such an allocation is to obtain the node id of the
+node to which the "current CPU" is attached using one of the kernel's
+numa_node_id() or CPU_to_node() functions and then request memory from only
+the node id returned. When such an allocation fails, the requesting subsystem
+may revert to its own fallback path. The slab kernel memory allocator is an
+example of this. Or, the subsystem may choose to disable or not to enable
+itself on allocation failure. The kernel profiling subsystem is an example of
+this.
+
+If the architecture supports--does not hide--memoryless nodes, then CPUs
+attached to memoryless nodes would always incur the fallback path overhead
+or some subsystems would fail to initialize if they attempted to allocated
+memory exclusively from a node without memory. To support such
+architectures transparently, kernel subsystems can use the numa_mem_id()
+or cpu_to_mem() function to locate the "local memory node" for the calling or
+specified CPU. Again, this is the same node from which default, local page
+allocations will be attempted.