On systems with huge amount of physical memory, VFS cache and memory memmap
may eat all available system memory under 4G, then the system may fail to
allocate swiotlb bounce buffer.
There was a fix for this issue in arch/x86_64/mm/numa.c, but that fix dose
not cover sparsemem model.
This patch add fix to sparsemem model by first try to allocate memmap above
4G.
Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
Acked-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Andi Kleen <ak@suse.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Enable system hashtable memory to be distributed among nodes on x86_64 NUMA
Forcing the kernel to use node interleaved vmalloc instead of bootmem for
the system hashtable memory (alloc_large_system_hash) reduces the memory
imbalance on node 0 by around 40MB on a 8 node x86_64 NUMA box:
Before the following patch, on bootup of a 8 node box:
Node 0 MemTotal: 3407488 kB
Node 0 MemFree: 3206296 kB
Node 0 MemUsed: 201192 kB
Node 0 Active: 7012 kB
Node 0 Inactive: 512 kB
Node 0 Dirty: 0 kB
Node 0 Writeback: 0 kB
Node 0 FilePages: 1912 kB
Node 0 Mapped: 420 kB
Node 0 AnonPages: 5612 kB
Node 0 PageTables: 468 kB
Node 0 NFS_Unstable: 0 kB
Node 0 Bounce: 0 kB
Node 0 Slab: 5408 kB
Node 0 SReclaimable: 644 kB
Node 0 SUnreclaim: 4764 kB
After the patch (or using hashdist=1 on the kernel command line):
Node 0 MemTotal: 3407488 kB
Node 0 MemFree: 3247608 kB
Node 0 MemUsed: 159880 kB
Node 0 Active: 3012 kB
Node 0 Inactive: 616 kB
Node 0 Dirty: 0 kB
Node 0 Writeback: 0 kB
Node 0 FilePages: 2424 kB
Node 0 Mapped: 380 kB
Node 0 AnonPages: 1200 kB
Node 0 PageTables: 396 kB
Node 0 NFS_Unstable: 0 kB
Node 0 Bounce: 0 kB
Node 0 Slab: 6304 kB
Node 0 SReclaimable: 1596 kB
Node 0 SUnreclaim: 4708 kB
I guess it is a good idea to keep HASHDIST_DEFAULT "on" for x86_64 NUMA
since x86_64 has no dearth of vmalloc space? Or maybe enable hash
distribution for all 64bit NUMA arches? The following patch does it only
for x86_64.
I ran a HPC MPI benchmark -- 'Ansys wingsolid', which takes up quite a bit of
memory and uses up tlb entries. This was on a 4 way, 2 socket
Tyan AMD box (non vsmp), with 8G total memory (4G pernode).
The results with and without hash distribution are:
1. Vanilla - runtime of 1188.000s
2. With hashdist=1 runtime of 1154.000s
Oprofile output for the duration of run is:
1. Vanilla:
PU: AMD64 processors, speed 2411.16 MHz (estimated)
Counted L1_AND_L2_DTLB_MISSES events (L1 and L2 DTLB misses) with a unit
mask of 0x00 (No unit mask) count 500
samples % app name symbol name
163054 6.5513 libansys1.so MultiFront::decompose(int, int,
Elemset *, int *, int, int, int)
162061 6.5114 libansys3.so blockSaxpy6L_fd
162042 6.5107 libansys3.so blockInnerProduct6L_fd
156286 6.2794 libansys3.so maxb33_
87879 3.5309 libansys1.so elmatrixmultpcg_
84857 3.4095 libansys4.so saxpy_pcg
58637 2.3560 libansys4.so .st4560
46612 1.8728 libansys4.so .st4282
43043 1.7294 vmlinux-t copy_user_generic_string
41326 1.6604 libansys3.so blockSaxpyBackSolve6L_fd
41288 1.6589 libansys3.so blockInnerProductBackSolve6L_fd
2. With hashdist=1
CPU: AMD64 processors, speed 2411.13 MHz (estimated)
Counted L1_AND_L2_DTLB_MISSES events (L1 and L2 DTLB misses) with a unit
mask of 0x00 (No unit mask) count 500
samples % app name symbol name
162993 6.9814 libansys1.so MultiFront::decompose(int, int,
Elemset *, int *, int, int, int)
160799 6.8874 libansys3.so blockInnerProduct6L_fd
160459 6.8729 libansys3.so blockSaxpy6L_fd
156018 6.6826 libansys3.so maxb33_
84700 3.6279 libansys4.so saxpy_pcg
83434 3.5737 libansys1.so elmatrixmultpcg_
58074 2.4875 libansys4.so .st4560
46000 1.9703 libansys4.so .st4282
41166 1.7632 libansys3.so blockSaxpyBackSolve6L_fd
41033 1.7575 libansys3.so blockInnerProductBackSolve6L_fd
35762 1.5318 libansys1.so inner_product_sub
35591 1.5245 libansys1.so inner_product_sub2
28259 1.2104 libansys4.so addVectors
Signed-off-by: Pravin B. Shelar <pravin.shelar@calsoftinc.com>
Signed-off-by: Ravikiran Thirumalai <kiran@scalex86.org>
Signed-off-by: Shai Fultheim <shai@scalex86.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Acked-by: Christoph Lameter <clameter@engr.sgi.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Fix unannotated variable declarations. Variables that have allocation
section annotations (such as __meminitdata) on their definitions must also
have them on their declarations as not doing so may affect the addressing
mode used by the compiler and may result in a linker error.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It has no users and it's doubtful that we'll need it again.
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
It fixes various coding style issues, specially when spaces are useless. For
example '*' go next to the function name.
Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
__init in headers is pretty useless because the compiler doesn't check it, and
they get out of sync relatively frequently. So if you see an __init in a
header file, it's quite unreliable and you need to check the definition
anyway.
Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The grow algorithm is simple, we grow if:
1) we see a hash chain collision at insert, and
2) we haven't hit the hash size limit (currently 1*1024*1024 slots), and
3) the number of xfrm_state objects is > the current hash mask
All of this needs some tweaking.
Remove __initdata from "hashdist" so we can use it safely at run time.
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix some FRV arch compile errors, including:
(*) Marking nr_kernel_pages as __meminitdata so that references to it end up
being properly calculated rather than being assumed to be in the small
data section (and thus calculated wrt the GP register). Not doing this
causes the linker to emit errors as the offset is too big to fit into the
load instruction.
(*) Move pm_power_off into an unconditionally compiled .c file as it's now
unconditionally accessed.
(*) Declare frv_change_cmode() in a header file rather than in a .c file, and
declare it asmlinkage.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Change definitions of some functions and data from __init to __meminit.
These functions and data can be used after bootup by this patch to be used for
hot-add codes.
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The node setup code would try to allocate the node metadata in the node
itself, but that fails if there is no memory in there.
This can happen with memory hotplug when the hotplug area defines an so
far empty node.
Now use bootmem to try to allocate the mem_map in other nodes.
And if it fails don't panic, but just ignore the node.
To make this work I added a new __alloc_bootmem_nopanic function that
does what its name implies.
TBD should try to use nearby nodes here. Currently we just use any.
It's hard to do it better because bootmem doesn't have proper fallback
lists yet.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add a list_head to bootmem_data_t and make bootmems use it. bootmem list is
sorted by node_boot_start.
Only nodes against which init_bootmem() is called are linked to the list.
(i386 allocates bootmem only from one node(0) not from all online nodes.)
A summary:
1. for_each_online_pgdat() traverses all *online* nodes.
2. alloc_bootmem() allocates memory only from initialized-for-bootmem nodes.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This fixes problems with very large nodes (over 128GB) filling up all of
the first 4GB with their mem_map and not leaving enough space for the
swiotlb.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Patch cleans up the alloc_bootmem fix for swiotlb. Patch removes
alloc_bootmem_*_limit api and fixes alloc_boot_*low api to do the right
thing -- allocate from low32 memory.
Signed-off-by: Ravikiran Thirumalai <kiran@scalex86.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This introduces a limit parameter to the core bootmem allocator; The new
parameter indicates that physical memory allocated by the bootmem
allocator should be within the requested limit.
We also introduce alloc_bootmem_low_pages_limit, alloc_bootmem_node_limit,
alloc_bootmem_low_pages_node_limit apis, but alloc_bootmem_low_pages_limit
is the only api used for swiotlb.
The existing alloc_bootmem_low_pages() api could instead have been
changed and made to pass right limit to the core allocator. But that
would make the patch more intrusive for 2.6.14, as other arches use
alloc_bootmem_low_pages(). We may be done that post 2.6.14 as a
cleanup.
With this, swiotlb gets memory within 4G for both x86_64 and ia64
arches.
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: Ravikiran G Thirumalai <kiran@scalex86.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch retrieves the max_pfn being used by previous kernel and stores it
in a safe location (saved_max_pfn) before it is overwritten due to user
defined memory map. This pfn is used to make sure that user does not try to
read the physical memory beyond saved_max_pfn.
Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Introduce a simple allocator for the NUMA remap space. This space is very
scarce, used for structures which are best allocated node local.
This mechanism is also used on non-NUMA ia64 systems with a vmem_map to keep
the pgdat->node_mem_map initialized in a consistent place for all
architectures.
Issues:
o alloc_remap takes a node_id where we might expect a pgdat which was intended
to allow us to allocate the pgdat's using this mechanism; which we do not yet
do. Could have alloc_remap_node() and alloc_remap_nid() for this purpose.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Zou Nan hai
Ravikiran G Thirumalai
David Howells
Andrew Morton
Franck Bui-Huu
David S. Miller
Yasunori Goto
Andi Kleen
KAMEZAWA Hiroyuki
Vivek Goyal
Dave Hansen
Linus Torvalds