Impact: fix crash during hibernation on 32-bit NUMA
The NUMA code on x86_32 creates special memory mapping that allows
each node's pgdat to be located in this node's memory. For this
purpose it allocates a memory area at the end of each node's memory
and maps this area so that it is accessible with virtual addresses
belonging to low memory. As a result, if there is high memory,
these NUMA-allocated areas are physically located in high memory,
although they are mapped to low memory addresses.
Our hibernation code does not take that into account and for this
reason hibernation fails on all x86_32 systems with CONFIG_NUMA=y and
with high memory present. Fix this by adding a special mapping for
the NUMA-allocated memory areas to the temporary page tables created
during the last phase of resume.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
There are a lot of places that define either a single bootmem descriptor or an
array of them. Use only one central array with MAX_NUMNODES items instead.
Signed-off-by: Johannes Weiner <hannes@saeurebad.de>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Kyle McMartin <kyle@parisc-linux.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Yinghai Lu <yhlu.kernel@gmail.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
we already have the same srat handling interface for 32bit.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
move out e820_register_active_regions from non numa zones_sizes_init()
and remove numa version zones_sizes_init().
and let 32 bit call remove_all_active_ranges() in setup_arch() directly
like 64-bit
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kva ram already mapped right after away, so don't need to get that for low ram.
avoid wasting one copy of pgdat.
also add node id in early_res name in case we get it from find_e820_area.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
so that max_low_pfn is not changed after it is set.
so we can move that early and out of initmem_init.
could call find_low_pfn_range just after max_pfn is set.
also could move reserve_initrd out of setup_bootmem_allocator
so 32bit is more like 64bit.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
so don't punish all other cpus without that problem when init highmem
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
use early_node_map to init high pages, so we can remove page_is_ram() and
page_is_reserved_early() in the big loop with add_one_highpage
also remove page_is_reserved_early(), it is not needed anymore.
v2: fix the build of other platforms
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
in case we have kva before ramdisk on a node, we still need to use
those ranges.
v2: reserve_early kva ram area, in case there are holes in highmem, to avoid
those area could be treat as free high pages.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
don't assume we can use RAM near the end of every node.
Esp systems that have few memory and they could have
kva address and kva RAM all below max_low_pfn.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Now we are using register_e820_active_regions() instead of
add_active_range() directly. So end_pfn could be different between the
value in early_node_map to node_end_pfn.
So we need to make shrink_active_range() smarter.
shrink_active_range() is a generic MM function in mm/page_alloc.c but
it is only used on 32-bit x86. Should we move it back to some file in
arch/x86?
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
this way 32-bit is more similar to 64-bit, and smarter e820 and numa.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
when 1/3 user/kernel split is used, and less memory is installed, or if
we have a big hole below 4g, max_low_pfn is still using 3g-128m
try to go down from max_low_pfn until we get it. otherwise will panic.
need to make 32-bit code to use register_e820_active_regions ... later.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
on 32-bit in head_32.S after initial page table is done, we get initial
max_pfn_mapped, and then kernel_physical_mapping_init will give us
a final one.
We need to use that to make sure find_e820_area will get valid addresses
for boot_map and for NODE_DATA(0) on numa32.
XEN PV and lguest may need to assign max_pfn_mapped too.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
we don't need to call memory_present that early.
numa and sparse will call memory_present later and might
even fail, it will call memory_present for the full range.
also for sparse it will call alloc_bootmem ... before we set up bootmem.
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
... otherwise alloc_remap will not get node_mem_map from kva area, and
alloc_node_mem_map has to alloc_bootmem_node to get mem_map.
It will use two low address copies ...
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
so every element will represent 64M instead of 256M.
AMD opteron could have HW memory hole remapping, so could have
[0, 8g + 64M) on node0. Reduce element size to 64M to keep that on node 0
Later we need to use find_e820_area() to allocate memory_node_map like
on 64-bit. But need to move memory_present out of populate_mem_map...
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
on Summit it's possible to have:
CONFIG_ACPI_SRAT=y
CONFIG_HAVE_ARCH_PARSE_SRAT=y
in which case acpi.h defines the acpi_numa_slit_init() and
acpi_numa_processor_affinity_init() methods as a macro.
reserve early numa kva, so it will not clash with new RAMDISK
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When allocating the pgdat's for numa nodes on x86_32 we attempt to place
them in the numa remap space for that node. However should the node not
have any remap space allocated (such as due to having non-ram pages in
the remap location in the node) then we will incorrectly place the pgdat
at zero. Check we have remap available, falling back to node 0 memory
where we do not.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Recent kernels have been panic'ing trying to allocate memory early in boot,
in __alloc_pages:
BUG: unable to handle kernel paging request at 00001568
IP: [<c10407b6>] __alloc_pages+0x33/0x2cc
*pdpt = 00000000013a5001 *pde = 0000000000000000
Oops: 0000 [#1] SMP
Modules linked in:
Pid: 1, comm: swapper Not tainted (2.6.25 #78)
EIP: 0060:[<c10407b6>] EFLAGS: 00010246 CPU: 0
EIP is at __alloc_pages+0x33/0x2cc
EAX: 00001564 EBX: 000412d0 ECX: 00001564 EDX: 000005c3
ESI: f78012a0 EDI: 00000001 EBP: 00001564 ESP: f7871e50
DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
Process swapper (pid: 1, ti=f7870000 task=f786f670 task.ti=f7870000)
Stack: 00000000 f786f670 00000010 00000000 0000b700 000412d0 f78012a0 00000001
00000000 c105b64d 00000000 000412d0 f78012a0 f7803120 00000000 c105c1c5
00000010 f7803144 000412d0 00000001 f7803130 f7803120 f78012a0 00000001
Call Trace:
[<c105b64d>] kmem_getpages+0x94/0x129
[<c105c1c5>] cache_grow+0x8f/0x123
[<c105c689>] ____cache_alloc_node+0xb9/0xe4
[<c105c999>] kmem_cache_alloc_node+0x92/0xd2
[<c1018929>] build_sched_domains+0x536/0x70d
[<c100b63c>] do_flush_tlb_all+0x0/0x3f
[<c100b63c>] do_flush_tlb_all+0x0/0x3f
[<c10572d6>] interleave_nodes+0x23/0x5a
[<c105c44f>] alternate_node_alloc+0x43/0x5b
[<c1018b47>] arch_init_sched_domains+0x46/0x51
[<c136e85e>] kernel_init+0x0/0x82
[<c137ac19>] sched_init_smp+0x10/0xbb
[<c136e8a1>] kernel_init+0x43/0x82
[<c10035cf>] kernel_thread_helper+0x7/0x10
Debugging this showed that the NODE_DATA() for nodes other than node 0
were all NULL. Tracing this back showed that the NODE_DATA() pointers
were being initialised to each nodes remap space. However under
SPARSEMEM remap is disabled which leads to the pgdat's being placed
incorrectly at kernel virtual address 0. Leading to the panic when
attempting to allocate memory from these nodes.
Numa remap was disabled in the commit below. This occured while fixing
problems triggered when attempting to boot x86_32 NUMA SPARSEMEM kernels
on non-numa hardware.
x86: make NUMA work on 32-bit
commit 1b000a5dbe
The real problem is believed to be related to other alignment issues in
the regions blocked out from the bootmem allocator for small memory
systems, and has been fixed separately. Therefore re-enable remap for
SPARSMEM, which fixes pgdat allocation issues. Testing confirms that
SPARSMEM NUMA kernels will boot correctly with this part of the change
reverted.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
arch/x86/pci/Makefile_32 has a nasty detail. VISWS and NUMAQ build
override the generic pci-y rules. This needs a proper cleanup, but
that needs more thoughts. Undo
commit 895d30935e
x86: numaq fix
do not override the existing pci-y rule when adding visws or
numaq rules.
There is also a stupid init function ordering problem vs. acpi.o
Add comments to the Makefile to avoid tripping over this again.
Remove the srat stub code in discontig_32.c to allow a proper NUMAQ
build.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
this function doesnt just 'find' the max_pfn - it also has
other side-effects such as registering sparse memory maps.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
we could call find_max_pfn() directly instead of setup_memory() to get
max_pfn needed for mtrr trimming.
otherwise setup_memory() is called two times... that is duplicated...
[ mingo@elte.hu: both Thomas and me simulated a double call to
setup_bootmem_allocator() and can confirm that it is a real bug
which can hang in certain configs. It's not been reported yet but
that is probably due to the relatively scarce nature of
MTRR-trimming systems. ]
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patchset adds a flags variable to reserve_bootmem() and uses the
BOOTMEM_EXCLUSIVE flag in crashkernel reservation code to detect collisions
between crashkernel area and already used memory.
This patch:
Change the reserve_bootmem() function to accept a new flag BOOTMEM_EXCLUSIVE.
If that flag is set, the function returns with -EBUSY if the memory already
has been reserved in the past. This is to avoid conflicts.
Because that code runs before SMP initialisation, there's no race condition
inside reserve_bootmem_core().
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix powerpc build]
Signed-off-by: Bernhard Walle <bwalle@suse.de>
Cc: <linux-arch@vger.kernel.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The DISCONTIG memory model on x86 32 bit uses a remap allocator early
in boot. The objective is that portions of every node are mapped in to
the kernel virtual area (KVA) in place of ZONE_NORMAL so that node-local
allocations can be made for pgdat and mem_map structures.
With SPARSEMEM, the amount that is set aside is insufficient for all the
mem_maps to be allocated. During the boot process, it falls back to using
the bootmem allocator. This breaks assumptions that SPARSEMEM makes about
the layout of the mem_map in memory and results in a VM_BUG_ON triggering
due to pfn_to_page() returning garbage values.
This patch only enables the remap allocator for use with DISCONTIG.
Without SRAT support, a compile-error occurs because ACPI table parsing
functions are only available in x86-64. This patch also adds no-op stubs
and prints a warning message. What likely needs to be done is sharing
the table parsing functions between 32 and 64 bit if they are
compatible.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
On 32-bit NUMA, the memmap representing struct pages on each node is
allocated from node-local memory if possible. As only node-0 has memory from
ZONE_NORMAL, the memmap must be mapped into low memory. This is done by
reserving space in the Kernel Virtual Area (KVA) for the memmap belonging
to other nodes by taking pages from the end of ZONE_NORMAL and remapping
the other nodes memmap into those virtual addresses. The node boundaries
are then adjusted so that the region of pages is not used and it is marked
as reserved in the bootmem allocator.
This reserved portion of the KVA is PMD aligned althought
strictly speaking that requirement could be lifted (see thread at
http://lkml.org/lkml/2007/8/24/220). The problem is that when aligned, there
may be a portion of ZONE_NORMAL at the end that is not used for memmap and
does not have an initialised memmap nor is it marked reserved in the bootmem
allocator. Later in the boot process, these pages are freed and a storm of
Bad page state messages result.
This patch marks these pages reserved that are wasted due to alignment
in the bootmem allocator so they are not accidently freed. It is worth
noting that memory from node-0 is wasted where it could have been put into
ZONE_HIGHMEM on NUMA machines. Worse, the KVA is always reserved from the
location of real memory even when there is plenty of spare virtual address
space.
This patch also makes sure that reserve_bootmem() is not called with a
0-length size in numa_kva_reserve(). When this happens, it usually means
that a kernel built for Summit is being booted on a normal machine. The
resulting BUG_ON() is misleading so it is caught here.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The boot protocol has until now required that the initrd be located in
lowmem, which makes the lowmem/highmem boundary visible to the boot
loader. This was exported to the bootloader via a compile-time
field. Unfortunately, the vmalloc= command-line option breaks this
part of the protocol; instead of adding yet another hack that affects
the bootloader, have the kernel relocate the initrd down below the
lowmem boundary inside the kernel itself.
Note that this does not rely on HIGHMEM being enabled in the kernel.
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
node0_bdata and paddr_to_nid() can become static.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch makes some needlessly global variables static.
[ tglx: arch/x86 adaptation ]
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
It's not necessarily a very sane configuration, but people running "make
randconfig" noticed it wouldn't compile. This fixes some obvious
problems in discontig.c to allow a clean compile.
Acked-by: andrew hendry <andrew.hendry@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Extern declarations belong in headers. Times, they are a'changin.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andi Kleen <ak@suse.de>
===================================================================
Fix compile error when config memory hotplug with numa on i386.
The cause of compile error was missing of arch_add_memory(),
remove_memory(), and memory_add_physaddr_to_nid().
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Acked-by: David Rientjes <rientjes@cs.washington.edu>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Harvey Harrison