memblock_add_region() carefully checked for merge and overlap
conditions while adding a new region, which is complicated and makes
it difficult to allow arbitrary overlaps or add more merge conditions
(e.g. node ID).

This re-implements memblock_add_region() such that insertion is done
in two steps - all non-overlapping portions of new area are inserted
as separate regions first and then memblock_merge_regions() scan and
merge all neighbouring compatible regions.

This makes addition logic simpler and more versatile and enables
adding node information to memblock.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310462166-31469-3-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Arch could implement memblock_memor_can_coalesce() to veto merging of
adjacent or overlapping memblock regions; however, no arch did and any
vetoing would trigger WARN_ON().  Memblock regions are supposed to
deal with proper memory anyway.  Remove the unused hook.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310462166-31469-2-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Node affine memblock allocation logic is currently implemented across
memblock_alloc_nid() and memblock_alloc_nid_region().  This
reorganizes it such that it resembles that of non-NUMA allocation API.

Area finding is collected and moved into new exported function
memblock_find_in_range_node() which is symmetrical to non-NUMA
counterpart - it handles @start/@end and understands ANYWHERE and
ACCESSIBLE.  memblock_alloc_nid() now simply calls
memblock_find_in_range_node() and reserves the returned area.

This makes memblock_alloc[_try]_nid() observe ACCESSIBLE limit on node
affine allocations too (again, this doesn't make any difference for
the current sole user - sparc64).
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310460395-30913-8-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

NUMA aware memblock alloc functions - memblock_alloc_[try_]nid() -
weren't properly top-down because memblock_nid_range() scanned
forward.  This patch reverses memblock_nid_range(), renames it to
memblock_nid_range_rev() and updates related functions to implement
proper top-down allocation.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310460395-30913-7-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

memblock_nid_range() is used to implement memblock_[try_]alloc_nid().
The generic version determines the range by walking early_node_map
with for_each_mem_pfn_range().  The generic version is defined __weak
to allow arch override.

Currently, only sparc overrides it; however, with the previous update
to the generic implementation, there isn't much to be gained with arch
override.  Sparc would behave exactly the same with the generic
implementation.

This patch disallows arch override for memblock_nid_range() and make
both generic and sparc versions static.

sparc is only compile tested.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310460395-30913-6-git-send-email-tj@kernel.org
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Given an address range, memblock_nid_range() determines the node the
start of the range belongs to and upto where the range stays in the
same node.

It's implemented by calling get_pfn_range_for_nid(), which determines
min and max pfns for a given node, for each node and testing whether
start address falls in there.  This is not only inefficient but also
incorrect when nodes interleave as min-max ranges for nodes overlap.

This patch reimplements memblock_nid_range() using
for_each_mem_pfn_range().  It's simpler, walks the mem ranges once and
can find the exact range the start address falls in.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310460395-30913-5-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

memblock_find_base() is a static function with two callers in
memblock.c and memblock_find_in_range() is a wrapper around it which
just changes the types and order of parameters.

Make memblock_find_in_range() take phys_addr_t instead of u64 for
consistency and replace memblock_find_base() with it.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310457490-3356-7-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

25818f0f (memblock: Make MEMBLOCK_ERROR be 0) thankfully made
MEMBLOCK_ERROR 0 and there already are codes which expect error return
to be 0.  There's no point in keeping MEMBLOCK_ERROR around.  End its
misery.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310457490-3356-6-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310457490-3356-5-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

After node affine allocation fails, memblock_alloc_try_nid() calls
memblock_alloc_base() with @max_addr set to MEMBLOCK_ALLOC_ANYWHERE.
This is inconsistent with memblock_alloc() and what the function's
sole user - sparc/mm/init_64 - expects, although it doesn't make any
difference as sparc64 doesn't have highmem and ACCESSIBLE equals
ANYWHERE.

This patch makes memblock_alloc_try_nid() use ACCESSIBLE instead of
ANYWHERE.  This isn't complete as node affine allocation doesn't
consider memblock.current_limit.  It will be handled with future
changes.

This patch doesn't introduce any behavior difference.
Signed-off-by: NTejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/1310457490-3356-4-git-send-email-tj@kernel.org
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Currently memblock_reserve() or memblock_free() don't handle overlaps of
any kind.  There is some special casing for coalescing exactly adjacent
regions but that's about it.

This is annoying because typically memblock_reserve() is used to mark
regions passed by the firmware as reserved and we all know how much we can
trust our firmwares...

Also, with the current code, if we do something it doesn't handle right
such as trying to memblock_reserve() a large range spanning multiple
existing smaller reserved regions for example, or doing overlapping
reservations, it can silently corrupt the internal region array, causing
odd errors much later on, such as allocations returning reserved regions
etc...

This patch rewrites the underlying functions that add or remove a region
to the arrays.  The new code is a lot more robust as it fully handles
overlapping regions.  It's also, imho, simpler than the previous
implementation.

In addition, while doing so, I found a bug where if we fail to double the
array while adding a region, we would remove the last region of the array
rather than the region we just allocated.  This fixes it too.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: NYinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

While applying patch to use memblock to find aperture for 64bit x86.
Ingo found system with 1g + force_iommu

> No AGP bridge found
> Node 0: aperture @ 38000000 size 32 MB
> Aperture pointing to e820 RAM. Ignoring.
> Your BIOS doesn't leave a aperture memory hole
> Please enable the IOMMU option in the BIOS setup
> This costs you 64 MB of RAM
> Cannot allocate aperture memory hole (0,65536K)

the corresponding code:

	addr = memblock_find_in_range(0, 1ULL<<32, aper_size, 512ULL<<20);
	if (addr == MEMBLOCK_ERROR || addr + aper_size > 0xffffffff) {
		printk(KERN_ERR
			"Cannot allocate aperture memory hole (%lx,%uK)\n",
				addr, aper_size>>10);
		return 0;
	}
	memblock_x86_reserve_range(addr, addr + aper_size, "aperture64")

fails because memblock core code align the size with 512M.  That could
make size way too big.

So don't align the size in that case.

actually __memblock_alloc_base, the another caller already align that
before calling that function.

BTW. x86 does not use __memblock_alloc_base...
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Dave Airlie <airlied@linux.ie>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

memblock_is_region_memory() uses reserved memblocks to search for the
given region, while it should use the memory memblocks.

I encountered the problem with OMAP's framebuffer ram allocation.
Normally the ram is allocated dynamically, and this function is not
called.  However, if we want to pass the framebuffer from the bootloader
to the kernel (to retain the boot image), this function is used to check
the validity of the kernel parameters for the framebuffer ram area.
Signed-off-by: NTomi Valkeinen <tomi.valkeinen@nokia.com>
Acked-by: NYinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Stephen found

WARNING: mm/built-in.o(.text+0x25ab8): Section mismatch in reference from the function memblock_find_base() to the function .init.text:memblock_find_region()
The function memblock_find_base() references
the function __init memblock_find_region().
This is often because memblock_find_base lacks a __init
annotation or the annotation of memblock_find_region is wrong.

So let memblock_find_region() to use __init_memblock instead of __init
directly.

Also fix one function that did not have __init* to be __init_memblock.
Reported-by: NStephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
LKML-Reference: <4CB366B1.40405@kernel.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

The Xen setup code needs to call memblock_x86_reserve_range() very early,
so allow it to initialize the memblock subsystem before doing so.  The
second memblock_init() is ignored.
Signed-off-by: NJeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
LKML-Reference: <4CACFDAD.3090900@goop.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

When trying to find huge range for crashkernel, get

[    0.000000] ------------[ cut here ]------------
[    0.000000] WARNING: at arch/x86/mm/memblock.c:248 memblock_x86_reserve_range+0x40/0x7a()
[    0.000000] Hardware name: Sun Fire x4800
[    0.000000] memblock_x86_reserve_range: wrong range [0xffffffff37000000, 0x137000000)
[    0.000000] Modules linked in:
[    0.000000] Pid: 0, comm: swapper Not tainted 2.6.36-rc5-tip-yh-01876-g1cac214-dirty #59
[    0.000000] Call Trace:
[    0.000000]  [<ffffffff82816f7e>] ? memblock_x86_reserve_range+0x40/0x7a
[    0.000000]  [<ffffffff81078c2d>] warn_slowpath_common+0x85/0x9e
[    0.000000]  [<ffffffff81078d38>] warn_slowpath_fmt+0x6e/0x70
[    0.000000]  [<ffffffff8281e77c>] ? memblock_find_region+0x40/0x78
[    0.000000]  [<ffffffff8281eb1f>] ? memblock_find_base+0x9a/0xb9
[    0.000000]  [<ffffffff82816f7e>] memblock_x86_reserve_range+0x40/0x7a
[    0.000000]  [<ffffffff8280452c>] setup_arch+0x99d/0xb2a
[    0.000000]  [<ffffffff810a3e02>] ? trace_hardirqs_off+0xd/0xf
[    0.000000]  [<ffffffff81cec7d8>] ? _raw_spin_unlock_irqrestore+0x3d/0x4c
[    0.000000]  [<ffffffff827ffcec>] start_kernel+0xde/0x3f1
[    0.000000]  [<ffffffff827ff2d4>] x86_64_start_reservations+0xa0/0xa4
[    0.000000]  [<ffffffff827ff3de>] x86_64_start_kernel+0x106/0x10d
[    0.000000] ---[ end trace a7919e7f17c0a725 ]---
[    0.000000] Reserving 8192MB of memory at 17592186041200MB for crashkernel (System RAM: 526336MB)

This is caused by a wraparound in the test due to size > end;
explicitly check for this condition and fail.
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
LKML-Reference: <4CAA4DD3.1080401@kernel.org>
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>

Stephen found a bunch of section mismatch warnings with the
new memblock changes.

Use __init_memblock to replace __init in memblock.c and remove
__init in memblock.h. We should not use __init in header files.
Reported-by: NStephen Rothwell <sfr@canb.auug.org.au>
Tested-by: NStephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: NYinghai Lu <Yinghai@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
LKML-Reference: <4C912709.2090201@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

So we can avoid export memblock_reserved_init_regions()
Suggested by Ben.

-v2: use __init_memblock attribute
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>

All callers expect a boolean result which is true if the region
overlaps a reserved region. However, the implementation actually
returns -1 if there is no overlap, and a region index (0 based)
if there is.

Make it behave as callers (and common sense) expect.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This is a wrapper for memblock_find_base() using slightly different
arguments (start,end instead of start,size for example) in order to
make it easier to convert existing arch/x86 code.
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Arch code can define ARCH_DISCARD_MEMBLOCK in asm/memblock.h,
which in turns causes memblock code and data to go respectively
into the .init and .initdata sections. This will be used by the
x86 architecture.

If ARCH_DISCARD_MEMBLOCK is defined, the debugfs files to inspect
the memblock arrays after boot are not created.
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

And ensure we don't hand out 0 as a valid allocation. We put the
low limit at PAGE_SIZE arbitrarily.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

will used by x86 memblock_x86_find_in_range_node and nobootmem replacement
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Print out the location info in addition to which array is being
resized. Also use memblocK_dbg() to put that under control of
the memblock_debug flag.
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This exposes memblock_debug and associated memblock_dbg() macro,
along with memblock_can_resize so that x86 can use these when
ported to use memblock
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

memblock_alloc_nid() used to fallback to allocating anywhere by using
memblock_alloc() as a fallback.

However, some of my previous patches limit memblock_alloc() to the region
covered by MEMBLOCK_ALLOC_ACCESSIBLE which is not quite what we want
for memblock_alloc_try_nid().

So we fix it by explicitely using MEMBLOCK_ALLOC_ANYWHERE.

Not that so far only sparc uses memblock_alloc_nid() and it hasn't been updated
to clamp the accessible zone yet. Thus the temporary "breakage" should have
no effect.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

The former is now strict, it will fail if it cannot honor the allocation
within the node, while the later implements the previous semantic which
falls back to allocating anywhere.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

We now provide a default (weak) implementation of memblock_nid_range()
which uses the early_pfn_map[] if CONFIG_ARCH_POPULATES_NODE_MAP
is set. Sparc still needs to use its own method due to the way
the pages can be scattered between nodes.

This implementation is inefficient due to our main algorithm and
callback construct wanting to work on an ascending addresses bases
while early_pfn_map[] would rather work with nid's (it's unsorted
at that stage). But it should work and we can look into improving
it subsequently, possibly using arch compile options to chose a
different algorithm alltogether.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

To constraint the search of a region between two boundaries,
which will be used by the new NUMA aware allocator among others.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Some archs such as ARM want to avoid coalescing accross things such
as the lowmem/highmem boundary or similar. This provides the option
to control it via an arch callback for which a weak default is provided
which always allows coalescing.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

When one of the array gets full, we resize it. After much thinking and
a few iterations of that code, I went back to on-demand resizing using
the (new) internal memblock_find_base() function, which is pretty much what
Yinghai initially proposed, though there some differences in the details.

To work this relies on the default alloc limit being set sensibly by
the architecture.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Some shuffling is needed for doing array resize so we may as well
put some sense into the ordering of the functions in the whole memblock.c
file. No code change. Added some comments.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This will be used by the array resize code and might prove useful
to some arch code as well at which point it can be made non-static.

Also add comment as to why aligning size is important
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
---

v2. Fix loss of size alignment
v3. Fix result code

It's a real PITA to have to search for it in the middle
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This function will be used to locate a free area to put the new memblock
arrays when attempting to resize them. memblock_alloc_region() is gone,
the two callsites now call memblock_add_region().
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
---
v2. Fix membase_alloc_nid_region() conversion

Since we allocate one more than needed, why not do a bit of sanity checking
here to ensure we don't walk past the end of the array ?
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This is in preparation for having resizable arrays.

Note that we still allocate one more than needed, this is unchanged from
the previous implementation.
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Right now, both the "memory" and "reserved" memblock_type structures have
a "size" member. It represents the calculated memory size in the former
case and is unused in the latter.

This moves it out to the main memblock structure instead
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>