Embedding first chunk allocator maintains the distances between units
in the vmalloc area and thus needs vmalloc space to be larger than the
maximum distances between units; otherwise, it wouldn't be able to
create any dynamic chunks.  This patch makes the embedding first chunk
allocator check vmalloc space size and if the maximum distance between
units is larger than 75% of it, print warning and, if page mapping
allocator is available, fail initialization so that the system falls
back onto it.

This should work around percpu allocation failure problems on certain
sparc64 configurations where distances between NUMA nodes are larger
than the vmalloc area and makes percpu allocator more robust for
future configurations.
Signed-off-by: NTejun Heo <tj@kernel.org>

pcpu_build_alloc_info() may be called multiple times when percpu is
falling back to different first chunk allocator.  Make it clear static
buffers so that they don't contain values from previous runs.
Signed-off-by: NTejun Heo <tj@kernel.org>

pcpu_setup_first_chunk() incorrectly used NR_CPUS as the impossible
unit number while unit number can equal and go over NR_CPUS with
sparse unit map.  This triggers BUG_ON() spuriously on machines which
have non-power-of-two number of cpus.  Use UINT_MAX instead.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reported-and-tested-by: NTony Vroon <tony@linx.net>

percpu incorrectly assumed that cpu0 was always there which led to the
following warning and eventual oops on sparc machines w/o cpu0.

  WARNING: at mm/percpu.c:651 pcpu_map+0xdc/0x100()
  Modules linked in:
  Call Trace:
    [000000000045eb70] warn_slowpath_common+0x50/0xa0
    [000000000045ebdc] warn_slowpath_null+0x1c/0x40
    [00000000004d493c] pcpu_map+0xdc/0x100
    [00000000004d59a4] pcpu_alloc+0x3e4/0x4e0
    [00000000004d5af8] __alloc_percpu+0x18/0x40
    [00000000005b112c] __percpu_counter_init+0x4c/0xc0
  ...
  Unable to handle kernel NULL pointer dereference
  ...
   I7: <sysfs_new_dirent+0x30/0x120>
   Disabling lock debugging due to kernel taint
   Caller[000000000053c1b0]: sysfs_new_dirent+0x30/0x120
   Caller[000000000053c7a4]: create_dir+0x24/0xc0
   Caller[000000000053c870]: sysfs_create_dir+0x30/0x80
   Caller[00000000005990e8]: kobject_add_internal+0xc8/0x200
  ...
   Kernel panic - not syncing: Attempted to kill the idle task!

This patch fixes the problem by backporting parts from devel branch to
make percpu core not depend on the existence of cpu0.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reported-by: NMeelis Roos <mroos@linux.ee>
Cc: David Miller <davem@davemloft.net>

With x86 converted to embedding allocator, lpage doesn't have any user
left.  Kill it along with cpa handling code.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Jan Beulich <JBeulich@novell.com>

Now that percpu core can handle very sparse units, given that vmalloc
space is large enough, embedding first chunk allocator can use any
memory to build the first chunk.  This patch teaches
pcpu_embed_first_chunk() about distances between cpus and to use
alloc/free callbacks to allocate node specific areas for each group
and use them for the first chunk.

This brings the benefits of embedding allocator to NUMA configurations
- no extra TLB pressure with the flexibility of unified dynamic
allocator and no need to restructure arch code to build memory layout
suitable for percpu.  With units put into atom_size aligned groups
according to cpu distances, using large page for dynamic chunks is
also easily possible with falling back to reuglar pages if large
allocation fails.

Embedding allocator users are converted to specify NULL
cpu_distance_fn, so this patch doesn't cause any visible behavior
difference.  Following patches will convert them.
Signed-off-by: NTejun Heo <tj@kernel.org>

ai->groups[] contains which units need to be put consecutively and at
what offset from the chunk base address.  Compile this information
into pcpu_group_offsets[] and pcpu_group_sizes[] in
pcpu_setup_first_chunk() and use them to allocate sparse vm areas
using pcpu_get_vm_areas().

This will be used to allow directly using sparse NUMA memories as
percpu areas.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Nick Piggin <npiggin@suse.de>

The only thing percpu allocator wants to know about a vmalloc area is
the base address.  Instead of requiring chunk->vm, add
chunk->base_addr which contains the necessary value.  This simplifies
the code a bit and makes the dummy first_vm unnecessary.  This change
will ease allowing a chunk to be mapped by multiple vms.
Signed-off-by: NTejun Heo <tj@kernel.org>

Currently units are mapped sequentially into address space.  This
patch adds pcpu_unit_offsets[] which allows units to be mapped to
arbitrary offsets from the chunk base address.  This is necessary to
allow sparse embedding which might would need to allocate address
ranges and memory areas which aren't aligned to unit size but
allocation atom size (page or large page size).  This also simplifies
things a bit by removing the need to calculate offset from unit
number.

With this change, there's no need for the arch code to know
pcpu_unit_size.  Update pcpu_setup_first_chunk() and first chunk
allocators to return regular 0 or -errno return code instead of unit
size or -errno.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: David S. Miller <davem@davemloft.net>

Till now, non-linear cpu->unit map was expressed using an integer
array which maps each cpu to a unit and used only by lpage allocator.
Although how many units have been placed in a single contiguos area
(group) is known while building unit_map, the information is lost when
the result is recorded into the unit_map array.  For lpage allocator,
as all allocations are done by lpages and whether two adjacent lpages
are in the same group or not is irrelevant, this didn't cause any
problem.  Non-linear cpu->unit mapping will be used for sparse
embedding and this grouping information is necessary for that.

This patch introduces pcpu_alloc_info which contains all the
information necessary for initializing percpu allocator.
pcpu_alloc_info contains array of pcpu_group_info which describes how
units are grouped and mapped to cpus.  pcpu_group_info also has
base_offset field to specify its offset from the chunk's base address.
pcpu_build_alloc_info() initializes this field as if all groups are
allocated back-to-back as is currently done but this will be used to
sparsely place groups.

pcpu_alloc_info is a rather complex data structure which contains a
flexible array which in turn points to nested cpu_map arrays.

* pcpu_alloc_alloc_info() and pcpu_free_alloc_info() are provided to
  help dealing with pcpu_alloc_info.

* pcpu_lpage_build_unit_map() is updated to build pcpu_alloc_info,
  generalized and renamed to pcpu_build_alloc_info().
  @cpu_distance_fn may be NULL indicating that all cpus are of
  LOCAL_DISTANCE.

* pcpul_lpage_dump_cfg() is updated to process pcpu_alloc_info,
  generalized and renamed to pcpu_dump_alloc_info().  It now also
  prints which group each alloc unit belongs to.

* pcpu_setup_first_chunk() now takes pcpu_alloc_info instead of the
  separate parameters.  All first chunk allocators are updated to use
  pcpu_build_alloc_info() to build alloc_info and call
  pcpu_setup_first_chunk() with it.  This has the side effect of
  packing units for sparse possible cpus.  ie. if cpus 0, 2 and 4 are
  possible, they'll be assigned unit 0, 1 and 2 instead of 0, 2 and 4.

* x86 setup_pcpu_lpage() is updated to deal with alloc_info.

* sparc64 setup_per_cpu_areas() is updated to build alloc_info.

Although the changes made by this patch are pretty pervasive, it
doesn't cause any behavior difference other than packing of sparse
cpus.  It mostly changes how information is passed among
initialization functions and makes room for more flexibility.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>

Unit map handling will be generalized and extended and used for
embedding sparse first chunk and other purposes.  Relocate two
unit_map related functions upward in preparation.  This patch just
moves the code without any actual change.
Signed-off-by: NTejun Heo <tj@kernel.org>

pcpu_fc_alloc_fn_t is about to see more interesting usage, add @align
parameter.
Signed-off-by: NTejun Heo <tj@kernel.org>

Now that all actual first chunk allocation and copying happen in the
first chunk allocators and helpers, there's no reason for
pcpu_setup_first_chunk() to try to determine @dyn_size automatically.
The only left user is page first chunk allocator.  Make it determine
dyn_size like other allocators and make @dyn_size mandatory for
pcpu_setup_first_chunk().
Signed-off-by: NTejun Heo <tj@kernel.org>

First chunk allocators assume percpu areas have been linked using one
of PERCPU_*() macros and depend on __per_cpu_load symbol defined by
those macros, so there isn't much point in passing in static area size
explicitly when it can be easily calculated from __per_cpu_start and
__per_cpu_end.  Drop @static_size from all percpu first chunk
allocators and helpers.
Signed-off-by: NTejun Heo <tj@kernel.org>

Now that all first chunk allocators are in mm/percpu.c, it makes sense
to make generalize percpu_alloc kernel parameter.  Define PCPU_FC_*
and set pcpu_chosen_fc using early_param() in mm/percpu.c.  Arch code
can use the set value to determine which first chunk allocator to use.
Signed-off-by: NTejun Heo <tj@kernel.org>

There's no need to build unused first chunk allocators in.  Define
CONFIG_NEED_PER_CPU_*_FIRST_CHUNK and let archs enable them
selectively.
Signed-off-by: NTejun Heo <tj@kernel.org>

Page size isn't always 4k depending on arch and configuration.  Rename
4k first chunk allocator to page.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: David Howells <dhowells@redhat.com>

Improve percpu boot messages such that they're uniform and contain
more information.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NChristoph Lameter <cl@linux-foundation.org>

pcpu_reclaim() calls pcpu_depopulate_chunk() which makes use of pages
array and bitmap returned by pcpu_get_pages_and_bitmap() and thus
should be called under pcpu_alloc_mutex.  pcpu_reclaim() released the
mutex before calling depopulate leading to double free and other
strange problems caused by the unexpected concurrent usages of pages
array and bitmap.  Fix it.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NChristoph Lameter <cl@linux-foundation.org>

get_vm_area() only accepts VM_* flags, not GFP_*.

And according to the doc of get_vm_area(), here should be
VM_ALLOC.
Signed-off-by: NWANG Cong <amwang@redhat.com>
Acked-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>

percpu code has been assuming num_possible_cpus() == nr_cpu_ids which
is incorrect if cpu_possible_map contains holes.  This causes percpu
code to access beyond allocated memories and vmalloc areas.  On a
sparc64 machine with cpus 0 and 2 (u60), this triggers the following
warning or fails boot.

 WARNING: at /devel/tj/os/work/mm/vmalloc.c:106 vmap_page_range_noflush+0x1f0/0x240()
 Modules linked in:
 Call Trace:
  [00000000004b17d0] vmap_page_range_noflush+0x1f0/0x240
  [00000000004b1840] map_vm_area+0x20/0x60
  [00000000004b1950] __vmalloc_area_node+0xd0/0x160
  [0000000000593434] deflate_init+0x14/0xe0
  [0000000000583b94] __crypto_alloc_tfm+0xd4/0x1e0
  [00000000005844f0] crypto_alloc_base+0x50/0xa0
  [000000000058b898] alg_test_comp+0x18/0x80
  [000000000058dad4] alg_test+0x54/0x180
  [000000000058af00] cryptomgr_test+0x40/0x60
  [0000000000473098] kthread+0x58/0x80
  [000000000042b590] kernel_thread+0x30/0x60
  [0000000000472fd0] kthreadd+0xf0/0x160
 ---[ end trace 429b268a213317ba ]---

This patch fixes generic percpu functions and sparc64
setup_per_cpu_areas() so that they handle sparse cpu_possible_map
properly.

Please note that on x86, cpu_possible_map() doesn't contain holes and
thus num_possible_cpus() == nr_cpu_ids and this patch doesn't cause
any behavior difference.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NDavid S. Miller <davem@davemloft.net>
Cc: Ingo Molnar <mingo@elte.hu>

Large page first chunk allocator is primarily used for NUMA machines;
however, its NUMA handling is extremely simplistic.  Regardless of
their proximity, each cpu is put into separate large page just to
return most of the allocated space back wasting large amount of
vmalloc space and increasing cache footprint.

This patch teachs NUMA details to large page allocator.  Given
processor proximity information, pcpu_lpage_build_unit_map() will find
fitting cpu -> unit mapping in which cpus in LOCAL_DISTANCE share the
same large page and not too much virtual address space is wasted.

This greatly reduces the unit and thus chunk size and wastes much less
address space for the first chunk.  For example, on 4/4 NUMA machine,
the original code occupied 16MB of virtual space for the first chunk
while the new code only uses 4MB - one 2MB page for each node.

[ Impact: much better space efficiency on NUMA machines ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jan Beulich <JBeulich@novell.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Miller <davem@davemloft.net>

Currently cpu and unit are always identity mapped.  To allow more
efficient large page support on NUMA and lazy allocation for possible
but offline cpus, cpu -> unit mapping needs to be non-linear and/or
sparse.  This can be easily implemented by adding a cpu -> unit
mapping array and using it whenever looking up the matching unit for a
cpu.

The only unusal conversion is in pcpu_chunk_addr_search().  The passed
in address is unit0 based and unit0 might not be in use so it needs to
be converted to address of an in-use unit.  This is easily done by
adding the unit offset for the current processor.

[ Impact: allows non-linear/sparse cpu -> unit mapping, no visible change yet ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>

percpu core doesn't need to tack all the allocated pages.  It needs to
know whether certain pages are populated and a way to reverse map
address to page when freeing.  This patch drops pcpu_chunk->page[] and
use populated bitmap and vmalloc_to_page() lookup instead.  Using
vmalloc_to_page() exclusively is also possible but complicates first
chunk handling, inflates cache footprint and prevents non-standard
memory allocation for percpu memory.

pcpu_chunk->page[] was used to track each page's allocation and
allowed asymmetric population which happens during failure path;
however, with single bitmap for all units, this is no longer possible.
Bite the bullet and rewrite (de)populate functions so that things are
done in clearly separated steps such that asymmetric population
doesn't happen.  This makes the (de)population process much more
modular and will also ease implementing non-standard memory usage in
the future (e.g. large pages).

This makes @get_page_fn parameter to pcpu_setup_first_chunk()
unnecessary.  The parameter is dropped and all first chunk helpers are
updated accordingly.  Please note that despite the volume most changes
to first chunk helpers are symbol renames for variables which don't
need to be referenced outside of the helper anymore.

This change reduces memory usage and cache footprint of pcpu_chunk.
Now only #unit_pages bits are necessary per chunk.

[ Impact: reduced memory usage and cache footprint for bookkeeping ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>

(de)populate functions are about to be reimplemented to drop
pcpu_chunk->page array.  Move a few functions so that the rewrite
patch doesn't have code movement making it more difficult to read.

[ Impact: code movement ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>

Now that all first chunk allocator helpers allocate and map the first
chunk themselves, there's no need to have optional default alloc/map
in pcpu_setup_first_chunk().  Drop @populate_pte_fn and only leave
@dyn_size optional and make all other params mandatory.

This makes it much easier to follow what pcpu_setup_first_chunk() is
doing and what actual differences tweaking each parameter results in.

[ Impact: drop unused code path ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>

Generalize and move x86 setup_pcpu_lpage() into
pcpu_lpage_first_chunk().  setup_pcpu_lpage() now is a simple wrapper
around the generalized version.  Other than taking size parameters and
using arch supplied callbacks to allocate/free/map memory,
pcpu_lpage_first_chunk() is identical to the original implementation.

This simplifies arch code and will help converting more archs to
dynamic percpu allocator.

While at it, factor out pcpu_calc_fc_sizes() which is common to
pcpu_embed_first_chunk() and pcpu_lpage_first_chunk().

[ Impact: code reorganization and generalization ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>

At first, percpu first chunk was always setup page-by-page by the
generic code.  To add other allocators, different parts of the generic
initialization was made optional.  Now we have three allocators -
embed, remap and 4k.  embed and remap fully handle allocation and
mapping of the first chunk while 4k still depends on generic code for
those.  This makes the generic alloc/map paths specifci to 4k and
makes the code unnecessary complicated with optional generic
behaviors.

This patch makes the 4k allocator to allocate and map memory directly
instead of depending on the generic code.  The only outside visible
change is that now dynamic area in the first chunk is allocated
up-front instead of on-demand.  This doesn't make any meaningful
difference as the area is minimal (usually less than a page, just
enough to fill the alignment) on 4k allocator.  Plus, dynamic area in
the first chunk usually gets fully used anyway.

This will allow simplification of pcpu_setpu_first_chunk() and removal
of chunk->page array.

[ Impact: no outside visible change other than up-front allocation of dyn area ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>

Generalize and move x86 setup_pcpu_4k() into pcpu_4k_first_chunk().
setup_pcpu_4k() now is a simple wrapper around the generalized
version.  Other than taking size parameters and using arch supplied
callbacks to allocate/free memory, pcpu_4k_first_chunk() is identical
to the original implementation.

This simplifies arch code and will help converting more archs to
dynamic percpu allocator.

While at it, s/pcpu_populate_pte_fn_t/pcpu_fc_populate_pte_fn_t/ for
consistency.

[ Impact: code reorganization and generalization ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>

The only extra feature @unit_size provides is making dead space at the
end of the first chunk which doesn't have any valid usecase.  Drop the
parameter.  This will increase consistency with generalized 4k
allocator.

James Bottomley spotted missing conversion for the default
setup_per_cpu_areas() which caused build breakage on all arcsh which
use it.

[ Impact: drop unused code path ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Ingo Molnar <mingo@elte.hu>

The @addr passed into pcpu_chunk_addr_search() is unit0 based address
and thus should be matched inside unit0 area.  Currently, when it uses
chunk size when determining whether the address falls in the first
chunk.  Addresses in unitN where N>0 shouldn't be passed in anyway, so
this doesn't cause any malfunction but fix it for consistency.

[ Impact: mostly cleanup ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>

This patch makes most !CONFIG_HAVE_SETUP_PER_CPU_AREA archs use
dynamic percpu allocator.  The first chunk is allocated using
embedding helper and 8k is reserved for modules.  This ensures that
the new allocator behaves almost identically to the original allocator
as long as static percpu variables are concerned, so it shouldn't
introduce much breakage.

s390 and alpha use custom SHIFT_PERCPU_PTR() to work around addressing
range limit the addressing model imposes.  Unfortunately, this breaks
if the address is specified using a variable, so for now, the two
archs aren't converted.

The following architectures are affected by this change.

* sh
* arm
* cris
* mips
* sparc(32)
* blackfin
* avr32
* parisc (broken, under investigation)
* m32r
* powerpc(32)

As this change makes the dynamic allocator the default one,
CONFIG_HAVE_DYNAMIC_PER_CPU_AREA is replaced with its invert -
CONFIG_HAVE_LEGACY_PER_CPU_AREA, which is added to yet-to-be converted
archs.  These archs implement their own setup_per_cpu_areas() and the
conversion is not trivial.

* powerpc(64)
* sparc(64)
* ia64
* alpha
* s390

Boot and batch alloc/free tests on x86_32 with debug code (x86_32
doesn't use default first chunk initialization).  Compile tested on
sparc(32), powerpc(32), arm and alpha.

Kyle McMartin reported that this change breaks parisc.  The problem is
still under investigation and he is okay with pushing this patch
forward and fixing parisc later.

[ Impact: use dynamic allocator for most archs w/o custom percpu setup ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NRusty Russell <rusty@rustcorp.com.au>
Acked-by: NDavid S. Miller <davem@davemloft.net>
Acked-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>
Reviewed-by: NChristoph Lameter <cl@linux.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Bryan Wu <cooloney@kernel.org>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Matthew Wilcox <matthew@wil.cx>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>

According to Andi, it isn't clear whether lpage allocator is worth the
trouble as there are many processors where PMD TLB is far scarcer than
PTE TLB.  The advantage or disadvantage probably depends on the actual
size of percpu area and specific processor.  As performance
degradation due to TLB pressure tends to be highly workload specific
and subtle, it is difficult to decide which way to go without more
data.

This patch implements percpu_alloc kernel parameter to allow selecting
which first chunk allocator to use to ease debugging and testing.

While at it, make sure all the failure paths report why something
failed to help determining why certain allocator isn't working.  Also,
kill the "Great future plan" comment which had already been realized
quite some time ago.

[ Impact: allow explicit percpu first chunk allocator selection ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Reported-by: NJan Beulich <JBeulich@novell.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Ingo Molnar <mingo@elte.hu>

In pcpu_unmap(), flushing virtual cache on vunmap can't be delayed as
the page is going to be returned to the page allocator.  Only TLB
flushing can be put off such that vmalloc code can handle it lazily.
Fix it.

[ Impact: fix subtle virtual cache flush bug ]
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>

Impact: use page->index for addr to chunk mapping instead of dedicated rbtree

The rbtree is used to determine the chunk from the virtual address.
However, we can already determine the page struct from a virtual
address and there are several unused fields in page struct used by
vmalloc.  Use the index field to store a pointer to the chunk. Then
there is no need anymore for an rbtree.

tj: * s/(set|get)_chunk/pcpu_\1_page_chunk/

    * Drop inline from the above two functions and moved them upwards
      so that they are with other simple helpers.

    * Initial pages might not (actually most of the time don't) live
      in the vmalloc area.  With the previous patch to manually
      reverse-map both first chunks, this is no longer an issue.
      Removed pcpu_set_chunk() call on initial pages.
Signed-off-by: NChristoph Lameter <cl@linux.com>
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: rusty@rustcorp.com.au
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: rmk@arm.linux.org.uk
Cc: starvik@axis.com
Cc: ralf@linux-mips.org
Cc: davem@davemloft.net
Cc: cooloney@kernel.org
Cc: kyle@mcmartin.ca
Cc: matthew@wil.cx
Cc: grundler@parisc-linux.org
Cc: takata@linux-m32r.org
Cc: benh@kernel.crashing.org
Cc: rth@twiddle.net
Cc: ink@jurassic.park.msu.ru
Cc: heiko.carstens@de.ibm.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Nick Piggin <npiggin@suse.de>
LKML-Reference: <49D43D58.4050102@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: both first chunks don't use rbtree, no functional change

There can be two first chunks - reserved and dynamic with the former
one being optional.  Dynamic first chunk was linked on reverse-mapping
rbtree while the reserved one was mapped manually using the start
address and reserved offset limit.

This patch makes both first chunks to be looked up manually without
using the rbtree.  This is to help getting rid of the rbtree.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: rusty@rustcorp.com.au
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: rmk@arm.linux.org.uk
Cc: starvik@axis.com
Cc: ralf@linux-mips.org
Cc: davem@davemloft.net
Cc: cooloney@kernel.org
Cc: kyle@mcmartin.ca
Cc: matthew@wil.cx
Cc: grundler@parisc-linux.org
Cc: takata@linux-m32r.org
Cc: benh@kernel.crashing.org
Cc: rth@twiddle.net
Cc: ink@jurassic.park.msu.ru
Cc: heiko.carstens@de.ibm.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Christoph Lameter <cl@linux.com>
LKML-Reference: <49D43CEA.3040609@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Impact: code reorganization

Separate out embedding first chunk setup helper from x86 embedding
first chunk allocator and put it in mm/percpu.c.  This will be used by
the default percpu first chunk allocator and possibly by other archs.
Signed-off-by: NTejun Heo <tj@kernel.org>

Impact: cleanup, more flexibility for first chunk init

Non-negative @dyn_size used to be allowed iff @unit_size wasn't auto.
This restriction stemmed from implementation detail and made things a
bit less intuitive.  This patch allows @dyn_size to be specified
regardless of @unit_size and swaps the positions of @dyn_size and
@unit_size so that the parameter order makes more sense (static,
reserved and dyn sizes followed by enclosing unit_size).

While at it, add @unit_size >= PCPU_MIN_UNIT_SIZE sanity check.
Signed-off-by: NTejun Heo <tj@kernel.org>

Impact: generic addr <-> pcpu ptr conversion macros

There's nothing arch specific about x86 __addr_to_pcpu_ptr() and
__pcpu_ptr_to_addr().  With proper __per_cpu_load and __per_cpu_start
defined, they'll do the right thing regardless of actual layout.

Move these macros from arch/x86/include/asm/percpu.h to mm/percpu.c
and allow archs to override it as necessary.
Signed-off-by: NTejun Heo <tj@kernel.org>

Impact: fix deadlock and allow atomic free

Percpu allocation always uses GFP_KERNEL and whole alloc/free paths
were protected by single mutex.  All percpu allocations have been from
GFP_KERNEL-safe context and the original allocator had this assumption
too.  However, by protecting both alloc and free paths with the same
mutex, the new allocator creates free -> alloc -> GFP_KERNEL
dependency which the original allocator didn't have.  This can lead to
deadlock if free is called from FS or IO paths.  Also, in general,
allocators are expected to allow free to be called from atomic
context.

This patch implements finer grained locking to break the deadlock and
allow atomic free.  For details, please read the "Synchronization
rules" comment.

While at it, also add CONTEXT: to function comments to describe which
context they expect to be called from and what they do to it.

This problem was reported by Thomas Gleixner and Peter Zijlstra.

  http://thread.gmane.org/gmane.linux.kernel/802384Signed-off-by: NTejun Heo <tj@kernel.org>
Reported-by: NThomas Gleixner <tglx@linutronix.de>
Reported-by: NPeter Zijlstra <peterz@infradead.org>