On OMAP platforms, some people want to declare to segment up the memory
between the kernel and a separate application such that there is a hole
in the middle of the memory as far as Linux is concerned.  However,
they want to be able to mmap() the hole.

This currently causes problems, because update_mmu_cache() thinks that
there are valid struct pages for the "hole".  Fix this by making
pfn_valid() slightly more expensive, by checking whether the PFN is
contained within the meminfo array.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>
Tested-by: NKhasim Syed Mohammed <khasim@ti.com>

If a machine class has a custom __virt_to_bus() implementation then it
must provide a __arch_page_to_dma() implementation as well which is
_not_ based on page_address() to support highmem.

This patch fixes existing __arch_page_to_dma() and provide a default
implementation otherwise.  The default implementation for highmem is
based on __pfn_to_bus() which is defined only when no custom
__virt_to_bus() is provided by the machine class.

That leaves only ebsa110 and footbridge which cannot support highmem
until they provide their own __arch_page_to_dma() implementation.
But highmem support on those legacy platforms with limited memory is
certainly not a priority.
Signed-off-by: NNicolas Pitre <nico@marvell.com>

The kmap virtual area borrows a 2MB range at the top of the 16MB area
below PAGE_OFFSET currently reserved for kernel modules and/or the
XIP kernel.  This 2MB corresponds to the range covered by 2 consecutive
second-level page tables, or a single pmd entry as seen by the Linux
page table abstraction.  Because XIP kernels are unlikely to be seen
on systems needing highmem support, there shouldn't be any shortage of
VM space for modules (14 MB for modules is still way more than twice the
typical usage).

Because the virtual mapping of highmem pages can go away at any moment
after kunmap() is called on them, we need to bypass the delayed cache
flushing provided by flush_dcache_page() in that case.

The atomic kmap versions are based on fixmaps, and
__cpuc_flush_dcache_page() is used directly in that case.
Signed-off-by: NNicolas Pitre <nico@marvell.com>

Let's provide an overridable default instead of having every machine
class define __virt_to_bus and __bus_to_virt to the same thing.  What
most platforms are using is bus_addr == phys_addr so such is the default.

One exception is ebsa110 which has no DMA what so ever, so the actual
definition is not important except only for proper compilation.  Also
added a comment about the special footbridge bus translation.

Let's also remove comments alluding to set_dma_addr which is not
(and should not) be commonly used.
Signed-off-by: NNicolas Pitre <nico@marvell.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

There is no machine class overriding this.  If non linear translations
are implemented again for some machines then this could be restored at
that time.
Signed-off-by: NNicolas Pitre <nico@marvell.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

As of 73bdf0a6, the kernel needs
to know where modules are located in the virtual address space.
On ARM, we located this region between MODULE_START and MODULE_END.
Unfortunately, everyone else calls it MODULES_VADDR and MODULES_END.
Update ARM to use the same naming, so is_vmalloc_or_module_addr()
can work properly.  Also update the comment on mm/vmalloc.c to
reflect that ARM also places modules in a separate region from the
vmalloc space.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Most ARM machines don't need a special "DMA" memory zone, and
when configured out, the kernel becomes a bit smaller:

|   text    data     bss     dec     hex filename
|3826182  102384  111700 4040266  3da64a vmlinux
|3823593  101616  111700 4036909  3d992d vmlinux.nodmazone

This is because the system now has only one zone total which effect is
to optimize away many conditionals in page allocation paths.

So let's configure this zone only on machines that need split zones.
Signed-off-by: NNicolas Pitre <nico@marvell.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

There's no point scattering this around the tree, the parsing
of the parameter might as well live beside the code which uses
it.  That also means we can make vmalloc_reserve a static
variable.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

This patch adds a config option (CONFIG_VMSPLIT_*) to allow choosing
between 3:1, 2:2 and 1:3 user:kernel memory splits.
Tested-by: NRiku Voipio <riku.voipio@iki.fi>
Signed-off-by: NLennert Buytenhek <buytenh@marvell.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

OMAP at least gets the return type(s) for the DMA translation functions
wrong, which can lead to subtle errors.  Avoid this by moving the DMA
translation functions to asm/dma-mapping.h, and converting them to
inline functions.

Fix the OMAP DMA translation macros to use the correct argument and
result types.

Also, remove the unnecessary casts in dmabounce.c.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

This patch will truncate and/or ignore memory banks if their kernel
direct mappings would (partially) overlap with the vmalloc area or
the mappings between the vmalloc area and the address space top, to
prevent crashing during early boot if there happens to be more RAM
installed than we are expecting.

Since the start of the vmalloc area is not at a fixed address (but
the vmalloc end address is, via the per-platform VMALLOC_END define),
a default area of 128M is reserved for vmalloc mappings, which can
be shrunk or enlarged by passing an appropriate vmalloc= command line
option as it is done on x86.

On a board with a 3:1 user:kernel split, VMALLOC_END at 0xfe000000,
two 512M RAM banks and vmalloc=128M (the default), this patch gives:

	Truncating RAM at 20000000-3fffffff to -35ffffff (vmalloc region overlap).
	Memory: 512MB 352MB = 864MB total

On a board with a 3:1 user:kernel split, VMALLOC_END at 0xfe800000,
two 256M RAM banks and vmalloc=768M, this patch gives:

	Truncating RAM at 00000000-0fffffff to -0e7fffff (vmalloc region overlap).
	Ignoring RAM at 10000000-1fffffff (vmalloc region overlap).
Signed-off-by: NLennert Buytenhek <buytenh@marvell.com>
Tested-by: NRiku Voipio <riku.voipio@iki.fi>

This just leaves include/asm-arm/plat-* to deal with.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Move platform independent header files to arch/arm/include/asm, leaving
those in asm/arch* and asm/plat* alone.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Most architectures have fairly simple discontiguous memory - a
simple set of successive regions each containing some memory.
These can be described simply as a log2 of their maximum size,
along with the base address of the first region and the number
of regions.

The base address is already described by PHYS_PFN_OFFSET, and
the number of regions via the MAX_NUMNODES and the number of
online nodes.

If we then supply the log2 of their maximum size, all the other
discontigmem macros can move into generic code.

There is one exception: lh7a40x seems to have a more complicated
setup; this is left alone.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Patch from Lennert Buytenhek

Analogous to the previous patch that allows ioremap() to use section
mappings, this patch allows ioremap() to use supersection mappings.
Original patch by Deepak Saxena.
Signed-off-by: NLennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Allow section mappings to be setup using ioremap() and torn down
with iounmap().  This requires additional support in the MM
context switch to ensure that mappings are properly synchronised
when mapped in.

Based an original implementation by Deepak Saxena, reworked and
ARMv6 support added by rmk.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Majorily based on Hyok Choi's patches, this fixes up the asm-arm
header files for mmuless systems.  Over and above Hyok's patches:

- nommu.h merged into mmu.h (it's only a structure)
- nommu_context.h is essentially the same as mmu_context.h, but
  without the MM switching code.

so there's no point having separate files.  Also, in memory.h,
there's no point #ifndef'ing PHYS_OFFSET and END_MEM - both
CONFIG_DRAM_BASE and CONFIG_DRAM_SIZE will always be set by the
configuration scripts.

Other files have minor formatting changes, but are essentially
the same.  Hyok's original patches were signed off thusly:
Signed-off-by: NHyok S. Choi <hyok.choi@samsung.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Signed-off-by: NDavid Woodhouse <dwmw2@infradead.org>

Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

This patch fixes arch_local_page_offset(pfn,nid) in arm.
This new one (added by unify_pfn_to_page patches) is obviously buggy.

This macro calculate page offset in a node.

Note: about LOCAL_MAP_NR()
comment in arm's sub-archs says...

 /*
  * Given a kaddr, LOCAL_MAP_NR finds the owning node of the memory
  * and returns the index corresponding to the appropriate page in the
  * node's mem_map.
  */

but LOCAL_MAP_NR() is designed to be able to take both paddr and kaddr.
In this case, paddr is better.
Signed-off-by: NKAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitu.com>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Patch from Lennert Buytenhek

This patch adds support for the I/O coherent cache available on the
xsc3.  The approach is to provide a simple API to determine whether the
chipset supports coherency by calling arch_is_coherent() and then
setting the appropriate system memory PTE and PMD bits.  In addition,
we call this API on dma_alloc_coherent() and dma_map_single() calls.
A generic version exists that will compile out all the coherency-related
code that is not needed on the majority of ARM systems.

Note that we do not check for coherency in the dma_alloc_writecombine()
function as that still requires a special PTE setting.  We also don't
touch dma_mmap_coherent() as that is a special ARM-only API that is by
definition only used on non-coherent system.
Signed-off-by: NDeepak Saxena <dsaxena@plexity.net>
Signed-off-by: NLennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

ARM can use generic funcs.
PFN_TO_NID, LOCAL_MAP_NR are defined by sub-archs.
Signed-off-by: NKAMEZAWA Hirotuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Patch from Kevin Hilman

This patch increase available DMA-consistent memory allocated by dma_coherent_alloc(). The default remains at 2M (defined in asm/memory.h) and each platform has the ability to override in asm/arch-foo/memory.h.
Signed-off-by: NKevin Hilman <kevin@hilman.org>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Patch from Deepak Saxena

In working on adding 36-bit addressed supersection support to ioremap(),
I came to the conclusion that it would be far simpler to do so by just
splitting __ioremap() into a main external interface and adding an
__ioremap_pfn() function that takes a pfn + offset into the page that
__ioremap() can call. This way existing callers of __ioremap() won't have
to change their code and 36-bit systems will just call __ioremap_pfn()
and we will not have to deal with unsigned long long variables.

Note that __ioremap_pfn() should _NOT_ be called directly by drivers
but is reserved for use by arch_ioremap() implementations that map
32-bit resource regions into the real 36-bit address and then call
this new function.
Signed-off-by: NDeepak Saxena <dsaxena@plexity.net>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Patch from Hiroki Kaminaga

This patch defines a new macro: pfn_to_kaddr(pfn).
Same macro is already defined on other arch, such as i386.
Signed-off-by: NHiroki Kaminaga <kaminaga@sm.sony.co.jp>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Patch from Nicolas Pitre

Since vmlinux.lds.S is preprocessed, we can use the defines already
present in asm/memory.h (allowed by patch #3060) for the XIP kernel link
address instead of relying on a duplicated Makefile hardcoded value, and
also get rid of its dependency on awk to handle it at the same time.

While at it let's clean XIP stuff even further and make things clearer
in head.S with a nice code reduction.
Signed-off-by: NNicolas Pitre <nico@cam.org>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Patch from Nicolas Pitre

This patch allows for assorted type of cleanups by letting assembly code
use the same set of defines for constant values and avoid duplicated
definitions that might not always be in sync, or that might simply be
confusing due to the different names for the same thing.
Signed-off-by: NNicolas Pitre <nico@cam.org>
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

Thomas Gleixner reported that mmaping and unmapping each physical
page in turn eventually caused the kernel to oops.  It appears
that pfn_valid() in the discontigmem case was too simplistic for
proper operation.

Tighten the logic so we also check if the PFN is within the range
of the selected memory node.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>

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!