When CONFIG_DEBUG_USER is enabled, it's possible for a user to
deliberately trigger dump_instr() with a chosen kernel address.

Let's avoid problems resulting from this by using get_user() rather than
__get_user(), ensuring that we don't erroneously access kernel memory.

So that we can use the same code to dump user instructions and kernel
instructions, the common dumping code is factored out to __dump_instr(),
with the fs manipulated appropriately in dump_instr() around calls to
this.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

The reworked MPU code produces a new warning in some configurations,
presumably starting with the code move after the compiler now makes
different inlining decisions:

arch/arm/mm/pmsa-v7.c: In function 'adjust_lowmem_bounds_mpu':
arch/arm/mm/pmsa-v7.c:310:5: error: 'specified_mem_size' may be used uninitialized in this function [-Werror=maybe-uninitialized]

This appears to be harmless, as we know that there is always at
least one memblock, and the only way this could get triggered is
if the for_each_memblock() loop was never entered.

I could not come up with a better workaround than initializing
the specified_mem_size to zero, but at least that is the value
that the variable would have in the hypothetical case of no
memblocks.

Fixes: 877ec119 ("ARM: 8706/1: NOMMU: Move out MPU setup in separate module")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Some terminals apparently have issues with "\n\r" and mess up the
display. Let's use the traditional "\r\n" ordering.

Signed-off-by: Nicolas Pitre <nico@linaro.org>
Reported-by: Chris Brandt <Chris.Brandt@renesas.com>
Tested-by: Chris Brandt <Chris.Brandt@renesas.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Add an additional symbol to the decompressor image, which will allow
future debugging of non-bootable problems similar to the one encountered
with the EFI stub.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

ARM depends on the macros '__ARMEL__' & '__ARMEB__' being defined
or not to correctly select or define endian-specific macros,
structures or pieces of code.

These macros are predefined by the compiler but sparse knows
nothing about them and thus may pre-process files differently
from what gcc would.

Fix this by passing '-D__ARMEL__' or '-D__ARMEB__' to sparse,
depending on the endianness of the kernel, like defined by GCC.

Note: In most case it won't change anything since most ARMs use
      little-endian (but an allyesconfig would use big-endian!).

To: Russell King <linux@armlinux.org.uk>

Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

ARM shares its EFI stub implementation with arm64, which has some
special handling in the virtual remapping code to
a) make sure that we can map everything even if the OS executes
   with 64k page size, and
b) make sure that adjacent regions with the same attributes are not
   reordered or moved apart in memory.

The latter is a workaround for a 'feature' that was shortly recommended
by UEFI spec v2.5, but deprecated shortly after, due to the fact that
it broke many OS installers, including non-Linux ones, and it was never
widely implemented for ARM systems. Before implementing b), the arm64
code simply rounded up all regions to 64 KB granularity, but given that
that results in moving adjacent regions apart, it had to be refined when
b) was implemented.

The adjacency check requires a sort() pass, due to the fact that the
UEFI spec does not mandate any ordering, and the inclusion of the
lib/sort.c code into the ARM EFI stub is causing some trouble with
the decompressor build due to the fact that its EXPORT_SYMBOL() call
triggers the creation of ksymtab/kcrctab sections.

So let's simply do away with the adjacency check for ARM, and simply put
all UEFI runtime regions together if they have the same memory attributes.
This is guaranteed to work, given that ARM only supports 4 KB pages,
and allows us to remove the sort() call entirely.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Tested-by: Jeffy Chen <jeffy.chen@rock-chips.com>
Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Tested-by: Matthias Brugger <matthias.bgg@gmail.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

The asm-generic/unaligned.h header provides two different implementations
for accessing unaligned variables: the access_ok.h version used when
CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS is set pretends that all pointers
are in fact aligned, while the le_struct.h version convinces gcc that the
alignment of a pointer is '1', to make it issue the correct load/store
instructions depending on the architecture flags.

On ARMv5 and older, we always use the second version, to let the compiler
use byte accesses. On ARMv6 and newer, we currently use the access_ok.h
version, so the compiler can use any instruction including stm/ldm and
ldrd/strd that will cause an alignment trap. This trap can significantly
impact performance when we have to do a lot of fixups and, worse, has
led to crashes in the LZ4 decompressor code that does not have a trap
handler.

This adds an ARM specific version of asm/unaligned.h that uses the
le_struct.h/be_struct.h implementation unconditionally. This should lead
to essentially the same code on ARMv6+ as before, with the exception of
using regular load/store instructions instead of the trapping instructions
multi-register variants.

The crash in the LZ4 decompressor code was probably introduced by the
patch replacing the LZ4 implementation, commit 4e1a33b1 ("lib: update
LZ4 compressor module"), so linux-4.11 and higher would be affected most.
However, we probably want to have this backported to all older stable
kernels as well, to help with the performance issues.

There are two follow-ups that I think we should also work on, but not
backport to stable kernels, first to change the asm-generic version of
the header to remove the ARM special case, and second to review all
other uses of CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS to see if they
might be affected by the same problem on ARM.

Cc: stable@vger.kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Currently, there is assumption in early MPU setup code that kernel
image is located in RAM, which is obviously not true for XIP. To run
code from ROM we need to make sure that it is covered by MPU. However,
due to we allocate regions (semi-)dynamically we can run into issue of
trimming region we are running from in case ROM spawns several MPU
regions. To help deal with that we enforce minimum alignments for start
end end of XIP address space as 1MB and 128Kb correspondingly.

Tested-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

PMSAv7 defines curious alignment requirements to the regions:
- size must be power of 2, and
- region start must be aligned to the region size

Because of that we currently adjust lowmem bounds plus we assign
only one MPU region to cover memory all these lead to significant amount of
memory could be wasted. As an example, consider 64Mb of memory at
0x70000000 - it fits alignment requirements nicely; now, imagine that
2Mb of memory is reserved for coherent DMA allocation, so now Linux is
expected to see 62Mb of memory... and here annoying thing happens -
memory gets truncated to 32Mb (we've lost 30Mb!), i.e. MPU layout
looks like:

0: base 0x70000000, size 0x2000000

This patch tries to allocate as much as possible MPU slots to minimise
amount of truncated memory. Moreover, with this patch MPU subregions
starting to get used. MPU subregions allow us reduce the number of MPU
slots used. For example given above, MPU layout looks like:

0: base 0x70000000, size 0x2000000
1: base 0x72000000, size 0x1000000
2: base 0x73000000, size 0x1000000, disable subreg 7 (0x73e00000 - 0x73ffffff)

Where without subregions we'd get:

0: base 0x70000000, size 0x2000000
1: base 0x72000000, size 0x1000000
2: base 0x73000000, size 0x800000
3: base 0x73800000, size 0x400000
4: base 0x73c00000, size 0x200000

To achieve better layout we fist try to cover specified memory as is
(maybe with help of subregions) and if we failed, we truncate memory
to fit alignment requirements (so it occupies one MPU slot) and
perform one more attempt with the reminder, and so on till we either
cover all memory or run out of MPU slots.

Tested-by: Szemző András <sza@esh.hu>
Tested-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

This patch makes it possible to use MPU with v7M cores.

Tested-by: Szemző András <sza@esh.hu>
Tested-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

The last user of CONFIG_VECTORS_BASE has gone, so kill it.

Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Reported-by: Afzal Mohammed <afzal.mohd.ma@gmail.com>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

It seems that MPU never worked with XIP, so we just disallow such
combination.

Tested-by: Szemző András <sza@esh.hu>
Tested-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Currently, there are several issues with how MPU is setup:

 1. We won't boot if MPU is missing
 2. We won't boot if use XIP
 3. Further extension of MPU setup requires asm skills

The 1st point can be relaxed, so we can continue with boot CPU even if
MPU is missed and fail boot for secondaries only. To address the 2nd
point we could create region covering CONFIG_XIP_PHYS_ADDR - _end and
that might work for the first stage of MPU enable, but due to MPU's
alignment requirement we could cover too much, IOW we need more
flexibility in how we're partitioning memory regions... and it'd be
hardly possible to archive because of the 3rd point.

This patch is trying to address 1st and 3rd issues and paves the path
for 2nd and further improvements.

The most visible change introduced with this patch is that we start
using mpu_rgn_info array (as it was supposed?), so change in MPU setup
done by boot CPU is recorded there and feed to secondaries. It
allows us to keep minimal region setup for boot CPU and do the rest in
C. Since we start programming MPU regions in C evaluation of MPU
constrains (number of regions supported and minimal region order) can
be done once, which in turn open possibility to free-up "probe"
region early.

Tested-by: Szemző András <sza@esh.hu>
Tested-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Currently, inline assembly for accessing to MPU's cp15 lacks volatile
keyword which opens possibility to compiler to optimise such accesses
as soon as we start using them more intensively. Rather than fixing
inline asm, lets move MPU accessors to use cp15 helpers which do the
right thing.

Tested-by: Szemző András <sza@esh.hu>
Tested-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Having MPU handling code in dedicated module makes it easier to
enhance/maintain it.

Tested-by: Szemző András <sza@esh.hu>
Tested-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

If CONFIG_DEBUG_LOCK_ALLOC=y, the kernel log is spammed with a few
hundred identical messages:

    unwind: Unknown symbol address c0800300
    unwind: Index not found c0800300

c0800300 is the return address from the last subroutine call (to
__memzero()) in __mmap_switched().  Apparently having this address in
the link register confuses the unwinder.

To fix this, reset the link register to zero before jumping to
start_kernel().

Fixes: 9520b1a1 ("ARM: head-common.S: speed up startup code")
Suggested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

With printch() the console messages are sent out one character at a time
which is agonizingly slow especially with semihosting as the whole trap
intercept, remote byte access, and system resume danse is performed for
every single character across a relatively slow remote debug connection.

Let's use printascii() to send a whole string at once. This is also going
to be more efficient, albeit to a quite lesser extent, with serial ports
as well.

Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

The svc instruction doesn't exist on v7m processors. Semihosting ops are
invoked with the bkpt instruction instead.

Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

This was located in .text which is meant to be read-only. And in the XIP
case this shortcut simply doesn't work and may trigger a Flash controller
mode switch and crash the kernel. Move it to the .bss area.

Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

By default sparse uses the characteristics of the build
machine to infer things like the wordsize.
This is fine when doing native builds but for ARM it's,
I suspect, very rarely the case and if the build are done
on a 64bit machine we get a bunch of warnings like:
  'cast truncates bits from constant value (... becomes ...)'

Fix this by adding the -m32 flags for sparse.

Reported-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Luc Van Oostenryck <luc.vanoostenryck@gmail.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Some nommu systems have RAM at address 0. When vectors are not located
there, the very beginning of memory remains available for dynamic
allocations. The memblock allocator explicitly skips the first page
but the standard page allocator does not, and while it correctly returns
a non-null struct page pointer for that page, page_address() gives 0
which gets confused with NULL (out of memory) by callers despite having
plenty of free memory left.

Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Add an additional extendable table to the compressed kernel so that we
can provide further information to boot loaders regarding the properties
of the image contained within.

This is necessary for correct behaviour of kexec.

Tested-by: Tony Lindgren <tony@atomide.com>
Tested-by: Keerthy <j-keerthy@ti.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Assuming size(1) gives the size of the BSS is a mistake - it reports
the size of the .bss section in the ELF image, which may not be the
same as the region we mark with the __bss_start..__bss_stop symbols.

We use the size of the BSS in the decompressor to know whether the
kernel will overwrite the appended dtb, by adding the BSS size to the
size of the Image (stored at the end of the compressed data) and adding
the desired address of the decompressed image.

If the BSS size is smaller than it really is, the decompressor can
incorrectly assume that the BSS clearance will not overwrite the DTB.

Here is an illustration:

$ arm-linux-size vmlinux
   text    data      bss      dec     hex filename
8136972 3098076 10240348 21475396 147b044 vmlinux
$ arm-linux-nm vmlinux | grep __bss_
c0ac0e34 B __bss_start
c1484f9c B __bss_stop
$ stat -c %s arch/arm/boot/Image
11243060

In the above case, we are 12 bytes short.  This is caused by the BSS
section being aligned by one of its input sections:

Idx Name          Size      VMA       LMA       File off  Algn
 23 __bug_table   00005d3c  c0abb0f8  c0abb0f8  00acb0f8  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 24 .bss          009c415c  c0ac0e40  c0ac0e40  00ad0e34  2**6
                  ALLOC

Note that there's an additional 12 bytes difference between the file
offset and LMA compared with the bug table - this occurs because one
of the input sections for the .bss section requires a 64 byte
alignment.

Fix this by using 'nm' and perl to obtain the address of the __bss_start
and __bss_stop symbols, using their difference for the size of the BSS.

Tested-by: Tony Lindgren <tony@atomide.com>
Tested-by: Keerthy <j-keerthy@ti.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Remove the special SA1111 MMIO accessors from the SA1111 PCMCIA driver
as their definition will be removed shortly.  The SA1111 accessors are
barrierless, so use the _relaxed variants.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Use the newly provided sa1111_get_irq() to fetch the IRQ resources for
the SA1111 PCMCIA driver.

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

This series provides the needed changes to suport the ELF_FDPIC binary
format on ARM. Both MMU and non-MMU systems are supported. This format
has many advantages over the BFLT format used on MMU-less systems, such
as being real ELF that can be parsed by standard tools, can support
shared dynamic libs, etc.

With a kernel containing both DT and atag support, the diagnostics
output when the dtb is missing or corrupt assume that we're trying
to boot using atags and the machine ID, and only print the machine
ID.  This is not useful for diagnosing a missing or corrupt dtb.

Move the message into arch/arm/kernel/setup.c, and print the address
of the dtb/atag list, and the first 16 bytes of memory of the dtb or
atag list.

This allows us to see whether the dtb was corrupted in some way,
causing the fallback to the machine ID / atag list.

Tested-by: Keerthy <j-keerthy@ti.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

This contains important fixes to the XIP linker script, some more linker
script cleanups, .bss clearing and .data copying speedups related to the
above, and an opt-in config option for XIP kernels that allows for
compressing .data in ROM that depend on those other patches to work
properly.

There are no users of init_dma_coherent_pool_size() left due to
387870f2 ("mm: dmapool: use provided gfp flags for all
dma_alloc_coherent() calls"), so remove it.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

atomic_pool is setup once while init stage and never changed after
that, so it is good candidate for __ro_after_init.

Since we are here mark atomic_pool_size with __init_data.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

gen_pool_first_fit_order_align() does not make use of additional data,
so pass plain NULL there.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

The code in question checks memory constrains to set default policy for
overcommit; however we support page size of 4K only thus condition is
always evaluated to false. Remove that dead code.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

We support page size of 4K only, remove dead code.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

fixmap_page_table was removed by commit 836a2418 (ARM: expand
fixmap region to 3MB), but some traces are still there - get rid of
them.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

On ARM the generic pfn_valid() version is used with some configurations
such as SA1100 based devices. In that case the memblock arrays are no
longer used after boot and can be discarded.

Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>

Pull UBI updates from Richard Weinberger:
 "Minor improvements"

* tag 'upstream-4.14-rc1' of git://git.infradead.org/linux-ubifs:
  UBI: Fix two typos in comments
  ubi: fastmap: fix spelling mistake: "invalidiate" -> "invalidate"
  ubi: pr_err() strings should end with newlines
  ubi: pr_err() strings should end with newlines
  ubi: pr_err() strings should end with newlines

Pull UML updates from Richard Weinberger:

 - minor improvements

 - fixes for Debian's new gcc defaults (pie enabled by default)

 - fixes for XSTATE/XSAVE to make UML work again on modern systems

* 'for-linus-4.14-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml:
  um: return negative in tuntap_open_tramp()
  um: remove a stray tab
  um: Use relative modversions with LD_SCRIPT_DYN
  um: link vmlinux with -no-pie
  um: Fix CONFIG_GCOV for modules.
  Fix minor typos and grammar in UML start_up help
  um: defconfig: Cleanup from old Kconfig options
  um: Fix FP register size for XSTATE/XSAVE