Merge branch 'for-next/misc'

sandbox: reinstate cooked terminal mode on sanitizer-induced death ...

If we exit due to AddressSanitizer error, we remain in raw mode, which
is not best user experience. Currently every exit is an AddressSanitizer
death due to leaks. We want to encourage users to always have ASan
enabled, to catch more errors, thus call cookmode() on exit.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: add libc memory allocator ...

While we typically want to reuse as much barebox functionality as
possible in sandbox, using system malloc(3) instead can be very
useful when using external memory integrity tools. This is even
useful for AddressSanitizer, because the reports resulting from
the memory poisoning API are less detailed than the built-in
support for the libc malloc(3).

Note that a barebox "heap" is still allocated upfront. It's only
used for request_sdram_region now though. Whatever is allocated by
means of malloc and memalign will be ina disjunct heap.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: os: common: fix compiler warning in add_image() ...

GCC reports:

  ./arch/sandbox/os/common.c: In function ‘add_image’:
  ./arch/sandbox/os/common.c:271:6: warning: cast to pointer from integer of
  	different size [-Wint-to-pointer-cast]

This is because hf->base is an unsigned long long, with the upper 32 bit
all-zeroes on 32-bit systems. The compiler doesn't see that though.

Change the cast, so we no longer warn.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: include: <asm/types.h>: don't define INTERNAL_SIZE_T ...

The INTERNAL_SIZE_T default is size_t, which already is 64-bit on 64-bit
platforms and 32-bit on 32-bit ones. We can thus drop the #define and go
back to the minimal <asm/types.h>.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: specify sizeof(dma_addr_t) == sizeof(phys_addr_t) == 8 on 64BIT ...

sizeof(dma_addr_t) == 8 shouldn't introduce any functional change, because
we can't have DMA on sandbox. For now it suppresses benign warnings about
mismatched pointer and integer sizes when some headers are included.

sizeof(phys_addr_t) == 8 was already the case on systems with
__x86_64__. As CONFIG_64BIT now is set according to the bitness
of the compiler, we can migrate this fully to Kconfig.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: fix generation of broken binaries ...

Code under arch/sandbox/os is meant to link against the libc primitives
of the OS, not the barebox counterparts (e.g. for malloc).

54d47eb7e912 ("sandbox: support forcing 32-bit x86") broke this
rendering the sandbox architecture unusable. Fix this.

Reported-by: Peter Mamonov <pmamonov@gmail.com>
Fixes: 54d47eb7e912 ("sandbox: support forcing 32-bit x86")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-next/sandbox-compile-test'

Merge branch 'for-next/misc'

sandbox: os: add_image: fix memory leak ...

devname is strdup'd few lines later, so remove the earlier strdup.
The pointer isn't stored anywhere persistent in between.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: asm: implement <asm/atomic.h> ...

For easier code-sharing, the UBIFS code still uses the Kernel's
atomic accessors. Provide a sandbox <asm/atomic.h> header, so we
can compile-test UBI under sandbox.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: select HAS_DMA ...

With the DMA stubs in place, we can indicate HAS_DMA in good conscience.
Do so.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: asm: implement stub DMA functions ...

For compile-testing drivers that use DMA operations under sandbox,
we want stub implementations of these. Provide them.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: implement stub physical virtual translation ...

For compile-testing drivers that use DMA operations under sandbox,
we want stub implementations of these. Provide them.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: <asm/io.h>: include bitio accessors ...

ARM <asm/io.h> defines the U-Boot style setbits/clrbits helpers.
Have sandbox follow suit to allow more drivers to be compile tested
there.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: support forcing 32-bit x86 ...

If gcc supports multiple architectures, barebox uses only the default
when compiling for ARCH=sandbox, this is e.g. the case with compilers
that generate both 32- and 64-bit x86 executables.

There can be good reasons to force 32-bit though, e.g. to reduce memory
consumption while fuzzing or to temporarily avoid 32-to-64-bit warnings,
while running the static analyzer. Add an option for this.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: asm: bitsperlong.h: detect bitness according to 64BIT symbol ...

Now that we have an accurate CONFIG_64BIT describing sandbox bitness,
we don't need to assume everything except __x86_64__ as 32-bit.

This assumes that sizeof(long) == sizeof(void *), which is assumed
anyway throughout the code base.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: define CONFIG_64BIT as appropriate ...

All 64-bit architectures are supposed to define CONFIG_64BIT to support
the relevant 64-bit MMIO accessors. The sandbox architecture is a bit
of a special case, because barebox uses the toolchain default and
doesn't force a bitness. Add 64BIT as promptless symbol, which reflects
the pointer size of the target platform.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

kbuild: switch over to thin archive ...

Starting v4.13, Linux always uses thin archive instead of relocatable
ELF to combine builtin objects.

The thin archive is basically a text file that only contains paths to
object files.

As Linux commit 98ced886dd79 mentioned, this has a lot of benefits:

 - save disk space for builds

 - speed-up building a little

 - fix some issues when liking a giant executable (for example,
   allyesconfig on ARM Linux)

 - work better with dead code elimination

The last one is important for barebox because pbl highly relies on the
dead code elimination (-f{function,data}-sections and --gc-sections)
dropping symbols that are unreachable from the linker's entry point.
The dead code elimination does not work if the same symbol names are
used in the pevious incremental link mechanism because the same name
symbols all go into the same section.

Commit a83c97f2a406 ("ARM: socfpga: generate smaller images when
multiple boards are selected") worked around it by giving a dedicate
section to each of the same name symbols. This workaround can go away.

built-in.o was renamed to built-in.a since it is now an archive.
built-in-pbl.o was renamed to built-in.pbl.a for consistency.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

arch: include <asm-generic/int-ll64.h> from <asm/types.h> ...

Many architectures duplicate the same fixed type definitions
as in <asm-generic/int-ll64.h>.

Include <asm-generic/int-ll64.h> from <asm/types.h>, and remove
the duplicated code.

Note:
Linux commit 0c79a8e29b5f ("asm/types.h: Remove
include/asm-generic/int-l64.h") unified the kernel space
definition into int-ll64.h. Barebox also adopts int-ll64.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Merge branch 'for-next/misc'

arch: remove KBUILD_SRC checks from arch Makefile ...

You can always add $(srctree)/ to the mach include path.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

kbuild: prefix compiler flag variables with KBUILD_ ...

In old days, Linux also used to use well-known variables such CFLAGS,
CPPFLAGS, etc.

They were prefixed with KBUILD_ presumably for preventing users from
overriding them accidentally.

Rename as follows:
  CFLAGS          ->  KBUILD_CFLAGS
  AFLAGS          ->  KBUILD_AFLAGS
  CPPFLAGS        ->  KBUILD_CPPFLAGS
  LDFLAGS         ->  KBUILD_LDFLAGS
  HOSTCFLAGS      ->  KBUILD_HOSTCFLAGS
  HOSTCXXFLAGS    ->  KBUILD_HOSTCXXFLAGS
  HOSTLDFLAGS     ->  KBUILD_HOSTLDFLAGS
  HOST_LOADLIBES  ->  KBUILD_HOSTLDLIBS

HOSTCFLAGS, HOSTCXXFLAGS, HOSTLDFLAGS, HOSTLDLIBS are re-used to allow
users to pass-in additional flags to the host compiler.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

treewide: remove references to CREDITS ...

The CREDITS file was removed from barebox in 2015 by commit 6570288f2d97
("Remove the CREDITS file"). Remove references to it from several files.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

kbuild: rename barebox-{common, lds} to BAREBOX_{OBJS, LDS} ...

The previous commit renamed barebox-pbl-common because you cannot
export a variable that contain hyphens.

These variables are not exported for now. I am just renaming them for
consistency.

  barebox-common -> BAREBOX_OBJS
  barebox-lds    -> BAREBOX_LDS

If barebox moves the link rules into scripts/link-barebox.sh like
Linux did, these two must be exported anyway.

Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: hostfile: fix UB-inducing shift ...

On 32-bit hosts, UBSan reports:

====================================================================
UBSAN: Undefined behaviour in ./arch/sandbox/board/hostfile.c:135:12
shift exponent 32 is too large for 32-bit type 'long unsigned int'
====================================================================

Fix this by using a unsigned long long for base. The upper 32-bit
won't be set, but device tree fixups can now read them.

Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: fix signed integer overflow ...

On a 32-bit host system, UBSan reports:

===============================================================
UBSAN: Undefined behaviour in ./arch/sandbox/os/common.c:115:32
signed integer overflow:
83598 * 1000000000 cannot be represented in type 'long int'
===============================================================

Fix this.

Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: add_image: mmap block devices ...

This makes it possible to mount block devices from the host machine,
which have been passed as arguments to --image

Signed-off-by: Robert Karszniewicz <r.karszniewicz@phytec.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

sandbox: add_image: output errno string on mmap failure ...

Signed-off-by: Robert Karszniewicz <r.karszniewicz@phytec.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

hwrng: dev-random: always use /dev/urandom ...

/dev/random can block long after boot time. It seems there's a consensus
that /dev/urandom is safe to use except for very early boot, which isn't
when barebox sandbox is usually run. To make the HWRNG more useful,
always use /dev/urandom.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>