| Commit message (Collapse) | Author | Age | Files | Lines |
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
There is value in testing LED trigger conditions in sandbox,
especially with the coming bthread work that should replace pollers,
like the one used by the heartbeat trigger.
To make this testable on sandbox as well, add a simple LED driver.
It supports 5 brightness levels, which are reflected as dots in
the terminal title. This requires a Xterm compatible terminal,
but the LED is inactive by default, so other terminals aren't
affected if they lack support.
Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
For the sandbox architecture, we use __sanitizer_start_switch_fiber
and __sanitizer_finish_switch_fiber to tell ASan when we switch stacks.
If we don't, ASan complains that:
==2472828==WARNING: ASan is ignoring requested __asan_handle_no_return:
stack top: 0xff9fc000; bottom 0xf3be8000; size: 0x0be14000 (199311360)
False positive error reports may follow
For details see https://github.com/google/sanitizers/issues/189
This works on 64-bit sandbox, but 32-bit sandbox currently crashes on
bthread -v:
==2469590==AddressSanitizer CHECK failed: ../../../../../src/libsanitizer/asan/asan_poisoning.cpp:37 "((AddrIsAlignedByGranularity(addr + size))) != (0)" (0x0, 0x0)
#0 0xf7a4aa46 in AsanCheckFailed ../../../../../src/libsanitizer/asan/asan_rtl.cpp:73
#1 0xf7a6b5cf in __sanitizer::CheckFailed(char const*, int, char const*, unsigned long long, unsigned long long) ../../../../../src/libsanitizer/sanitizer_common/sanitizer_termination.cpp:78
#2 0xf7a4489f in __asan::PoisonShadow(unsigned long, unsigned long, unsigned char) ../../../../../src/libsanitizer/asan/asan_poisoning.cpp:37
#3 0xf7a4c81b in __asan_handle_no_return ../../../../../src/libsanitizer/asan/asan_rtl.cpp:595
#4 0x566a1ce7 in bthread_schedule /home/a3f/dl/barebox-stm32mp/common/bthread.c:178
#5 0x566a1d54 in bthread_reschedule /home/a3f/dl/barebox-stm32mp/common/bthread.c:165
#6 0x566a1d80 in bthread_trampoline /home/a3f/dl/barebox-stm32mp/common/bthread.c:56
#7 0x567f5bfb in coroutine_bootstrap (/home/a3f/dl/build/barebox/sandbox/barebox+0x1bdbfb)
#8 0x567f5c4b in coroutine_trampoline (/home/a3f/dl/build/barebox/sandbox/barebox+0x1bdc4b)
#9 0xf7f7056f (linux-gate.so.1+0x56f)
#10 0xf7f70558 (linux-gate.so.1+0x558)
#11 0x56892fff (/home/a3f/dl/build/barebox/sandbox/barebox+0x25afff)
Just disable the special ASan accounting there until this is figured
out. bthreads still function there, but ASan may yield false positives
according to the message. This does not affect non-sandbox platforms.
Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
To extend incoming bthread support to sandbox, implement setjmp, longjmp
and initjmp. Unlike bare metal platforms, setjmp() and longjmp() are
readily provided on standard-conforming hosted platforms. initjmp() on
the other hand requires us to be able to invoke a function with a
user-supplied stack pointer, which isn't possible in standard C.
For POSIX systems, there are two methods to portably achieve this though:
- Use makecontext(2) to set up a new context. makecontext(2) was however
removed in POSIX.1-2008 and at least GCC 10.2.1 ASan complains that it
"doesn't fully support makecontext/swapcontext functions and may
produce false positives in some cases!"
- Use sigaltstack to set a new signal stack, raise the signal call,
setjmp in the signal handler to store the new stack pointer, return
regularly from signal handler and then longjmp back
Both methods are implemented in QEMU. While QEMU uses the makecontext
method by default, for the reasons described, import the second implementation
and use it implement initjmp.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|/
|
|
|
|
|
|
| |
fd could be uninitialized in some error paths. Give it a value that
close can be called on without adverse effect.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
SDL 2.0.4 features SDL_QueueAudio, which we could use to just queue
buffers for audio playback. Use it to implement the new sound API for
the sandbox architecture.
Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
SDL 2.0 has been released more than 7 years ago and should be available
everywhere. Replace barebox's SDL 1.2 with SDL 2.0 instead:
- better maintained
- At least on Debian, you can't install x86_64 and i386 libsdl-dev at
the same time, which makes a quick use of CONFIG_SANDBOX_LINUX_I386
harder as the user needs to reinstall libraries. With SDL 2.0, it
works
- SDL 2.0 has easier audio API, which will be used in a later commit
- Wayland support for sandbox video driver
Port to SDL 2.0 and as we touch everything anyway, rename the
sdl_ functions to start with sdl_video_ to differentiate from upcoming
sdl_sound_.
Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This allows an easy way to test upcoming changes to the barebox block
layer to support 64-bit block IO.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
4M of malloc space is quite limited when comparing this to the systems
that run Linux and use barebox for bootstrap. Increase the malloc size
in sandbox to be better able to test code that does bigger allocations.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
barebox' use of int for the sector offset puts an upper bound of 1TB on
the size of supported block devices, which is already exceeded by common
place USB mass storage. Increasing the sizes involved to 64 bit
like Linux does won't magically add missing driver support, but it gives
us at least a fighting chance. Do so.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |/
|/|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If barebox detects that an --image file is a block device, it will try
to represent it as a block device internally as well, provided that
the user didn't explicitly specify that it should be mapped as a
character device instead. The hf_info::is_cdev member used to indicate
this override wasn't initialized though, so detection was unreliable,
Fix this.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|/
|
|
|
|
|
|
|
| |
`make hosttools_defconfig` shows that PARTITION_DISK was selected
without all its dependencies being enabled as well.
Select the missing PARTITION symbol to fix this.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
5211e428cbab ("image: Convert the IH_... values to enums") removed the
IH_ARCH_LINUX enumeration value leading to breakage of sandbox
configurations that enable bootm. Instead of reinstating IH_ARCH_LINUX,
just use IH_ARCH_SANDBOX and remove mention of CONFIG_LINUX altogether,
it's always true anyway when CONFIG_SANDBOX is true.
Fixes: 5211e428cbab ("image: Convert the IH_... values to enums")
Cc: Antony Pavlov <antonynpavlov@gmail.com>
Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
CONFIG_MXS_HOSTTOOLS currently only builds the tools mxsimage and
mxsboot, which are both only needed to create bootable images for the
MXS platform, and are not really useful as stand-alone host tools.
Signed-off-by: Roland Hieber <rhi@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|/
|
|
|
|
|
|
|
|
|
|
|
| |
dc7f1fce6747 ("sandbox: fix SANDBOX_UNWIND dependency to be KASAN
only") already restricted ARCH_HAS_STACK_DUMP to be dependent
only on AddressSanitizer being available.
This change got lost by an erroneous merge conflict resolution in
39bdcdfb814a ("Merge branch 'for-next/misc' into master"). Fix it up.
Fixes: 39bdcdfb814a ("Merge branch 'for-next/misc' into master")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Recent commits have modified sandbox for easier support of many common
features:
- block devices
- watchdogs
- reset source
- reboot mode
- barebox state
Enable those in defconfig along with some other useful options.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
For testing barebox-state, it would be nice to have a state
pre-configured out-of-the-box. Add one to sandbox.dts.
Because the barebox message on an non-configured state can looks
quite scary, add an init script that tells the user that all
is well.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Add SIGALRM based watchdog driver. This can reset barebox if stuck and
plays nicely with $global.system.reset.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Having a default reboot mode in the device tree allows testing the
functionality out of sandbox without prior setup.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| | |
We can differentiate between POR and RST by explicitly storing RST as
reset reason when we invoke the reset handler. Do so.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Follow-up will extend the poweroff driver to support system reset source.
Set the stage by renaming the driver to power (as it does reset as well)
and make it probe from device tree, so it can point at the system reset
source syscon via phandle.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Now that we can define a hostfile that's persistent over barebox resets,
but does not rely on a hard coded filename, we can have an always-on
environment node.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Currently, hostfils without a barebox,filename are assigned a new
temporary file on every start.
For use of a hostfile syscon for reboot mode or reset source, we will
need to persist the changes over a regular reset. This can easily be
done by passing through all automatically created temporary file with
an explicit --image option on subsequent runs. Do so.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
So far use of hostfile in the device tree required us to hard code a file
name. If we instead create a temporary file on demand, we can support:
- environment
- barebox-state
- syscon for reset source and reboot mode
out of the box with no dependency on external files.
Do the necessary, so a hostfile device tree node without a
barebox,filename gets a temporary file generated with the appropriate
size.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
801b7cff0210 ("sandbox: hostfile: allow probing from device tree") added
initial support for probing hostfiles out of a non-fixed up device tree.
This was achieved by having the driver probe call linux_open if the device
tree node doesn't contain a valid barebox,fd property. A mmap did not
happen in that case and as such the node couldn't be used as a syscon.
Fix this by unifying the binding for hostfiles added by --image and ones
already in the device tree. Both will result in a device node with just
a barebox,filename property. On pure init level, both kinds of hostfiles
will be iterated through and filenames will be opened and mmaped.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Follow up commits will flesh out the sandbox.dts and make most of what's
in the skeleton.dtsi redundant. Merge the .dtsi with the .dts to have
all the code at one place.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
hf->base may remain uninitialized if an image is > 4G, but sandbox
barebox is compiled for 32-bit. Fix this.
Reported-by: clang-analyzer-10
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The sandbox architecture is device tree only and so far when no --dtb
was passed, it created a basic one itself and used that.
Instead of creating the default device tree in C, just build the dtb
into barebox like we do on other platforms.
When --dtb is specified, that device tree will be used instead.
This results in functional change: Model and hostname are now more
accurate.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| | |
The different architectures duplicate some code around unflattening and
registering the device tree. Add common functions to reduce this
duplication.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Super I/O devices on x86 have their base usually located at one of two
I/O ports. They are probed by writing a "passcode" to the I/O port
and then a value is read back.
When COMPILE_TESTing these drivers and running them on sandbox,
this fails with a segfault because all I/O port access lead to
zero page dereference on that architecture. Even without running
barebox, static analyzers correctly flag these accesses as invoking
undefined behavior. Adding a dummy I/O region solves both issues.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| | |
Indentation was unintentionally doubled. Fix this.
Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
While we can mount file systems on cdevs in barebox, partition table
parsing only works for block devices. Allow for
--image=argument,blkdev to try to mount an image as block device.
This will fail for files that aren't of a multiple of the 512 byte
block size. Host OS block devices are suitable for use as barebox
block devices always, so that's the default unless overridden with
a ,cdev suffix.
The initcall level has been changed to occur after fs initcall level.
This is required, because we can't have automounts without / mounted.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
On Linux, /proc/self/exe is a symlink to the originally exec(2)d
executable. We can exec that with the original argv to simulate
a reset. This is useful for shorter development cycles on sandbox
and in future, could be used to test barebox behavior around resets
(e.g. reset reason can be passed through via libc environment).
We leave the original hanging reset in place though, because:
- Many boards have multiple reset providers and incoming patches
will allow users to select a specific one. Having this on
sandbox as well makes testing easier.
- /proc/self/exe is Linux-specific and wouldn't work when being
run on e.g. BSDs or macOS
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Typing reset in sandbox results in hang() while the terminal is not
cooked and ^C is ineffective. Only way to terminate barebox then is
via kill. Reinstate cooked mode on reset, so ^C termination is
possible.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
gcc v9.3.0's libubsan does not export a __sanitizer_print_stack_trace
symbol. Play it safe and avoid possible linker errors by having the
optional SANDBOX_UNWIND depend only on CONFIG_KASAN.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Some tools like afl-fuzz generate file names containing commas.
Allow escaping the commas in the file names, so they can be passed
to barebox.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
BLKGETSIZE64 writes 64-bit to the address pointed at by the ioctl argument.
As hf->size is a 32-bit size_t on 32-bit systems, on such systems,
the adjacent member might be corrupted. Fix this.
Fixes: 8d6da6462b12 ("sandbox: add_image: mmap block devices")
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| | |
The file descriptor is mandatory for doing anything useful.
Error out if we don't have one.
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
| |
| |
| |
| |
| |
| |
| |
| | |
For distro packaging, make it possible to build all host tools in one go
(and a very minimal barebox image as an unimportant side artifact).
Signed-off-by: Roland Hieber <rhi@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ \ |
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Unlike Linux, Barebox does not support 'make dtbs_install'.
There is no good reason to use dtb-y, which is not natively
supported by the Barebox build system.
Sandbox is the special architecture that compiles DTB as a
separate file. Add .dtb to always-y directly.
CONFIG_OFTREE is a bool option. Use a more Kbuild-ish style
instead of ifeq ($(CONFIG_OFTREE),y).
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|/
|
|
|
|
|
|
|
| |
sandbox really has ASan support, that is address sanitizer with the help
of the userspace library libasan. In contrast KASan is used on real
hardware where we have to implement our own support code. Rename sandbox
KASan to ASan to not clash with upcoming KASan support.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
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>
|