Testing zero_page_access only makes sense if we actually have RAM at
address 0 that we may read from. If it happens to I/O memory, the SoC
may hang or a data abort may be raised, which would make the test fail.
Avoid this by explicitly checking for that case and skipping the test
without running.
Once we start annotating malloc with __attribute((alloc_size)), GCC will
detect that we intentionally buffers that are too big.
Avoid the warning by hiding the size using the RELOC_HIDE macro.
Signed-off-by: Ahmad Fatoum
---
test/self/malloc.c | 8
1 file changed, 4 insertions(+), 4 de
The board is called STM32MP135 and there is currently no STM32MP137.
The typo was likely caused by copy-pasting from the STM32MP15x, which
has STM32MP157.
Signed-off-by: Ahmad Fatoum
---
arch/arm/mach-stm32mp/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/arm/ma
We should do away altogether with , so prepare for doing
that in the future by moing the only content it has on sandbox into
Kconfig.
Signed-off-by: Ahmad Fatoum
---
Kconfig | 2 +-
arch/Kconfig | 13 +
arch/sandbox/Kconfig
The spinlock stubs don't actually do anything with the spinlock_t.
To avoid the compiler warning about it add an __always_unused annotation
to the spinlock.
Signed-off-by: Ahmad Fatoum
---
include/linux/spinlock.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/include/linux/
Drivers have spaces inside their names, which makes tab completion
tricky as it restarts for every separate argument.
Support for doing this in readline and hush landed in commits
1498093ccd11 ("readline: Complete strings containing whitespaces correctly")
and 70e0885229d2 ("hush: Fix handling '\
The implementation of the glob and fnmatch functions is taken from glibc
and it shows in the amount of #ifdefs. We don't need that in barebox and
can even get the glob_t size down by dropping unused members, so let's
do that.
Cc: Christian Melki
Signed-off-by: Ahmad Fatoum
---
v1 -> v2:
- new
drvinfo can be very long especially for the in-tree defconfigs.
Make it more convenient to use by add optional filtering support:
barebox@board:/ drvinfo '*imx7d*'
Driver Device(s)
imx7d-src
3039.reset-control...@3039.of
imx7d_adc
306100
pm_domain.h so far included no headers and only worked, because it was
included in a "correct" order in other files. Fix this by including the
header it depends on.
Signed-off-by: Ahmad Fatoum
---
include/pm_domain.h | 3 +++
1 file changed, 3 insertions(+)
diff --git a/include/pm_domain.h b/in
On the STM32MP1, shutting down barebox can result in a
ERROR: phy1: phy exit failed --> -22
message printed to the console. This is because the regulator is disabled
more often than it was enabled, because stm32_usbphyc_pll_disable
is called twice:
- Once from stm32_usbphyc_remove calling st
We already stub out a number of MODULE_* macros to save porters the
hassle of deleting them. Let's add MODULE_VERSION to the list as well.
Signed-off-by: Ahmad Fatoum
---
include/linux/barebox-wrapper.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/include/linux/barebox-wrapper.h b/include
entry_count has type size_t and not long as the format string wrongly
suggested.
Signed-off-by: Ahmad Fatoum
---
drivers/bus/acpi.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/bus/acpi.c b/drivers/bus/acpi.c
index f69789200633..ac034cce4569 100644
--- a/drivers/bu
When of_syscon_register is called with check_clk=true, it will attempt
to attach the clock. If that fails, the newly allocated syscon is freed,
but a dangling pointer remained in the syscon_list.
Fix this by only registering the syscon in the successful case.
Fixes: b36b7b727192 ("mfd: syscon: cl
We already have dummy definitions for common spin lock operations,
so reuse them instead of defining them in the driver.
Signed-off-by: Ahmad Fatoum
---
drivers/usb/dwc2/gadget.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/drivers/usb/dwc2/gadget.c b/drivers/usb/dwc2/
The STM32MP135F-DK support in barebox has bitrotted, because of changes
to the upstream device tree: The root of the clock tree is now provided
by OP-TEE over SCMI for which no support exists in barebox, so drivers
requesting clocks will defer indefinitely, including the console device.
A series a
From: Ahmad Fatoum
clk_get will return -EPROBE_DEFER if clock provider hasn't yet been
probed. In a system with deep probe enabled, dependencies are probed
on demand, so a -EPROBE_DEFER return is final and the console probe
will never succeed.
CONFIG_DEBUG_LL is often used to debug this, but bec
We are using an older version of setlocalversion that sources the
config. This is wrong as there's no guarantee that it's ok to source the
config, e.g.:
CONFIG_DEFAULT_ENVIRONMENT_PATH="$(objtree)/.ptxdist-defaultenv"
when sourced will lead to:
scripts/setlocalversion: 15: include/config/aut
Adorning functions allocating dynamic memory with __alloc_size allows
GCC to warn about heap overflows it notices during normal compilation.
Import the definitions from Linux and switch over barebox
to make use of it.
Signed-off-by: Ahmad Fatoum
---
include/linux/compiler-gcc.h | 8
Once we add __alloc_size attributes to allocations, GCC will complain
about violation of memory safety in test_kasan.c.
That memory violation is intended though as test_kasan is meant to
trigger kasan at runtime to verify correct operation.
Silence the warnings by hiding the origin of ptr, so the
barebox remaps all reserved entries as uncached to avoid speculative
access into the described regions using remap_range.
The ARM implementation requires buffers to be page aligned, which we
can't assume unconditionally. For this reason, reserve_sdram_region
will align region start and size before
20 matches
Mail list logo
