Match the Application.mk install rule: keep the application attribute
symbols (nx_stacksize etc.) through strip as NX_KEEP does, and mark
the installed binaries executable, since ld -r output is not and
filesystem images built from bin/ refuse to exec.
Signed-off-by: raiden00pl <raiden00@railab.me>
Assisted-by: Claude Code
CMAKE_LD was never set, so application link commands were invalid, and
CMAKE_STRIP defaulted to plain 'strip', which removes the symbol and
relocation tables required to load CONFIG_BINFMT_ELF_RELOCATABLE
binaries. Use the host linker and 'strip --strip-unneeded' as on the
other architectures.
Signed-off-by: raiden00pl <raiden00@railab.me>
Assisted-by: Claude Code
arch_interface exists only for CONFIG_BUILD_PROTECTED, but x86_64 and
arm64 referenced it for any non-flat build, breaking CMake
configuration of kernel builds. Use the CONFIG_BUILD_PROTECTED guard
as arm and risc-v do.
Signed-off-by: raiden00pl <raiden00@railab.me>
Assisted-by: Claude Code
mknxflat is built from tools/nxflat by the NuttX build itself, so the
toolchain section no longer sends the reader to buildroot for it; only
ldnxflat still comes from there. Bring the mknxflat usage text in line
with the tool, note that MKNXFLAT and LDNXFLAT are supplied by the ARM
Toolchain.defs, and correct the r10 references left in the PIC
descriptions.
Describe the module ABI marker, so that a user whose prebuilt module
starts failing exec() with ENOEXEC finds out that the loader refuses a
module whose import table does not name __nxflat_abi_v2, and that
rebuilding the module is the fix.
Assisted-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Marco Casaroli <marco.casaroli@gmail.com>
ARM PIC has used r10 as the base register, but the tree has never been
consistent about it. Toolchain.defs gives CONFIG_BUILD_PIC
-mpic-register=r9 and CONFIG_PIC -mpic-register=r10, twenty-five lines
apart, and arm_initialstate.c sets REG_R9 from inline assembly under one
and REG_PIC under the other, with a comment reading "Set the PIC base
register (probably R10)". This settles it on r9 for all of PIC: NXFLAT,
ELF PIC and CONFIG_BUILD_PIC alike.
r9 is the right choice rather than an arbitrary one. It is the AAPCS
platform register, the "static base", and it is what GCC itself picks
for -msingle-pic-base on an EABI target; r10 is the non-EABI default.
It also removes a combination that cannot build today. Stack checking
adds -ffixed-r10 in armv7-m/Toolchain.defs and armv8-m/Toolchain.defs,
while CONFIG_PIC adds -mpic-register=r10, and GCC rejects the pair with
"unable to use 'r10' for PIC register". The comment above REG_PIC has
always said the register "can be R9 if stack checking is enabled", but
the definition was unconditionally REG_R10, so it would have named the
wrong register even had the build succeeded.
The thunk generator moves with the firmware. NXFLAT import stubs had
the register baked in as "add ip,ip,sl", so a module built for r9 would
load and then branch to a wild address on its first call out. The stubs
now come from NXFLAT_PIC_REG in the in-tree tool, which is built only
when CONFIG_NXFLAT is set, following the
CONFIG_BOARD_ETC_ROMFS_PASSWD_ENABLE precedent in tools/Unix.mk.
That leaves modules built before this change, and they are the reason
for the ABI marker. The NXFLAT header cannot carry a version: h_magic
is written by ldnxflat, which is GPL, derived from elf2flt, and stays
out of this repository, so it can never be changed in step with the
loader. The import table can, because both of its ends are in-tree --
mknxflat emits it and nxflat_bindimports() reads it -- and ldnxflat
passes it through untouched. So every module now imports
__nxflat_abi_v2, the base firmware defines it, and a module that does
not import it is refused.
Making the marker a real exported symbol rather than a name the loader
special-cases is what keeps it out of the build system's way: a board's
symbol table picks it up exactly as it picks up printf, so mksymtab.sh
and its equivalents need no change. It also gives the reverse direction
a diagnosis for free -- a module built against a newer ABI than its
firmware fails with "Exported symbol __nxflat_abi_v2 not found".
Most of the remaining churn is boards restating a default. ARCHPICFLAGS
is a "?=" default so that a board only speaks up when it differs, and
twenty-six were assigning the value the default already had. MKNXFLAT
gets the same treatment: thirteen boards named the same tool, and the
only thing that varies is ARM versus Thumb-2, which falls out of
CONFIG_ARM_THUMB. LDNXFLAT gains a default too -- it stays an
out-of-tree PATH lookup, but naming it centrally fixes boards that never
assigned it, where it expanded to nothing and handed make a recipe
beginning "-e", whose leading dash make ate as "ignore errors".
The non-ARM boards carrying -mpic-register=r10 lose it: it is an
ARM-only option, reachable only through CPICFLAGS, which is only used to
build NXFLAT modules, and no non-ARM board enables NXFLAT.
Boards keep nothing about PIC flags any more. ARCHPICFLAGS was set by
sixty-three of them and only ever fed CPICFLAGS, which is only used to
build NXFLAT modules; no board outside arch/arm enables NXFLAT, so every
non-ARM copy was setting a variable nothing read. Those are removed
rather than moved somewhere more central, which would only make dead
text look load-bearing. LDNXFLAT goes the same way as MKNXFLAT, for the
same reason: thirteen boards named the same tool that Toolchain.defs now
names once.
One of them was not merely redundant. am67/t3-gem-o1 asked for
"-mpic-register=r10 -ffixed-r10", which GCC refuses outright with
"unable to use 'r10' for PIC register" -- the very combination the
filter-out machinery in Toolchain.defs exists to prevent. It has
survived because that board does not build NXFLAT modules, so the flags
are never handed to a compiler. Renaming the register would have
carried the fault forward unchanged, so the line goes.
Tested on lm3s6965-ek:qemu-nxflat under QEMU, configured and built with
no overrides. The nxflat example runs the errno, hello and struct
modules with output identical to the same config built from master.
Built with the old out-of-tree thunk generator instead, the same
firmware refuses all three with ENOEXEC rather than locking up in a
HardFault, which is what this change is for. mps3-an547:picostest,
which is CONFIG_PIC without CONFIG_NXFLAT, builds clean and does not
build the thunk generator.
The .def files pick up two cosmetic changes here alongside the register:
a "Dyanamic" typo that codespell rejects, and a reworded comment in each
thunk_*.c. Neither appears in the emitted thunk -- both are in C
comments -- so the generated text is still what the upstream tool
produces, modulo the register itself.
BREAKING CHANGE: ARM PIC moves from r10 to r9. An NXFLAT module built
before this change has r10 baked into its import stubs and will not run
against a firmware carrying it; the two cannot be mixed. The module is
refused with ENOEXEC rather than branching to a wild address, by way of the
__nxflat_abi_v2 marker described below.
Quick fix: rebuild the module against this tree. Its source needs no
change. A board that reserved r10 by hand, or that assigned ARCHPICFLAGS
or MKNXFLAT to restate a default, should drop those assignments; nothing
else is affected, and CONFIG_PIC without CONFIG_NXFLAT needs no action.
Assisted-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Marco Casaroli <marco.casaroli@gmail.com>
The tool arrived from the buildroot NXFLAT toolchain under BSD-3-Clause,
jointly copyright Gregory Nutt and Cadenux, LLC. Gregory Nutt owned
Cadenux and was its only developer on this code, and has agreed to the
conversion, so the six files take the ASF header like the rest of the
NuttX code he donated. Copyright attribution moves to NOTICE, which is
where the donation put it for everything else of his in the tree.
This covers only what was imported: mknxflat and the thunk skeletons it
emits from. ldnxflat is the file with an elf2flt lineage, and it is not
here -- it stays out of tree in buildroot, and NuttX keeps calling it as
an external tool.
The .def files also gain their in-tree path on the first line, which the
import had left pointing at the buildroot layout.
Assisted-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Marco Casaroli <marco.casaroli@gmail.com>
An NXFLAT module reaches the base firmware through a "thunk" file: one
assembly stub per imported function, generated by mknxflat. That tool
has always lived outside this repository, in the NuttX buildroot NXFLAT
toolchain, so building an NXFLAT module needs a separate checkout and a
separate build of a tool that links against libbfd.
libbfd is why it stayed out. It is GPL, which an Apache project cannot
depend on, and it is awkward to obtain besides -- a stock binutils
install often ships libbfd without the libiberty it needs to link. But
the dependency was never deep. mknxflat used libbfd for eight calls,
all of them opening the file and walking the symbol table; it never
relocates or rewrites anything. That is replaced here by reading the
ELF symbol table directly, which removes the dependency outright and
costs about a hundred lines.
The emitted text is unchanged. The format strings live in the .def
files, which are carried here byte-for-byte from upstream, and the
selection rule for what becomes a thunk is the upstream one: everything
undefined that is not explicitly an object. Symbol typing cannot be
trusted for this -- imported functions are routinely emitted as
STT_NOTYPE rather than STT_FUNC, while a weakly defined object does
appear as an undefined object -- so the test is on what a symbol is not.
Upstream chose the instruction set at compile time through an "arch"
symlink pointing at either arm/ or thumb2/. A symlink cannot be carried
in the repository, and one host binary has to serve boards of both
flavours, since lpc31xx is ARM while lpc17xx, tiva, stm32f1 and rp23xx
are Thumb-2. That choice becomes a runtime "-a" option. The "-f"
option, which read further command line arguments from a file, is
dropped; nothing in the tree used it.
This commit changes no output. Against the upstream tool, for both
architectures, with and without -w, over modules exercising the plain,
weak and non-returning thunk paths, the generated thunk files are
byte-identical.
Assisted-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Marco Casaroli <marco.casaroli@gmail.com>
Include STM32_HAVE_USART_H5 in the UART7/UART8 driver choice
depends so CONFIG_STM32_UART*_SERIALDRIVER can be selected on
STM32H5.
Signed-off-by: Joao Mario Lago <joao.mario.lago@hotmail.com>
g_rtc_lock is used by the up_rtc_settime, a base RTC function.
Therefore, it should be available even for this procedure, not just when
CONFIG_RTC_HIRES.
Signed-off-by: Jiri Vlasak <jvlasak@elektroline.cz>
Several risc-v assembly labels are .global but untyped, so mkallsyms.py
(which only collects STT_FUNC symbols) silently drops them from the
ALLSYMS table, and backtraces/%pS print raw addresses instead of names.
Add .type <name>, function to the affected labels, matching existing
convention elsewhere in the tree.
Signed-off-by: liang.huang <liang.huang@houmo.ai>
Assisted-by: Claude Code:claude-sonnet-5
allsyms_lookup() derived a symbol's size from the physically next table
entry, assuming address order. Under CONFIG_SYMTAB_ORDEREDBYNAME the
table is sorted by name instead, producing a huge bogus size in
%pS/backtrace output.
Scan for the closest larger address instead of relying on table order.
Signed-off-by: liang.huang <liang.huang@houmo.ai>
Assisted-by: Claude Code:claude-sonnet-5
allsyms_findbyvalue()/%pS printed a bogus name/offset for addresses
outside the real symbol table's coverage, due to the boundary sentinels
being matchable as real symbols.
Compute the high sentinel from the actual symbol range and treat a
sentinel match as "not found".
Signed-off-by: liang.huang <liang.huang@houmo.ai>
Assisted-by: Claude Code:claude-sonnet-5
g_allsyms/g_nallsyms only exist in the kernel image, but symtab_allsyms.c
is unconditionally built into libc.a, so user-mode code under
CONFIG_BUILD_PROTECTED/CONFIG_BUILD_KERNEL fails to link.
Guard the affected code so user-mode libc.a no longer references these
kernel-only symbols.
Signed-off-by: liang.huang <liang.huang@houmo.ai>
Assisted-by: Claude Code:claude-sonnet-5
When a running task changes its affinity on SMP and is no longer eligible to run on the current CPU, merely delivering an equal-priority scheduling request can leave the task in g_readytorun while the target CPU remains idle. The task then never runs again.
Remove the task from its current CPU, add it back to the ready-to-run list so a suitable CPU is selected, and perform the context switch unconditionally. This ensures the task is migrated according to its updated affinity.
Fixes: https://github.com/apache/nuttx/issues/19680
Assisted-by: GitHub Copilot:ppio/pa/gpt-5.6-sol
Signed-off-by: hujun5 <hujun5@xiaomi.com>
Two level errors corrupted kernel memory when a user process extended
its heap with sbrk:
- PGT_LAST was X86_MMU_PT_LEVELS (4), but valid levels are 0-3, so the
final level entry was written with an out-of-range index.
- x86_64_get_pgtable indexed the PD (level 2) with level 3, installing
newly allocated page tables into the wrong PD slot.
Signed-off-by: raiden00pl <raiden00@railab.me>
Assisted-by: Claude Code
A fetch() only lower half is always ready, so a subscriber that asked for
a rate with SNIOC_SET_INTERVAL got no pacing from poll(): the descriptor
reported POLLIN on every pass and the application had to sleep out the
period itself. That does not compose. An application polling several
topics reads them sequentially from one thread, so per read sleeps
serialize: three topics at 10 Hz sleeping 100 ms each yield 3.3 Hz per
topic rather than 10.
Pace it where poll() can act on it instead. A subscriber that never
requested a rate stays always ready, and one that did becomes ready once
per its own interval, driven by a watchdog armed in sensor_poll(). This
is the fetch() side of what sensor_is_updated() already does for a
pushing lower half, so both models now honor a requested rate the same
way.
The wdog_s lives in sensor_user_s rather than in the device, so each
subscriber is paced at its own interval instead of at the minimum across
all of them, and the timer only runs while somebody is polling. The
expiry runs in timer context and takes no lock: poll_notify() is safe
from an interrupt handler, and a teardown that raced it has already
cleared fds, which makes both the notify and the re-arm no-ops. Teardown
therefore just cancels the watchdog where it clears fds, and a watchdog
keeps this off the work queue entirely, which a fetch() only sensor
exists to avoid.
sensor_close() needs nothing of its own: poll_setup() holds a reference
on the file for the duration of the poll, so file_close() cannot run
until poll_teardown() has called sensor_poll() with setup false, and that
already cancelled the watchdog.
Assisted-by: Claude:claude-sonnet-5
Signed-off-by: Felipe Moura <moura.fmo@gmail.com>
A fetch() only lower half reads the device on demand, so its data is
always available and there is never anything to wait for. The upper half
did not reflect that: poll() only reported POLLIN when the descriptor was
opened O_NONBLOCK, and a blocking read() waited on buffersem, which is
only posted when the lower half drives notify_event from an interrupt of
its own.
A fetch() only sensor with no interrupt therefore never satisfied
poll()/read() at all. This is not hypothetical: in the in tree
nucleo-h563zi:dts configuration CONFIG_STM32_DTS_TRIGGER defaults to 0,
which selects stm32_dts_fetch(), and no CONFIG_STM32_DTS_ITEN_* option is
enabled, so the DTS interrupt never fires. A blocking read() on that
sensor waits forever, even though stm32_dts_fetch() performs a complete
software triggered measurement on its own and needs no interrupt at all.
Applications had to work around this by forcing O_NONBLOCK on the
descriptor themselves, see apache/nuttx-apps#3686.
Drop the O_NONBLOCK special case in both paths: sensor_poll() now always
reports POLLIN for a fetch only sensor and sensor_read() calls fetch()
directly instead of waiting. Update the sensor_ops_s::fetch
documentation, which described the old contract.
With the wait gone, buffersem has no waiters left. Its only two readers
were the ones removed here, both in the fetch path: the wait in
sensor_read() and the nxsem_get_value() in sensor_poll(). The remaining
nxsem_post() calls in sensor_push_event() and sensor_notify_event() had
nothing left to wake, so drop the semaphore and those posts as well.
Assisted-by: Claude:claude-sonnet-5
Signed-off-by: Felipe Moura <moura.fmo@gmail.com>
The driver had two modes selected by CONFIG_SENSORS_L3GD20_BUFFER_SIZE:
with a buffer it pushed samples from a work queue, and without one it
exposed fetch() while still using the data ready interrupt to signal
readiness through notify_event.
That second mode misuses the fetch interface. fetch() means the data is
read from the device on demand and is therefore always available, while
an interrupt driven sensor is exactly what push_event is for. Mixing the
two forces the upper half to guess whether a fetch() only lower half
will ever notify, and it makes poll() unusable in a multi descriptor
loop, because the descriptor reports ready while the read still has to
wait for the next interrupt.
Drop the fetch path and always use the work queue and push_event, which
is what the driver already did by default since BUFFER_SIZE defaults to
1. CONFIG_SENSORS_L3GD20_BUFFER_SIZE gains a range of 1 to 32, as a zero
sized buffer no longer has a meaning, and SCHED_HPWORK is now selected
unconditionally because the work queue is always used.
No in tree configuration enables this driver and the previous default
already took the push path, so no defconfig changes are needed.
Assisted-by: Claude:claude-sonnet-5
Signed-off-by: Felipe Moura <moura.fmo@gmail.com>
The condition work_available(&conn->work) && tx_unacked != 0
prevented tcp_update_retrantimer from being called when the work
queue was still busy, leaving conn->timer stale or zero on
subsequent sends. This caused the RTT estimation to compute a
false RTT (m = rto - 0 = rto), creating a positive feedback loop
that inflated the RTO to extreme values (e.g., 232 half-seconds
= ~116 seconds).
Fix: remove the work_available check so that tcp_update_retrantimer
is always called when there is unacknowledged data. The decision to
re-queue the work is handled internally by tcp_update_timer.
Signed-off-by: zhekunren <zhekunren@qq.com>
The audio subsystem page listed the source files and the configuration
options, but nothing about the interface the upper half presents to
applications. PR #18348 added a device state machine, a second buffer
allocation mode, poll and mmap support and several new ioctls, none of
which were described anywhere, so the only way to learn the expected
call sequence was to read audio/audio.c.
Document what the upper half now guarantees:
- the device state machine, and the fact that AUDIOIOC_START is
rejected until AUDIOIOC_CONFIGURE has moved the device out of
AUDIO_STATE_OPEN;
- the normal open/configure/allocate/enqueue/start sequence;
- the two AUDIOIOC_ALLOCBUFFER modes selected by u.pbuffer, who owns
the buffers in each, and that a shared ring request may return zero
when the ring is already populated;
- that AUDIOIOC_GETBUFFERINFO also establishes the shared ring depth,
so a lower half which does not implement it disables that mode;
- the poll event semantics and how mmap() selects between a ring
buffer and the device status by requested length;
- all ioctls handled by the upper half, grouped by purpose;
- how per-open state is aggregated into the device state when several
applications share one device.
No functional change.
Signed-off-by: fangyibo <fangyibo@xiaomi.com>
The upper->periods >= upper->nbuffers check sat at the top of
audio_allocbuffer(), but upper->periods is only incremented for shared
ring requests (u.pbuffer == NULL), so for private buffer callers the
check degenerated into "nbuffers == 0" and rejected every allocation
when the lower half does not implement AUDIOIOC_GETBUFFERINFO, which is
the only place nbuffers is ever assigned.
Move the guard inside the shared ring branch so private buffers, which
never enter upper->apbs[] and are unrelated to the ring depth, stay
allocatable. The zero return value is kept as-is because a second
application attaching to the same device relies on it to skip
allocation and go straight to AUDIOIOC_ENQUEUEBUFFER.
Signed-off-by: fangyibo <fangyibo@xiaomi.com>
RGBTO8 shifted each component up before masking:
(((uint8_t)(r) << 5) & 0xe0)
The cast is promoted to int before the shift, so the mask keeps bits 5:7
of the shifted value, which are bits 0:2 of r. The macro therefore
encoded the three least significant bits of red and green and the two
least significant bits of blue, rather than the most significant.
This disagrees with RGBTO16 in the same file, which correctly takes the
high bits, and with RGB8RED/RGB8GREEN/RGB8BLUE immediately below it,
which are documented as the inverse transformation but read the result
as high bits.
All in-tree callers pass full 8-bit components, so all were affected:
RGBTO8(39, 64, 139) in apps/examples/nxterm, intended as midnight blue,
evaluates to 0xe3 -- full red plus full blue, i.e. magenta.
Take the high bits instead, so that RGBTO8 matches RGBTO16 and the
RGB8xxx macros become its true inverse.
Tested on a RISC-V LiteX/VexRiscv target with an 8bpp RGB332 frame
buffer, and with a host round-trip check over all 256 representable
colours.
Assisted-by: Claude:claude-opus-5
Signed-off-by: William Byatt <william@byatt.io>
The registration example showed the three argument form, which no longer
compiles, and fetch() as the only way samples are taken. Update it, add
a section on the two acquisition modes and the attach() a board provides
for the interrupt one, and state the 100 Hz sample rate.
Signed-off-by: Felipe Moura <moura.fmo@gmail.com>
fetch() timestamps a sample when the application asks for it, not when
the device measured it, and reads accel and gyro separately so the two
topics never share an instant. Add an optional push mode behind
CONFIG_SENSORS_MPU6050_INT: the board supplies mpu6050_config_s::attach,
the handler timestamps and defers to HPWORK, and the worker reads once
and pushes both topics. The I2C read cannot run in the interrupt.
The mode is chosen at build time, so fetch() is simply left out of the
ops table and out of the build when the option is set: an instance uses
one model or the other, never the mixture that made poll() unusable on
l3gd20. A board that enables it without attach fails with -EINVAL.
Also set CONFIG so the DLPF is on. Left at reset the gyroscope output is
8 kHz, not 1 kHz, so SMPLRT_DIV 9 gave 800 Hz rather than the documented
100 Hz; measured 833 Hz before and 101 Hz after. fetch() hid this since
the application set the pace.
Signed-off-by: Felipe Moura <moura.fmo@gmail.com>
RNDIS reports Ethernet packet sizing to the host through NDIS OIDs.
OID_GEN_MAXIMUM_FRAME_SIZE is the MTU-style value and excludes the
link-layer header, while CONFIG_NET_ETH_PKTSIZE includes the Ethernet
header.
Report the frame size as CONFIG_NET_ETH_PKTSIZE - ETH_HDRLEN, and
report OID_GEN_MAXIMUM_TOTAL_SIZE as CONFIG_NET_ETH_PKTSIZE instead of
a hardcoded 2048.
Assisted-by: OpenAI Codex:GPT-5
Signed-off-by: shichunma <shichunma@bestechnic.com>
The documentation job installs with `pipenv install`, which does not honour
the committed Pipfile.lock. Every run in the logs prints "Locking
dependencies..." and "Updated Pipfile.lock" and then installs from the set it
has just re-resolved, so each build takes whatever PyPI resolves that day
rather than what the lock file names.
On the evening of 2026-08-03 that resolution produced a virtualenv without
packaging, and four unrelated pull requests failed identically, before Sphinx
had read a single file:
File ".../sphinx/extension.py", line 7, in <module>
from packaging.version import InvalidVersion, Version
ModuleNotFoundError: No module named 'packaging'
`pipenv sync` installs exactly what Pipfile.lock names and never re-resolves,
which is what the lock file is for. The committed lock covers all fourteen
packages the Pipfile asks for, packaging included, so it is complete enough to
install from as it stands.
The workflow also ran only for changes under Documentation/, so a change to the
documentation build was never exercised by the build it changed. It now
triggers on its own path as well, which is what tests this commit.
Assisted-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Marco Casaroli <marco.casaroli@gmail.com>
Adds LDO support for RISC-V Espressif devices. In general, should
only affect ESP32-P4 on a few boards.
ESP32-P4 has 4 channels of low-dropout voltage regulators, which
are programmable. Can be used to power external devices.
Signed-off-by: Filipe Cavalcanti <filipe.cavalcanti@espressif.com>
- Zero initialize union in s32k3xx_transmit()
- Allow setting BRS from frame flags
Signed-off-by: Takashi Furuya <takashi.furuya@nttedt.co.jp>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Zero initialize union in s32k1xx_transmit()
- Allow setting BRS from frame flags
Signed-off-by: Takashi Furuya <takashi.furuya@nttedt.co.jp>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Zero initialize union in kinetis_transmit()
- Allow setting BRS from frame flags
Signed-off-by: Takashi Furuya <takashi.furuya@nttedt.co.jp>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Zero initialize union in imx9_transmit()
- Allow setting BRS from frame flags
Signed-off-by: Takashi Furuya <takashi.furuya@nttedt.co.jp>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Clear NXSEM_MBLOCKING_BIT in nxsem_wait_irq() when a mutex waiter is
removed and the wait queue becomes empty.
Signed-off-by: Martin Krasula <mkrasula@elektroline.cz>
Describe the shared library open semantics in the FLAT and PROTECTED
builds: dlopen() of a library that is already loaded returns a handle
to it and takes an additional reference, and the library is unloaded
only when the last handle is closed. Note the consequences that follow
from having a single instance: libraries are matched by basename, data
is shared by all users, and constructors and destructors run once.
Contrast this with insmod(), which still rejects a duplicate module
name, and note that dlopen() is not implemented in the KERNEL build.
Assisted-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Marco Casaroli <marco.casaroli@gmail.com>
dlopen() of a library that is already loaded fails. libelf_insert()
rejects a name that is already in the module registry with EEXIST, and
dlinsert() passes that straight out, so the second caller gets NULL.
POSIX says dlopen() shall return a handle to the object, and there is no
way today for two modules to hold the same library at once -- which is
what a shared library is for.
So dlopen() now takes another reference on a library that is already
there, and dlclose() only tears it down when the last handle goes. The
count lives in the dlfcn layer rather than in libelf_insert() so that
insmod keeps its own behaviour: a second insmod of the same name still
fails with EEXIST, which is right for a kernel module.
The module name is what makes any of this possible, and a PROTECTED build
did not have one. Names were defined for CONFIG_BUILD_FLAT or the kernel
side of a split build, on the reasoning that only the kernel needed them,
which predates dlopen() being usable from user space. Without a name the
user-space copy of libelf cannot recognise a second open of a library,
cannot count opens, and cannot make dlclose() mean anything -- two
dlopen()s there produce two independent copies of the library and lose
track of the first. Names are therefore defined wherever CONFIG_LIBC_DLFCN
is, which costs NAME_MAX per loaded module in that configuration.
The path no longer has to be copied either. The module name is the
basename of the file and libelf_insert() takes it as a const string, so
dlinsert() finds it with strrchr() instead of handing a writable
duplicate of the whole path to basename().
BUILD_KERNEL is deliberately untouched. dlopen() returns NULL there
unconditionally: dlinsert() is a stub, because sharing a library between
processes with separate address spaces needs the text in a shared region
and the data per process at a matching virtual address, which is a
different problem from this one.
Built for mps3-an547:picostest with and without CONFIG_LIBC_DLFCN, and
for stm32f4discovery:kostest, a PROTECTED configuration, with it enabled.
Assisted-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Signed-off-by: Marco Casaroli <marco.casaroli@gmail.com>
USDHC_PROCTL selects whether the usdhc uses SDMA or ADMA. This driver
only uses the SDMA or no DMA at all, so select that option. There might be a
wrong register value left by ROM code or an earlier bootloader which uses
ADMA.
Signed-off-by: Jukka Laitinen <jukka.laitinen@tii.ae>
The options that let a board trade memory for drawing speed were only
described in their Kconfig help, where someone bringing the game up on a
new board is unlikely to find them. Describe what each one does and
when it is worth enabling.
Signed-off-by: Jorge Guzman <jorge.gzm@gmail.com>
Runs NXDoom on the board's LCD, played with a USB HID keyboard and
reading the game data from the microSD card, so it exercises the LTDC
framebuffer, the OTG FS host and the SDMMC peripheral at once.
Three settings are needed that are not obvious:
CONFIG_FAT_FORCE_INDIRECT, because the FAT layer otherwise reads whole
sectors straight into the caller's buffer and the SDMMC IDMA cannot
reach the caller's buffer when it lives in external SDRAM. The
failure appears part way through startup, once the internal RAM has
filled and allocations start coming from SDRAM.
CONFIG_HIDKBD_NOGETREPORT, because the keyboard answers GET_REPORT on
the control pipe with an empty report and only delivers key data on
its interrupt endpoint. Without it the keyboard enumerates, reports
no error, and no key is ever seen.
CONFIG_STM32_LTDC_L1_L8 with the frame buffer colour map, because DOOM
is natively palettised: letting the display convert the palette while
it scans out removes the conversion from the blit and halves the
amount of data written per frame.
Signed-off-by: Jorge Guzman <jorge.gzm@gmail.com>
chromakey is a uint32_t, so printing it with %d warns. The line is only
compiled when CONFIG_STM32_FB_CMAP is enabled, which is why it has gone
unnoticed.
Signed-off-by: Jorge Guzman <jorge.gzm@gmail.com>
Document inode_checkperm / inode_checkpathperm, mountpoint traverse vs open
semantics, and the optional mountpt_operations.permission hook in
file_permission.rst.
Signed-off-by: Abhishek Mishra <mishra.abhishek2808@gmail.com>
Require X_OK on path ancestors and mountpoint gates via
inode_checkpathperm(), and use inode_checkperm() for inode mode checks.
inode_checkpathperm() takes the inode tree read lock (or INODE_CHECK_LOCKED
when the caller already holds it) and can fold a non-zero amode check onto
the target inode in the same call.
When CONFIG_FS_PERMISSION is disabled, inode_checkperm/inode_checkpathperm
are empty macros returning 0 so there is no stub .text cost.
Skip mountpoint R/W open-mode checks in inode_checkopenperm(); keep
traversal separate from open. Add an optional mountpt_operations.permission
hook (tmpfs when CONFIG_FS_PERMISSION) for in-volume DAC without making the
VFS mount-crossing path depend on it.
Signed-off-by: Abhishek Mishra <mishra.abhishek2808@gmail.com>
The terminal no longer has separate matrix and USB input sources: it
reads any keyboard through one path.
Left as it was, the unknown symbol would have been dropped and the
configuration would have fallen back to the touch variant, which is the
default. This board has no touchscreen, so the terminal would have had
no input at all and the build would have said nothing.
Signed-off-by: Jorge Guzman <jorge.gzm@gmail.com>
The lvglterm page still told the reader to pick the input source that
matches the hardware, which was the whole thing being removed: it has
one physical keyboard variant now, and reads whatever is registered.
The hidkbd page pointed at HIDKBD_ENCODED, which no longer exists. What
produces that byte stream now is INPUT_KEYBOARD_BYTESTREAM.
system/kbd had no page at all.
Signed-off-by: Jorge Guzman <jorge.gzm@gmail.com>
The terminal reads a keyboard device and does not care which one, so it
can be driven with no keyboard plugged in at all. Enable UINPUT_KEYBOARD
and system/kbd so that the configuration can do it out of the box, and
document how, including how to have the USB keyboard and the console feed
the terminal at the same time.
The terminal no longer has a USB specific input source either, so the
device path has to be spelled out: the USB HID driver names its devices
/dev/kbda onwards while the option defaults to /dev/kbd0.
INPUT_KEYBOARD_BYTESTREAM is not needed here. The terminal reads events
now, and this configuration has no other keyboard consumer.
UINPUT_KEYBOARD_BUFNUMBER is raised to 128. It counts events rather than
keys, so the default of eight holds four keystrokes, and a console hands
over a whole line at once. The upper half overwrites the oldest event
when the buffer is full, so a typed line arrived with its beginning
silently missing.
HIDKBD_NOGETREPORT is enabled as well. Sampling the keyboard over the
control pipe every 40 ms loses any key pressed and released between two
samples, which on this board meant most of them.
Signed-off-by: Jorge Guzman <jorge.gzm@gmail.com>
On K3 SoCs the A53 cluster does not own device power, clocks, resets or
interrupt routing; these are managed by the DM/TIFS firmware and requested
over TISCI through the secure-proxy mailbox. Add:
- a TISCI client (secure-proxy transport plus device/clock/reset helpers)
- a GPIO driver for banks 0/1 with pad mux and TISCI power-on
- an OMAP-I2C driver for the main-domain I2C controllers
and wire them into the PocketBeagle2 and BeaglePlay bring-up.
Tested on PocketBeagle2 (SYSFW 11.2.5) with the TechLab cape attached:
nsh> i2c bus
nsh> i2c dev -b 0 0x03 0x77 # I2C0 detects onboard 0x20 and 0x50
nsh> gpio -o 1 /dev/gpio3 # drives a user LED; read back with:
nsh> gpio /dev/gpio3
Signed-off-by: Piyush Patle <piyushpatle228@gmail.com>
IPv6 input currently lets fragmented packets continue into the IP filter path before reassembly. Forwarded IP filter rules can inspect L4 fields, but non-first IPv6 fragments do not carry the transport header.
Consume IPv6 fragments through the existing reassembly/drop helper before NAT66 or IP filter processing. This keeps plain IPv6 forwarding unchanged when neither L4-dependent feature is enabled, while avoiding filtering incomplete fragments.
Signed-off-by: shichunma <shichunma@bestechnic.com>
NAT66 needs transport headers to create or match address and port mappings. Consume IPv6 fragments through local reassembly before NAT66 processing so forwarded fragmented traffic is not translated fragment-by-fragment without L4 context.
Also dispatch inbound NAT66 using the parsed IPv6 next-header value, matching the outbound path and allowing packets with extension headers before the transport header to be translated.
Signed-off-by: shichunma <shichunma@bestechnic.com>
