Fix security issue where uninitialized kernel stack contents could be
leaked to userspace when mfrc522_picc_select() fails.
In mfrc522_read(), the local variable 'uid' was not initialized before
being passed to mfrc522_picc_select(). If the function fails (e.g., due
to bad data on the SPI bus), the uninitialized uid.sak value could pass
the PICC_TYPE_NOT_COMPLETE check, causing snprintf() to copy
uninitialized kernel stack data to the userspace buffer.
Fixes#19417
Signed-off-by: hanzhijian <hanzhijian@zepp.com>
Author: hanzhijian <hanzhijian@zepp.com>
Provide per-instance capture automonitor lookup for watchdog lower halves
that pass callback context, and avoid selecting an unrelated watchdog when
legacy lower halves provide no context. Update STM32 WWDG lower halves to pass
their instance context so multiple watchdog devices remain distinguishable.
Signed-off-by: hanzhijian <hanzhijian@zepp.com>
Assisted-by: OpenAI Codex
Fix two latent defects in the W5500 Ethernet driver:
1. NETDEV_RXERRORS() references non-existent variables (line 1351):
In w5500_receive(), the error path uses &priv->dev but the function
parameter is named self and the device field is w_dev. This compiles
only because NETDEV_RXERRORS() expands to nothing without
CONFIG_NETDEV_STATISTICS; enabling statistics breaks the build.
2. d_private set to the device array instead of the instance (line 2069):
In w5500_initialize(), d_private was set to g_w5500 (the global array)
instead of self (the current instance). This is harmless for device 0
(g_w5500 == &g_w5500[0]) but wrong for any devno > 0 — every callback
that recovers the driver state via dev->d_private would operate on
device 0's state.
Fixes#19306
Signed-off-by: hanzhijian <hanzhijian@zepp.com>
Author: hanzhijian <hanzhijian@zepp.com>
ice40_endwrite() computes how many dummy SPI bytes to clock out after
the bitstream to finish FPGA configuration with:
for (size_t i = 0; i < ICE40_SPI_FINAL_CLK_CYCLES + 7 / 8; i++)
`/` binds tighter than `+` in C, so this parses as
ICE40_SPI_FINAL_CLK_CYCLES + (7 / 8) = 160 + 0 = 160, i.e. the "+ 7 / 8"
is a silent no-op. The macro name and the classic `(n + 7) / 8`
ceiling-division idiom (used elsewhere in embedded code to convert a
bit/cycle count into a byte count) make clear the intent was to send
ceil(ICE40_SPI_FINAL_CLK_CYCLES / 8) = 20 bytes (160 SPI clock cycles,
matching the macro name). Instead the unmodified code sends 160 bytes,
i.e. 1280 clock cycles - 8x more than intended.
Fix by parenthesizing the ceiling-division: (ICE40_SPI_FINAL_CLK_CYCLES
+ 7) / 8, which evaluates to 20, restoring the intended 160-clock-cycle
finalization sequence.
Fixes#19367
Assisted-by: Claude Code:claude-sonnet-5
Signed-off-by: yi chen <94xhn1@gmail.com>
uart_putxmitchar() manipulates dev->xmit.head/buffer directly with
no internal locking - every other caller (uart_write()) takes
dev->xmit.lock first. uart_readv()'s ECHO handling (both the
backspace/delete erase sequence and the normal character echo)
calls uart_putxmitchar() without taking that lock, so a concurrent
uart_write() and a local echo can race on the same circular buffer
state, corrupting it.
Take dev->xmit.lock around each echo's uart_putxmitchar() calls,
matching what uart_write() already does. uart_readv() holds
dev->recv.lock for its own duration, but no other code path ever
acquires recv.lock while holding xmit.lock, so nesting xmit.lock
inside the existing recv.lock scope here doesn't introduce a new
lock-ordering cycle.
Fixes#14845
Signed-off-by: yi chen <94xhn1@gmail.com>
GS2200M response parsing copies device-provided text fields into fixed-size local buffers.
Add field widths to the sscanf string conversions so the parsed address, port, and command fields stay within their destination arrays.
Signed-off-by: Old-Ding <35417409+Old-Ding@users.noreply.github.com>
PWMIOC_ENABLE_LED_BANK_MODE uses the provided LED number to index the LP503X led_mode array. Reject values outside the RGB LED range before writing that array.
Generated-by: OpenAI Codex
Signed-off-by: aineoae86-sys <ai.neo.ae86@gmail.com>
pci_epf_device_register() and pci_epf_unregister_driver() used || in DEBUGASSERT expressions that validate a pointer and a required field or callback. If the pointer is NULL, the right-hand side dereferences it; if the pointer is valid but the required member is NULL, the assertion passes.
Require both conditions in each assertion so the debug checks match the preconditions used later in the functions.
Generated-by: OpenAI Codex
Signed-off-by: Old-Ding <35417409+Old-Ding@users.noreply.github.com>
When several RNDIS responses are queued, the control request handler sends one complete response if the host wLength is smaller than the whole queue. If the first queued response is also larger than wLength, copying hdr->msglen bytes would overrun the requested transfer and the completion path would consume a partial message.
Return EMSGSIZE before copying in that case so the queued response remains intact for a retry with a large enough wLength.
Signed-off-by: Old-Ding <35417409+Old-Ding@users.noreply.github.com>
pci_epc_map_msi_irq() used && when checking the EPC pointer and map_msi_irq callback. If epc is NULL, the right-hand side dereferences it; if the callback is NULL on a valid EPC, the guard does not reject it before the call.
Match the surrounding EPC wrappers by rejecting a NULL EPC, an out-of-range function number, or a missing map_msi_irq callback before taking the lock and invoking the operation.
Generated-by: OpenAI Codex
Signed-off-by: Old-Ding <35417409+Old-Ding@users.noreply.github.com>
EEPROM_DS_COUNT is the number of supported DS2XXX device types and is used as the size of the EEPROM geometry tables. Reject it before storing the device type for later table indexing.
Generated-by: OpenAI Codex
Signed-off-by: aineoae86-sys <ai.neo.ae86@gmail.com>
The existing first-order LCG (seed = 470001 * seed % 999563) has several
quality problems:
- Output range limited to [1, 999562] (~20 bits) instead of full 32-bit
- Lower bits have very short periods (8-bit period = 1, 16-bit = 105)
- Overall period only ~1M, far too short for many applications
- Causes mbedtls_rsa_gen_key to loop forever when rand() consumption
aligns with the cycle length (issue #16760)
Replace the entire order-based LCG implementation (CONFIG_LIBC_RAND_ORDER
0-3) with Marsaglia's xorshift32:
- Full 32-bit output range
- Period 2^32 - 1 (~4.29 billion)
- Fast: just three XOR/shift operations
- No floating-point math needed
- No CONFIG_LIBC_RAND_ORDER configuration required
Remove the CONFIG_LIBC_RAND_ORDER Kconfig option and clean up all
defconfig references (12 boards) and related comments.
Signed-off-by: hanzhijian <hanzhijian@zepp.com>
An unchecked integer assignment in the CAN driver may result in a division-by-zero which would result in a kernel crash (denial of service).
Tested locally, builds fine.
An unchecked integer assignment in the CAN driver may result in a division-by-zero which would result in a kernel crash (denial of service).
The CAN driver `can_ioctl` function, shown below (drivers/can/can.c), receives commands from user processes. Its received arguments `cmd` and content of `arg` are under attacker's control. In the CAN driver, some `ioctl` commands are hardware specific and are processed by calling `dev_ioctl()`. Subsequently, depending on the hardware (STM32 or AT32), this function calls `fdcan_ioctl` or `at32can_ioctl`, as shown in the snippets that follow.
```c
static int can_ioctl(FAR struct file *filep, int cmd, unsigned long arg)
{
FAR struct inode *inode = filep->f_inode;
FAR struct can_dev_s *dev = inode->i_private;
FAR struct can_reader_s *reader = filep->f_priv;
...
flags = enter_critical_section();
/* Handle built-in ioctl commands */
switch (cmd)
{
...
/* Not a "built-in" ioctl command.. perhaps it is unique to this
* lower-half, device driver. */
default:
{
ret = dev_ioctl(dev, cmd, arg);
}
break;
}
leave_critical_section(flags);
return ret;
}
```
There are a few instances where the user can trigger a kernel crash, caused by a division by zero on `CANIOC_SET_BITTIMING` command. On STM32 platforms (arch/arm/src/stm32/stm32_fdcan.c), while there is a `DEBUGASSERT` assertion for `bt->bt_baud`, the check does not cover value `0`.
```c
static const struct can_ops_s g_fdcanops =
{
...
.co_ioctl = fdcan_ioctl,
...
};
static int fdcan_ioctl(struct can_dev_s *dev, int cmd, unsigned long arg)
{
...
switch (cmd)
{
...
case CANIOC_SET_BITTIMING:
{
const struct canioc_bittiming_s *bt = (const struct canioc_bittiming_s *)arg;
uint32_t nbrp;
uint32_t ntseg1;
uint32_t ntseg2;
uint32_t nsjw;
uint32_t ie;
uint8_t state;
DEBUGASSERT(bt != NULL);
DEBUGASSERT(bt->bt_baud < STM32_FDCANCLK_FREQUENCY); // <- not valid
DEBUGASSERT(bt->bt_sjw > 0 && bt->bt_sjw <= 16);
DEBUGASSERT(bt->bt_tseg1 > 1 && bt->bt_tseg1 <= 64);
DEBUGASSERT(bt->bt_tseg2 > 0 && bt->bt_tseg2 <= 16);
/* Extract bit timing data */
ntseg1 = bt->bt_tseg1 - 1;
ntseg2 = bt->bt_tseg2 - 1;
nsjw = bt->bt_sjw - 1;
nbrp = (uint32_t)
( ((float) STM32_FDCANCLK_FREQUENCY /
((float)(ntseg1 + ntseg2 + 3) * (float)bt->bt_baud)) - 1 ); // <- div by 0
```
Similarly, another division by zero was found in Artery Technology AT32 (arch/arm/src/stm32/stm32_can.c) driver:
```c
static const struct can_ops_s g_canops =
{
...
.co_ioctl = at32can_ioctl,
...
};
static int at32can_ioctl(struct can_dev_s *dev, int cmd, unsigned long arg)
{
...
/* Handle the command */
switch (cmd)
{
...
case CANIOC_SET_BITTIMING:
{
const struct canioc_bittiming_s *bt = (const struct canioc_bittiming_s *)arg;
...
uint32_t tmp;
uint32_t regval;
DEBUGASSERT(bt != NULL);
DEBUGASSERT(bt->bt_baud < AT32_PCLK1_FREQUENCY); // <- not valid
DEBUGASSERT(bt->bt_sjw > 0 && bt->bt_sjw <= 4);
DEBUGASSERT(bt->bt_tseg1 > 0 && bt->bt_tseg1 <= 16);
DEBUGASSERT(bt->bt_tseg2 > 0 && bt->bt_tseg2 <= 8);
regval = at32can_getreg(priv, AT32_CAN_BTMG_OFFSET);
/* Extract bit timing data tmp is in clocks per bit time */
tmp = AT32_PCLK1_FREQUENCY / bt->bt_baud; // <- div by 0
```
Instances found are listed in the *Location* section below. They are not shown in detail to reduce the length of the issue.
Ensure the attacker-controlled data is properly validated before use, to stop division by zero situations. For instance:
```c
DEBUGASSERT(bt->bt_baud > 0 && bt->bt_baud < AT32_PCLK1_FREQUENCY);
```
* arch/arm/src/stm32/stm32_fdcan.c
* arch/arm/src/stm32/stm32_can.c
* arch/arm/src/sama5/sam_mcan.c
* arch/arm/src/at32/at32_can.c
* arch/arm/src/samv7/sam_mcan.c
* arch/arm/src/stm32f0l0g0/stm32_fdcan.c
* arch/arm/src/stm32f7/stm32_can.c
* arch/arm/src/stm32h5/stm32_fdcan.c
* arch/arm/src/stm32l4/stm32l4_can.c
* arch/arm/src/tiva/common/tiva_can.c
* drivers/can/mcp2515.c
Signed-off-by: Catalin Visinescu <catalin_visinescu@yahoo.com>
Align the NuttX open(2) flag constants with the Linux asm-generic
values so that the FUSE wire protocol and other cross-platform
interfaces work without conversion.
All code that used '(flags & O_RDONLY)' as a bitmask check (always 0
now that O_RDONLY=0) has been updated to use '(flags & O_ACCMODE)'
comparisons.
The NUTTX_O_* constants in include/nuttx/fs/hostfs.h are updated to
match, and the sim hostfs open flag mapping is fixed.
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
O_RDOK and O_WROK are non-standard aliases for O_RDONLY and O_WRONLY
respectively. Having two names for the same flag creates confusion,
especially when aligning the flag values with Linux. Remove the
aliases and replace all uses with the standard O_RDONLY/O_WRONLY.
No functional change — O_RDOK was defined as O_RDONLY and O_WROK as
O_WRONLY, so the replacement is a pure text substitution.
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
An attacker can specify an arbitrary page address when reading MIFARE tags.
Without validation, this could read beyond intended memory regions on the
tag, potentially causing a crash.
The mfrc522_mifare_read() command is also not robust and does not check
that the page address is valid. From section 7.6.5 of the *MIFARE Ultralight
contactless single-ticket IC
(https://www.nxp.com/docs/en/data-sheet/MF0ICU1.pdf) document:
>> The READ command needs the page address as a parameter. Only addresses
00h to 0Fh are decoded.
Testing: Builds fine.
Signed-off-by: Catalin Visinescu <catalin_visinescu@yahoo.com>
Add QSPI mode to the GD25 MTD driver alongside the existing SPI path.
When CONFIG_GD25_QSPI is selected, sector erase, chip erase, byte read,
page write, and byte write all use the QSPI command/memory interfaces
instead of SPI. Reads use the quad I/O fast-read command (1-4-4) and
writes use the quad page-program command (1-1-4) via QSPIMEM_QUADDATA.
A new Kconfig option enables the QE bit in SR2 at initialisation so the
quad I/O pins are active.
Signed-off-by: Sammy Tran <sammytran@geotab.com>
Invalid data is passed to the NXP Plug & Trust Nano Package used by the
NuttX secure element driver. If the NXP code is not handling the malformed
data, a corruption can occur. Alternatively, if the attacker is able to
point create_signature_args->algorithm in memory at an address that is not
accessible, a crash can occur.
Applicable to:
* `signature_algorithm_mapping[create_signature_args->algorithm]`
* `signature_algorithm_mapping[verify_signature_args->algorithm]`
Signed-off-by: Catalin Visinescu <catalin_visinescu@yahoo.com>
The audio tone generator stopped working with the 'echo' command
since https://github.com/apache/nuttx-apps/pull/1559
Before that PR:
nsh> echo "t120o1l16b9n0baan0bn0bn0baaan0b9n0baan0b" > /dev/tone0
tone_write: Received 41 bytes
nsh>
After that PR:
nsh> echo "t120o1l16b9n0baan0bn0bn0baaan0b9n0baan0b" > /dev/tone0
tone_write: Received 40 bytes
tone_write: Received 1 bytes
nsh>
Unfortunately the Audio Tone was not block new write attempts even
when it was already playing a melody.
This commit fix it and avoids the issue caused by that PR.
Signed-off-by: Alan C. Assis <acassis@gmail.com>
PR https://github.com/apache/nuttx/pull/19139 addresses the issue, but there
is one minor problem. In the for loop the element `i+1` is written which
means there can still be an overflow by one element (uint32_t or 4 bytes).
Addressing here with this PR.
Tested locally, builds fine.
Signed-off-by: Catalin Visinescu <catalin_visinescu@yahoo.com>
The current implementation exits syslog_write_foreach function
if write to one channel fails, causing other channels not being written
and returning negated errno. libc syslog functions then stay in
an infinite loop, because error is returned and the same bytes
are still passed to syslograwstream_flush and syslog_write_foreach.
The channel write may fail for many reasons - disconnected USB if
CDC ACM syslog is enabled, lost networking if telnet syslog is enabled,
error on NOR flash etc. This shouldn't lead to an ininite loop in the
code though.
The solution ensures all channels in syslog_write_foreach are tried,
therefore the user get the output to the working channels even if
the first one is broken. It also updates syslograwstream_addchar and
syslograwstream_addstring to skip the bytes if all channels fails. This
ensures syslog call won't result in an infinite loop, but the user may
lost the debugging output.
Co-authored-by: Martin Krasula <mkrasula@elektroline.cz>
Signed-off-by: Michal Lenc <michallenc@seznam.cz>
The function seek which allows the user to move the cursor to a particular
offset in order to read and write from EEPROM storage does not validate the
offset is valid. Later, this can cause an out-of-bounds reads or writes.
Note that newpos may store a large value, larger than the size of the EEPROM.
Similar change in the SPI driver.
Tested locally, builds fine.
Signed-off-by: Catalin Visinescu <catalin_visinescu@yahoo.com>
Update the function definitions in all 6 clk implementation files to
match the uintptr_t parameter type already declared in clk_provider.h.
Fixes CI error:
error: conflicting types for 'clk_register_divider'
Signed-off-by: hanzhijian <hanzhijian@zepp.com>
Change the 'reg' field type from uint32_t to uintptr_t in all clock
provider structs (clk_gate_s, clk_divider_s, clk_phase_s,
clk_fractional_divider_s, clk_multiplier_s, clk_mux_s) and their
corresponding clk_register_*() function prototypes.
Also update clk_write() and clk_read() inline functions to take
uintptr_t parameter and remove the now-redundant (uintptr_t) cast.
On 32-bit embedded platforms uintptr_t equals uint32_t so there is
no functional change. On 64-bit targets (e.g. sim) this fixes
-Wint-to-pointer-cast warnings that GCC15 promotes to errors.
Fixes: #16896
Signed-off-by: hanzhijian <hanzhijian@zepp.com>
This removes the DEBUGASSERT in ftl_initialize_by_path. Instead, check
compile time that FTL is enabled in case some of the drivers implement
the isbad and markbad functions.
Also select the FTL_BBM for those drivers as they require it.
Signed-off-by: Jukka Laitinen <jukka.laitinen@tii.ae>
When calling Set RF Configuration command, a compromised user
process can trigger memory corruption in the kernel. This can
lead to a system crash or potentially arbitrary code execution
in the kernel.
It addresses an earlier incomplete fix.
Tested locally.
Signed-off-by: Your Name <catalin_visinescu@yahoo.com>
A malformed packet can trigger memory corruption in the kernel leading to a
system crash or potentially arbitrary code execution in the kernel.
The CAN driver for the CTU CAN FD IP Core connected to the NuttX device
via a PCI / PCI Express (PCIe) bus shows a lack of consideration for
malformed data, assuming the CAN frames are always correct.
Ensure `frame->fmt.rwcnt` is 21 or less before it is used in the `for` loop.
A similar change was done in ctucanfd_sock_recv().
Tested locally, builds fine.
Signed-off-by: Catalin Visinescu <catalin_visinescu@yahoo.com>
up_timer_gettime() computed tv_sec and tv_nsec from two separate calls
to current_usec(), which returns a free-running microsecond counter.
The counter advances between the two calls, so a read that straddles a
second boundary yields an inconsistent (possibly backwards) timespec.
Read current_usec() once into a local variable and derive both fields
from that single snapshot.
Signed-off-by: yushuailong <yyyusl@qq.com>
While attempting to create architecture-specific implementation
of up_udelay, it was discovered that the overriding function is not
included in the final binary, the weak implementation was used instead.
Further investigation and experimentation showed that the linker
only overrides the weak implementation with the custom one if
the custom one is present in a .c source file that contains at least
one other function that is called from somewhere. Some additional
testing revealed that at least one other already present up_udelay
override (rv32m1-vega:nsh) is affected by this.
In a short mailing list discussion it was determined that this
is a likely result of using static libraries during the build process
and it was suggested to introduce configuration option that will
exclude weak implementations of the function from the build altogether.
This patch does that.
This patch does not enable this configuration option for any existing
board/chip because doing so would change its behaviour and needs
to be tested by users of the hardware.
Also changed is the static assertion in sched/clock/clock_delay.c
to not prevent building the code when architecture declares that
it does not use BOARD_LOOPSPERMSEC to determine required loop count.
BOARD_LOOPSPERMSEC is made undefined in such case.
Patch was tested by building breadxavr:nsh (identical binary by SHA256),
rv32m1-vega:nsh (identical text section) and rv-virt:nsh (text section
differs because of different ordering of functions in the binary, ostest
passed though.)
Signed-off-by: Kerogit <kr.git@kerogit.eu>
When the CONFIG_MMCSD_MMCSUPPORT is disabled, we can remove the
mmc partition support, saving ~300+ bytes of flash on a 32-bit Arm
target. These partitions don't exist on SD cards.
Signed-off-by: Jukka Laitinen <jukka.laitinen@tii.ae>
When CONFIG_FB_UPDATE is defined the missing updatearea() trigger caused
the splashscreen to not be displayed.
Signed-off-by: Martin Vajnar <martin.vajnar@gmail.com>
BREAKING CHANGE: bme680_register() takes an additional "*config" argument.
When config is NULL - driver behavior is the same as before.
With this change bme680 can be configured during registration in board logic.
This way we don't have to callibrate sensor from user-space but sensor is ready
to use after registration.
This change makes the registration the same as for bme688, which is a similar
sensor.
Signed-off-by: raiden00pl <raiden00@railab.me>
SDIO_REGISTERCALLBACK is only defined when both CONFIG_SCHED_WORKQUEUE and CONFIG_SCHED_HPWORK are enabled.
Guard the callback registration call in mmcsd_sdio.c so the source matches the SDIO callback API availability and avoids
build issues when HPWORK support is not configured.
Signed-off-by: Jerry Ma <shichunma@bestechnic.com>
uart_tcdrain() takes a caller-supplied timeout (e.g. 10s from the
TCDRN ioctl path), but the timeout was only applied to the final
TX-FIFO polling loop. The earlier xmit-buffer drain loop called
nxsem_wait(&dev->xmitsem) with no timeout, so any condition that
prevents the lower half from posting xmitsem (e.g. a stuck DMA
completion path, a wedged hardware-flow-control stall) would block
tcdrain() indefinitely, regardless of the timeout the caller asked
for. The pre-existing comment ("NOTE: There is no timeout on the
following loop. ... the caller should call tcflush() first") openly
acknowledged this hang.
Move the start timestamp before both phases and replace the bare
nxsem_wait() with nxsem_tickwait() using the remaining time, so the
total time spent in tcdrain() honors the caller's timeout regardless
of which phase stalls. When the remaining time is already exhausted,
short-circuit to -ETIMEDOUT without calling into the scheduler. The
existing exit path (drop critical section, skip the FIFO polling
loop, unlock the xmit mutex, leave the cancellation point) handles
the new -ETIMEDOUT propagation correctly without further changes.
Also fold the "Set up for the timeout" comment into the kludge
REVISIT comment, since the timestamp is no longer set up at that
point.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
uart_tcdrain() registers a cancellation point on entry via
enter_cancellation_point() (when called with cancelable=true), and the
normal exit path calls leave_cancellation_point() before returning.
However the timeout path inside the FIFO drain loop returns -ETIMEDOUT
directly without going through the normal exit path, leaking one
cancellation point reference (tcb->cpcount is left incremented). Over
repeated timeouts this counter will desync and prevent
pthread_cancel() / pthread_setcancelstate() from behaving correctly
for the calling thread.
Fix by calling leave_cancellation_point() before the early return,
matching the existing exit path.
Signed-off-by: Lingao Meng <menglingao@xiaomi.com>
Bad block management and the logical->physical mapping is only needed
with raw NAND type memories, so we can compile that in conditionally,
saving ~600 bytes of flash on a 32-bit arm when NAND is not used.
Signed-off-by: Jukka Laitinen <jukka.laitinen@tii.ae>
dac7554_initialize() calls kmm_malloc twice without checking the return
value. If either allocation fails, the subsequent pointer dereferences
lead to a NULL pointer access and crash.
Add NULL checks for both allocations, following the pattern already used
in mcp3008.c, mcp48xx.c, and mcp47x6.c. When the second allocation
fails, free the first allocation before returning NULL.
Signed-off-by: hanzj <hanzjian@zepp.com>
Rename rpmsg_device_destory() to rpmsg_device_destroy() to fix a
spelling error in the function name. The function is declared in the
private header drivers/rpmsg/rpmsg.h and used only within the
drivers/rpmsg/ subsystem (rpmsg.c, rpmsg_virtio.c,
rpmsg_router_edge.c, rpmsg_port.c), so there is no public API or
ABI impact.
Signed-off-by: hanzj <hanzjian@zepp.com>
Rename static function lan9250_set_txavailabe() to
lan9250_set_txavailable() — missing letter 'l' in 'available'.
This is a static function used only within drivers/net/lan9250.c,
so there is no API or ABI impact.
Signed-off-by: hanzj <hanzjian@zepp.com>
Fix two issues in ads1115_initialize():
1. Add missing kmm_free(priv) when adcdev allocation fails:
If the second kmm_malloc() for adcdev returns NULL, the function
returns without freeing the already-allocated priv structure,
causing a memory leak.
2. Use kmm_free() instead of free() for consistency:
The error path after cmdbyte_init() failure used free(priv) to
release memory allocated by kmm_malloc(). Use kmm_free() instead
to match the allocation API.
Signed-off-by: hanzj <hanzjian@zepp.com>
On some boards, the PCF85263 RTC does not count between reboots. Due to STOP_ENABLE (register 0x2E), bit=0 = 1, which freezes the RTC counter.
The exact trigger is unknown - not all boards exhibit the issue. The bit is battery-backed and
persists across reboots, so once set (e.g. by a power glitch or undefined hardware state) the
RTC stays frozen until explicitly cleared. The old driver never did this.
Fix: write `0x00` to `STOP_ENABLE` on init, which is the correct reset value per the datasheet.
Fix: set time properly:
Due to datasheet the set_time should be as follow:
1. set stop_enable
2. clear prescaler
3. set time
4. clear stop_enable
Signed-off-by: Marin Doetterer <marin@auterion.com>
Fix two bugs in mcp3008_initialize():
1. Remove dead free(priv) when priv is NULL (line 382):
The first allocation checks if priv == NULL, then calls free(priv)
which is a no-op since priv is NULL. Remove the dead call.
2. Add missing kmm_free(priv) when adcdev allocation fails (line 396):
If the second kmm_malloc() for adcdev fails, the function returns
NULL without freeing the already-allocated priv, causing a memory
leak. Add kmm_free(priv) before the return.
Signed-off-by: hanzj <hanzjian@zepp.com>
Adjust the generic 16550 driver for the AM62x console path. Preserve
the FIFO programming sequence needed by the TI UART, keep the bootloader
owned early console state when requested, and drain the transmit buffer
correctly when polling mode is enabled.
Signed-off-by: Piyush Patle <piyushpatle228@gmail.com>
