Openamp/libmetal's metal io operation used the atomic operation, but
in QEMU 8.1.2, the risc-v IO region is not allow to use atomic instruction
in SMP case.
Signed-off-by: Bowen Wang <wangbowen6@xiaomi.com>
Previous adjtime() implementation was limited to adjusting system
timer tick period. This commit reimplements the internals to use
a kernel watchdog timer. Platform-independent part of the code now
works also for adjusting hires RTC and tickless timer rate.
User code facing API is unchanged. Architecture code API has changed:
up_adj_timer_period() is replaced by up_adjtime().
Other improvements:
- Support query of remaining adjustment by passing NULL to first
argument of adjtime(). This matches Linux behavior.
- Improve resolution available for architecture driver, previously
limited to 1 microsecond per tick. Now 1 nanosecond per second.
In order to achieve better scalability, change the stride
from pixel mode to byte mode.For example, in the case of RGB888
mode with 466 pixels in width and a 4-byte aligned buffer,it is
only necessary to extend the buffer of one line from 1398 bytes
to 1400 bytes, instead of extending it to 1404 bytes.
Signed-off-by: rongyichang <rongyichang@xiaomi.com>
These values are board-specific properties that must be known on the application side.
Until now, these values had to be hardcoded on the application side.
When we record data and dump data, they are all executed within the system,
and there is no need to consider the issue of big or small endianness.
Signed-off-by: yinshengkai <yinshengkai@xiaomi.com>
1. Supporting `SIOCSIFADDR` and `SIOCDIFADDR` with Linux in6_ifreq struct to manage ipv6 addresses.
Ref: https://man7.org/linux/man-pages/man7/netdevice.7.html
2. Supporting alias like 'eth0:0' for multiple IPv6 addresses, to keep previous ioctl `SIOCGLIFADDR`, `SIOCSLIFADDR`, `SIOCGLIFNETMASK` and `SIOCSLIFNETMASK` working.
Ref: https://man7.org/linux/man-pages/man8/ifconfig.8.html
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
Race condition if the remote proc is stoped during initialization phase
| #0 0x0249f959 in rpmsg_destroy_ept (ept=0xffffffc0) at open-amp/lib/rpmsg/rpmsg.c:376
| #1 0x024a938c in rpmsg_deinit_vdev (rvdev=0xf2303a48) at open-amp/lib/rpmsg/rpmsg_virtio.c:971
| #2 0x02117e33 in rptun_dev_stop (rproc=0xf2303a04, stop_ns=true) at rptun/rptun.c:891
| #3 0x021181d8 in rptun_do_ioctl (priv=0xf2303a00, cmd=11010, arg=0) at rptun/rptun.c:922
| #4 0x02119722 in rptun_ioctl_foreach (cpuname=0x0, cmd=11010, value=0) at rptun/rptun.c:1086
| #5 0x0211b9df in rptun_poweroff (cpuname=0x0) at rptun/rptun.c:1378
| #6 0x02053aa6 in board_power_off (status=0) at sim/sim_head.c:206
| #7 0x0253d65c in boardctl (cmd=65283, arg=0) at boardctl.c:400
| #8 0x021eb497 in cmd_poweroff (vtbl=0xef606280, argc=1, argv=0xef9b73e0) at nsh_syscmds.c:356
| #9 0x021cdb4d in nsh_command (vtbl=0xef606280, argc=1, argv=0xef9b73e0) at nsh_command.c:1164
| #10 0x021baa72 in nsh_execute (vtbl=0xef606280, argc=1, argv=0xef9b73e0, redirfile=0x0, oflags=0) at nsh_parse.c:845
| #11 0x021c6b0a in nsh_parse_command (vtbl=0xef606280, cmdline=0xef606708 "poweroff") at nsh_parse.c:2744
| #12 0x021c7166 in nsh_parse (vtbl=0xef606280, cmdline=0xef606708 "poweroff") at nsh_parse.c:2828
| #13 0x0221fa2f in nsh_session (pstate=0xef606280, login=1, argc=1, argv=0xef7a7860) at nsh_session.c:245
| #14 0x021f8c04 in nsh_consolemain (argc=1, argv=0xef7a7860) at nsh_consolemain.c:75
| #15 0x021b77eb in nsh_main (argc=1, argv=0xef7a7860) at nsh_main.c:74
| #16 0x02166ddf in nxtask_startup (entrypt=0x21b76ca <nsh_main>, argc=1, argv=0xef7a7860) at sched/task_startup.c:70
| #17 0x020b363c in nxtask_start () at task/task_start.c:134
Signed-off-by: chao an <anchao@xiaomi.com>
In `drivers/rptun/rptun.c`, we have a `rptun_is_recursive` function,
which lets rptun thread run recursively. Then the `usrsock_rpmsg_ept_cb`
may be called inside `usrsock_rpmsg_ept_cb`, which can cause some
unexpected behavior, so we add a queue to keep the order of incoming
messages to reduce the complexity.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
When tun_txavail_work is running, switch to tun_close thread and priv->lock will be destroyed, then switch back to tun_txavail_work thread, an error will occur when nxmutex_unlock(&priv->lock)
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
There is a risk that interfaces such as psock_ioctl/psock_setsockopt
will cause paramete to be not four-byte aligned after calculating the
offset, so align the length of the structure with parameters by four
bytes.
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
When tun_txavail_work is running, switch to tun_close thread and priv->lock will be destroyed, then switch back to tun_txavail_work thread, an error will occur when nxmutex_unlock(&priv->lock)
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
In previous versions, during the teardown phase, the "open_blockdriver" would call the "mtd proxy" causing the "file mtd" node to be registered in the ftl. " node is registered in ftl.
Therefore, we changed the behavior to find the corresponding inode by node name.
Signed-off-by: chenrun1 <chenrun1@xiaomi.com>
Add the Sim WiFi function, which can provide the wifi operating on nuttx sim emulator,
and support two modes that simulate wifi, HWSIM and RNC(real network card).
- In the HWSIM mode, we simulates two wlan interfaces. The wlan0 is STA and
the wlan1 is AP. The wlan0 can connect to the wlan1 in the nuttx simulator.
- In the RNC mode, we can use the same wlan interface name on the nuttx simulator
to control the connection behavior of the real wireless card.
Signed-off-by: liqinhui <liqinhui@xiaomi.com>
The BIT macro is widely used in NuttX,
and to achieve a unified strategy,
we have placed the implementation of the BIT macro
in bits.h to simplify code implementation.
Signed-off-by: hujun5 <hujun5@xiaomi.com>
so the user could register fb device directly instead
go through the sequence of up_fbinitialize/up_fbgetvplane
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
forget from patch:
commit bbf5b0bb6d
Author: jianglianfang <jianglianfang@xiaomi.com>
Date: Tue Aug 15 19:47:14 2023 +0800
driver/video: adapting Goldfish FB to VSync optimized FB driver
The circbuf part is implemented by the fb driver, and goldfish needs to
delete the part related to circbuf and adapt to the interface of the fb
driver.
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
==2787216==ERROR: AddressSanitizer: heap-use-after-free on address 0xf4702448 at pc 0x0136c544 bp 0x9f54d358 sp 0x9f54d348
READ of size 4 at 0xf4702448 thread T0
0 0x136c543 in rpmsg_destroy_ept open-amp/lib/rpmsg/rpmsg.c:357
1 0x1252b29 in sensor_rpmsg_ns_unbind_cb sensors/sensor_rpmsg.c:1216
2 0x1371de5 in rpmsg_virtio_ns_callback open-amp/lib/rpmsg/rpmsg_virtio.c:663
3 0x1371026 in rpmsg_virtio_rx_callback open-amp/lib/rpmsg/rpmsg_virtio.c:583
4 0x137c105 in virtqueue_notification open-amp/lib/virtio/virtqueue.c:711
5 0x1368604 in rproc_virtio_notified open-amp/lib/remoteproc/remoteproc_virtio.c:433
6 0x1363abe in remoteproc_get_notification open-amp/lib/remoteproc/remoteproc.c:10
Signed-off-by: yintao <yintao@xiaomi.com>
convert from body coordinate system to right-hand
coordinate system.
Example:
Compared to the standard coordinate system,
the x-axis and y-axis are interchanged and have opposite directions,
the z-axis remains normal.
body coordinate ----> right-hand coordinate
+x +y
| |
| |
| |
| |
-y<-------. .------>+x
/ /
/ /
/ /
/ /
+z +z
So for the above conversion, using "P3" to represent transformation relationships
The front is 1 0 2, which represents the y x z axis.
The standard order is 0 1 2, so y and x are interchanged.
The following -1 1 1 indicates the direction of the axis.
The standard is 1 1 1. Because the current y-axis is opposite to
the standard x-axis, it is -1.
static const struct sensor_axis_map_s g_remap_tbl[] =
{
{ 0, 1, 2, 1, 1, 1 }, /* P0 */
{ 1, 0, 2, 1, -1, 1 }, /* P1 */
{ 0, 1, 2, -1, -1, 1 }, /* P2 */
{ 1, 0, 2, -1, 1, 1 }, /* P3 */
{ 0, 1, 2, -1, 1, -1 }, /* P4 */
{ 1, 0, 2, -1, -1, -1 }, /* P5 */
{ 0, 1, 2, 1, -1, -1 }, /* P6 */
{ 1, 0, 2, 1, 1, -1 }, /* P7 */
};
you can call the function sensor_remap_vector_raw16 and pass P3 parameters
to perform the conversion.
Signed-off-by: Jiuzhu Dong <dongjiuzhu1@xiaomi.com>
When the application calls poll and returns the POLLIN event,
QPOLL will be called again during the read operation,
causing rx_length to change and data to be lost.
Therefore, read only need to actively call qpoll to collect driver
data when rx length is 0.
Signed-off-by: dongjiuzhu1 <dongjiuzhu1@xiaomi.com>
I added a function that I didn't use when I upstream another patch. Delete it
https://github.com/apache/nuttx/pull/10709
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
1.check the rpmsg name when clk_register.
2.clk_get rpmsg clk print error:
clk/clk.c:1231:3: runtime error: null pointer passed as argument 2, which is declared to never be null
3.fix some function miss keyword FAR.
4.clk_disable_unused may disable rpmsg clk which disable the clk that maybe used by others
Signed-off-by: dulibo1 <dulibo1@xiaomi.com>
1.client get the regualtor which is enabled by server, will disable the regualtor.
2.regulator_rpmsg_server_unbind will disable regualtor which maybe used by other client.
3.regulator_rpmsg_server_unbind will be deadloop when the regulator is always_on;
4.regulator_rpmsg_client_destroy does not match server cpu name,may destory by stop other rptun dev
Signed-off-by: dulibo1 <dulibo1@xiaomi.com>
These structures need to be aligned with the firmware, packed_struct will cause the scan information from Wi-Fi firmware not to be parsed by broadcom Wi-Fi driver. Therefore, we need to remove the packed_struct of the structure associated with the scan information.
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
Solve the problem that when a user modifies the sampling rate,
the difference between the generation and the mainline generation
may be greater than the own sampling interval.
After this problem occurs, sensor_is_update() will continue to
return true, resulting in a busy loop in the upper-layer application,
so, need update user generation according to mainline generation.
Signed-off-by: dongjiuzhu1 <dongjiuzhu1@xiaomi.com>
When the upper layer calls goldfish_camera_data_uninit, priv->capture_cb=NULL, but when there is data transmission in goldfish_camera_thread, priv->capture_cb will be called, which will cause a crash.
Signed-off-by: yangsen5 <yangsen5@xiaomi.com>
1.Change priv to client or server
2.Allocate memory for name in regulator_rpmsg_get
3.Lock/unlock regulator_list in regulator_rpmsg_server_s
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
It uses the memory block as the serial communication medium, which can communicate between different processes or different CPUs
Using the following configuration, the cross-core communication rate of 200MHz cortex-M55 is about 461KB/s
RAM_UART_BUFSIZE=1024
RAM_UART_POLLING_INTERVAL=100
Signed-off-by: yinshengkai <yinshengkai@xiaomi.com>
A dead lock may happen before this patch:
rptun thread (usrsock_rpmsg_recvfrom_handler):
Gets all tx payload but waiting net_lock in psock_recvfrom
net driver thread (higher priority):
Receives a packet, holding net_lock, calling usrsock_rpmsg_send_event
But no tx buffer left, so rpmsg_send blocks, and won't release net_lock
In short:
Rptun: Hold all `tx_payload_buffer` -> want `net_lock`
Driver: Hold `net_lock` -> want `tx_payload_buffer`
This patch use net_lock to combine get_tx_payload and recvfrom together,
then:
- If it's waiting net_lock, tx payload will not be held, other threads
may get tx buffer and do their work.
- If net_lock is got, it won't be disturbed before finish receiving,
then tx payload won't have chance to be blocked before releasing.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
Note:
tcp_poll_eventhandler have logic after poll_notify, if we teardown poll
inside poll callback without clearing revents, use-after-free will
happen.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
The circbuf part is implemented by the fb driver, and goldfish needs to
delete the part related to circbuf and adapt to the interface of the fb
driver.
Signed-off-by: jianglianfang <jianglianfang@xiaomi.com>
refact vsync queue to support multi fb poll and overlay poll.
Signed-off-by: pengyiqiang <pengyiqiang@xiaomi.com>
Signed-off-by: rongyichang <rongyichang@xiaomi.com>
Summary
The following compilation error occurs after configuring CONFIG_DEBUG_OPTLEVEL="-O3"
CC: ping.c video/video.c: In function 'video_ioctl':
video/video.c:2347:22: error: 'control.size' may be used uninitialized [-Werror=maybe-uninitialized]
2347 | control->size,
| ^~
video/video.c:2273:28: note: 'control' declared here
2273 | struct v4l2_ext_control control;
| ^~~~~~~
Signed-off-by: wangming9 <wangming9@xiaomi.com>
Add usb_fs driver so that userspace can directly transfer USB packets
through the EP node. ADB, Fastboot, MTP will use usb_fs, these class
driver only need to provide the descriptors and register the usb_fs device.
Signed-off-by: zhangyuan21 <zhangyuan21@xiaomi.com>
Signed-off-by: zhangyuan21 <zhangyuan21@xiaomi.com>
The RPMB partition cannot be accessed via standard block layer, but by a set of specific commands: WRITE, READ, GET_WRITE_COUNTER, and PROGRAM_KEY. Such a partition provides authenticated and replay protected access, hence suitable as a secure storage.
Signed-off-by: wanggang26 <wanggang26@xiaomi.com>
In sensor_rpmsg_ioctl, when rpmsg_get_tx_payload_buffer is block,
The rptun thread will continue to process rx_buffers from other cores,
So the proxy may be released in sensor_unadv_handler, then if using
the proxy, It causes a crash.
backtrace:
0 0x06007c2e in is_rpmsg_ept_ready (ept=0x5a5a5a5a) at nuttx/include/openamp/rpmsg.h:172
1 0x06013354 in rpmsg_send_nocopy (len=<optimized out>,
data=<optimized out>, ept=<optimized out>) at nuttx/include/openamp/rpmsg.h:521
2 rpmsg_send_nocopy (ept=0x5a5a5a5a, data=data@entry=0x201facf0,
len=len@entry=40) at nuttx/include/openamp/rpmsg.h:512
3 0x06013b34 in sensor_rpmsg_ioctl (dev=dev@entry=0x201b7388,
cmd=538701816, cmd@entry=2690, arg=0, len=len@entry=0,
wait=<error reading variable: dwarf2_find_location_expression:
Corrupted DWARF expression.>) at ../../../drivers/sensors/sensor_rpmsg.c:395
4 0x06013cce in sensor_rpmsg_alloc_proxy (dev=dev@entry=0x201b7388,
ept=ept@entry=0x202073d0, msg=msg@entry=0x201684f0) at ../../../drivers/sensors/sensor_rpmsg.c:498
5 0x06013d12 in sensor_rpmsg_adv_handler (ept=0x202073d0,
data=0x201684f0, len=49, src=<optimized out>, priv=0x202073c8) at ../../../drivers/sensors/sensor_rpmsg.c:905
6 0x0600d45e in sensor_rpmsg_ept_cb (ept=<optimized out>,
data=<optimized out>, len=<optimized out>, src=<optimized out>,
priv=0x202073c8) at ../../../drivers/sensors/sensor_rpmsg.c:1168
Signed-off-by: dongjiuzhu1 <dongjiuzhu1@xiaomi.com>
vfork use waitpid hang father process,
but waitpid release child processs information by default.
So when user call wait, it return errno 10.
Signed-off-by: yangyalei <yangyalei@xiaomi.com>
When the machine is cold started, the psram area where the syslog
buffer is located contains some random values, so it will cause
loss of valid log.
Signed-off-by: dongjiuzhu1 <dongjiuzhu1@xiaomi.com>
When the system starts up, the hw count may not start counting from 0,
which can cause the processing time of the first oneshot handler to be
too long. Therefore, the current tick value should be set based on the
hw count during initialization.
Signed-off-by: zhangyuan21 <zhangyuan21@xiaomi.com>
The class device only handles descriptor information specific to the class,
and shared descriptor information is passed through parameters and
handled by the composite driver.
Signed-off-by: zhangyuan21 <zhangyuan21@xiaomi.com>
Whevenever the bit rate is going to be set, it's necessary to first
set the master clock if the underlying device supports it. If it
fails, just return the error code.
According to the struct `mclk_rate_s`, set the default sample rate
to its lowest valid value. This prevents the underlying driver from
setting unsupported values.
Create bmi160_base.h, bmi160_base.c bmi160_uorb.c files, the bmi160_base file stores
public function interfaces, and the bmi160_uorb file stores functions related to the
uorb framework. Switch the character interface and UORB interface through the
macro CONFIG_SENSORS_BMI160_UORB.
Signed-off-by: likun17 <likun17@xiaomi.com>
A usbdev has only one serial string, so use a unique macro to control it.
For boards that enable board serial string using COMPOSITE_BOARD_SERIALSTR,
PL2303_BOARD_SERIALSTR, CDCACM_BOARD_SERIALSTR, USBADB_BOARD_SERIALSTR,
USBMSC_BOARD_SERIALSTR, and RNDIS_BOARD_SERIALSTR, they need to be replaced
with BOARD_USBDEV_SERIALSTR.
Signed-off-by: zhangyuan21 <zhangyuan21@xiaomi.com>
Commit 50a8ec6 broke many mmc devices. Only if the flag
priv->caps & SDIO_CAPS_4BIT_ONLY was set, it migth work.
Without the flag, the mmc clock is never set (mmcsd_widebus()
call is terminated early stopping the clock). This flag
is probably not very generic because most mmc hw support
1, 4 and 8 bit modes.
JEDEC specifies a bus width selection procedure, but it's
not implemented in this mmcsd_sdio.c driver. Thus, it's
not known whether the hw supports 1, 4 anf 8 bit modes or
a combination of them.
However, with priv->caps & SDIO_CAPS_4BIT_ONLY the driver
suddenly assigns priv->buswidth = MMCSD_SCR_BUSWIDTH_4BIT
making it the only way to have the driver working.
Fix this by relaxing the above mentioned restrictions.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
This commit adds basic support for electrically isolated 8 bit expander.
The expander communicates with the MCU via SPI interface. Both single
and multiple pin read are supported.
Signed-off-by: Michal Lenc <michallenc@seznam.cz>
isx019 driver freezes by I2C access in the interrupt context.
This I2C access is intended to stop data output from FPGA.
Delete this I2C access for the following reasons.
- Data output from FPGA does not affect power consumption
- There are no problems in capture restart without data output stop
since restart is done by image data block first.
This commit adds support for galvanic isolated ISO1H812G SPI expander.
Both single pin and multiple pin writes are supported.
Signed-off-by: Michal Lenc <michallenc@seznam.cz>
Documentation for PWMIOC_SETCHARACTERISTICS ioctl command mentioned
that this command will neither start nor stop the pulsed output. This
however is incorrect as PWMIOC_SETCHARACTERISTICS command leads to
pwm_start() function which starts the pulsed output.
While this might not be the correct behaviour (I would probably welcome
the option to set PWM characteristics without starting the pulsed output)
it is the way the PWM driver is coded for many architectures. Future
enhancement might be to add function pwm_setchar() for example to just
set characteristics without starting the PWM output.
Signed-off-by: Michal Lenc <michallenc@seznam.cz>
TCP receive tested with different IOB_BUFSIZE:
| | 256B vs 1534B | 512B vs 1534B | 768B vs 1534B |
| :---------: | :-----------: | :-----------: | :-----------: |
| Non-SMP | ~85% | ~93% | ~96% |
| armv8a-SMP | ~66% | ~84% | ~92% |
| rv32/64-SMP | ~52% | ~72% | ~83% |
It seems we still get performance penalty on smaller IOBs, and may be
affected more under SMP mode. It may be caused by critical sections in
IOB operations.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
Some driver like virtio-net can offload fragmented IOBs, so an interface
to support this feature is needed.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
Previously, the CONFIG_NET_LL_GUARDSIZE is fixed to 32 (64-Bit) or
28 (32-Bit), it's a little bit tricky.
Now add support to any value greater than minimal size.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
