This commit reverts 19f269e54b and
substitutes the semaphore to a mutex, used as a locking mechanism
to avoid syslog messages being messed.
Considering SMP, threads running on different CPUs could call
syslog_default_write simultaneously. It then calls `up_nputs` that,
in most of the implementations, calls `up_putc` multiple times to
write to the serial. So, calling it at the same would cause syslog
messages to interfere with each other.
`up_nputs` and `up_putc` are low-level functions that could be used
during initialization and by the interrupt handler. Hence, it isn't
advisable to implement any locking mechanism on them. On the other
hand, syslog can also be used from the interrupt: if the interrupt
buffer is selected (`CONFIG_SYSLOG_INTBUFFER=y`), the syslog
messages will be flushed when we aren't serving an interrupt.
On the contrary, if it isn't set, it would call the registered
`sc_force` callback, which calls `syslog_default_putc` instead of
the lock-protected `syslog_default_write`.
Summary:
- I noticed that the driver sends incorrect packets sometimes.
- This commit fixes this issue.
Impact:
- None
Testing:
- Tested with qemu-7.1
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
This is a refactored version of the SLIP network driver. Updates
include:
1. The original design started two kernel threads per SLIP device.
The refactored version uses file_poll to essentially be driven
by the UART RX and TX interrupts and pushes work to the low
priority work queue.
2. The SLIP byte un-/stuffing is more efficient now, using memcpy
instead of handling each byte individually.
3. The switch of the old SLIP driver to IOBs caused buffer overwrites
if packets were sent that would not fit into a single IOB. This is
fixed now.
Signed-off-by: Michael Jung <michael.jung@secore.ly>
When updating the full-screen data, sq_ init() will clear the updqueue and add just one new full-screen data to the updqueue. So when the vnc_updater thread is awakened, it may read a null data due to multiple reading, leading to vnc_updater thread exited , which is not expected.
Signed-off-by: jianglianfang <jianglianfang@xiaomi.com>
set touch.maxpoint to 1 in vnc_server.c for circbuf_init，otherwise
touch_event will circbuf_overwrite a wrong value.
test step:vnc run lvgldemo，test input event
Signed-off-by: jianglianfang <jianglianfang@xiaomi.com>
USB drivers should not try control devices in the usbhost_disconnected
function. The control endpoint is NULL at this point because the device
has already been disconnected.
Before this change, the hub driver crashed when the child called
usbhost_hubpwr after the parent had already deactivated the control
pipe and set hport->ep0 to NULL.
1 Gbit W25Q01JV flash uses two 512 Mbit dies. This requires specific
operations in the driver in order to support the entire 1 Gbit region.
This commit adds dual die support for W25Q01JV flash.
Generally the driver has to switch active dies when die specific read from
status register is required (therefore before read, write, erase and when
checking if write status register was successful).
Signed-off-by: Michal Lenc <michallenc@seznam.cz>
if multiple threads are doing serial read/write at the same time,
the driver will only wake up one of the thread, which will cause
other threads fail to be woken up in time and cause blocking
Signed-off-by: chao an <anchao@xiaomi.com>
Before atomic is ready on every platform, we need a way to protect
quota, previously it's protected by net_lock, but this makes netpkt_xxx
interfaces unable to be called under interrupt context, so changed to
spin lock. Also, we can easily change them into atomic under this
implementation.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
All interrupts must be disabled to prevent re-entrancy and to prevent
interrupts from firing in the serial driver code.
Signed-off-by: hujun5 <hujun5@xiaomi.com>
W25QXXXJV_WRITE_STATUS_2 register uses just first byte therefore all
operations has to be done in priv->cmdbuf[0]. Previous priv->cmdbuf[1]
caused QuadSPI mode not being enabled.
Signed-off-by: Michal Lenc <michallenc@seznam.cz>
Summary:
- This commit adds virtio-mmio-blk driver
Impact:
- None
Testing:
- Tested with rv-virt:netnsh which will be updated later
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
If the current poll setup only has POLLIN, adding POLLOUT does not cause the
protocol stack to refocus on flags such as TCP_ACK or UDP_POLL, the user is
not notified when flags for POLLOUT relationships appear and vice versa,
so we have to call poll_setup again.
Signed-off-by: zhanghongyu <zhanghongyu@xiaomi.com>
An off-by-one error caused invalid memory to be accessed when the
last port was used. The first entry, at index 0, was never used.
The USB standard uses 1-based numbering for hub ports. This number
was used to index an array. This change converts the port number to
a zero-based port index when the array is accessed. The zero-based
value in the port field of the usbhost_hubport_s structure is also
converted before printing.
For testing, this bug was exposed on a 4 port hub when port 4 was
used after changing USBHUB_MAX_PORTS to 4 in hub.h. The bug should
also be triggered without changing hub.h if a 7 port hub is
available.
Considering that the future fb needs to achieve dynamic resolution, the address and size of fbmem will be redistributed on the bottom driver side, so remove fbmem, fblen, and bpp cached in fb_chardev_s, and change it to call getplaneinfo every time to re-acquire panel info from the bottom driver. Avoid cache coherency problems.
Signed-off-by: pengyiqiang <pengyiqiang@xiaomi.com>
Signed-off-by: jianglianfang <jianglianfang@xiaomi.com>
Devices connected to the same USB bus should have unique function addresses.
This was not true for root hubs with multiple ports. After this change,
enumeration is more reliable on the sama5d3-xplained board when both root hub
ports are used.
This change amounts to using one usbhost_devaddr_s object per root hub
instead of one per root hub port. For the majority of boards only one
root hub port is available so no change in behavior should be expected.
