Remove the deprecated HCI driver API which was provided by the hci_driver.h
header file. The deprecation happened in Zephyr 3.7, so the API can now be
removed for Zephyr 4.1.
Signed-off-by: Johan Hedberg <johan.hedberg@silabs.com>
This prevents recursive logging loop when USB CDC ACM is used for shell
with logging.
Change affects both USB stacks.
Signed-off-by: Xudong Zheng <7pkvm5aw@slicealias.com>
The check for the device being configured or suspended in the fifo_fill
implementation can starve other threads, because the early return does
not change the state of the TX path, and fifo_fill is called again
from the callback loop. The ringbuffer that emulates the TX FIFO can be
filled and marked ready as long as the space is available.
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
Update the interrupt driver UART drivers that use `struct ring_buf`
internally to report the number of bytes that can be pushed in
`uart_fifo_fill` without fragmentation.
Signed-off-by: Jordan Yates <jordan@embeint.com>
cdc_acm_fifo_fill() logged warnings about dropped data
when they in fact just couldn't be sent (due to connection
of filled buffer). Not being able to send (all) data is
a valid behaviour and data isn't dropped. Return value
shows how many data was accepted (which can be zero).
Logging will now instead only inform data can't be sent
due to the respective issue (connection not ready/ buffer full).
Signed-off-by: Martin Koehler <koehler@metratec.com>
Move the network buffer header file from zephyr/net/buf.h to
zephyr/net_buf.h as the implementation now lives outside of the networking
subsystem.
Add (deprecated) zephyr/net/buf.h header to maintain compatibility with old
file path.
Signed-off-by: Henrik Brix Andersen <henrik@brixandersen.dk>
Usb "write" method never copies the input data buffer, it uses the
buffer pointer directly. The fact that it was on the stack made
the buffer ephemeral, and could disappear before the USB transfer
actually occurred. The data transmitted was then random. This commit
converts the transfer buffer to a static buffer, so that it always
exists for the duration of the USB transfer.
Known side-effect: if "read capacity" command is read twice
simultaneously, the same buffer will be used twice. As the capacity of a
USB drive should not change between 2 calls, this side effect can be
ignored.
Signed-off-by: Alexi Demers <alexi.demers@axceta.com>
Convert users of net_buf_put() and net_buf_get() functions to use
non-wrapped putters and getters k_fifo_put() and k_fifo_get().
Special handling of net_bufs in k_fifos is no longer needed after commit
3d306c181f, since these actions are now
atomic regardless of any net_buf fragments.
Signed-off-by: Henrik Brix Andersen <henrik@brixandersen.dk>
Change updates log level only if UART log backend is enabled in Kconfig
configuration. The log level update is not needed in case logs are
provided over other backend (e.g. RTT).
Change affects both USB stacks.
Signed-off-by: Marek Pieta <Marek.Pieta@nordicsemi.no>
The ring buffer API is explicitly not thread safe, with users needing to
implement their own locking. As `poll_out` and `fifo_fill` are operating
on the same ringbuffer, these locks are needed.
Signed-off-by: Jordan Yates <jordan@embeint.com>
Generating separate log entry at INFO level for every single character
dropped is excessive and leads to log flood. Logging dropped character
in no way helps end user and is really a delayed performance killer that
triggers when CDC ACM buffer gets full.
If user does not want to lose outgoing characters then the solution is
to enable hardware flow control which properly blocks in the case the
output buffer is full.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
In some relatively rare conditions, the DFU detach/attach happens to
fast, which cause the host to not notice it. It seems to to be the case
for instance on STM32 when no endpoint are being used.
Adding a 1 ms delay (i.e. one SOF period) between the detach and the
attach fixes the issue.
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
It might happens that DT(_INST)_PROP_OR is used with boolean properties.
For instance:
.single_wire = DT_INST_PROP_OR(index, single_wire, false), \
.tx_rx_swap = DT_INST_PROP_OR(index, tx_rx_swap, false), \
This is not required as boolean properties are generated with false
value when not present, so the _OR macro extension is superflous
and the above code can be replaced by:
.single_wire = DT_INST_PROP(index, single_wire), \
.tx_rx_swap = DT_INST_PROP(index, tx_rx_swap), \
Signed-off-by: Roman Studenikin <srv@meta.com>
audio_receive_cb is only used by headphones and headset, it is unused in
microphone-only configurations. Gate compilation based on the device
tree.
Signed-off-by: James Zipperer <jzipperer@fb.com>
Queue characters in TX ring buffer regardless of interface state. This
allows simple applications that print basic output to CDC ACM port to
not care about connection or DTR state. Note that depending on actual
application requirements it might be still wise to wait for DTR before
trying to send data.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
Prevent ECHO on Linux by arming IN endpoint with ZLP when interface is
configured and making sure that actual payload is only sent after
initialization timeout. The ZLP is not visible to host side applications
because the applications are really accessing tty buffer and received
ZLP does not modify tty buffer in any way.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
According to the UART API documentation, implementation must block when
the transceiver is full. For CDC ACM UART, TX ringbuffer can be
considered as transceiver buffer/FIFO. Blocking when the USB subsystem
is not ready is considered highly undesirable behavior. Blocking may
also be undesirable when CDC ACM UART is used as a logging backend.
Change the behavior of CDC ACM poll out to:
- Block if the TX ring buffer is full, hw_flow_control property is
enabled, and called from a non-ISR context.
- Do not block if the USB subsystem is not ready, poll out
implementation is called from an ISR context, or hw_flow_control
property is disabled.
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
When a usb_transfer is cancelled, the callback is finally invoked.
Silly, there is no transfer status passed, and in the callback we can
only check the size and device status.
Reported-by: Alex Kaiser
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
Fix few instances of delayable work handlers using the k_work pointer
directly in a CONTAINER_OF pointing to a k_work_delayable.
This is harmless since the k_work is the first element in
k_work_delayable, but using k_work_delayable_from_work is the right way
of handling it.
Change a couple of explicit CONTAINER_OF doing the same work as the
macro in the process.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
The usb_msc.h header does not provide any API. The content is only
used by the MSC implementation for the current USB device stack and
is not required for any use of MSC by the application.
The content has no proper namespace and must not be reused for
anything else in (new) USB support.
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
Effectively, this option changes code triple in device descriptor.
Although the name is misleading, renaming it would again lead
to negative user experiences. Instead, clarify what the option does
and always select it where it is required.
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
Every USB device must buffer isochronous data for at least 1 frame in
order to remove packet jitter within a frame. USB Specification Revision
2.0 describes necessary buffering in 5.12.5 Data Prebuffering. USB Audio
1.0 specification mentions buffering in 3.4 Inter Channel Synchronization.
Set bDelay to 1 instead of 0, because bDelay 1 is both the minimum
allowed value and matches the actual device behavior.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
Instead of assuming pointers are 32bits wide,
cast them to the appropriate pointer arithmetic type.
Signed-off-by: Alberto Escolar Piedras <alberto.escolar.piedras@nordicsemi.no>
Prevent potential buffer overflow when encapsulated response is more
than CONFIG_USB_REQUEST_BUFFER_SIZE. Log error and truncate response if
USB control transfer request buffer is not large enough.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
The init infrastructure, found in `init.h`, is currently used by:
- `SYS_INIT`: to call functions before `main`
- `DEVICE_*`: to initialize devices
They are all sorted according to an initialization level + a priority.
`SYS_INIT` calls are really orthogonal to devices, however, the required
function signature requires a `const struct device *dev` as a first
argument. The only reason for that is because the same init machinery is
used by devices, so we have something like:
```c
struct init_entry {
int (*init)(const struct device *dev);
/* only set by DEVICE_*, otherwise NULL */
const struct device *dev;
}
```
As a result, we end up with such weird/ugly pattern:
```c
static int my_init(const struct device *dev)
{
/* always NULL! add ARG_UNUSED to avoid compiler warning */
ARG_UNUSED(dev);
...
}
```
This is really a result of poor internals isolation. This patch proposes
a to make init entries more flexible so that they can accept sytem
initialization calls like this:
```c
static int my_init(void)
{
...
}
```
This is achieved using a union:
```c
union init_function {
/* for SYS_INIT, used when init_entry.dev == NULL */
int (*sys)(void);
/* for DEVICE*, used when init_entry.dev != NULL */
int (*dev)(const struct device *dev);
};
struct init_entry {
/* stores init function (either for SYS_INIT or DEVICE*)
union init_function init_fn;
/* stores device pointer for DEVICE*, NULL for SYS_INIT. Allows
* to know which union entry to call.
*/
const struct device *dev;
}
```
This solution **does not increase ROM usage**, and allows to offer clean
public APIs for both SYS_INIT and DEVICE*. Note that however, init
machinery keeps a coupling with devices.
**NOTE**: This is a breaking change! All `SYS_INIT` functions will need
to be converted to the new signature. See the script offered in the
following commit.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
init: convert SYS_INIT functions to the new signature
Conversion scripted using scripts/utils/migrate_sys_init.py.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
manifest: update projects for SYS_INIT changes
Update modules with updated SYS_INIT calls:
- hal_ti
- lvgl
- sof
- TraceRecorderSource
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
tests: devicetree: devices: adjust test
Adjust test according to the recently introduced SYS_INIT
infrastructure.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
tests: kernel: threads: adjust SYS_INIT call
Adjust to the new signature: int (*init_fn)(void);
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Track LBA rather than byte offset in order to simplify the code and
support devices larger than 4GiB.
Signed-off-by: JP Sugarbroad <jpsugar@amazon.com>
Allows receive packets that are bigger than
CONFIG_RNDIS_BULK_EP_MPS.
Considering:
* MaxPacketsPerTransfer is 1
* MaxTransferSize is 1558
* PacketAlignmentFactor is 0 (20=1)
The rndis_bulk_out shall be able to receive transfer
up to 1558 bytes (inclusive).
Signed-off-by: Georgij Cernysiov <geo.cgv@gmail.com>
Cancelling transfers on suspend contradicts Universal Serial Bus
Specification Revision 2.0, 9.1.1.6 Suspended:
* When suspended, the USB device maintains any internal status,
including its address and configuration.
The internal status definitely includes any pending USB transfers.
If there is a class that wants to cancel transfer on suspend, then the
cancel should be initiated by the class, not the device stack itself.
Update hal_nordic to a version that does not abort all endpoints at
suspend. It seems that aborting endpoints on suspend in nrfx driver was
the actual reason why transfers were canceled on suspend.
Remove transfer retriggering on resume from CDC ACM and Bluetooth class
implementations because transfers are no longer cancelled on suspend.
Other classes do not have any suspend related workarounds implemented.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
USB_TRANS_NO_ZLP flag has no meaning for usb_transfer() in
host-to-device direction (USB_TRANS_READ).
Signed-off-by: Johann Fischer <johann.fischer@nordicsemi.no>
Use k_work_delayable instead of k_timer in order to execute timeout in
USB workqueue context instead of ISR context. This fixes Will-Detach on
targets where usb_dc_detach() uses functions not allowed in ISR context,
e.g. nrfx usb_dc_detach() acquires mutex.
Fixes: c27d48c89a ("usb: dfu: Support DFU with WinUSB on Windows")
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
bcdHID is intended for the spec version of USB HID. It was pointing to the
v1.1 bcdUSB define which happens to be the same value for the v1.10 hid
spec version. This corrects it to use the v1.11 HID spec.
Signed-off-by: Ryan McClelland <ryanmcclelland@meta.com>
Clean up occurrences of "#if IS_ENABLED(CONFIG_FOO)" an replace
with classical "#if defined(CONFIG_FOO)".
Signed-off-by: Erwan Gouriou <erwan.gouriou@linaro.org>
Allow enabling CDC ACM logging only if CDC ACM is not used as logging
backend. This prevents endless recursive logging loop, especially
visible when minimal footprint logging is enabled.
Fixes: #52981
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
Schedule CDC ACM tx work to happen 1 ms after first unprocessed data
write. This gives enough leeway for the task writing to the queue to
write more than 1 byte before USB workqueue preempts and schedules IN
transfer. Sending more than 1 byte at a time increases data throughput
because transaction overhead remains the same regardless of packet size.
Prior to this change, virtually every IN transaction carried only one
character when CDC ACM was used as a console.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
The bwPollTimeout name is somewhat confusing as it refers to minimum
polling period. The bwPollTImeout is essentially a mechanism allowing
device to ratelimit DFU_GETSTATUS requests from host.
Setting bwPollTimeout to relatively high values effectively slows down
DFU download process. The slowdown is not observed if the time between
DFU_DNLOAD and DFU_GETSTATUS is enough for the device to process the
write. That is, the bwPollTimeout does not effect DFU download if the
first DFU_GETSTATUS after DFU_DNLOAD reports dfuDNLOAD-IDLE state.
Otherwise the host must wait bwPollTimeout ms before issuing next
DFU_GETSTATUS, which slows the communicaiton to not more than 1 download
block (CONFIG_USB_REQUEST_BUFFER_SIZE) every bwPollTimeout ms.
The bwPollTimeout ideally should report an estimate how much longer the
download operation will take. Zephyr does not have such estimate and
therefore defaults to using fixed value. Reduce default bwPollTimeout
from 256 to 8 ms to allow significantly faster DFU downloads on devices
where the time between DFU_DNLOAD and DFU_GETSTATUS is too short to
process download block at the expense of unnecessary bus traffic if
processing download block takes longer than 8 ms.
Signed-off-by: Tomasz Moń <tomasz.mon@nordicsemi.no>
Pass result of usb_transfer() as return value to the user,
and allow user to react accordingly, e.g. release net_buf on error.
Signed-off-by: Chun-Chieh Li <ccli8@nuvoton.com>
