Qspi patch #1
Open
Qspi patch #1
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The backend currently returns the pipe closed event immediately after calling uart_rx_disable() which is not the correct behavior. the pipe closed event should be called when the UART_RX_DISABLED event is raised by the UART driver. With this fix, back-to-back open/close/open... will work as expected, where before the second open would often fail since the UART was not actually disabled yet. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit improves the usage of the bus pipe (connected to UART) to only open it when the modem is actually powered and ready, not when leaving the idle state. This ensures the pipe is flushed before sending the init script, and re-enables the UART driver if it is disabled due to errors. While building a test platform based on the nRF9160 and a Quectel BG95, it was discovered that the nRF9160 correctly throws UART errors if the RX is enabled while the UART RX line is low (which was due to the modem being powered down). The improvements should also help help remove the "<wrn> modem_chat: receive buffer overrun" warning which would occur during startup as the pipe was opened, but nothing was receiving the data, causing the buffer to overflow. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit delegates the modem_pipe_notify_closed() call resulting from the UART async API UART_RX_DISABLED event to the workqueue. This is neccesary as the async UART callback may be called from ISR context. modem_pipe_notify_closed() must be called from outside of the ISR context as it takes a mutex. The commit also adds a missing break to the async UART callback, and adds a missing dependency to the Kconfig for the UART backends, RING_BUFFER=y Signed-off-by: Bjarki Arge Andreasen <[email protected]>
Modem backed configs should be used instead of uart. Signed-off-by: Wojciech Slenska <[email protected]>
This commit adds support for partial matches for the modem_chat module. A match is a combination of an expected response to a request along with delimiters to use and a handler for the parsed response. The usual behavior of the modem_chat script is to continue to the next step when any expected response is received. The partial flag indicates that the script should not proceed to the next step if the response matches the match. This is useful for commands which respond with an unspecified number of lines, followed by an "OK". This flag allows the script to essentially run in a "while" loop until OK is received. Along with this addition, a more scalable macro for initializing the modem_chat match struct, MODEM_CHAT_MATCH_INITIALIZER(). Without this macro, we will need 4 different macros to initialize the 4 variants of a chat_match, and 8 when the next feature is added... Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This PR makes the modem_pipe instances track if they have data ready to receive, and invoke the RECEIVE_READY event every time they are attached if the backend implementing the pipe has notified that receive is ready. This mechanism ensures that modules attaching to a pipe get the async RECEIVE_READY event immediately after attaching to a pipe if there is data ready, instead of having to poll the pipe, or worse, wait until newer data becomes available. The addition revealed a timing issue in the cmux test suite. Specifically the CMUX instance now immediately receives the response to a command which the CMUX instance has not sent yet, causing it to drop the response. The CMUX test suite now uses the transaction mechanism of the mock_pipe to wait for the command before sending the response. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
Added ppp net stats to modem subsys. Signed-off-by: Wojciech Slenska <[email protected]>
This commit adds one new feature, modem_chat_run_script(), which synchronously runs a script. The existing function modem_chat_script_run() is async, and doesn't follow the naming convention created for the other modem modules, specifically, all functions are sync, unless appended with _async. To preserve backwards compatibility, the existing function modem_chat_script_run() will remain until deprecated. It simply calls modem_chat_run_script_async(). Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit updates the modem_cellular driver to use the new naming scheme for the modem_chat functions. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit optimizes the performance of the script chats by storing the size of requests with the chat script chat, negating the need to use strlen() within the modem_chat module every time a chat script chat request is sent. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This PR adds a mechanism to avoid calling open() or close() on pipes which are already opened or closed respectively. This optimization can help simplify backends implementing the modem_pipe API by avoiding duplicated boilerplate code. The TTY backend test suite has been updated to match the new behavior. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit adds a timeout of 300ms to the generic (gsm_ppp) init chat script. This delay is required for some modems (discovered on a Telit ME910G1-WW) to allow it to enter CMUX mode. Without this delay, the modem simply refuses to respond to any CMUX commands. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit sets the C/R (command/response) bit when UIH CMUX frames are sent from the modem_cmux module. This bit is ignored by some modems like the Quectel BG95, as there is no defined response to this specific CMUX frame type. However, other modems, like the TELIT ME910, require the bit to be set (command). If the bit is not set, the modem will simply ignore the frame completely. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit increases the buffer used for commands in the control channel within an instance of the modem_cmux module. The buffer was not large enough to store an MSC command if the optional break signals where included. This commit fixes the issue and updates the test suite to use the max size MSC message. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit adjust the chat scripts for the simcom sim7080 and gsm_ppp compatibles to fix an issue found when trying to use a sim7080 modem. The CMUX command includes optional parameters which are not identical for all modems, so the AT+CMUX command has been adjusted to only contain the mandatory parameters. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
There is no fallthrough statement or comment so I assume break is missing. Fixes build time warning from some tools. Signed-off-by: Andrei Emeltchenko <[email protected]>
This commit adds a check to the async connect and disconnect functions to validate the CMUX is not already connected or disconnected respectively. This was already part of the sync connect and disconnect functions, so the sync functions now simply wrap the async functions to avoid duplicate code. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit resets the CMUX events to match the initial value of disconnected. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit replaces the k_event_wait calls using K_NO_WAIT with k_event_test. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit resets the receive ring buffer for each DLCI pipe when they are opened. They may have old data stored from last time they where opened. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit adds logging for CMUX frames and commands and their data, for both transmit and receive. It also removes the superseded LOG_DBG() lines like "Received frame". Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit resets the state of the CMUX receive state machine when the CMUX instance is attached to a pipe. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit reuses the string "Unknown %s frame type" for two log messages, as suggested by ycsin. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
Fix fault on modem_cellular disconnection/reconnection Issue #64291 Signed-off-by: romain pelletant <[email protected]>
Refactor device macro to be custom for each modem to allow custom configurations. Signed-off-by: Jeff Welder <[email protected]>
Added chat script and initial config to support Telit ME910G1. Needed to edit power_pulse pin timing in accordance to datasheet. Signed-off-by: Jeff Welder <[email protected]>
The modem subsystem is novel and should therefore be marked experimental. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
The init scripts need a missing wait for OK after sending AT+CGMM and the AT+CMUX command specifies more parameters than nessesary, and in some cases incorrectly. This commit adds the missing wait for OK and fixes the AT+CMUX commands in the init scripts. Signed-off-by: Bjarki Arge Andreasen <[email protected]>
This commit introduces support for the modem EG25-G from Quectel. Signed-off-by: Lucas Denefle <[email protected]>
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hello @ldenefle, and thank you very much for your first pull request to the Zephyr project!
A project maintainer just triggered our CI pipeline to run it against your PR and ensure it's compliant and doesn't cause any issues. You might want to take this opportunity to review the project's Contributor Expectations and make any updates to your pull request if necessary. 😊
The way that the QSPI peripheral is activated has been changed in nrfx 3.2.0. Now the peripheral is not activated during the driver initialization. Instead, the driver activates the peripheral when the first operation is requested or when `nrfx_qspi_activate()` is called. In case of XIP, the latter needs to be used, as there may be no standard operation request. Signed-off-by: Andrzej Głąbek <[email protected]>
After entering the Deep Power-down mode, some flash chips ignore all commands except from the one that releases the chip from the DP mode and it is not possible to successfully read their Status Register then. Since the QSPI peripheral tries to read this register when it is being activated, it consequently fails to send the actual command that would release the flash chip from the DP mode if that is to be done right after QSPI initialization. Prevent this problem by performing the QSPI activation with all pins disconnected. This causes that the Status Register value is read as all zeros and allows the activation to always finish successfully, and the RDPD command to be properly sent. Signed-off-by: Andrzej Głąbek <[email protected]>
Consistently use `res` for results of calls to nrfx functions and `rc` for Zephyr return codes, to avoid mixing up those two and for example calling `qspi_get_zephyr_ret_code()` for a value that is already a Zephyr return code. Correct also such call in `qspi_nor_write()`. Signed-off-by: Andrzej Głąbek <[email protected]>
After integration of nrfx 3.2.0, it is no longer needed to deinitialize the nrfx_qspi driver to avoid increased power consumption when the QSPI peripheral is idle. Now it is enough to call `nrfx_qspi_dectivate()` when a given operation is done. The driver will automatically activate the QSPI peripheral again when a next operation is requested. This commit applies the following changes: - `qspi_device_init` and `qspi_device_uninit` functions are replaced by `qspi_acquire` and `qspi_release`, respectively; those handle exclusive access to the QSPI peripheral and deactivation of it or runtime device power management - locking is removed from `qspi_send_cmd` as it is the resposibility of the caller of that function - `trans_lock` and `trans_unlock` functions are removed together with the related semaphore as they are no longer needed - checking of input parameters is moved from `qspi_erase` to its caller, `qspi_nor_erase` - `qspi_nor_pm_action` is refactored to properly handle locking of the QSPI peripheral; checking of the `xip_enabled` flag is removed from that function as now the call to `pm_device_is_busy()` covers that (when XIP is enabled, the device is kept indicated as busy) Signed-off-by: Andrzej Głąbek <[email protected]>
So far the driver first changed the configuration of the flash chip and after that checked the signature of that chip. This could lead to improper change of the chip configuration if the actually found one was different than that specified in devicetree. This commit reverses the order of these two initialization steps and also restructures a bit the initialization code. Signed-off-by: Andrzej Głąbek <[email protected]>
ldenefle
force-pushed
the
main
branch
2 times, most recently
from
6c5b388 to
8fb601a
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.