LeafLabs Garden » Topic: I2C Slave code implemented

LeafLabs Garden » Topic: I2C Slave code implemented http://forums.leaflabs.com/topic.php?id=2630 A place to share, learn, and grow... en-US Fri, 22 Jan 2016 00:06:05 +0000 http://bbpress.org/?v=1.0.2 <![CDATA[Search]]> q http://forums.leaflabs.com/search.php mbolivar on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13104 Mon, 03 Sep 2012 13:23:02 +0000 mbolivar 13104@http://forums.leaflabs.com/ Got it, thanks. I will review and test tomorrow (Tuesday) and get back to you. Let's move discussion of the patch over to the pull request page: <a href="https://github.com/leaflabs/libmaple/pull/54/" rel="nofollow">https://github.com/leaflabs/libmaple/pull/54/</a> Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13102 Mon, 03 Sep 2012 11:52:54 +0000 Anonymous 13102@http://forums.leaflabs.com/ OK so I put it up on my github I have requested a pull from you. Commit msg <a href="https://github.com/ginge/libmaple/commit/da01c80f24dc9c43ff951ea9fd283abb4628ceb7" rel="nofollow">https://github.com/ginge/libmaple/commit/da01c80f24dc9c43ff951ea9fd283abb4628ceb7</a> Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13095 Mon, 03 Sep 2012 03:05:44 +0000 Anonymous 13095@http://forums.leaflabs.com/ It would be nice to get it into the tree. I will follow the submission guidelines and also write something that resembles documentation for the feature. ginge mbolivar on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13084 Sun, 02 Sep 2012 20:12:37 +0000 mbolivar 13084@http://forums.leaflabs.com/ thanks for the updates and bugfixes, as well as for the advice on what makes sense regarding the API names and code structure. are you posting this patch as a proof of concept, or would you like us to apply it to the official leaflabs tree? if you're looking for this to get into our tree, we need a sign-off line along with the patch. see the libmaple contributing doc: <a href="http://leaflabs.com/docs/libmaple/contributing.html#submitting-via-email" rel="nofollow">http://leaflabs.com/docs/libmaple/contributing.html#submitting-via-email</a> Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13083 Sun, 02 Sep 2012 14:36:32 +0000 Anonymous 13083@http://forums.leaflabs.com/ Another patch! Fixed a corner case where the tx callback was called before the rx callback on a repeated start. Added new I2C state in the i2c_state enum I2C_STATE_SL_RX = 4, /**< Slave receiving */ when transmitting it is I2C_STATE_BUSY <a href="http://pastebin.com/9mk2bssk" rel="nofollow">http://pastebin.com/9mk2bssk</a> ginge Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13054 Sat, 01 Sep 2012 08:18:50 +0000 Anonymous 13054@http://forums.leaflabs.com/ new diff. fixed tabs to spaces fixed 2 small syntactical bugs added i2c_slave_init() <a href="http://pastebin.com/cjza1CnD" rel="nofollow">http://pastebin.com/cjza1CnD</a> Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13053 Sat, 01 Sep 2012 07:48:27 +0000 Anonymous 13053@http://forums.leaflabs.com/ please ignore the fact the pasted snipped has a bug ;) Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13050 Sat, 01 Sep 2012 07:36:14 +0000 Anonymous 13050@http://forums.leaflabs.com/ So I implemented an i2c_slave_enable function. It calls i2c_master_disable and then re-enables it passing the flags. i2c_slave_enable(i2c_dev *dev, uint32 flags) { i2c_master_disable(i2c_dev *dev); i2c_master_enable(dev->config_flags & flags); } I don't like it. it seems wrong. I would much prefer i2c_master_enable to be renamed to i2c_enable or somesuch. why? 1) slave depends on master code 2) enabling slave only would require duplicating a few of code paths. Unacceptable. 3) if you want to use slave only, don't call master_xfer :) 4) slave extensions to IRQ handlers are very small. separating them to accommodate i2c_reconfigure or somesuch is code duplication, and I feel a bit overkill. 5) barely any of the patch will be optimised out by the compiler because it is mostly in interrupt callbacks well... I will let you decide what is best. I will happily make some changes to the code to better fit the rest. back to implementing CAN bus! ginge Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13035 Fri, 31 Aug 2012 19:32:58 +0000 Anonymous 13035@http://forums.leaflabs.com/ Ok you are only really talking a minor change to the code for this. * Enabling the master and the slave at the same time saves having to check the master is enabled and then initialising/re-initialising. * Slave code is dependent on the master code. (clocks, irq setup, interrupts etc) to separate them completely would be a nightmare. I will add a slave_init function, but it will basically check the module is enabled, and if not call master_init with the same flags. if the module is enabled, the simplest and most code efficient route would be to disable the I2C module and then call master_init with the new flags. Let me see what I can do :) mbolivar on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13034 Fri, 31 Aug 2012 19:11:43 +0000 mbolivar 13034@http://forums.leaflabs.com/ Awesome! Looks cool; thanks for posting. One issue, though: <blockquote> Under the circumstances, I don't think i2c_master_init is a true representation of what it now does, but creating a slave_init function just added bloat. </blockquote> We can't call it i2c_master_init() if it also enables slave mode ;). Adding another function won't add to the code size of the final program unless it gets used (due to the -ffunction-sections -fdata-sections GCC flags used when compiling and --gc-sections flag passed when linking). You may be able to get master versus slave enabling sharing most of their code in a third function that's hidden from view. That's the approach taken in libmaple/spi.c: spi_master_enable() and spi_slave_enable() both defer to a single function, spi_reconfigure(): <a href="https://github.com/leaflabs/libmaple/blob/master/libmaple/spi.c#L159" rel="nofollow">https://github.com/leaflabs/libmaple/blob/master/libmaple/spi.c#L159</a> Anonymous on "I2C Slave code implemented" http://forums.leaflabs.com/topic.php?id=2630#post-13032 Fri, 31 Aug 2012 18:43:27 +0000 Anonymous 13032@http://forums.leaflabs.com/ Hi All, I am new here! Long story short: I needed I2C slave support in the libmaple core. I saw some half working and half implemented attempts scattered around, so I gave up and extended libmaple. So: I have a solid working I2C code implementation. Features: *Allows STM to run as a slave AND as a master. Multi-master tested OK (apart from the usual multi-master issues) *Allows each I2C module to run as a master and a slave *Each I2C module can accept dual i2C bus addresses *Supports General call write only addressing (on address 0x00) *Has configurable callbacks for each i2C module *Can callback on each incoming/outgoing byte *supports buffered IO. Single RX callback and TX callback for each request *Requests use the same dev_msg machanism I have extended a few chunks of code, but kept to the spirit of the current (excellent) design. notably each feature now has a new flag that is passed to i2c_master_init. Under the circumstances, I don't think i2c_master_init is a true representation of what it now does, but creating a slave_init function just added bloat. Resources: Diff file <a href="http://pastebin.com/k0vTt5rT" rel="nofollow">http://pastebin.com/k0vTt5rT</a> Example: <code> setup() { msgs.data = buf; i2c_master_enable(I2C1, I2C_FAST_MODE | I2C_SLAVE_DUAL_ADDRESS | I2C_SLAVE_GENERAL_CALL | I2C_SLAVE_USE_RX_BUFFER | I2C_SLAVE_USE_TX_BUFFER); i2c_slave_attach_recv_handler(I2C1, &msgs, funcrx); i2c_slave_attach_transmit_handler(I2C1, &msgs, functx); i2c_slave_set_own_address(I2C1, 0x0010); i2c_slave_set_own_address2(I2C1, 0x0020); } void funcrx(i2c_msg *msg) { someval = msg->data[0]; } void functx(i2c_msg *msg) { msg->data[0] = 0x01; msg->data[1] = 0x02; msg->data[2] = 0x03; msg->data[3] = 0x04; msg->data[4] = 0x05; msg->length = 5; } i2c_msg msgs; uint8 buf[255]; </code> Questions? Ginge