assert RTS signal during transmit in EIA-485 half duplex mode for 16550A compatible controllers.

From 0aa88d5d4fe5ab43f3069ab4dc8125d4a866f24c Mon Sep 17 00:00:00 2001

Hi,

I make PATCH for the 16550A serial drivers.
It allow to assert RTS signal during transmit in EIA-485 half duplex mode for 16550A compatible controllers.

I have also modified the libmodbus-2.0.3 to be used with xenomai if anyone is intesested// in this lib, please let me know

Thanks, Alexandre

From 0aa88d5d4fe5ab43f3069ab4dc8125d4a866f24c Mon Sep 17 00:00:00 2001
From: Alexandre Coffignal <[email protected]>
Date: Wed, 20 Jan 2010 15:13:39 +0100
Subject: [PATCH] rtdm : Allow to assert RTS signal during transmit in EIA-485 half duplex
mode for 16550A compatible controllers.

---
 include/rtdm/rtserial.h      |   34 ++++++++++++
 ksrc/drivers/serial/16550A.c |  115 ++++++++++++++++++++++++++++++++++++++++-
 ksrc/drivers/serial/Kconfig  |    7 +++
 3 files changed, 153 insertions(+), 3 deletions(-)

include/rtdm/rtserial.h +0 -0

index 48712b2..3a7b091 100644

include/rtdm/rtserial.h +2 -0

 /** @} */
 
+
+
 /*!
  * @anchor RTSER_FIFO_xxx   @name RTSER_FIFO_xxx
  * Reception FIFO interrupt threshold

include/rtdm/rtserial.h +13 -0

+
+#define RS485_DEF_DELAY_AFTER_SEND	100000	//100uS
+
+/**
+ * Set config for rs485 half duplex mode
+ *
+ * @param[in] arg New rs485 config content
+ *
+ * @return 0 on success, otherwise negative error code
+ *
+ * Environments:
+ *
+ * This service can be called from:

include/rtdm/rtserial.h +0 -0

 #endif /* _RTSERIAL_H */

ksrc/drivers/serial/16550A.c +15 -1

index db8a515..45c5cc3 100644
@@ -107,6 +107,12 @@ struct rt_16550_context {
 	int mcr_status;			/* MCR cache */
 	int status;			/* cache for LSR + soft-states */
 	int saved_errors;		/* error cache for RTIOC_GET_STATUS */
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	size_t bytes_send;		/* bytes sent on last tx it*/
+	rtdm_task_t rts_task;	/* task used to assert RTS signal during transmit(rs485)*/
+ 	rtdm_sem_t itTx;		/* raised to unblock rts task (rs485) */
+ 	rs485_config_t rs485_config;		/* delay after TX frame before rts down in nanosecond(rs485) */
+#endif
 };
 
 static const struct rtser_config default_config = {

@@ -204,8 +210,16 @@ static inline void rt_16550_tx_interrupt(struct rt_16550_context *ctx)
 			ctx->out_head &= (OUT_BUFFER_SIZE - 1);
 		}
 	}
-}
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	if(ctx->config.handshake==RTSER_RTS_RS485_HALF)
+	{
+		/* up to unblock rts task in rs485 half duplex mode */
+		ctx->bytes_send=ctx->tx_fifo-count;
+		rtdm_sem_up(&ctx->itTx);
+	}
+#endif
 
+}
 static inline void rt_16550_stat_interrupt(struct rt_16550_context *ctx)
 {
 	unsigned long base = ctx->base_addr;

ksrc/drivers/serial/16550A.c +38 -0

 		ctx->config.handshake = config->handshake;
 
 		switch (ctx->config.handshake) {
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+		case RTSER_RTS_RS485_HALF:
+			ctx->mcr_status =
+			    RTSER_MCR_DTR | RTSER_MCR_RTS | RTSER_MCR_OUT2;
+			break;
+#endif
 		case RTSER_RTSCTS_HAND:
 			// ...?
 

@@ -437,6 +456,51 @@ void rt_16550_cleanup_ctx(struct rt_16550_context *ctx)
 	rtdm_mutex_destroy(&ctx->out_lock);
 }
 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+#define RTS_STD_PRIO 1
+
+/* RS485
+ * allow to put in transmit mode explicitly by asserting RTC signal to the driver during
+ * transmission
+ *
+ * This allows EIA-485 to implement linear topologies using only two wires
+ *
+ */
+void rts_task(void *arg)
+{
+	struct rt_16550_context *ctx= (struct rt_16550_context *)arg;
+ 	int ret;
+ 	nanosecs_rel_t delay ;
+ 	char event=1;
+
+ 	while(1){
+ 		//wait sem up from itTx
+ 		ret =rtdm_sem_down (&ctx->itTx);
+ 		if(ret)
+ 			printk("error rts_task... sem down error %d\n",ret);
+
+ 		do{
+ 			//wait end of transmission nbit + * fifo size / baud rate
+ 			delay=11*ctx->bytes_send*1000000/ctx->config.baud_rate;
+ 			ret=rtdm_sem_timeddown(&ctx->itTx, delay*1000+ctx->rs485_config.delay_after_send,NULL);
+  			if(ret){
+ 				if(ret!=-ETIMEDOUT){
+ 					printk("error rts_task rtdm_event_timedwait... error %d\n",ret);
+ 				}
+ 				//no new rtdm event itTx receive during the transfert

ksrc/drivers/serial/16550A.c +41 -1

+}
+#endif
+
 int rt_16550_open(struct rtdm_dev_context *context,
 		  rtdm_user_info_t * user_info, int oflags)
 {

@@ -458,7 +522,9 @@ int rt_16550_open(struct rtdm_dev_context *context,
 	rt_16550_init_io_ctx(dev_id, ctx);
 
 	ctx->tx_fifo = tx_fifo[dev_id];
-
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	ctx->rs485_config.delay_after_send=RS485_DEF_DELAY_AFTER_SEND;
+#endif
 	ctx->in_head = 0;
 	ctx->in_tail = 0;
 	ctx->in_npend = 0;

@@ -475,6 +541,7 @@ int rt_16550_open(struct rtdm_dev_context *context,
 	ctx->status = 0;
 	ctx->saved_errors = 0;
 
+
 	rt_16550_set_config(ctx, &default_config, &dummy);
 
 	err = rtdm_irq_request(&ctx->irq_handle, irq[dev_id],

@@ -493,6 +560,17 @@ int rt_16550_open(struct rtdm_dev_context *context,
 
 	rtdm_lock_get_irqsave(&ctx->lock, lock_ctx);
 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	/* create semaphore used to unblock rts_task (rs485)*/
+	rtdm_sem_init(&ctx->itTx, 0);
+	/* create task used to assert RTS signal during transmit in (rs485)*/
+	err = rtdm_task_init(&ctx->rts_task, "rts_task", &rts_task,  ctx, RTS_STD_PRIO, 0);
+	if (err) {
+		printk("error rtdm_task_init\n");
+		return 0;
+	}
+#endif
+
 	/* enable interrupts */
 	ctx->ier_status = IER_RX;
 	rt_16550_reg_out(rt_16550_io_mode_from_ctx(ctx), ctx->base_addr, IER,

@@ -535,6 +613,11 @@ int rt_16550_close(struct rtdm_dev_context *context,
 
 	rtdm_irq_free(&ctx->irq_handle);
 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	rtdm_sem_destroy(&ctx->itTx);
+	rtdm_task_destroy(&ctx->rts_task);
+#endif
+
 	rt_16550_cleanup_ctx(ctx);
 
 	if (in_history) {

@@ -802,6 +885,24 @@ int rt_16550_ioctl(struct rtdm_dev_context *context,
 		rtdm_lock_put_irqrestore(&ctx->lock, lock_ctx);
 		break;
 	}
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	case RTSER_RTIOC_RS485_SET_CONFIG: {
+		struct rs485_config *config;
+		struct rs485_config config_buf;
+		config = (struct rtser_config *)arg;
+		if (user_info) {
+			err = rtdm_safe_copy_from_user(user_info, &config_buf,
+						     arg,
+						     sizeof(struct rs485_config));
+			if (err)
+				return err;
+
+			config = &config_buf;
+			ctx->rs485_config.delay_after_send=config->delay_after_send;
+		}
+		break;
+#endif
+	}
 
 	default:
 		err = -ENOTTY;

@@ -1005,6 +1106,14 @@ ssize_t rt_16550_write(struct rtdm_dev_context * context,
 	if (ret)

ksrc/drivers/serial/16550A.c +0 -0

 	while (nbyte > 0) {
 		rtdm_lock_get_irqsave(&ctx->lock, lock_ctx);
 

ksrc/drivers/serial/Kconfig +0 -0

index afbaabf..95337bf 100644
@@ -39,4 +39,11 @@ config XENO_DRIVERS_16550A_ANY
 
 endchoice

Quoted 9 lines · Alexandre Coffignal

Hi,

I make PATCH for the 16550A serial drivers.
It allow to assert RTS signal during transmit in EIA-485 half duplex mode for 16550A compatible controllers.

I have also modified the libmodbus-2.0.3 to be used with xenomai if anyone is intesested// in this lib, please let me know

Thanks for sharing your work! In general, I'm open to enhance the driver to support this, but find my question regarding the necessity below.

Quoted 8 lines

Thanks, Alexandre

From 0aa88d5d4fe5ab43f3069ab4dc8125d4a866f24c Mon Sep 17 00:00:00 2001
From: Alexandre Coffignal <[email protected]>
Date: Wed, 20 Jan 2010 15:13:39 +0100
Subject: [PATCH] rtdm : Allow to assert RTS signal during transmit in EIA-485 half duplex
mode for 16550A compatible controllers.

Short subject line + some separate description lines, please. If you are interesting in getting something integrated upstream, you would also have to add a signed-off line.

Quoted 11 lines

---
 include/rtdm/rtserial.h      |   34 ++++++++++++
 ksrc/drivers/serial/16550A.c |  115 ++++++++++++++++++++++++++++++++++++++++-
 ksrc/drivers/serial/Kconfig  |    7 +++
 3 files changed, 153 insertions(+), 3 deletions(-)

include/rtdm/rtserial.h +0 -0

index 48712b2..3a7b091 100644

include/rtdm/rtserial.h line 133

Quoted 7 lines

include/rtdm/rtserial.h +3 -0

@@ -127,8 +127,13 @@
 #define RTSER_NO_HAND			0x00
 #define RTSER_RTSCTS_HAND		0x01
 #define RTSER_DEF_HAND			RTSER_NO_HAND
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+#define RTSER_RTS_RS485_HALF	0x02
+#endif
 

A conditionally API in a user space header makes no sense. There is no technical reason to hide this in the !CONFIG_XENO_DRIVERS_16550A_RS485_HALF case, and user space will not even have this define set - so won't be able to use it.

Quoted 8 lines

include/rtdm/rtserial.h +2 -0

 /** @} */
 
+
+
 /*!
  * @anchor RTSER_FIFO_xxx   @name RTSER_FIFO_xxx

* Reception FIFO interrupt threshold

include/rtdm/rtserial.h typedef struct rtser_event { · line 517

Quoted 10 lines

include/rtdm/rtserial.h +6 -0

@@ -503,4 +508,33 @@ typedef struct rtser_event {
 
 /** @} */
 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+
+typedef struct rs485_config
+{
+	nanosecs_rel_t delay_after_send;
+}rs485_config_t;
 

In general no typedef, please, just a struct. But we could easily fold this field into existing rtser_config as we have flags to indicate which field is valid.

Quoted 14 lines

include/rtdm/rtserial.h +13 -0

+
+#define RS485_DEF_DELAY_AFTER_SEND	100000	//100uS
+
+/**
+ * Set config for rs485 half duplex mode
+ *
+ * @param[in] arg New rs485 config content
+ *
+ * @return 0 on success, otherwise negative error code
+ *
+ * Environments:
+ *
+ * This service can be called from:

include/rtdm/rtserial.h typedef struct rtser_event { · line 539

Quoted 10 lines

include/rtdm/rtserial.h +10 -0

+ *
+ * - Kernel module initialization/cleanup code
+ * - Kernel-based task
+ * - User-space task (RT, non-RT)
+ *
+ * Rescheduling: never.
+ */
+#define RTSER_RTIOC_RS485_SET_CONFIG \
+	_IOW(RTIOC_TYPE_SERIAL, 0x07, struct rs485_config)
+#endif

No need for this then.

Quoted 37 lines

#endif /* _RTSERIAL_H */

ksrc/drivers/serial/16550A.c +15 -1

index db8a515..45c5cc3 100644
@@ -107,6 +107,12 @@ struct rt_16550_context {
 	int mcr_status;			/* MCR cache */
 	int status;			/* cache for LSR + soft-states */
 	int saved_errors;		/* error cache for RTIOC_GET_STATUS */
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	size_t bytes_send;		/* bytes sent on last tx it*/
+	rtdm_task_t rts_task;	/* task used to assert RTS signal during transmit(rs485)*/
+ 	rtdm_sem_t itTx;		/* raised to unblock rts task (rs485) */
+ 	rs485_config_t rs485_config;		/* delay after TX frame before rts down in nanosecond(rs485) */
+#endif
 };
 
 static const struct rtser_config default_config = {

@@ -204,8 +210,16 @@ static inline void rt_16550_tx_interrupt(struct rt_16550_context *ctx)
 			ctx->out_head &= (OUT_BUFFER_SIZE - 1);
 		}
 	}
-}
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	if(ctx->config.handshake==RTSER_RTS_RS485_HALF)
+	{
+		/* up to unblock rts task in rs485 half duplex mode */
+		ctx->bytes_send=ctx->tx_fifo-count;
+		rtdm_sem_up(&ctx->itTx);
+	}
+#endif
 
+}
 static inline void rt_16550_stat_interrupt(struct rt_16550_context *ctx)
 {
 	unsigned long base = ctx->base_addr;

ksrc/drivers/serial/16550A.c static int rt_16550_set_config(struct rt_16550_context *ctx, · line 426

Quoted 5 lines

ksrc/drivers/serial/16550A.c +0 -1

@@ -409,10 +423,15 @@ static int rt_16550_set_config(struct rt_16550_context *ctx,
 	if (testbits(config->config_mask, RTSER_SET_HANDSHAKE)) {
 		/* change handshake atomically */
 		rtdm_lock_get_irqsave(&ctx->lock, lock_ctx);
-
 

Take care to avoid spurious whitespace changes.

Quoted 49 lines

ksrc/drivers/serial/16550A.c +39 -0

 		ctx->config.handshake = config->handshake;
 
 		switch (ctx->config.handshake) {
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+		case RTSER_RTS_RS485_HALF:
+			ctx->mcr_status =
+			    RTSER_MCR_DTR | RTSER_MCR_RTS | RTSER_MCR_OUT2;
+			break;
+#endif
 		case RTSER_RTSCTS_HAND:
 			// ...?

@@ -437,6 +456,51 @@ void rt_16550_cleanup_ctx(struct rt_16550_context *ctx)
 	rtdm_mutex_destroy(&ctx->out_lock);
 }
 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+#define RTS_STD_PRIO 1
+
+/* RS485
+ * allow to put in transmit mode explicitly by asserting RTC signal to the driver during
+ * transmission
+ *
+ * This allows EIA-485 to implement linear topologies using only two wires
+ *
+ */
+void rts_task(void *arg)
+{
+	struct rt_16550_context *ctx= (struct rt_16550_context *)arg;
+ 	int ret;
+ 	nanosecs_rel_t delay ;
+ 	char event=1;
+
+ 	while(1){
+ 		//wait sem up from itTx
+ 		ret =rtdm_sem_down (&ctx->itTx);
+ 		if(ret)
+ 			printk("error rts_task... sem down error %d\n",ret);
+
+ 		do{
+ 			//wait end of transmission nbit + * fifo size / baud rate
+ 			delay=11*ctx->bytes_send*1000000/ctx->config.baud_rate;
+ 			ret=rtdm_sem_timeddown(&ctx->itTx, delay*1000+ctx->rs485_config.delay_after_send,NULL);
+  			if(ret){
+ 				if(ret!=-ETIMEDOUT){
+ 					printk("error rts_task rtdm_event_timedwait... error %d\n",ret);
+ 				}
+ 				//no new rtdm event itTx receive during the transfert
+ 				event=0;

ksrc/drivers/serial/16550A.c void rt_16550_cleanup_ctx(struct rt_16550_context *ctx) · line 501

Quoted 10 lines

ksrc/drivers/serial/16550A.c +9 -0

+ 			}
+ 			else//new rtdm event itTx receive during the transfert, still waiting
+ 				event=1;
+
+ 		}while(event);
+		//down RTS
+		ctx->mcr_status &= ~RTSER_MCR_RTS;
+		rt_16550_reg_out(rt_16550_io_mode_from_ctx(ctx),ctx->base_addr, MCR, ctx->mcr_status);
+	}
 

This task can't be terminated properly. So you risk all kinds of corruptions by enforcing this on close via rtdm_task_destroy. Most
problematic: Deleting the sem and not breaking out of the loop will freeze your box (without an active Xenomai watchdog).

However, can you describe to me as someone not really familiar with 485 what it is doing? Also, what prevents running this job in user space, maybe extending the existing rtserial API a bit if something is missing to do so? Do you need some xmit completing indication? That would be an extension to RTSER_RTIOC_WAIT_EVENT then.

Quoted 83 lines

ksrc/drivers/serial/16550A.c +3 -0

+}
+#endif
+
 int rt_16550_open(struct rtdm_dev_context *context,
 		  rtdm_user_info_t * user_info, int oflags)

{

ksrc/drivers/serial/16550A.c +38 -1

@@ -458,7 +522,9 @@ int rt_16550_open(struct rtdm_dev_context *context,
 	rt_16550_init_io_ctx(dev_id, ctx);
 
 	ctx->tx_fifo = tx_fifo[dev_id];
-
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	ctx->rs485_config.delay_after_send=RS485_DEF_DELAY_AFTER_SEND;
+#endif
 	ctx->in_head = 0;
 	ctx->in_tail = 0;
 	ctx->in_npend = 0;

@@ -475,6 +541,7 @@ int rt_16550_open(struct rtdm_dev_context *context,
 	ctx->status = 0;
 	ctx->saved_errors = 0;
 
+
 	rt_16550_set_config(ctx, &default_config, &dummy);
 
 	err = rtdm_irq_request(&ctx->irq_handle, irq[dev_id],

@@ -493,6 +560,17 @@ int rt_16550_open(struct rtdm_dev_context *context,
 
 	rtdm_lock_get_irqsave(&ctx->lock, lock_ctx);
 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	/* create semaphore used to unblock rts_task (rs485)*/
+	rtdm_sem_init(&ctx->itTx, 0);
+	/* create task used to assert RTS signal during transmit in (rs485)*/
+	err = rtdm_task_init(&ctx->rts_task, "rts_task", &rts_task,  ctx, RTS_STD_PRIO, 0);
+	if (err) {
+		printk("error rtdm_task_init\n");
+		return 0;
+	}
+#endif
+
 	/* enable interrupts */
 	ctx->ier_status = IER_RX;
 	rt_16550_reg_out(rt_16550_io_mode_from_ctx(ctx), ctx->base_addr, IER,

@@ -535,6 +613,11 @@ int rt_16550_close(struct rtdm_dev_context *context,
 
 	rtdm_irq_free(&ctx->irq_handle);
 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	rtdm_sem_destroy(&ctx->itTx);
+	rtdm_task_destroy(&ctx->rts_task);
+#endif
+
 	rt_16550_cleanup_ctx(ctx);
 
 	if (in_history) {

@@ -802,6 +885,24 @@ int rt_16550_ioctl(struct rtdm_dev_context *context,
 		rtdm_lock_put_irqrestore(&ctx->lock, lock_ctx);
 		break;
 	}
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	case RTSER_RTIOC_RS485_SET_CONFIG: {
+		struct rs485_config *config;
+		struct rs485_config config_buf;
+		config = (struct rtser_config *)arg;
+		if (user_info) {
+			err = rtdm_safe_copy_from_user(user_info, &config_buf,
+						     arg,
+						     sizeof(struct rs485_config));
+			if (err)
+				return err;
+
+			config = &config_buf;
+			ctx->rs485_config.delay_after_send=config->delay_after_send;
+		}
+		break;
+#endif
+	}
 
 	default:
 		err = -ENOTTY;

@@ -1005,6 +1106,14 @@ ssize_t rt_16550_write(struct rtdm_dev_context * context,
 	if (ret)
 		return ret;

ksrc/drivers/serial/16550A.c ssize_t rt_16550_write(struct rtdm_dev_context * context, · line 1117

Quoted 10 lines

ksrc/drivers/serial/16550A.c +8 -0

 
+#ifdef CONFIG_XENO_DRIVERS_16550A_RS485_HALF
+	if(ctx->config.handshake==RTSER_RTS_RS485_HALF)
+	{
+		//assert RTS signal during transmit in rs485 half duplex mode
+		ctx->mcr_status |= RTSER_MCR_RTS;
+		rt_16550_reg_out(rt_16550_io_mode_from_ctx(ctx),ctx->base_addr, MCR, ctx->mcr_status);
+	}
+#endif
 

Again my question: Does this have to be in the driver?

Quoted 11 lines

ksrc/drivers/serial/16550A.c +0 -0

 	while (nbyte > 0) {
 		rtdm_lock_get_irqsave(&ctx->lock, lock_ctx);

ksrc/drivers/serial/Kconfig +0 -0

index afbaabf..95337bf 100644
@@ -39,4 +39,11 @@ config XENO_DRIVERS_16550A_ANY
 
 endchoice
 

ksrc/drivers/serial/Kconfig config XENO_DRIVERS_16550A_ANY

Quoted 10 lines

ksrc/drivers/serial/Kconfig +7 -1

+config XENO_DRIVERS_16550A_RS485_HALF
+	depends on XENO_DRIVERS_16550A
+	bool "EIA-485 half duplex"
+	help
+	Assert RTS signal during transmit in EIA-485 half duplex mode
+	for 16550A compatible controllers.
+
 endmenu
-- 1.6.0.4

Jan

Hello,

Thanks you for your response.

I used this patch to communicate with a slave equipment in rs485 (modbus protocol). The data is transmitted over a 2-wire twisted pair bus. Master initiates all communication activity, by sending data. To do this it must set RTS line to high state during all it's transmission. Before slave answer, master must lower the RTS line.

my slave equipment answer in less than 200us, so i must know when xmit is completing to lower the RTS line

An extension to RTSER_RTIOC_WAIT_EVENT may be a good solution but do you know if an extension to RTSER_RTIOC_WAIT_EVENT can guarantee this timing?

Quoted 10 lines · Alexandre Coffignal

Hello,

Thanks you for your response.

I used this patch to communicate with a slave equipment in rs485 (modbus protocol). The data is transmitted over a 2-wire twisted pair bus. Master initiates all communication activity, by sending data. To do this it must set RTS line to high state during all it's transmission. Before slave answer, master must lower the RTS line.

Will the slave wait on the deassertion of RTS? Or what will happen if RTS is still high and the slave starts its answer?

Quoted 8 lines

my slave equipment answer in less than 200us, so i must know when xmit is completing to lower the RTS line

An extension to RTSER_RTIOC_WAIT_EVENT may be a good solution but do you know if an extension to RTSER_RTIOC_WAIT_EVENT can guarantee this timing?

Given a proper platform (check its capabilities with the latency or the irqbench test), yes. It's a myth that such things can only be done in kernel space.

Jan

Quoted 6 lines · Jan Kiszka

Quoted 4 lines · Alexandre Coffignal

Hello,

Thanks you for your response.

Quoted 10 lines

Quoted 6 lines

I used this patch to communicate with a slave equipment in rs485 (modbus protocol). The data is transmitted over a 2-wire twisted pair bus. Master initiates all communication activity, by sending data. To do this it must set RTS line to high state during all it's transmission. Before slave answer, master must lower the RTS line.

Will the slave wait on the deassertion of RTS? Or what will happen if RTS is still high and the slave starts its answer?

No slave does not wait on the deassertion of RTS, if RTS is still high during slave answer, it happen a collision and
the master does not receive the frame.

Quoted 12 lines

Quoted 8 lines

my slave equipment answer in less than 200us, so i must know when xmit is completing to lower the RTS line

An extension to RTSER_RTIOC_WAIT_EVENT may be a good solution but do you know if an extension to RTSER_RTIOC_WAIT_EVENT can guarantee this timing?

Given a proper platform (check its capabilities with the latency or the irqbench test), yes. It's a myth that such things can only be done in kernel space.

Quoted 3 lines

Jan

Ok I take note of this, but do you think this function should not be in kernel space in half duplex RS485 ?