login
Search
For serial over ethernet communication, /dev/ptmx will be used to
create a virtual device.
Due to the lack of a modem status register and a modem control
register in the ptmx driver,
it won't be possible to transfer handshake signaling.I wrote this patch to add these registers to the ptmx driver.
The modem status register, which controls CD,CTS,DSR,RI is normally
read-only. To be able to
pass these signals from the remote port to the local ptmx device, I've
added an IOCTL routine.
The result of a remote TIOCMGET IOCTL call , can be passed directly
through this IOCTL routine.
The IOCTL value I used for this is, 0x5426 which was formerly known as
TIOCTTYGSTRUCT.I've tested this patch at an 2.6.16.27 kernel.
Suggestions and/or other test results are welcome.
Sander
Signed-off-by: Sander van Ginkel <sander@van-ginkel.eu>
===================================================================
--- drivers/char/pty.c 2008-03-03 20:02:09.000000000 +0100
+++ drivers/char/pty-new.c 2008-03-02 21:55:53.000000000 +0100
@@ -7,6 +7,8 @@
* -- C. Scott Ananian <cananian@alumni.princeton.edu>, 14-Jan-1998
* Added TTY_DO_WRITE_WAKEUP to enable n_tty to send POLL_OUT to
* waiting writers -- Sapan Bhatia <sapan@corewars.org>
+ * Added support for MCR/MSR, used for serial over ethernet
+ * -- Sander van Ginkel <sander@van-ginkel.eu>
*
*
*/
@@ -32,6 +34,16 @@
#include <linux/bitops.h>
#include <linux/devpts_fs.h>+struct pty_mcrmsr {
+
+ struct semaphore sem; /* locks this structure */
+
+ /* for tiocmget and tiocmset functions */
+
+ int msr; /* MSR shadow */
+ int mcr; /* MCR shadow */
+};
+
/* These are global because they are accessed in tty_io.c */
#ifdef CONFIG_UNIX98_PTYS
struct tty_driver *ptm_driver;
@@ -199,6 +211,7 @@static int pty_open(struct tty_struct *tty, struct file * filp)
{
+ ...
This weekend I upgraded my kernel to 2.6.22.5 and tried my patch.
Unfortunately pty.c seems to be a slightly different, so generated a new
patch:Signed-off-by: Sander van Ginkel <sander@van-ginkel.eu>
===================================================================
--- linux-2.6.22.5.orig/drivers/char/pty.c 2007-08-23 01:23:54.000000000 +0200
+++ linux-2.6.22.5.new/drivers/char/pty.c 2008-03-23 15:25:24.000000000 +0100
@@ -7,6 +7,8 @@
* -- C. Scott Ananian <cananian@alumni.princeton.edu>, 14-Jan-1998
* Added TTY_DO_WRITE_WAKEUP to enable n_tty to send POLL_OUT to
* waiting writers -- Sapan Bhatia <sapan@corewars.org>
+ * Added support for MCR/MSR, used for serial over ethernet
+ * -- Sander van Ginkel <sander@van-ginkel.eu>
*
*
*/
@@ -29,6 +31,16 @@
#include <linux/bitops.h>
#include <linux/devpts_fs.h>+struct pty_mcrmsr {
+
+ struct semaphore sem; /* locks this structure */
+
+ /* for tiocmget and tiocmset functions */
+
+ int msr; /* MSR shadow */
+ int mcr; /* MCR shadow */
+};
+
/* These are global because they are accessed in tty_io.c */
#ifdef CONFIG_UNIX98_PTYS
struct tty_driver *ptm_driver;
@@ -196,6 +208,7 @@static int pty_open(struct tty_struct *tty, struct file * filp)
{
+ struct pty_mcrmsr *mcrmsr;
int retval = -ENODEV;if (!tty || !tty->link)
@@ -213,10 +226,90 @@
set_bit(TTY_THROTTLED, &tty->flags);
set_bit(TTY_DO_WRITE_WAKEUP, &tty->flags);
retval = 0;
+
+ /* initialize the pointer in case something fails */
+
+ tty->driver_data = NULL;
+ tty->link->driver_data = NULL;
+
+ /* first time accessing this device, let's create it */
+
+ mcrmsr = kmalloc(sizeof(*mcrmsr), GFP_KERNEL);
+
+ init_MUTEX(&mcrmsr->sem);
+
+ if (mcrmsr != NULL) {
+ down(&mcrmsr->sem);
+
+ /* save our structure within the tty structure...
On Sun, 30 Mar 2008 14:09:48 +0200
Really you want to be working against 2.6.25-rc5-mm1 or later as there
are big changes in the tty layer pending, some of which affect theYou create a lock but don't use it. Also btw the -mm tty layer has a
That may cause wakeups, open, hangup etc to occur - does that need to be
Use a proper value as I said before.
--
I've created a new patch, build against linux-2.6.25-rc8.
Signed-off-by: Sander van Ginkel <sander@van-ginkel.eu>
===================================================================
diff -pru linux-2.6.25-rc8.orig/drivers/char/pty.c
linux-2.6.25-rc8/drivers/char/pty.c
--- linux-2.6.25-rc8.orig/drivers/char/pty.c 2008-04-01
21:44:26.000000000 +0200
+++ linux-2.6.25-rc8/drivers/char/pty.c 2008-04-13 16:40:03.000000000
+0200
@@ -7,6 +7,8 @@
* -- C. Scott Ananian <cananian@alumni.princeton.edu>, 14-Jan-1998
* Added TTY_DO_WRITE_WAKEUP to enable n_tty to send POLL_OUT to
* waiting writers -- Sapan Bhatia <sapan@corewars.org>
+ * Added support for MCR/MSR, used for serial over ethernet
+ * -- Sander van Ginkel <sander@van-ginkel.eu>
*
*
*/
@@ -196,6 +198,7 @@ static void pty_flush_buffer(struct tty_static int pty_open(struct tty_struct *tty, struct file * filp)
{
+ unsigned int *mcrmsr;
int retval = -ENODEV;if (!tty || !tty->link)
@@ -212,11 +215,56 @@ static int pty_open(struct tty_struct *t
clear_bit(TTY_OTHER_CLOSED, &tty->link->flags);
set_bit(TTY_THROTTLED, &tty->flags);
set_bit(TTY_DO_WRITE_WAKEUP, &tty->flags);
+
+ /* initialize the pointer in case something fails */
+
+ tty->driver_data = NULL;
+ tty->link->driver_data = NULL;
+
+ /* first time accessing this device, let's create it */
+
+ mcrmsr = kmalloc(sizeof(*mcrmsr), GFP_KERNEL);
+
+ if (mcrmsr != NULL) {
+
+ /* save our data within the tty structure */
+
+ *mcrmsr=0;
+
+ tty->driver_data = mcrmsr;
+ tty->link->driver_data = mcrmsr;
+ }
+ else
+ {
+ goto out;
+ }
+
retval = 0;
+
out:
return retval;
}+static int pty_tiocmget(struct tty_struct *tty, struct file *file)
+{
+ unsigned int *mcrmsr;
+
+ mcrmsr = tty->driver_data;...
We've been trying to get rid of these long lists in the code and put them
Not needed
Ok - possibly we shouldn't allow people to set undefined bits but I'm not
Why ??
I am curious how this is handled by other Unix systems and if there is an
ioctl we can follow from other systems ?Looks basically ok, coding style is wrong, some odd extra assignments but
I agree entirely with the idea of adding this functionality to keep
remote serial drivers in user space.I'll try and find out if other Unixes have similar features we can use to
keep API consistency tidy it up and fold it at some point in the next
couple of weeks.Alan
--
Corrected the last issues
Signed-off-by: Sander van Ginkel <sander@van-ginkel.eu>
===================================================================
diff -pru linux-2.6.25-rc8.orig/drivers/char/pty.c
linux-2.6.25-rc8/drivers/char/pty.c
--- linux-2.6.25-rc8.orig/drivers/char/pty.c 2008-04-01
21:44:26.000000000 +0200
+++ linux-2.6.25-rc8/drivers/char/pty.c 2008-04-14 23:01:56.000000000
+0200
@@ -196,6 +196,7 @@ static void pty_flush_buffer(struct tty_static int pty_open(struct tty_struct *tty, struct file * filp)
{
+ unsigned int *mcrmsr;
int retval = -ENODEV;if (!tty || !tty->link)
@@ -212,11 +213,50 @@ static int pty_open(struct tty_struct *t
clear_bit(TTY_OTHER_CLOSED, &tty->link->flags);
set_bit(TTY_THROTTLED, &tty->flags);
set_bit(TTY_DO_WRITE_WAKEUP, &tty->flags);
+
+ /* first time accessing this device, let's create it */
+
+ mcrmsr = kzalloc(sizeof(*mcrmsr), GFP_KERNEL);
+
+ if (mcrmsr != NULL) {
+
+ /* save our data within the tty structure */
+
+ *mcrmsr=0;
+
+ tty->driver_data = mcrmsr;
+ tty->link->driver_data = mcrmsr;
+ }
+ else
+ {
+ goto out;
+ }
+
retval = 0;
+
out:
return retval;
}+static int pty_tiocmget(struct tty_struct *tty, struct file *file)
+{
+ unsigned int *mcrmsr;
+
+ mcrmsr = tty->driver_data;
+
+ return *mcrmsr;
+}
+
+static int pty_tiocmset(struct tty_struct *tty, struct file
*file,unsigned int set, unsigned int clear)
+{
+ unsigned int *mcrmsr;
+
+ mcrmsr = tty->driver_data;
+ *mcrmsr=set;
+
+ return 0;
+}
+
static void pty_set_termios(struct tty_struct *tty, struct ktermios
*old_termios)
{
tty->termios->c_cflag &= ~(CSIZE | PARENB);
@@ -232,6 +272,8 @@ static const struct tty_operations pty_o
.chars_in_buffer = pty_chars_in_buffer,
.unthrottle = pty_unthrottle,
.set_termios...
On Mon, 14 Apr 2008 22:39:21 +0200
NAK
Having played with this (and fixed up further issues) it simply doesn't
work in the real world. pty/tty pairs are always 8bit, don't support
CREAD and can't pass break signals, emulate multiple flow control types
or parity.A virtual serial device is still a good idea, but a quick hack on pty
devices doesn't actually work - the problem is far more complicated.Alan
--
Not AFAICT. There's no full changelog/description here...
---
~Randy
--
Ok, I will test this kernel if its stable enough for my needs, and
later this week I will move on to the latest kernel.
Further I shall look at your comments and make some corrections to my code.I may pick the last value + 1 defined in the lib?
thanks for your input
--
| Tarkan Erimer | Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 |
| Greg KH | [GIT PATCH] driver core patches against 2.6.24 |
| holzheu | Re: [RFC/PATCH] Documentation of kernel messages |
| FUJITA Tomonori | Re: Integration of SCST in the mainstream Linux kernel |
git: | |
| Jarek Poplawski | [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Gerrit Renker | [PATCH 13/37] dccp: Deprecate Ack Ratio sysctl |
| Arjan van de Ven | Re: [GIT]: Networking |
| Evgeniy Polyakov | Re: [BUG] New Kernel Bugs |
