linux/drivers/usb/gadget/u_serial.c

1323 lines
33 KiB
C
Raw Normal View History

usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/*
* u_serial.c - utilities for USB gadget "serial port"/TTY support
*
* Copyright (C) 2003 Al Borchers (alborchers@steinerpoint.com)
* Copyright (C) 2008 David Brownell
* Copyright (C) 2008 by Nokia Corporation
*
* This code also borrows from usbserial.c, which is
* Copyright (C) 1999 - 2002 Greg Kroah-Hartman (greg@kroah.com)
* Copyright (C) 2000 Peter Berger (pberger@brimson.com)
* Copyright (C) 2000 Al Borchers (alborchers@steinerpoint.com)
*
* This software is distributed under the terms of the GNU General
* Public License ("GPL") as published by the Free Software Foundation,
* either version 2 of that License or (at your option) any later version.
*/
/* #define VERBOSE_DEBUG */
#include <linux/kernel.h>
USB: gadget: compilation issue: missing TASK_INTERRUPTIBLE Here is the patch for the following issue: drivers/usb/gadget/u_serial.c: In function ‘gs_start_tx’: drivers/usb/gadget/u_serial.c:369: error: ‘TASK_INTERRUPTIBLE’ undeclared (first use in this function) drivers/usb/gadget/u_serial.c:369: error: (Each undeclared identifier is reported only once drivers/usb/gadget/u_serial.c:369: error: for each function it appears in.) drivers/usb/gadget/u_serial.c: In function ‘gs_rx_push’: drivers/usb/gadget/u_serial.c:546: error: ‘TASK_INTERRUPTIBLE’ undeclared (first use in this function) drivers/usb/gadget/u_serial.c: In function ‘gs_close’: drivers/usb/gadget/u_serial.c:857: error: ‘TASK_INTERRUPTIBLE’ undeclared (first use in this function) drivers/usb/gadget/u_serial.c:857: error: implicit declaration of function ‘signal_pending’ drivers/usb/gadget/u_serial.c:857: error: implicit declaration of function ‘schedule_timeout’ drivers/usb/gadget/u_serial.c: In function ‘gserial_cleanup’: drivers/usb/gadget/u_serial.c:1190: error: ‘TASK_UNINTERRUPTIBLE’ undeclared (first use in this function) drivers/usb/gadget/u_serial.c:1190: error: implicit declaration of function ‘schedule’ drivers/usb/gadget/u_serial.c: In function ‘gserial_disconnect’: drivers/usb/gadget/u_serial.c:1311: error: ‘TASK_INTERRUPTIBLE’ undeclared (first use in this function) Signed-off-by: Stephane Duverger <stephane.duverger@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2010-06-29 14:57:25 +00:00
#include <linux/sched.h>
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
#include <linux/interrupt.h>
#include <linux/device.h>
#include <linux/delay.h>
#include <linux/tty.h>
#include <linux/tty_flip.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
#include <linux/slab.h>
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
#include "u_serial.h"
/*
* This component encapsulates the TTY layer glue needed to provide basic
* "serial port" functionality through the USB gadget stack. Each such
* port is exposed through a /dev/ttyGS* node.
*
* After initialization (gserial_setup), these TTY port devices stay
* available until they are removed (gserial_cleanup). Each one may be
* connected to a USB function (gserial_connect), or disconnected (with
* gserial_disconnect) when the USB host issues a config change event.
* Data can only flow when the port is connected to the host.
*
* A given TTY port can be made available in multiple configurations.
* For example, each one might expose a ttyGS0 node which provides a
* login application. In one case that might use CDC ACM interface 0,
* while another configuration might use interface 3 for that. The
* work to handle that (including descriptor management) is not part
* of this component.
*
* Configurations may expose more than one TTY port. For example, if
* ttyGS0 provides login service, then ttyGS1 might provide dialer access
* for a telephone or fax link. And ttyGS2 might be something that just
* needs a simple byte stream interface for some messaging protocol that
* is managed in userspace ... OBEX, PTP, and MTP have been mentioned.
*/
#define PREFIX "ttyGS"
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/*
* gserial is the lifecycle interface, used by USB functions
* gs_port is the I/O nexus, used by the tty driver
* tty_struct links to the tty/filesystem framework
*
* gserial <---> gs_port ... links will be null when the USB link is
* inactive; managed by gserial_{connect,disconnect}(). each gserial
* instance can wrap its own USB control protocol.
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
* gserial->ioport == usb_ep->driver_data ... gs_port
* gs_port->port_usb ... gserial
*
* gs_port <---> tty_struct ... links will be null when the TTY file
* isn't opened; managed by gs_open()/gs_close()
* gserial->port_tty ... tty_struct
* tty_struct->driver_data ... gserial
*/
/* RX and TX queues can buffer QUEUE_SIZE packets before they hit the
* next layer of buffering. For TX that's a circular buffer; for RX
* consider it a NOP. A third layer is provided by the TTY code.
*/
#define QUEUE_SIZE 16
#define WRITE_BUF_SIZE 8192 /* TX only */
/* circular buffer */
struct gs_buf {
unsigned buf_size;
char *buf_buf;
char *buf_get;
char *buf_put;
};
/*
* The port structure holds info for each port, one for each minor number
* (and thus for each /dev/ node).
*/
struct gs_port {
spinlock_t port_lock; /* guard port_* access */
struct gserial *port_usb;
struct tty_struct *port_tty;
unsigned open_count;
bool openclose; /* open/close in progress */
u8 port_num;
wait_queue_head_t close_wait; /* wait for last close */
struct list_head read_pool;
struct list_head read_queue;
unsigned n_read;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
struct tasklet_struct push;
struct list_head write_pool;
struct gs_buf port_write_buf;
wait_queue_head_t drain_wait; /* wait while writes drain */
/* REVISIT this state ... */
struct usb_cdc_line_coding port_line_coding; /* 8-N-1 etc */
};
/* increase N_PORTS if you need more */
#define N_PORTS 4
static struct portmaster {
struct mutex lock; /* protect open/close */
struct gs_port *port;
} ports[N_PORTS];
static unsigned n_ports;
#define GS_CLOSE_TIMEOUT 15 /* seconds */
#ifdef VERBOSE_DEBUG
#define pr_vdebug(fmt, arg...) \
pr_debug(fmt, ##arg)
#else
#define pr_vdebug(fmt, arg...) \
({ if (0) pr_debug(fmt, ##arg); })
#endif
/*-------------------------------------------------------------------------*/
/* Circular Buffer */
/*
* gs_buf_alloc
*
* Allocate a circular buffer and all associated memory.
*/
static int gs_buf_alloc(struct gs_buf *gb, unsigned size)
{
gb->buf_buf = kmalloc(size, GFP_KERNEL);
if (gb->buf_buf == NULL)
return -ENOMEM;
gb->buf_size = size;
gb->buf_put = gb->buf_buf;
gb->buf_get = gb->buf_buf;
return 0;
}
/*
* gs_buf_free
*
* Free the buffer and all associated memory.
*/
static void gs_buf_free(struct gs_buf *gb)
{
kfree(gb->buf_buf);
gb->buf_buf = NULL;
}
/*
* gs_buf_clear
*
* Clear out all data in the circular buffer.
*/
static void gs_buf_clear(struct gs_buf *gb)
{
gb->buf_get = gb->buf_put;
/* equivalent to a get of all data available */
}
/*
* gs_buf_data_avail
*
* Return the number of bytes of data written into the circular
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
* buffer.
*/
static unsigned gs_buf_data_avail(struct gs_buf *gb)
{
return (gb->buf_size + gb->buf_put - gb->buf_get) % gb->buf_size;
}
/*
* gs_buf_space_avail
*
* Return the number of bytes of space available in the circular
* buffer.
*/
static unsigned gs_buf_space_avail(struct gs_buf *gb)
{
return (gb->buf_size + gb->buf_get - gb->buf_put - 1) % gb->buf_size;
}
/*
* gs_buf_put
*
* Copy data data from a user buffer and put it into the circular buffer.
* Restrict to the amount of space available.
*
* Return the number of bytes copied.
*/
static unsigned
gs_buf_put(struct gs_buf *gb, const char *buf, unsigned count)
{
unsigned len;
len = gs_buf_space_avail(gb);
if (count > len)
count = len;
if (count == 0)
return 0;
len = gb->buf_buf + gb->buf_size - gb->buf_put;
if (count > len) {
memcpy(gb->buf_put, buf, len);
memcpy(gb->buf_buf, buf+len, count - len);
gb->buf_put = gb->buf_buf + count - len;
} else {
memcpy(gb->buf_put, buf, count);
if (count < len)
gb->buf_put += count;
else /* count == len */
gb->buf_put = gb->buf_buf;
}
return count;
}
/*
* gs_buf_get
*
* Get data from the circular buffer and copy to the given buffer.
* Restrict to the amount of data available.
*
* Return the number of bytes copied.
*/
static unsigned
gs_buf_get(struct gs_buf *gb, char *buf, unsigned count)
{
unsigned len;
len = gs_buf_data_avail(gb);
if (count > len)
count = len;
if (count == 0)
return 0;
len = gb->buf_buf + gb->buf_size - gb->buf_get;
if (count > len) {
memcpy(buf, gb->buf_get, len);
memcpy(buf+len, gb->buf_buf, count - len);
gb->buf_get = gb->buf_buf + count - len;
} else {
memcpy(buf, gb->buf_get, count);
if (count < len)
gb->buf_get += count;
else /* count == len */
gb->buf_get = gb->buf_buf;
}
return count;
}
/*-------------------------------------------------------------------------*/
/* I/O glue between TTY (upper) and USB function (lower) driver layers */
/*
* gs_alloc_req
*
* Allocate a usb_request and its buffer. Returns a pointer to the
* usb_request or NULL if there is an error.
*/
struct usb_request *
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
gs_alloc_req(struct usb_ep *ep, unsigned len, gfp_t kmalloc_flags)
{
struct usb_request *req;
req = usb_ep_alloc_request(ep, kmalloc_flags);
if (req != NULL) {
req->length = len;
req->buf = kmalloc(len, kmalloc_flags);
if (req->buf == NULL) {
usb_ep_free_request(ep, req);
return NULL;
}
}
return req;
}
/*
* gs_free_req
*
* Free a usb_request and its buffer.
*/
void gs_free_req(struct usb_ep *ep, struct usb_request *req)
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
{
kfree(req->buf);
usb_ep_free_request(ep, req);
}
/*
* gs_send_packet
*
* If there is data to send, a packet is built in the given
* buffer and the size is returned. If there is no data to
* send, 0 is returned.
*
* Called with port_lock held.
*/
static unsigned
gs_send_packet(struct gs_port *port, char *packet, unsigned size)
{
unsigned len;
len = gs_buf_data_avail(&port->port_write_buf);
if (len < size)
size = len;
if (size != 0)
size = gs_buf_get(&port->port_write_buf, packet, size);
return size;
}
/*
* gs_start_tx
*
* This function finds available write requests, calls
* gs_send_packet to fill these packets with data, and
* continues until either there are no more write requests
* available or no more data to send. This function is
* run whenever data arrives or write requests are available.
*
* Context: caller owns port_lock; port_usb is non-null.
*/
static int gs_start_tx(struct gs_port *port)
/*
__releases(&port->port_lock)
__acquires(&port->port_lock)
*/
{
struct list_head *pool = &port->write_pool;
struct usb_ep *in = port->port_usb->in;
int status = 0;
bool do_tty_wake = false;
while (!list_empty(pool)) {
struct usb_request *req;
int len;
req = list_entry(pool->next, struct usb_request, list);
len = gs_send_packet(port, req->buf, in->maxpacket);
if (len == 0) {
wake_up_interruptible(&port->drain_wait);
break;
}
do_tty_wake = true;
req->length = len;
list_del(&req->list);
req->zero = (gs_buf_data_avail(&port->port_write_buf) == 0);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
pr_vdebug(PREFIX "%d: tx len=%d, 0x%02x 0x%02x 0x%02x ...\n",
port->port_num, len, *((u8 *)req->buf),
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
*((u8 *)req->buf+1), *((u8 *)req->buf+2));
/* Drop lock while we call out of driver; completions
* could be issued while we do so. Disconnection may
* happen too; maybe immediately before we queue this!
*
* NOTE that we may keep sending data for a while after
* the TTY closed (dev->ioport->port_tty is NULL).
*/
spin_unlock(&port->port_lock);
status = usb_ep_queue(in, req, GFP_ATOMIC);
spin_lock(&port->port_lock);
if (status) {
pr_debug("%s: %s %s err %d\n",
__func__, "queue", in->name, status);
list_add(&req->list, pool);
break;
}
/* abort immediately after disconnect */
if (!port->port_usb)
break;
}
if (do_tty_wake && port->port_tty)
tty_wakeup(port->port_tty);
return status;
}
/*
* Context: caller owns port_lock, and port_usb is set
*/
static unsigned gs_start_rx(struct gs_port *port)
/*
__releases(&port->port_lock)
__acquires(&port->port_lock)
*/
{
struct list_head *pool = &port->read_pool;
struct usb_ep *out = port->port_usb->out;
unsigned started = 0;
while (!list_empty(pool)) {
struct usb_request *req;
int status;
struct tty_struct *tty;
/* no more rx if closed */
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
tty = port->port_tty;
if (!tty)
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
break;
req = list_entry(pool->next, struct usb_request, list);
list_del(&req->list);
req->length = out->maxpacket;
/* drop lock while we call out; the controller driver
* may need to call us back (e.g. for disconnect)
*/
spin_unlock(&port->port_lock);
status = usb_ep_queue(out, req, GFP_ATOMIC);
spin_lock(&port->port_lock);
if (status) {
pr_debug("%s: %s %s err %d\n",
__func__, "queue", out->name, status);
list_add(&req->list, pool);
break;
}
started++;
/* abort immediately after disconnect */
if (!port->port_usb)
break;
}
return started;
}
/*
* RX tasklet takes data out of the RX queue and hands it up to the TTY
* layer until it refuses to take any more data (or is throttled back).
* Then it issues reads for any further data.
*
* If the RX queue becomes full enough that no usb_request is queued,
* the OUT endpoint may begin NAKing as soon as its FIFO fills up.
* So QUEUE_SIZE packets plus however many the FIFO holds (usually two)
* can be buffered before the TTY layer's buffers (currently 64 KB).
*/
static void gs_rx_push(unsigned long _port)
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
{
struct gs_port *port = (void *)_port;
struct tty_struct *tty;
struct list_head *queue = &port->read_queue;
bool disconnect = false;
bool do_push = false;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/* hand any queued data to the tty */
spin_lock_irq(&port->port_lock);
tty = port->port_tty;
while (!list_empty(queue)) {
struct usb_request *req;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
req = list_first_entry(queue, struct usb_request, list);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/* discard data if tty was closed */
if (!tty)
goto recycle;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/* leave data queued if tty was rx throttled */
if (test_bit(TTY_THROTTLED, &tty->flags))
break;
switch (req->status) {
case -ESHUTDOWN:
disconnect = true;
pr_vdebug(PREFIX "%d: shutdown\n", port->port_num);
break;
default:
/* presumably a transient fault */
pr_warning(PREFIX "%d: unexpected RX status %d\n",
port->port_num, req->status);
/* FALLTHROUGH */
case 0:
/* normal completion */
break;
}
/* push data to (open) tty */
if (req->actual) {
char *packet = req->buf;
unsigned size = req->actual;
unsigned n;
int count;
/* we may have pushed part of this packet already... */
n = port->n_read;
if (n) {
packet += n;
size -= n;
}
count = tty_insert_flip_string(tty, packet, size);
if (count)
do_push = true;
if (count != size) {
/* stop pushing; TTY layer can't handle more */
port->n_read += count;
pr_vdebug(PREFIX "%d: rx block %d/%d\n",
port->port_num,
count, req->actual);
break;
}
port->n_read = 0;
}
recycle:
list_move(&req->list, &port->read_pool);
}
/* Push from tty to ldisc; without low_latency set this is handled by
* a workqueue, so we won't get callbacks and can hold port_lock
*/
if (tty && do_push) {
tty_flip_buffer_push(tty);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
}
/* We want our data queue to become empty ASAP, keeping data
* in the tty and ldisc (not here). If we couldn't push any
* this time around, there may be trouble unless there's an
* implicit tty_unthrottle() call on its way...
*
* REVISIT we should probably add a timer to keep the tasklet
* from starving ... but it's not clear that case ever happens.
*/
if (!list_empty(queue) && tty) {
if (!test_bit(TTY_THROTTLED, &tty->flags)) {
if (do_push)
tasklet_schedule(&port->push);
else
pr_warning(PREFIX "%d: RX not scheduled?\n",
port->port_num);
}
}
/* If we're still connected, refill the USB RX queue. */
if (!disconnect && port->port_usb)
gs_start_rx(port);
spin_unlock_irq(&port->port_lock);
}
static void gs_read_complete(struct usb_ep *ep, struct usb_request *req)
{
struct gs_port *port = ep->driver_data;
/* Queue all received data until the tty layer is ready for it. */
spin_lock(&port->port_lock);
list_add_tail(&req->list, &port->read_queue);
tasklet_schedule(&port->push);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
spin_unlock(&port->port_lock);
}
static void gs_write_complete(struct usb_ep *ep, struct usb_request *req)
{
struct gs_port *port = ep->driver_data;
spin_lock(&port->port_lock);
list_add(&req->list, &port->write_pool);
switch (req->status) {
default:
/* presumably a transient fault */
pr_warning("%s: unexpected %s status %d\n",
__func__, ep->name, req->status);
/* FALL THROUGH */
case 0:
/* normal completion */
gs_start_tx(port);
break;
case -ESHUTDOWN:
/* disconnect */
pr_vdebug("%s: %s shutdown\n", __func__, ep->name);
break;
}
spin_unlock(&port->port_lock);
}
static void gs_free_requests(struct usb_ep *ep, struct list_head *head)
{
struct usb_request *req;
while (!list_empty(head)) {
req = list_entry(head->next, struct usb_request, list);
list_del(&req->list);
gs_free_req(ep, req);
}
}
static int gs_alloc_requests(struct usb_ep *ep, struct list_head *head,
void (*fn)(struct usb_ep *, struct usb_request *))
{
int i;
struct usb_request *req;
/* Pre-allocate up to QUEUE_SIZE transfers, but if we can't
* do quite that many this time, don't fail ... we just won't
* be as speedy as we might otherwise be.
*/
for (i = 0; i < QUEUE_SIZE; i++) {
req = gs_alloc_req(ep, ep->maxpacket, GFP_ATOMIC);
if (!req)
return list_empty(head) ? -ENOMEM : 0;
req->complete = fn;
list_add_tail(&req->list, head);
}
return 0;
}
/**
* gs_start_io - start USB I/O streams
* @dev: encapsulates endpoints to use
* Context: holding port_lock; port_tty and port_usb are non-null
*
* We only start I/O when something is connected to both sides of
* this port. If nothing is listening on the host side, we may
* be pointlessly filling up our TX buffers and FIFO.
*/
static int gs_start_io(struct gs_port *port)
{
struct list_head *head = &port->read_pool;
struct usb_ep *ep = port->port_usb->out;
int status;
unsigned started;
/* Allocate RX and TX I/O buffers. We can't easily do this much
* earlier (with GFP_KERNEL) because the requests are coupled to
* endpoints, as are the packet sizes we'll be using. Different
* configurations may use different endpoints with a given port;
* and high speed vs full speed changes packet sizes too.
*/
status = gs_alloc_requests(ep, head, gs_read_complete);
if (status)
return status;
status = gs_alloc_requests(port->port_usb->in, &port->write_pool,
gs_write_complete);
if (status) {
gs_free_requests(ep, head);
return status;
}
/* queue read requests */
port->n_read = 0;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
started = gs_start_rx(port);
/* unblock any pending writes into our circular buffer */
if (started) {
tty_wakeup(port->port_tty);
} else {
gs_free_requests(ep, head);
gs_free_requests(port->port_usb->in, &port->write_pool);
status = -EIO;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
}
return status;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
}
/*-------------------------------------------------------------------------*/
/* TTY Driver */
/*
* gs_open sets up the link between a gs_port and its associated TTY.
* That link is broken *only* by TTY close(), and all driver methods
* know that.
*/
static int gs_open(struct tty_struct *tty, struct file *file)
{
int port_num = tty->index;
struct gs_port *port;
int status;
if (port_num < 0 || port_num >= n_ports)
return -ENXIO;
do {
mutex_lock(&ports[port_num].lock);
port = ports[port_num].port;
if (!port)
status = -ENODEV;
else {
spin_lock_irq(&port->port_lock);
/* already open? Great. */
if (port->open_count) {
status = 0;
port->open_count++;
/* currently opening/closing? wait ... */
} else if (port->openclose) {
status = -EBUSY;
/* ... else we do the work */
} else {
status = -EAGAIN;
port->openclose = true;
}
spin_unlock_irq(&port->port_lock);
}
mutex_unlock(&ports[port_num].lock);
switch (status) {
default:
/* fully handled */
return status;
case -EAGAIN:
/* must do the work */
break;
case -EBUSY:
/* wait for EAGAIN task to finish */
msleep(1);
/* REVISIT could have a waitchannel here, if
* concurrent open performance is important
*/
break;
}
} while (status != -EAGAIN);
/* Do the "real open" */
spin_lock_irq(&port->port_lock);
/* allocate circular buffer on first open */
if (port->port_write_buf.buf_buf == NULL) {
spin_unlock_irq(&port->port_lock);
status = gs_buf_alloc(&port->port_write_buf, WRITE_BUF_SIZE);
spin_lock_irq(&port->port_lock);
if (status) {
pr_debug("gs_open: ttyGS%d (%p,%p) no buffer\n",
port->port_num, tty, file);
port->openclose = false;
goto exit_unlock_port;
}
}
/* REVISIT if REMOVED (ports[].port NULL), abort the open
* to let rmmod work faster (but this way isn't wrong).
*/
/* REVISIT maybe wait for "carrier detect" */
tty->driver_data = port;
port->port_tty = tty;
port->open_count = 1;
port->openclose = false;
/* if connected, start the I/O stream */
if (port->port_usb) {
struct gserial *gser = port->port_usb;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
pr_debug("gs_open: start ttyGS%d\n", port->port_num);
gs_start_io(port);
if (gser->connect)
gser->connect(gser);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
}
pr_debug("gs_open: ttyGS%d (%p,%p)\n", port->port_num, tty, file);
status = 0;
exit_unlock_port:
spin_unlock_irq(&port->port_lock);
return status;
}
static int gs_writes_finished(struct gs_port *p)
{
int cond;
/* return true on disconnect or empty buffer */
spin_lock_irq(&p->port_lock);
cond = (p->port_usb == NULL) || !gs_buf_data_avail(&p->port_write_buf);
spin_unlock_irq(&p->port_lock);
return cond;
}
static void gs_close(struct tty_struct *tty, struct file *file)
{
struct gs_port *port = tty->driver_data;
struct gserial *gser;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
spin_lock_irq(&port->port_lock);
if (port->open_count != 1) {
if (port->open_count == 0)
WARN_ON(1);
else
--port->open_count;
goto exit;
}
pr_debug("gs_close: ttyGS%d (%p,%p) ...\n", port->port_num, tty, file);
/* mark port as closing but in use; we can drop port lock
* and sleep if necessary
*/
port->openclose = true;
port->open_count = 0;
gser = port->port_usb;
if (gser && gser->disconnect)
gser->disconnect(gser);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/* wait for circular write buffer to drain, disconnect, or at
* most GS_CLOSE_TIMEOUT seconds; then discard the rest
*/
if (gs_buf_data_avail(&port->port_write_buf) > 0 && gser) {
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
spin_unlock_irq(&port->port_lock);
wait_event_interruptible_timeout(port->drain_wait,
gs_writes_finished(port),
GS_CLOSE_TIMEOUT * HZ);
spin_lock_irq(&port->port_lock);
gser = port->port_usb;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
}
/* Iff we're disconnected, there can be no I/O in flight so it's
* ok to free the circular buffer; else just scrub it. And don't
* let the push tasklet fire again until we're re-opened.
*/
if (gser == NULL)
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
gs_buf_free(&port->port_write_buf);
else
gs_buf_clear(&port->port_write_buf);
tty->driver_data = NULL;
port->port_tty = NULL;
port->openclose = false;
pr_debug("gs_close: ttyGS%d (%p,%p) done!\n",
port->port_num, tty, file);
wake_up_interruptible(&port->close_wait);
exit:
spin_unlock_irq(&port->port_lock);
}
static int gs_write(struct tty_struct *tty, const unsigned char *buf, int count)
{
struct gs_port *port = tty->driver_data;
unsigned long flags;
int status;
pr_vdebug("gs_write: ttyGS%d (%p) writing %d bytes\n",
port->port_num, tty, count);
spin_lock_irqsave(&port->port_lock, flags);
if (count)
count = gs_buf_put(&port->port_write_buf, buf, count);
/* treat count == 0 as flush_chars() */
if (port->port_usb)
status = gs_start_tx(port);
spin_unlock_irqrestore(&port->port_lock, flags);
return count;
}
static int gs_put_char(struct tty_struct *tty, unsigned char ch)
{
struct gs_port *port = tty->driver_data;
unsigned long flags;
int status;
pr_vdebug("gs_put_char: (%d,%p) char=0x%x, called from %p\n",
port->port_num, tty, ch, __builtin_return_address(0));
spin_lock_irqsave(&port->port_lock, flags);
status = gs_buf_put(&port->port_write_buf, &ch, 1);
spin_unlock_irqrestore(&port->port_lock, flags);
return status;
}
static void gs_flush_chars(struct tty_struct *tty)
{
struct gs_port *port = tty->driver_data;
unsigned long flags;
pr_vdebug("gs_flush_chars: (%d,%p)\n", port->port_num, tty);
spin_lock_irqsave(&port->port_lock, flags);
if (port->port_usb)
gs_start_tx(port);
spin_unlock_irqrestore(&port->port_lock, flags);
}
static int gs_write_room(struct tty_struct *tty)
{
struct gs_port *port = tty->driver_data;
unsigned long flags;
int room = 0;
spin_lock_irqsave(&port->port_lock, flags);
if (port->port_usb)
room = gs_buf_space_avail(&port->port_write_buf);
spin_unlock_irqrestore(&port->port_lock, flags);
pr_vdebug("gs_write_room: (%d,%p) room=%d\n",
port->port_num, tty, room);
return room;
}
static int gs_chars_in_buffer(struct tty_struct *tty)
{
struct gs_port *port = tty->driver_data;
unsigned long flags;
int chars = 0;
spin_lock_irqsave(&port->port_lock, flags);
chars = gs_buf_data_avail(&port->port_write_buf);
spin_unlock_irqrestore(&port->port_lock, flags);
pr_vdebug("gs_chars_in_buffer: (%d,%p) chars=%d\n",
port->port_num, tty, chars);
return chars;
}
/* undo side effects of setting TTY_THROTTLED */
static void gs_unthrottle(struct tty_struct *tty)
{
struct gs_port *port = tty->driver_data;
unsigned long flags;
spin_lock_irqsave(&port->port_lock, flags);
if (port->port_usb) {
/* Kickstart read queue processing. We don't do xon/xoff,
* rts/cts, or other handshaking with the host, but if the
* read queue backs up enough we'll be NAKing OUT packets.
*/
tasklet_schedule(&port->push);
pr_vdebug(PREFIX "%d: unthrottle\n", port->port_num);
}
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
spin_unlock_irqrestore(&port->port_lock, flags);
}
static int gs_break_ctl(struct tty_struct *tty, int duration)
{
struct gs_port *port = tty->driver_data;
int status = 0;
struct gserial *gser;
pr_vdebug("gs_break_ctl: ttyGS%d, send break (%d) \n",
port->port_num, duration);
spin_lock_irq(&port->port_lock);
gser = port->port_usb;
if (gser && gser->send_break)
status = gser->send_break(gser, duration);
spin_unlock_irq(&port->port_lock);
return status;
}
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
static const struct tty_operations gs_tty_ops = {
.open = gs_open,
.close = gs_close,
.write = gs_write,
.put_char = gs_put_char,
.flush_chars = gs_flush_chars,
.write_room = gs_write_room,
.chars_in_buffer = gs_chars_in_buffer,
.unthrottle = gs_unthrottle,
.break_ctl = gs_break_ctl,
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
};
/*-------------------------------------------------------------------------*/
static struct tty_driver *gs_tty_driver;
static int __init
gs_port_alloc(unsigned port_num, struct usb_cdc_line_coding *coding)
{
struct gs_port *port;
port = kzalloc(sizeof(struct gs_port), GFP_KERNEL);
if (port == NULL)
return -ENOMEM;
spin_lock_init(&port->port_lock);
init_waitqueue_head(&port->close_wait);
init_waitqueue_head(&port->drain_wait);
tasklet_init(&port->push, gs_rx_push, (unsigned long) port);
INIT_LIST_HEAD(&port->read_pool);
INIT_LIST_HEAD(&port->read_queue);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
INIT_LIST_HEAD(&port->write_pool);
port->port_num = port_num;
port->port_line_coding = *coding;
ports[port_num].port = port;
return 0;
}
/**
* gserial_setup - initialize TTY driver for one or more ports
* @g: gadget to associate with these ports
* @count: how many ports to support
* Context: may sleep
*
* The TTY stack needs to know in advance how many devices it should
* plan to manage. Use this call to set up the ports you will be
* exporting through USB. Later, connect them to functions based
* on what configuration is activated by the USB host; and disconnect
* them as appropriate.
*
* An example would be a two-configuration device in which both
* configurations expose port 0, but through different functions.
* One configuration could even expose port 1 while the other
* one doesn't.
*
* Returns negative errno or zero.
*/
int __init gserial_setup(struct usb_gadget *g, unsigned count)
{
unsigned i;
struct usb_cdc_line_coding coding;
int status;
if (count == 0 || count > N_PORTS)
return -EINVAL;
gs_tty_driver = alloc_tty_driver(count);
if (!gs_tty_driver)
return -ENOMEM;
gs_tty_driver->owner = THIS_MODULE;
gs_tty_driver->driver_name = "g_serial";
gs_tty_driver->name = PREFIX;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/* uses dynamically assigned dev_t values */
gs_tty_driver->type = TTY_DRIVER_TYPE_SERIAL;
gs_tty_driver->subtype = SERIAL_TYPE_NORMAL;
gs_tty_driver->flags = TTY_DRIVER_REAL_RAW | TTY_DRIVER_DYNAMIC_DEV;
gs_tty_driver->init_termios = tty_std_termios;
/* 9600-8-N-1 ... matches defaults expected by "usbser.sys" on
* MS-Windows. Otherwise, most of these flags shouldn't affect
* anything unless we were to actually hook up to a serial line.
*/
gs_tty_driver->init_termios.c_cflag =
B9600 | CS8 | CREAD | HUPCL | CLOCAL;
gs_tty_driver->init_termios.c_ispeed = 9600;
gs_tty_driver->init_termios.c_ospeed = 9600;
coding.dwDTERate = cpu_to_le32(9600);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
coding.bCharFormat = 8;
coding.bParityType = USB_CDC_NO_PARITY;
coding.bDataBits = USB_CDC_1_STOP_BITS;
tty_set_operations(gs_tty_driver, &gs_tty_ops);
/* make devices be openable */
for (i = 0; i < count; i++) {
mutex_init(&ports[i].lock);
status = gs_port_alloc(i, &coding);
if (status) {
count = i;
goto fail;
}
}
n_ports = count;
/* export the driver ... */
status = tty_register_driver(gs_tty_driver);
if (status) {
pr_err("%s: cannot register, err %d\n",
__func__, status);
goto fail;
}
/* ... and sysfs class devices, so mdev/udev make /dev/ttyGS* */
for (i = 0; i < count; i++) {
struct device *tty_dev;
tty_dev = tty_register_device(gs_tty_driver, i, &g->dev);
if (IS_ERR(tty_dev))
pr_warning("%s: no classdev for port %d, err %ld\n",
__func__, i, PTR_ERR(tty_dev));
}
pr_debug("%s: registered %d ttyGS* device%s\n", __func__,
count, (count == 1) ? "" : "s");
return status;
fail:
while (count--)
kfree(ports[count].port);
put_tty_driver(gs_tty_driver);
gs_tty_driver = NULL;
return status;
}
static int gs_closed(struct gs_port *port)
{
int cond;
spin_lock_irq(&port->port_lock);
cond = (port->open_count == 0) && !port->openclose;
spin_unlock_irq(&port->port_lock);
return cond;
}
/**
* gserial_cleanup - remove TTY-over-USB driver and devices
* Context: may sleep
*
* This is called to free all resources allocated by @gserial_setup().
* Accordingly, it may need to wait until some open /dev/ files have
* closed.
*
* The caller must have issued @gserial_disconnect() for any ports
* that had previously been connected, so that there is never any
* I/O pending when it's called.
*/
void gserial_cleanup(void)
{
unsigned i;
struct gs_port *port;
if (!gs_tty_driver)
return;
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/* start sysfs and /dev/ttyGS* node removal */
for (i = 0; i < n_ports; i++)
tty_unregister_device(gs_tty_driver, i);
for (i = 0; i < n_ports; i++) {
/* prevent new opens */
mutex_lock(&ports[i].lock);
port = ports[i].port;
ports[i].port = NULL;
mutex_unlock(&ports[i].lock);
tasklet_kill(&port->push);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
/* wait for old opens to finish */
wait_event(port->close_wait, gs_closed(port));
WARN_ON(port->port_usb != NULL);
kfree(port);
}
n_ports = 0;
tty_unregister_driver(gs_tty_driver);
put_tty_driver(gs_tty_driver);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
gs_tty_driver = NULL;
pr_debug("%s: cleaned up ttyGS* support\n", __func__);
}
/**
* gserial_connect - notify TTY I/O glue that USB link is active
* @gser: the function, set up with endpoints and descriptors
* @port_num: which port is active
* Context: any (usually from irq)
*
* This is called activate endpoints and let the TTY layer know that
* the connection is active ... not unlike "carrier detect". It won't
* necessarily start I/O queues; unless the TTY is held open by any
* task, there would be no point. However, the endpoints will be
* activated so the USB host can perform I/O, subject to basic USB
* hardware flow control.
*
* Caller needs to have set up the endpoints and USB function in @dev
* before calling this, as well as the appropriate (speed-specific)
* endpoint descriptors, and also have set up the TTY driver by calling
* @gserial_setup().
*
* Returns negative errno or zero.
* On success, ep->driver_data will be overwritten.
*/
int gserial_connect(struct gserial *gser, u8 port_num)
{
struct gs_port *port;
unsigned long flags;
int status;
if (!gs_tty_driver || port_num >= n_ports)
return -ENXIO;
/* we "know" gserial_cleanup() hasn't been called */
port = ports[port_num].port;
/* activate the endpoints */
status = usb_ep_enable(gser->in, gser->in_desc);
if (status < 0)
return status;
gser->in->driver_data = port;
status = usb_ep_enable(gser->out, gser->out_desc);
if (status < 0)
goto fail_out;
gser->out->driver_data = port;
/* then tell the tty glue that I/O can work */
spin_lock_irqsave(&port->port_lock, flags);
gser->ioport = port;
port->port_usb = gser;
/* REVISIT unclear how best to handle this state...
* we don't really couple it with the Linux TTY.
*/
gser->port_line_coding = port->port_line_coding;
/* REVISIT if waiting on "carrier detect", signal. */
/* if it's already open, start I/O ... and notify the serial
* protocol about open/close status (connect/disconnect).
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
*/
if (port->open_count) {
pr_debug("gserial_connect: start ttyGS%d\n", port->port_num);
gs_start_io(port);
if (gser->connect)
gser->connect(gser);
} else {
if (gser->disconnect)
gser->disconnect(gser);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
}
spin_unlock_irqrestore(&port->port_lock, flags);
return status;
fail_out:
usb_ep_disable(gser->in);
gser->in->driver_data = NULL;
return status;
}
/**
* gserial_disconnect - notify TTY I/O glue that USB link is inactive
* @gser: the function, on which gserial_connect() was called
* Context: any (usually from irq)
*
* This is called to deactivate endpoints and let the TTY layer know
* that the connection went inactive ... not unlike "hangup".
*
* On return, the state is as if gserial_connect() had never been called;
* there is no active USB I/O on these endpoints.
*/
void gserial_disconnect(struct gserial *gser)
{
struct gs_port *port = gser->ioport;
unsigned long flags;
if (!port)
return;
/* tell the TTY glue not to do I/O here any more */
spin_lock_irqsave(&port->port_lock, flags);
/* REVISIT as above: how best to track this? */
port->port_line_coding = gser->port_line_coding;
port->port_usb = NULL;
gser->ioport = NULL;
if (port->open_count > 0 || port->openclose) {
wake_up_interruptible(&port->drain_wait);
if (port->port_tty)
tty_hangup(port->port_tty);
}
spin_unlock_irqrestore(&port->port_lock, flags);
/* disable endpoints, aborting down any active I/O */
usb_ep_disable(gser->out);
gser->out->driver_data = NULL;
usb_ep_disable(gser->in);
gser->in->driver_data = NULL;
/* finally, free any unused/unusable I/O buffers */
spin_lock_irqsave(&port->port_lock, flags);
if (port->open_count == 0 && !port->openclose)
gs_buf_free(&port->port_write_buf);
gs_free_requests(gser->out, &port->read_pool);
gs_free_requests(gser->out, &port->read_queue);
usb gadget: split out serial core This abstracts the "gadget serial" driver TTY glue into a separate component, cleaning it up and disentangling it from connection state. It also changed some behaviors for the better: - Stops using "experimental" major #127, and switches over to having the TTY layer allocate the dev_t numbers. - Provides /sys/class/tty/ttyGS* nodes, thus mdev/udev support. (Note "mdev" hotplug bug in Busybox v1.7.2: /dev/ttyGS0 will be a *block* device without CONFIG_SYSFS_DEPRECATED_V2.) - The tty nodes no longer reject opens when there's no host. Now they can support normal getty configs in /etc/inttab... - Now implements RX throttling. When the line discipline says it doesn't want any more data, only packets in flight will be delivered (currently, max 1K/8K at full/high speeds) until it unthrottles the data. - Supports low_latency. This is a good policy for all USB serial adapters, since it eliminates scheduler overhead on RX paths. This also includes much cleanup including better comments, fixing memory leaks and other bugs (including some locking fixes), messaging cleanup, and an interface audit and tightening. This added up to a significant object code shrinkage, on the order of 20% (!) depending on CPU and compiler. A separate patch actually kicks in this new code, using the functions declared in this new header, and removes the previous glue. Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2008-06-20 00:51:44 +00:00
gs_free_requests(gser->in, &port->write_pool);
spin_unlock_irqrestore(&port->port_lock, flags);
}