Return immediately from generic process_read_urb if urb is empty.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Remove multi-urb write from the generic driver and simplify the
prepare_write_buffer prototype:

	int (*prepare_write_buffer)(struct usb_serial_port *port,
						void *dest, size_t size);

The default implementation simply fills dest with data from port write
fifo but drivers can override it if they need to process the outgoing
data (e.g. add headers).

Turn ftdi_sio into a generic fifo-based driver, which lowers CPU usage
significantly for small writes while retaining maximum throughput.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Reimplement fifo-based writes in the generic driver using a multiple
pre-allocated urb scheme.

In contrast to multi-urb writes, no allocations (of urbs or buffers) are
made during run-time and there is less pressure on the host stack
queues as currently only two urbs are used (implementation is generic
and can handle more than two urbs as well, though).

Initial tests using ftdi_sio show that the implementation achieves the
same (maximum) throughput at high baudrates as multi-urb writes. The CPU
usage is much lower than for multi-urb writes for small write requests
and only slightly higher for large (e.g. 2k) requests (due to extra copy
via fifo?).

Also outperforms multi-urb writes for small write requests on an
embedded arm-9 system, where multi-urb writes are CPU-bound at high
baudrates (perf reveals that a lot of time is spent in the host stack
enqueue function -- could perhaps be a bug as well).

Keeping the original write_urb, buffer and flag for now as there are
other drivers depending on them.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Generalise write buffer preparation.

This allows for drivers to manipulate (e.g. add headers) to bulk out
data before it is sent.

This adds a new function pointer to usb_serial_driver:

int (*prepare_write_buffer)(struct usb_serial_port *port,
		void **dest, size_t size, const void *src, size_t count);

The function is generic and can be used with either kfifo-based or
multi-urb writes:

If *dest is NULL the implementation should allocate dest.
If src is NULL the implementation should use the port write fifo.

If not set, a generic implementation is used which simply uses memcpy or
kfifo_out.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Use dynamic transfer buffer sizes since it is more efficient to let the
host controller do the partitioning to fit endpoint size. This way we
also do not use more than one urb per write request.

Replace max_in_flight_urbs with multi_urb_write flag in struct
usb_serial_driver to enable multi-urb writes.

Use MAX_TX_URBS=40 and a max buffer size of PAGE_SIZE to prevent DoS
attacks.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Allow drivers to implement their own multi-urb write bulk callbacks as
we do for single urb writes.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Submit write urb if it is not already in use and we have buffered data.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Clean up error messages on usb_submit_urb failure.
Remove debug message on zero-length writes.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Allow drivers to use the generic throttle and unthrottle implementation.
This makes sense for drivers using the generic read functionality.

Note that drivers need to set these explicitly in order to enable them
(i.e., we do not set them at port probe if not defined).
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Add process_read_urb to usb_serial_driver so that a driver can rely on
the generic read (and throttle) mechanism but still do device specific
processing of incoming data (such as adding tty_flags before pushing to
line discipline).

The default generic implementation handles sysrq for consoles but
otherwise simply pushes to tty.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Always process and flush read urb, but only resubmit when not throttled.
The new tty-layer supply plenty of slack so there is really no need to
cache and delay processing of a single urb while throttled.

Note that unthrottle now submits using GFP_KERNEL as we are not in
atomic context (so there is no need to save irq state either).

Note also that the process_read_urb function could be added to
usb_serial_driver should any driver need to do any device specific
processing.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The generic read and write bulk urbs are initialised when allocated in
usb_serial_probe. The only field that needs to be updated after that is
the transfer_buffer_length of the write urb.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Use the already exported function for submitting the read urb associated
with a usb_serial_port.

Make sure it returns the result of usb_submit_urb and rename to the
more descriptive usb_serial_generic_submit_read_urb.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Export usb_serial_generic_close so that drivers can easily kill the read
and write urb and make sure that the write fifo is reset.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Make sure fifo is emptied on close.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

On errors the fifo was reset without any locking. This could race with
write which do kfifo_put and perhaps also chars_in_buffer and write_room.

Every other access to the fifo is protected using the port lock so
better add it to the error path as well.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Make sure chars_in_buffer accounts also for data in host stack queues.

This fixes the problem with tty_wait_until_sent returning too soon at
close which could cause the final write urb to be cancelled.
Reported-by: NPete Zaitcev <zaitcev@redhat.com>
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Fix build error when CONFIG_MAGIC_SYSRQ is not enabled:

drivers/usb/serial/generic.c:566: error: implicit declaration of function 'handle_sysrq'
Signed-off-by: NRandy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Replace all instances of using the console variable in struct
usb_serial_port with the struct tty_port version.

CC: Alan Cox <alan@linux.intel.com>
CC: Alan Stern <stern@rowland.harvard.edu>
CC: Oliver Neukum <oliver@neukum.org>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: linux-usb@vger.kernel.org
CC: linux-kernel@vger.kernel.org
Signed-off-by: NJason Wessel <jason.wessel@windriver.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

It is possible to have a multi-port device with a port lacking an in or
out bulk endpoint. Only checking for num_bulk_in or num_bulk_out is thus not
sufficient to determine whether a specific port has an in or out bulk
endpoint.

This fixes potential null pointer dereferences in the generic open and
write routines, as well as access to uninitialised fifo in write_room
and chars_in_buffer.

Also let write fail with ENODEV (instead of 0) on missing out endpoint
(also on zero-length writes).
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Make sure usb_serial_port_softint is called on errors also when using
multi urb writes.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Resubmitting read urb fails with -EPERM if completion handler runs while
urb is being killed on close. This should not be reported as an error.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch (as1344) removes references to port->port.count from the
USB serial drivers.  Now that serial ports are properly reference
counted, port.count checking is unnecessary and incorrect.  Drivers
should assume that the port is in use from the time the open method
runs until the close method is called.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

#define dbg adds the newline, messages shouldn't.
Converted dbg("%s", "some string") to dbg("some string")
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The id_table field of the struct usb_device_id is constant in <linux/usb.h>
so it is worth to make the initialization data also constant.

The semantic match that finds this kind of pattern is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@r@
disable decl_init,const_decl_init;
identifier I1, I2, x;
@@
	struct I1 {
	  ...
	  const struct I2 *x;
	  ...
	};
@s@
identifier r.I1, y;
identifier r.x, E;
@@
	struct I1 y = {
	  .x = E,
	};
@c@
identifier r.I2;
identifier s.E;
@@
	const struct I2 E[] = ... ;
@depends on !c@
identifier r.I2;
identifier s.E;
@@
+	const
	struct I2 E[] = ...;
// </smpl>
Signed-off-by: NNémeth Márton <nm127@freemail.hu>
Cc: Julia Lawall <julia@diku.dk>
Cc: cocci@diku.dk
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Fix a regression introduced by commit
715b1dc0 ("USB: usb_debug,
usb_generic_serial: implement multi urb write").

URB transfer buffer was never freed when using multi-urb writes. 
Currently the only driver enabling multi-urb writes is usb_debug.
Signed-off-by: NJohan Hovold <jhovold@gmail.com>
Cc: Greg KH <greg@kroah.com>
Acked-by: NJason Wessel <jason.wessel@windriver.com>
Cc: stable <stable@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch fix a possible race bug in drivers/usb/serial/generic with
the new kfifo API.

Please apply it to the 2.6.33-rc* tree.
Signed-off-by: NStefani Seibold <stefani@seibold.net>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The USB serial code was a new user of the kfifo API, and it was missed
when porting things to the new kfifo API.

Please make the write_fifo in place.  Here is my patch to fix the
regression and full ported version.
Signed-off-by: NStefani Seibold <stefani@seibold.net>
Reported-and-tested-by: NRafael J. Wysocki <rjw@sisk.pl>
Cc: Greg KH <greg@kroah.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

rename kfifo_put...  into kfifo_in...  to prevent miss use of old non in
kernel-tree drivers

ditto for kfifo_get...  -> kfifo_out...

Improve the prototypes of kfifo_in and kfifo_out to make the kerneldoc
annotations more readable.

Add mini "howto porting to the new API" in kfifo.h
Signed-off-by: NStefani Seibold <stefani@seibold.net>
Acked-by: NGreg Kroah-Hartman <gregkh@suse.de>
Acked-by: NMauro Carvalho Chehab <mchehab@redhat.com>
Acked-by: NAndi Kleen <ak@linux.intel.com>
Acked-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

change name of __kfifo_* functions to kfifo_*, because the prefix __kfifo
should be reserved for internal functions only.
Signed-off-by: NStefani Seibold <stefani@seibold.net>
Acked-by: NGreg Kroah-Hartman <gregkh@suse.de>
Acked-by: NMauro Carvalho Chehab <mchehab@redhat.com>
Acked-by: NAndi Kleen <ak@linux.intel.com>
Acked-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Move the pointer to the spinlock out of struct kfifo.  Most users in
tree do not actually use a spinlock, so the few exceptions now have to
call kfifo_{get,put}_locked, which takes an extra argument to a
spinlock.
Signed-off-by: NStefani Seibold <stefani@seibold.net>
Acked-by: NGreg Kroah-Hartman <gregkh@suse.de>
Acked-by: NMauro Carvalho Chehab <mchehab@redhat.com>
Acked-by: NAndi Kleen <ak@linux.intel.com>
Acked-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The generic usbserial driver in Linux 2.6.31 halts its receiving
channel in response to throttle requests from the line discipline.
Unfortunately it drops the contents of the first URB received after
throttling takes effect. This patch corrects that problem.
Signed-off-by: NJoris van Rantwijk <jorispubl@xs4all.nl>
Acked-by: NJohan Hovold <jhovold@gmail.com>
Cc: stable <stable@kernel.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

When do_output_char() attempts to write a carriage return/line feed sequence,
it first checks to see how much buffer room is available. If there are at least
two characters free, it will write the carriage return/line feed with two calls
to tty_put_char(). It calls the tty_operation functions write() for devices that
don't support the tty_operations function put_char(). If the USB generic serial
device's write URB is not in use, it will return the buffer size when asked how
much room is available. The write() of the carriage return will cause it to mark
the write URB busy, so the subsequent write() of the line feed will be ignored.

This patch uses the kfifo infrastructure to implement a write FIFO that
accurately returns the amount of space available in the buffer.
Signed-off-by: NDavid VomLehn <dvomlehn@cisco.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

And indeed none of them use it. Clean this up as it will make moving to a
standard open method rather easier.
Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The sysrq code acquired a kref leak. Fix it by passing the tty separately
from the caller (thus effectively using the callers kref which all the
callers hold anyway)
Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

We can't go around calling all sorts of magic per character functions at
full rate 3G data speed.
Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patch (as1254) splits up the shutdown method of usb_serial_driver
into a disconnect and a release method.

The problem is that the usb-serial core was calling shutdown during
disconnect handling, but drivers didn't expect it to be called until
after all the open file references had been closed.  The result was an
oops when the close method tried to use memory that had been
deallocated by shutdown.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The only time a sysrq should get processed is if the attached device
is a console.  This is intended to protect sysrq execution on a host
connected with a terminal program.

Here is the problem scenario:

host A <-- rs232 link --> host B

Host A is using mincom and a usb pl2303 device to connect to host b
which is a linux system with a usb pl2303 device acting as the serial
console.  When host B is rebooted the pl2303 emits random junk
characters on reset.  These character sequences contain serial break
signals most of the time and when translated to a sysrq have caused
host A to get random processes killed, reboots or power down.

It is true that in this setup with this patch host B might still have
the same problem as host A if you reboot host A.  In most cases host A
is a development host which seldom gets rebooted, and you could turn
off sysrq temporarily on host B if you need to reboot host A.
Signed-off-by: NJason Wessel <jason.wessel@windriver.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The usb_debug driver was modified to implement serial break handling
by using a "magic" data packet comprised of the sequence:

       0x00 0xff 0x01 0xfe   0x00 0xfe 0x01 0xff

When the tty layer requests a serial break the usb_debug driver sends
the magic packet.  On the receiving side the magic packet is thrown
away or a sysrq is activated depending on what kernel .config options
have been set.

The generic serial driver was modified as well as the usb serial
headers to generically implement sysrq processing in the same way the
non usb uart based drivers implement the sysrq handling.  This will
allow other usb serial devices to implement sysrq handling as desired.

The new usb serial functions are named similarly and implemented
similarly to the uart functions as follows:

usb_serial_handle_break <-> uart_handle_break
usb_serial_handle_sysrq_char <-> uart_handle_sysrq_char
Signed-off-by: NJason Wessel <jason.wessel@windriver.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The usb_debug driver, when used as the console, will always fail to
insert the carriage return and new line sequence as well as randomly
drop console output.  This is a result of only having the single
write_urb and that the tty layer will have a lock that prevents the
processing of the back to back urb requests.

The solution is to allow more than one urb to be outstanding and have
a slightly deeper transmit queue.  The idea and some code is borrowed
from the ftdi_sio usb driver.

The generic usb serial driver was modified so as to allow the classic
method of 1 write urb, or a multi write urb scheme with N allowed
outstanding urbs where N is controlled by max_in_flight_urbs.  When
max_in_flight_urbs in a "struct usb_serial_driver" is non zero the
multi write urb scheme will be used.

The size of 4000 was selected for the usb_debug driver so that the
driver lowers possibility of losing the queued console messages during
the kernel startup.
Signed-off-by: NJason Wessel <jason.wessel@windriver.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>