This patch is based on the DMA-patch by Chip Coldwell for the AT91/AT32 serial
USARTS, with some tweaks to make it apply neatly on top of the other patches
in this series.

The RX and TX code has been moved to a tasklet and reworked a bit.  Instead of
depending on the ENDRX and TIMEOUT bits in CSR, we simply grab as much data as
we can from the DMA buffers.  I think this closes a race where the ENDRX bit
is set after we read CSR but before we read RPR, although I haven't confirmed
this.

Similarly, the two TX handlers (ENDTX and TXBUFE) have been combined into one.
 Since the current code only uses a single TX buffer, there's no point in
handling those interrupts separately.

This also fixes a DMA sync bug in the original patch.

[linux@bohmer.net: rebased onto irq-splitup patch]
[hskinnemoen@atmel.com: moved to tasklet, fixed dma bug, misc cleanups]
[hskinnemoen@atmel.com: atmel_serial dma: Misc fixes and cleanups]
Signed-off-by: NRemy Bohmer <linux@bohmer.net>
Signed-off-by: NHaavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Andrew Victor <linux@maxim.org.za>
Tested-by: NMarc Pignat <marc.pignat@hevs.ch>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Split up the interrupt handler of the serial port into a interrupt top-half
and a tasklet.

The goal is to get the interrupt top-half as short as possible to minimize
latencies on interrupts.  But the old code also does some calls in the
interrupt handler that are not allowed on preempt-RT in IRQF_NODELAY context.
This handler is executed in this context because of the interrupt sharing with
the timer interrupt.  The timer interrupt on Preempt-RT runs in IRQF_NODELAY
context.

The tasklet takes care of handling control status changes, pushing incoming
characters to the tty layer, handling break and other errors.  It also handles
pushing TX data into the data register.

Reading the complete receive queue is still done in the top-half because we
never want to miss any incoming character.

[hskinnemoen@atmel.com: misc cleanups and simplifications]
Signed-off-by: NRemy Bohmer <linux@bohmer.net>
Signed-off-by: NHaavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Andrew Victor <linux@maxim.org.za>
Tested-by: NMarc Pignat <marc.pignat@hevs.ch>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When an error happens in probe(), the clocks should be disabled, but
only if the port isn't already used as a console.

In remove(), the port struct shouldn't be freed because it's defined
statically.
Signed-off-by: NHaavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Andrew Victor <linux@maxim.org.za>
Tested-by: NMarc Pignat <marc.pignat@hevs.ch>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

If BRGR is zero, the baud rate generator isn't running, so the boot loader
can't have initialized the port.
Signed-off-by: NHaavard Skinnemoen <hskinnemoen@atmel.com>
Acked-by: NAndrew Victor <linux@maxim.org.za>
Tested-by: NMarc Pignat <marc.pignat@hevs.ch>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Replace two instances of barrier() with cpu_relax() since that's the right
thing to do when busy-waiting.  This does not actually change anything since
cpu_relax() is defined as barrier() on both ARM and AVR32.
Signed-off-by: NHaavard Skinnemoen <hskinnemoen@atmel.com>
Acked-by: NAndrew Victor <linux@maxim.org.za>
Tested-by: NMarc Pignat <marc.pignat@hevs.ch>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Clean up the atmel_serial driver to conform the coding rules.  It contains no
functional change.
Signed-off-by: NRemy Bohmer <linux@bohmer.net>
Signed-off-by: NHaavard Skinnemoen <hskinnemoen@atmel.com>
Acked-by: NAndrew Victor <linux@maxim.org.za>
Tested-by: NMarc Pignat <marc.pignat@hevs.ch>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NJiri Slaby <jirislaby@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This is a set of changes to implement proper resource management in the
driver, including iomem space reservation and operating on physical
addresses ioremap()ped appropriately using accessory functions rather than
unportable direct assignments.

Some adjustments to code are made to reflect the architecture of the
interface, which is a centrally controlled multiport (or, as referred to
from DEC documentation, a serial line multiplexer, going up to 8 lines
originally) rather than a bundle of separate ports.

Types are changed, where applicable, to specify the width of hardware
registers explicitly.  The interrupt handler is now managed in the
->startup() and ->shutdown() calls for consistency with other drivers and
also in preparation to handle the handover from the initial firmware-based
console gracefully.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Replace all casts from "struct uart_port *" to "struct dz_port *" with a
construct based on container_of().  This makes the conversion work
irrespective of where the former struct is located within the latter.

By popular request I have implemented it as an inline function rather than
a macro this time.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

A set of changes to the way termios settings are propagated to the serial
port hardware.  The DZ11 only supports a selection of fixed baud settings,
so some requests may not be fulfilled.  Keep the old setting in such a case
and failing that resort to 9600bps.  Also add a missing update of the
transmit timeout.  And remove the explicit encoding of the line selected
from writes to the Line Parameters Register as it has been preencoded by
the ->set_termios() call already.  Finally, remove a duplicate macro for
the Receiver Enable bit.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Handle the read and ignore status masks correctly.  Handle the BREAK condition
as expected: a framing error with a null character is a BREAK, any other
framing error is a framing error indeed.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The ->start_tx(), ->stop_tx() and ->stop_rx() backends are called with the
port's lock already taken.  Remove locking from within them and wrap around
calls as necessary.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Rename the serial console structure so that `modpost' does not complain about
a reference to an "init" section -- "_console" is magic.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Reformat the Kconfig entries and update descriptions for accuracy.  Select the
driver by default for configurations of interest.  For the curious: 32BIT
means only 32-bit DECstations support the device, not that the driver is not
64-bit clean; I have not checked that either though.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Sort the header inclusions, add a few that are needed but pulled indirectly
only and remove ones that are not really used.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Well, panic() is a little bit undue if request_irq() fails; there is probably
no need to justify it any further.  Handle the case gracefully, by
unregistering the driver.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Polled transmission is tricky enough with the DZ11 design.  While "loop" is
set to a high value, conceptually you are not allowed to transmit without
checking whether the device offers the right transmission line (yes, it is the
device that selects the line -- the driver has no control over it other than
disabling the transmitter offered if it is the wrong one), so the loop has to
be run at least once.

Well, the '1977 or PDP11 view of how serial lines should be handled...  Except
that the serial interface used to be quite an impressive board back then
rather than chip.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Remove unused module function prototypes that would not even build if enabled.
Signed-off-by: NMaciej W. Rozycki <macro@linux-mips.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When rebasing one of the mpc5200 psc UART patches I made a mistake and
damaged the patch.

This patch fixes the compile failure introduced in commit
25ae3a07Signed-off-by: NGrant Likely <grant.likely@secretlab.ca>

Add 512x support using the psc_ops framework established
with the previous patch.

All 512x PSCs share the same interrupt so add
IRQF_SHARED to irq flags.
Signed-off-by: NJohn Rigby <jrigby@freescale.com>
Signed-off-by: NGrant Likely <grant.likely@secretlab.ca>

PSC devices are different between the mpc5200 and the mpc5121
this patch localizes the differences in preparation for adding mpc5121
support to the psc uart driver.
Signed-off-by: NJohn Rigby <jrigby@freescale.com>
Signed-off-by: NGrant Likely <grant.likely@secretlab.ca>

New serial driver for SC2681/SC2691 uarts.  Older SNI RM400 machines are
using these chips for onboard serial ports.
Signed-off-by: NThomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Torben Mathiasen <device@lanana.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

- Use SGI_HAS_ZILOG for IP22_ZILOG depends
- remove IP22 from description, because the driver works on more than
  IP22 SGI machines
Signed-off-by: NThomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Mostly in and around irq handlers.
Signed-off-by: NJeff Garzik <jgarzik@redhat.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: "Luck Tony" <tony.luck@intel.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: NDmitry Torokhov <dtor@mail.ru>
Cc: Karsten Keil <kkeil@suse.de>
Acked-by: N"John W. Linville" <linville@tuxdriver.com>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Acked-by: NJosh Boyer <jwboyer@linux.vnet.ibm.com>
Acked-by: NHolger Schurig <hs4233@mail.mn-solutions.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Probing of the console at console_initcall time is broken.  It tries to
call memory allocation routines which aren't initialized yet.

Problem solved by removing the early probe entirely.  The console init
is called again anyway after the uartlite device is initialized and the
memory allocation routines can be called safely.
Signed-off-by: NGrant Likely <grant.likely@secretlab.ca>
Acked-by: NPeter Korsgaard <jacmet@sunsite.dk>

Mainly, this involves two changes:
1) xilinx->xlnx (recognized standard is to use the stock ticker)
2) In order to have the device tree focus on describing what the
hardware is as exactly as possible, the compatible strings contain the
full IP name and IP version.
Signed-off-by: NStephen Neuendorffer <stephen.neuendorffer@xilinx.com>
Acked-by: NPeter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: NGrant Likely <grant.likely@secretlab.ca>

Fix compile errors in the xilinxfb, xsysace and uartlite drivers used
by the Xilinx Virtex platform
Signed-off-by: NGrant Likely <grant.likely@secretlab.ca>
Acked-by: NPeter Korsgaard <jacmet@sunsite.dk>

Use ARRAY_SIZE macroto get maximum ports in ColdFire serial driver.
Signed-off-by: NGreg Ungerer <gerg@uclinux.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The clock to generate the desired baudrate with the MPSC is first divided
by the Baud Rate Generator (BRG) and then by the MPSC itself.  So, when the
BRG divider is changed, the MPSC divider must also be changed to generate
the correct baudrate.  During MPSC initialization, the BRG divider is
changed but the MPSC divider isn't changed until much later.  This results
in some printk's coming out garbled.  To fix that, set the MPSC divider at
the same time that the BRG divider is changed.
Signed-off-by: NMark A. Greer <mgreer@mvista.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Coding style tweaks and printk levels.
Signed-off-by: NAlan Cox <alan@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Invalid speeds are forced to 9600. Update the code for this to encode new
style baud rates properly.
Signed-off-by: NAlan Cox <alan@redhat.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Some ports seem to be unable to drain their transmitters on shut down.  Such a
problem can occur if the port is programmed for hardware imposed flow control,
characters are in the FIFO but the CTS signal is inactive.

Normally, this isn't a problem because most places where we wait for the
transmitter to drain have a time-out.  However, there is no timeout in the
suspend path.

Give a port 30ms to drain; this is an arbitary value chosen to avoid long
delays if there are many such ports in the system, while giving a reasonable
chance for a single port to drain.  Should a port not drain within this
timeout, issue a warning.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>
Acked-by: NAlan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>