Use msleep()/msleep_interruptible() [as appropriate]
instead of schedule_timeout() to guarantee the task delays
as expected.

Signed-off-by: Nishanth Aravamudan
Signed-off-by: Maximilian Attems
Signed-off-by: Russell King <rmk@arm.linux.org.uk>

Thomas Koeller points out that we're always returning IRQ_HANDLED
from our interrupt handler, even if we find no interrupt work to
be done.

We must unregister all serial ports before driver_unregister()
can complete.  This means that we must unregister all ports in
serial8250_remove, including our legacy ISA ports.  We flag this
special cleanup operation by setting serial8250_isa_devs to NULL
and not handling our own platform device any differently from
any others.

Use !iobase && !membase rather than !ops for console port
availability.

The curse of the missing __devexit_p() returns, and asm-*/ obviously
marks the end of the comment.

into flint.arm.linux.org.uk:/usr/src/bk/linux-2.6-serial

This patch conflicts with work to properly integrate the device model
into the serial layer, and provide architectures with a *clean* way
to tell 8250.c about their serial ports at run time.

This change allows platform devices named "serial8250" to provide
lists of serial ports to the 8250 driver at runtime, in addition to
the hard coded table in include/asm-*/serial.h.

The next step is to deprecate the tables in serial.h.

Add a platform device for ISA 8250-compatible serial devices listed
in the table in include/asm-*/serial.h.  Arrange for unregistered
serial devices to be owned by this device.

This enables power management for ISA 8250 devices, and starts to
opens the way for architectures to dynamically provide their own
lists of 8250 devices via platform device(s).

Only register the 8250 serial driver with the device model after
registering and setting up our internal uart ports.  Do the reverse
on module finalisation.

This moves the basic initialisation of 8250 ports from
serial8250_register_ports() into serial8250_isa_init_ports()

serial8250_request_std_resource() is now responsible for claiming
the standard resources, _and_ calling ioremap if necessary.
serial8250_release_std_resource() performs the complementary function
in its entirety.
serial8250_*_rsa_resource() perform the similar operations for RSA
ports, with the exception that RSA ports can only be mapped into IO
space.

Since the tty layer now takes care of user space writes,
__uart_user_write() and associated temporary buffer and temporary
buffer semaphore have all become unnecessary.  There's also little
point in having __uart_kern_write() separate from uart_write(), so
combine the two together.

Adrian Bunk kindly provided the patch to remove __uart_user_write().
The rest of the work is rmk's.

Signed-off-by: Adrian Bunk
Signed-off-by: Russell King <rmk@arm.linux.org.uk>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

Since bit 0 is only set when the MM is mapped onto the CPU, we can use
this rather than comparing the MM pointer with current->active_mm.
This simplifies the inline cache flushing and tlb code.

Christoph suggests letting the compiler choose.  No real compelling reason
to inline anyhow.  I had some vmlinux size numbers suggesting inline was
better, but re-running them on newer kernel is giving different results,
favoring uninline.  Best let compiler choose.  Un-inline __sigqueue_alloc.
Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

into ppc970.osdl.org:/home/torvalds/v2.6/linux

From: Jens Axboe <axboe@suse.de>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

destroy_proc_ide_interfaces is only used inside ide-proc.c;
so lets make it static.
Signed-off-by: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

cs5530 overrides hwif->autodma anyway.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Do not set hwif->autodma in ide_pci_setup_ports().
All DMA capable PCI host drivers override it.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

* hwif->dma_base is set in ->init_iops() for MMIO
* if !dma_base then dma_base is really equal to zero
* remove dead code from simplex check
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

DMA is always used by default (->autodma is never checked).
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

Original patch from Mark Lord <lkml@rtr.ca>.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

We need to k[un]map_atomic() the current page not the first page
of the scatterlist segment.  Fixes OOPS when using HIGHMEM.

Big thanks to Randy Dunlap for a lot of debugging/testing.
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

This avoids a bug where the compiler would overwrite the
stackframe that the caller also considered to be a register
save area.

It also shrinks the code segment by a tiny amount by moving
the failure case argument setup into the slow path. This not
only makes the fast path smaller, but it makes it easier on
gcc (gcc is not very good at generating code that uses fixed
register names).

This file is no longer reqired - the MAKEDEV program makes the ida/ nodes
automatically.
Signed-off-by: James Nelson <james4765@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>