Make sure we delay for a minimum desired length of time.

The semaphore was initialized and up'd for each command, but nobody
was listening.  Replace this with a completion, which may or may not
be present.

This is the much-discussed "speed up SATA" patch.  It limits requests
to 1MB as discussed, rather than the hardware maximum (32MB).

As soon as Jens Axboe's patch to dynamically determining request
size is merged, max_sectors becomes what it properly should be --
a description of the absolute hardware maximum.

Make sure to initialize PIO data xfer state.

Use queue_delayed_work() rather than manually calling schedule_timeout(),
then queue_work(), ourselves.

Remove a lot of redundant code in ATAPI packet submission.

ATAPI is still disabled, it doesn't work yet.

Integrate it into its caller.

Now that libata-scsi module no longer calls queue_work() directly,
we can localize the use of ata_wq.

This allows us to kill the katad thread itself, and several
thread-related variables in struct ata_port.

libata creates one thread per ata_port structure.  This is inadequate
for our needs, and also cumbersome to maintain, now that workqueues
and Rusty's thread work is available.

This patch begins to move libata away from doing its own per-port
thread, by moving the ATAPI command initiation code to work under
the workqueue system.

This patch also creates a private workqueue, global to all of libata.

When writing taskfile (an ATA command) to the controller, the exact
setup of the taskfile is dependent on the taskfile "protocol":
PIO, PIO Multiple, DMA, Non-data, etc.

So, we separate out the submission of the taskfile to hardware into
a separate function ata_qc_issue_prot(), which will later be the
home for more code.

Also, remove some dead code (#if 0'd).

into redhat.com:/spare/repo/libata-2.6

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

If a filesystem's ->writepage implementation repeatedly refuses to write the
page (it keeps on redirtying it instead) (reiserfs seems to do this) then the
writeback logic can get stuck repeately trying to write the same page.

Fix that up by correctly setting wbc->pages_skipped, to tell the writeback
logic that things aren't working out.

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

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

into nuts.davemloft.net:/disk1/BK/net-2.6

Since symlink.c uses "name" field of a kobj when it calculates the
length,  it gets a wrong value if the kobj's name  has more than 20
charathers.  A correct way to do that is to call kobject_name(kobj)
instead of using kobj->name directly.

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

this patch (2.4 & 2.6) fixes a bug about the timeout value. The formula
used to calculate jiffies from timeout is wrong.
The new formula is ok and takes care of integer computation/rounding.
There is the same bug in the tiglusb.c module which will be fixed by another
patch.

Handle some PL-2301/2302 devices better.

I just checked out the latest 2.6.6-rc1-mm1 to see that the ID patch for
ftdi has applied cleanly, but apparently someone was faster and in the
combined ID table our ID's were missing.
Most probably the patch program got confused and applied the hung at wrong
spot..

Attached is a patch that places that 2 lines in the correct spot.

Hi,

this patch (cumulative; 2.4 & 2.6) fixes another bug in the tiglusb
driver. The formula used to calculate jiffies from timeout is wrong.
The new formula is ok and takes care of integer computation/rounding.
This is the same kind of bug than in the tipar char driver.

Warnings aren't terribly important in and of themselves, but there
isn't really much the warning tells us to do here, so it would appear
that caving in to the compiler is the thing to do for now.

Various build fixes:  64bit (Andrew Morton), static linking,
broken on big-endian, etc.

Tighten up the integration with the main "ether" driver, so
state transitions and host ethernet addresses are shared too.
Add missing spinlock calls around RNDIS command outcall,
fix GET_INTERFACE issue, host mustn't clobber netdev flags.

Minor code cleanups.

Cope better when PCI misbehaves badly and registers misbehave:

    - terminate some loops before they get to infinity
       * capability scan
       * port reset
    - after init failure, memory may already be cleaned up

Some systems have been reporting such problems after ACPI resume.

This replaces current_thread_info()->cpu in i386' init_IRQ() by the
equivalent smp_processor_id().

Reduces overhead on UP, and makes the code cleaner.

This simplifies the Pentium M quirk code in nmi.c, and eliminates an
unnecessary apic_read().

Local APIC accesses are not zero-cycle; let's not inflict more damage
than we must.

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

We need some alignement of those structs for proper operations especially
with FP and Altivec, or SLAB_DEBUG can break us.

From: Stephen Rothwell <sfr@canb.auug.org.au>

This patch just adds some proc entries for the virtual tape and cdrom
drivers to allow mapping between linux devices and OS/400 ones.  This is
expected by existing users and there is no other way to do this
translation.

From: "J. Bruce Fields" <bfields@fieldses.org>

Unregister svcauth_gss caches on exit from gss module; fixes an oops on
rmmod.

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>

From: David Woodhouse <dwmw2@infradead.org>

Without this the task struct gets unaligned when using SLAB_DEBUG, causing
random problems with FP and Altivec.

From: Ralf Baechle <ralf@linux-mips.org>

MIPS now also uses the generic ioctl compat code.

From: Ralf Baechle <ralf@linux-mips.org>

MIPS never uses a.out

From: Ralf Baechle <ralf@linux-mips.org>

- Add HPC3 PS/2 driver bits for SGI IP22 aka Indy
- Add Mace PS/2 driver bits for SGI IP32 aka O2
- Add R4030 PS/2 driver bits for Jazz family
- Don't register I/O ports where we're using the I/O port memory window
   to access the i8042 registers