This patch fixes an oops discovered in test which can occur
on bad hardware if the ipr adapter times out coming operational. 

This patch adds additional error logging to abort, device reset,
and bus reset paths to help in diagnosing scsi problems on ipr.

This patch fixes an issue where ipr was including a kernel
data structure, list_head, in a packed structure, which causes
compile issues on some architectures, and is just a bad thing to do.

From: 	Alan Cox <alan@redhat.com>

Pretty minimal. queue_command is now called locked, this requires propogating
some small locking changes for send_s870

and here's a variant using #if 0

Here comes the next (trivial) patch. I just want to get it out of the way
to make the next (scary) patch smaller and simpler. This one doesn't
modify the object code either. Actually, well, it does - it makes a few
objects, that were previously declared extern static. And removes a couple
more defines.

From: garloff@suse.de

Feature

Make the timeout for INQUIRY during SCSI scan adjustable via boot parameter.
Note that the second INQUIRY does use a shorter timeout, as the long timeout
is for recovery from the initial reset, not because existing devices would
take so long to answer INQUIRY. SPC3 says that INQUIRY should be available
right away, but real life is different unfortunately.

From: garloff@suse.de

Cleanup/Feature

Remove CONFIG_SCSI_REPORT_LUNS config option.
Instead provide BLIST_NOREPORTLUN that can be passed as default_dev_flags
(but also per device if needed).
Provide BLIST_REPORTLUN2 that allows trying to use REPORT_LUNS for SCSI-2
devices, if they are connected to a host adapter supporting more than 8 LUNs
(and thus avoiding the usual USB crap to render this feature useless when
used with default_dev_flags).

 drivers/scsi/Kconfig        |   11 -----------
 drivers/scsi/scsi_scan.c    |   19 +++++++++----------
 include/scsi/scsi_devinfo.h |    3 +++
 3 files changed, 12 insertions(+), 21 deletions(-)

From: garloff@suse.de

Cleanup

Mark BLIST_FORCELUN as deprecated, as we don't want to collect a list
of perfectly working multi-LUN devices with BLIST_FORCELUN. Instead
document max_luns boot/module parameter.

From: garloff@suse.de

Optimization.

Tell the compiler that the SCSI LOG will not likely happen.

 From:  a.othieno@bluewin.ch (Arthur Othieno)

This one is purely cosmetic - it removes legacy (pre 2.4) and some void
(e.g. #defined as nothing locks) code. So, it doesn't affect the resulting
object file.

From: Christoph Hellwig <hch@lst.de>

When looking at the driver after reviewing the changes I stubled over
that code.  Almost all of the funcitonality (except DUMP) should go
through the midlayer and will at least cause some confusion if it
doesn't.

Also the implementation isn't exactly nice, like kmalloced scsi commands
and the tytpical 'parsing' macros cut & pasted over all historic scsi
drivers.  I think we're better of removing it completly and maybe readd
dump as a sysfs attribute if needed.

500 lines of code gone.

drivers/scsi/qlogicfas.c:190: warning: initialization from incompatible pointer type

Here comes the 2nd one. I wanted to get Christoph's patches now in, but
already the first his patch comes on the top of this one: remove internal
command queuing in the driver. And, in fact, it fixes some bugs in it. So,
they should go in together. Here's the original comment from Christoph's
email of 1 Feb 2004:

<start quote>

Patch looks mostly good for me.  But there's some fishyness in queuecommand,
mostly from before you patch:

 - many failure cases return one with the new EH code although we wouldn't
   want to requeue the midlayer in that case.  I removed the ifdef and added
   a failed goto to handle them.
 - we need to set cmd->result onlyh if we have an error instead of always
   an overriding it - else it will leak to the midlayer in the return 1
   case.
 - the check for ids out of range are superflous (this was one of the errors
   above, I decided to remove it instead of fixing it).
 - you don't do DC390_UNLOCK_ACB when failing.  While it's a noop I think
   it's bad to have locking macros in place and don't balance them.  You
   should probably remove it completly in one of the next patches.
 - the pDCB->pWaitingSRB looks a bit strange.  By unifying the the codepathes
   it becomes much more readable.

<end quote>

 - needs linux/delay.h for udelay
 - C doesn't allow taking addresses of register variables and gcc even
   complains these days

Ok, here comes the first one. I chose this one because it fixes an actual
bug in the driver. This bug was (partially) introduced by myself when
porting to 2.6. Partly the reason was that I disliked using
function-like macros as lvalues:

sg_dma_address(x) = ...
sg_dma_len(x) = ...

[OT] wouldn't it be better to introduce some macros like
set_sg_dma_{address|len}(x, y)?

A part of the original patch has already been merged (s/UINT/ULONG/), so,
the actual version is re-diffed against 2.6.6-bk6 and re-tested (on a
plain Pentium).

Also updates the driver-version, printed on startup.

From: 	Heiko Carstens <Heiko.Carstens@de.ibm.com>

This happened when sending out the inquiry command and the LLDD was in
recovery for the specific LUN and returned SCSI_MLQUEUE_DEVICE_BUSY.
Problem is that max_device_blocked gets set _after_ the inquiry command
finished. In this specific case max_device_blocked was 0 and thus the
device was never blocked resulting in the observed stack overflow.
I moved the initializazion of max_device_blocked from scsi_add_lun to
sdev_alloc_sdev and the problem is gone.

From: 	Matthew Wilcox <willy@debian.org>,
	Douglas Gilbert <dougg@torque.net>

  - Fix vmalloc,vfree with intrs disabled
  - bump version number
  - introduce MODULE_VERSION
  - increase over allocation of sg_dev_arr from 6 to 32

From: Go Taniguchi <go@turbolinux.co.jp>

Make it build (and work) again.

From: 	Leubner, Achim <Achim_Leubner@adaptec.com>

- Support for drives > 2 TB implemented
- 64-bit DMA support depending on the controller firmware version implemented
- Some important bug fixes made

A little more testing of st's SG_IO ioctl turned up a
small problem.

This is the corresponding patch that was applied to the
sd driver when it received the block layer SG_IO ioctl.

For least surprise of lk 2.4 utilities that use the
SCSI_IOCTL_GET_IDLUN and SCSI_IOCTL_GET_BUS_NUMBER
ioctls (e.g. sg_map) it is better to return the correct
values rather than 0.

Following a prior patch, we made cancel the sole
route into deleted, but forgot to update the 
state model checks to reflect this.

offline and created also go to cancel

From: 	Brian King <brking@us.ibm.com>

This is a fairly minimal fix for aacraid. It removes the happy cast
pointers to u32 garbage in the 2.6 code and replaces it with the working
2.4 equivalents. I've not backported any of the other changes from 2.4 to
2.6 yet, and some things have gone which are good to be gone (eg the
proc/scsi horror). There is a certain amount of white space noise caused
by realigning with the 2.4 code so I could see what was going on.

Tested on a dual opteron

In MAINTAINERS and in individual low-level drivers.

* bug fix: make sure 'nsect' member of struct ata_queued_cmd is
initialized each time a cmd is re-used.  Only affects PIO data xfers,
which nobody uses.
* slightly change the way a device's flags are printed out.  currently
the only flag is 'lba48', but soon 'wcache' will appear also.
* add WB-cache-related constants and macros to linux/ata.h

This fixes a corruption problem with overlapping copyarea()'s
in the radeon driver.

At the moment, on PPC64, the instruction we use for wmb() doesn't
order cacheable stores vs. non-cacheable stores.  (It does order
cacheable vs. cacheable and non-cacheable vs. non-cacheable.)  This
causes problems in the sort of driver code that writes stuff into
memory, does a wmb(), then a writel to the device to start a DMA
operation to read the stuff it has just written to memory.

This patch solves the problem by adding a sync instruction before the
store in the write* and out* macros.  The sync is a full barrier that
orders all loads and stores, cacheable or not.  The patch also moves
the eieio instruction that we had after the store to before the load
in the read* and in* macros.  With the sync before the store, we don't
need an eieio as well in a sequence of stores, but we still need an
eieio between a store and a load.

I think it is better to do this than to turn wmb() into a full memory
barrier (a sync instruction) because the full barrier is slow and
isn't needed with the sync in the write*/out* macros.  This way,
write*/out* are fully ordered with respect to preceding loads and
stores, which is what driver writers expect, and we avoid penalizing
users of wmb() who are only doing cacheable stores.

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

 - Split PA7300LC from PA7100LC (Matthew Wilcox)
 - Handle 32-bit firmware and 64-bit kernel at runtime (Ryan Bradetich)
 - Fix building in a separate tree (Matthew Wilcox)
 - Update defconfigs (Randolph Chung)
 - Make WCHAN work (Randolph Chung)
 - Initial support for SMP in 2.6 (Grant Grundler)
 - Use 8-byte PTEs on 32-bit kernels (James Bottomley)
 - Implement L2/L3 hybrid page tables for 64 bit kernels (James Bottomley)
 - Support 8TB of physical and virtual address space (James Bottomley)
 - Macro'ise the tlb miss handlers (James Bottomley)
 - Check the ptrace flags correctly in the syscall return path (Randolph Chung)
 - Eliminate many magic numbers (James Bottomley)
 - Work around linker bug in vmlinux.lds.S (James Bottomley)
 - Many cache flushing fixes (James Bottomley)
 - first baby step for PA8800 support (Grant Grundler)
 - Self-aligning spinlocks (Randolph Chung)

M68k: <asm/virtconvert.h> needs include <linux/compiler.h> for
__attribute_const__ (from Richard Zidlicky)

Sun3x: Like most other platforms, Sun3x needs conswitchp set if
CONFIG_DUMMY_CONSOLE is defined (from Sam Creasey)

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

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

into redhat.com:/spare/repo/net-drivers-2.6

into redhat.com:/spare/repo/net-drivers-2.6

	ntfs_fill_super() and ntfs_read_inode_mount() cleaned up.  Removed
the kludges around the first iget() on NTFS.  Instead of playing with
(re)setting ->s_op we have the MFT_FILE inode set up by explicit new_inode()/
set ->i_ino/insert_inode_hash()/call ntfs_read_inode_mount() directly.
That kills the need of second super_operations and it allows to return
error from ntfs_read_inode_mount() without resorting to ugly "poisoning"
tricks.

into redhat.com:/spare/repo/net-drivers-2.6

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