If the command is already completed, megaraid_queue_command() must
return 0.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

aac_read() and aac_write() may complete a command but return -1, which
is propagated through aac_scsi_cmd() to the mid layer through
->queuecommand. If the command has been completed, it must return 0.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The driver appropriately returns SCSI_MLQUEUE_HOST_BUSY, but it has
already completed the command. Just return 0 instead and let the mid
layer handle it through scsi_done().
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

One more case of returning non-zero with a completed command.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

In case of scb allocation failure, mega_build_cmd() can return NULL scb
but set busy, which causes a non-zero return from ->queuecommand() while
the scsi command has been completed.   Just returning busy without calling
done() should be enough for appropriate retries.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

ide-scsi completes a command but returns ownership to the mid layer.
Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

If your ->queuecommand() completes a command through done, it must not
return non-zero to the mid layer so that it takes ownership of the
command again.
Signed-off-by: Jens Axboe <axboe@suse.de>

corrected typo
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Because of the changes to add the target in to the
driver model, the cancellation method no-longer works
correctly.

Fix it by iterating using shost_for_each_device instead.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

To make spinlock/rwlock initialization consistent all over the kernel,
this patch converts explicit lock-initializers into spin_lock_init() and
rwlock_init() calls.

Currently, spinlocks and rwlocks are initialized in two different ways:

  lock = SPIN_LOCK_UNLOCKED
  spin_lock_init(&lock)

  rwlock = RW_LOCK_UNLOCKED
  rwlock_init(&rwlock)

this patch converts all explicit lock initializations to
spin_lock_init() or rwlock_init(). (Besides consistency this also helps
automatic lock validators and debugging code.)

The conversion was done with a script, it was verified manually and it
was reviewed, compiled and tested as far as possible on x86, ARM, PPC.

There is no runtime overhead or actual code change resulting out of this
patch, because spin_lock_init() and rwlock_init() are macros and are
thus equivalent to the explicit initialization method.

That's the second batch of the unifying patches.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Fix a missed piece of junk from a typedef conversion in st.h
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Call pci_disable_device if initialisation fails and during cleanup.
Signed-off-by: Jamie Lenehan <lenehan@twibble.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Store the port and irq in unsigned long and unsigned int's instead of
as u16's and u8's. The later was fine on x86 but not on sparc64.
Signed-off-by: Jamie Lenehan <lenehan@twibble.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Store the PCI device pointer into the adapter control block. This is
used in the pci_map_*/pci_unmap_* calls and previously it would have
been set to NULL. This appears to be no problem for x86 but is a
problem for sparc64.
Signed-off-by: Jamie Lenehan <lenehan@twibble.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

- make all this needlessly global code static
- remove the unused firmware_version variables
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The patch below makes a needlessly global struct static.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The patch below makes the needlessly global function adapter_init_chip
static.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The patch below makes the function is870 which has no external users
static.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The patch below makes some needlessly global code static.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Streamline EEPROM handling code

The patch below is a fixed (and slightly reworked) patch from Christoph.
Thanks to Chiaki Ishikawa for fixes and tests.

Original author: Christoph Hellwig
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

A number of USB mass storage devices return the wrong value in response to
READ CAPACITY: They give the total number of blocks rather than the
highest block number.  Currently the usb-storage driver tries to adjust
for this by altering the contents of the reply buffer.  That's not a very
good idea because it prevents users from seeing what the device actually
has to say; it violates the principle that a transport should do nothing
more than transfer data.

This patch moves the responsibility for adjusting the capacity to the sd
driver instead, by adding a new device flag.  After this has been merged
the corresponding change (to set the flag) will be made to the usb-storage
driver.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

Use scsi_track_queue_full(), don't lie about driver's queueing
abilities.
Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

This patch updates Documentation/kernel-parameters.txt with the proper
SCSI LUNs params and adds a description for 'max_luns'.
Signed-off-by: Cal Peake <cp@absolutedigital.net>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: Miles Bader <miles@gnu.org>
Signed-off-by: Miles Bader <miles@gnu.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: "Bagalkote, Sreenivas" <sreenib@lsil.com>

We are releasing megaraid 2.20.4.1 driver. This version fixes two issues -

- Handle IOCTL command timeouts properly

- Replace incorrectly introduced pci_dma_sync_{sg,single}_for_cpu with
  correct pci_dma_sync_{sg,single}_for_device.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

This patch updates the rx check health function.  My previous only
updated the rkt check health code.  This and the previous patch should
fix the kmalloc return checking bug:

http://bugme.osdl.org/show_bug.cgi?id=3699Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

On Tue, 2 Nov 2004, Marc Thomas wrote:
> I've noticed that in 2.6.10-rc1 the "mt tell" command (from mt-st) fails to
> report SCSI tape position, returning "No such device". Other tape (read)
> operations appear to be ok. It worked fine in 2.6.9-rc4.

The patch at the end of this message should fix this (passes basic tests).
Someone added the ioctl to reset the SCSI device but the error handling
is not quite correct. The code in 2.6.10-rc1 allows st_ioctl to continue
unless reset succeeded. The fix allows continuation only if
scsi_nonblockable_ioctl() does not handle the ioctl (returns -ENODEV).
retval is cleared because the following code assumes it is zero unless set
to something else.
Signed-off-by: Kai Makisara <kai.makisara@kolumbus.fi>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

This patch updates the adapter health check function and cleans up some
unused interrupt related functions and macros.
Signed-off-by: Mark Haverkamp <markh@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

On Sat, Oct 23, 2004 at 11:55:20AM +1000, Eyal Lebedinsky wrote:
> Should the second line be indented?
>
> Qlogic QLA 1240/1x80/1x160 SCSI support (SCSI_QLOGIC_1280) [M/n/?] m
> Qlogic QLA 1020/1040 SCSI support (SCSI_QLOGIC_1280_1040) [N/y/?] (NEW) y

Correct, SCSI_QLOGIC_1280_1040 should depend on SCSI_QLOGIC_1280.

The patch below does this.

Additionally, it prevents the case that both drivers are built
statically and both support the 1020/1040.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The attached minor patch to linux/drivers/scsi/pcmcia/sym53c500_cs.c
allows interrupt sharing, which is evidently a "must have" feature for
at least G4 PowerBooks (ppc architecture).  The other user of the New
Media Bus Toaster reports that his powerbook consistently assigns the
yenta CardBus controller IRQ to whatever card he inserts.
Signed-off-by: Bob Tracy <rct@frus.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

The following patch fixes a race condition in sg of sg_cmd_done racing
with sg_release. I've seen this bug hit several times on test machines
and the following patch fixes it. The race is that if srp->done is set
and the waiting thread gets a spurious wakeup immediately afterwards,
then the waiting thread can end up executing and completing, then getting
closed, freeing sfp before the wake_up_interruptible is called, which
then will result in an oops. The oops is fixed by locking around the
setting srp->done to 1 and the wake_up, and also locking around the
checking of srp->done, which guarantees that the wake_up_interruptible
will always occur before the sleeping thread gets a chance to run.
Signed-off-by: Brian King <brking@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

From: Mike Christie <mikenc@us.ibm.com>

> Oct 26 23:32:55 mina kernel: Unable to handle kernel paging request at 
> virtual address 6b6b6c7b
> Oct 26 23:32:55 mina kernel:  printing eip:
> Oct 26 23:32:55 mina kernel: f882b8ce
> Oct 26 23:32:55 mina kernel: *pde = 00000000
> Oct 26 23:32:55 mina kernel: Oops: 0000 [#1]
> Oct 26 23:32:55 mina kernel: PREEMPT
> Oct 26 23:32:55 mina kernel: Modules linked in: sd_mod usb_storage 
> ide_cd cdrom sg scsi_mod rd
> Oct 26 23:32:55 mina kernel: CPU:    0
> Oct 26 23:32:55 mina kernel: EIP:    0060:[<f882b8ce>]    Not tainted VLI
> Oct 26 23:32:55 mina kernel: EFLAGS: 00010296   (2.6.10-rc1-mm1y)
> Oct 26 23:32:55 mina kernel: EIP is at 
> scsi_block_when_processing_errors+0xe/0xe0 [scsi_mod]
> Oct 26 23:32:55 mina kernel: eax: 00000000   ebx: 6b6b6b6b   ecx: 
> f88ef640   edx: ec5b6578
> Oct 26 23:32:55 mina kernel: esi: e9baa4b8   edi: c17e9268   ebp: 
> e9677f0c   esp: e9677eb4
> Oct 26 23:32:55 mina kernel: ds: 007b   es: 007b   ss: 0068
> Oct 26 23:32:55 mina kernel: Process fdisk (pid: 2891, 
> threadinfo=e9676000 task=ea0c61f0)
> Oct 26 23:32:55 mina kernel: Stack: 00000000 00000001 00000000 00000000 
> 00000000 00000000 00000000 00000000
> Oct 26 23:32:55 mina kernel:        00000000 00000000 00000003 e9677ef0 
> c17e9268 0000006b c17e9268 e9677f0c
> Oct 26 23:32:55 mina kernel:        c0159761 c17e93e4 00000000 e9b98780 
> e9baa4b8 c17e9268 e9677f24 f88ef6a8
> Oct 26 23:32:55 mina kernel: Call Trace:
> Oct 26 23:32:55 mina kernel:  [<c01056cf>] show_stack+0x7f/0xa0
> Oct 26 23:32:55 mina kernel:  [<c0105876>] show_registers+0x156/0x1c0
> Oct 26 23:32:55 mina kernel:  [<c0105af6>] die+0x156/0x2e0
> Oct 26 23:32:55 mina kernel:  [<c011628d>] do_page_fault+0x36d/0x69c
> Oct 26 23:32:55 mina kernel:  [<c01051ed>] error_code+0x2d/0x38
> Oct 26 23:32:55 mina kernel:  [<f88ef6a8>] sd_release+0x68/0xa0 [sd_mod]
> Oct 26 23:32:55 mina kernel:  [<c0184c03>] blkdev_put+0x183/0x1b0
> Oct 26 23:32:55 mina kernel:  [<c01784ad>] __fput+0x14d/0x160
> Oct 26 23:32:55 mina kernel:  [<c0176797>] filp_close+0x57/0x90
> Oct 26 23:32:55 mina kernel:  [<c01768e3>] sys_close+0x113/0x240
> Oct 26 23:32:55 mina kernel:  [<c0104ff1>] sysenter_past_esp+0x52/0x71
> Oct 26 23:32:55 mina kernel: Code: 0c 8b 43 04 8b 00 89 5c 24 04 c7 04 
> 24 e4 d1 83 f8 89 44 24 08 e8 d3 21 8f c7 8d 76 00 55 89 e5 57 56 53 83 
> ec 4c 8b 75 08 8b 1e <8b> 83 10 01 00 00 a8 08 74 7c fc 31 c0 8d 7d b4 
> b9 05 00 00 00


The problem with using shost_for_each_device wrt to the above oops is
that scsi_forget_host sets the state to SDEV_CANCEL, so that when
scsi_host_cancel iterates over the devices using shost_for_each_device
it cannot get a handle to the sdev (scsi_device_get fails becuase the
state is set to SDEV_CANCEL). And, __scsi_iterate_devices does not clear
the next pointer if this happens, so I think this is needed to fix just
the refcount bug in shost_for_each_device.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

asked for it.

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

Patch from Ben Dooks

Update the s3c2410 machines for the new serial handling,
as well as add the clock sources the bast and vr1000
rely on.

Signed-off-by: Ben Dooks
Signed-off-by: Russell King

Patch from Ben Dooks

Update the arch/arm/mach-s3c2410 core files for the
new serial port bits, and fix the bugs with s3c2440
handling.

Signed-off-by: Ben Dooks
Signed-off-by: Russell King

Patch from Ben Dooks

Clock code now uses the supplied device to check for
the id of the clock, so uart clocks are now simply
"uart" instead of "uart0", "uart1" and "uart2"

Signed-off-by: Ben Dooks
Signed-off-by: Russell King