The rpciod shutdown code gives ugly sleep_on without BKL warnings in
-mm.  And it looks indeed somewhat racy.

The easy fix would be to simply use a completion as in the patch below,
but that removes all the signal fuzzing semantics the current code has.
I don't really understand why we want to cancel the operation by
signals, but I think it'd be better to leave that to people familar with
the code anyway..

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

From: mikem@beardog.cca.cpqcorp.net

This patch fixes an Oops when unloading the driver.  Bug fix.  Please
consider this for inclusion.

All of the patches sent out are needed to get the driver in the 2.6 tree up
to the level of the driver that is in the 2.4 tree, excluding this patch
which is not required in 2.4.

More patches will be coming.  They include multi-path failover support,
support for more than 8 controllers, and msi support.  Presently working on a
per logical volume queueing scheme.

From: mikem@beardog.cca.cpqcorp.net

This patch uses the pci_module_init wrapper for hot plug cases.  It is in the
2.4.  tree.

From: mikem@beardog.cca.cpqcorp.net

This patch changes the way we fill out the /proc files we create.  It now has
human readable volume sizes, RAID levels, etc.  Also removes some fields that
were orginally for debug purposes.  This is in the 2.4 tree.

From: mikem@beardog.cca.cpqcorp.net

This patch changes a format specifier to unsigned to prevent the number of
blocks being displayed as a negative value on very large volumes.

From: mikem@beardog.cca.cpqcorp.net

This patch replaces reading directly form PCI config space where possible.
Most of what we need is in the pdev struct.  This is in 2.4.

From: mikem@beardog.cca.cpqcorp.net

This patch moves the check of the controller to before trying to enable it.
If a controller is disabled the system will Oops without this fix.  This in
the 2.4 tree.

From: mikem@beardog.cca.cpqcorp.net

This patch addresses a bug in the ASIC on the 6400 series controllers.  When
prefetching from host memory we grab an extra 750 or so bytes of data.  If
this occurs on a memory boundary the machine will MCA.  This bug affects IPF
and Alpha based platforms.  It is not known to be a problem on x86.

Prefetch will be disabled via the f/w.  We need to enable it on x86 to
address a _big_ performance hit on RAID 1 operations.

It is in the 2.4 tree.

From: mikem@beardog.cca.cpqcorp.net

This patch fixes a bug when sharing IRQs with another controller that
receives a lot of interrupts.  Without this check we will panic the system
when unloading and reloading the driver.  This is in 2.4.

From: mikem@beardog.cca.cpqcorp.net

This patch adds support for the next generation embedded cciss controller.
It also bumps the version and changes the author to HP.  This patch is in
2.4.

From: mikem@beardog.cca.cpqcorp.net

This patch fixes a bug where under certain error conditions we bail and try
to free our I/O memory.  This patch is in the 2.4 tree.

From: <mikem@beardog.cca.cpqcorp.net>

This patch eliminates the bad assumption that all of our PCI BARs will always
be 32-bits.  Tested against the 2.6.2 kernel.

This is required to support the Pinnacles architecture.  It is already in the
2.4 tree.

(This needs to be converted to use pci_request_regions())

Print a size_t correctly.

Include the needed compiler.h.  Also swizze the return type and the attribute
around to match everyone else.

From: Joe Korty <joe.korty@ccur.com>

Now that bitmap_parse is part of bitmap.[ch], it is allowed to manipulate
bits directly as the other bitmap routines do.

It was old.

From: <viro@parcelfarce.linux.theplanet.co.uk>

Simplify the code, eliminate an unused variable warning.

From: Andreas Steinmetz <ast@domdv.de>

The attached patch adds the pci id 5961 to radeonfb.  Without the patch my
9200 displays only a blank screen.  lspci output below.

05:00.0 VGA compatible controller: ATI Technologies Inc Radeon RV280
[Radeon 9200] (rev 01) (prog-if 00 [VGA])
         Subsystem: Giga-byte Technology: Unknown device 4018
         Flags: bus master, 66Mhz, medium devsel, latency 64, IRQ 16
         Memory at e0000000 (32-bit, prefetchable) [size=128M]
         I/O ports at b800 [size=256]
         Memory at feaf0000 (32-bit, non-prefetchable) [size=64K]
         Expansion ROM at feac0000 [disabled] [size=128K]
         Capabilities: [58] AGP version 3.0
         Capabilities: [50] Power Management version 2

From: <miles@mcspd15.ucom.lsi.nec.co.jp> (Miles Bader)

make __delay function handle a loop count of zero.  In practice, this only
seems to occur in odd debugging situations, but it's quite annoying then.

From: <miles@mcspd15.ucom.lsi.nec.co.jp> (Miles Bader)

Define ARCH_HAS_*_EXTABLE macros for v850

From: <miles@mcspd15.ucom.lsi.nec.co.jp> (Miles Bader)

Add some #includes for the v850 to eliminate some compiler warnings

Always compile ide-dma.c if CONFIG_BLK_DEV_IDEDMA=y, mark PCI specific code
with CONFIG_BLK_DEV_IDEDMA_PCI for now (it should migrate to ide_pcidma.c
over a time).  This fixes a small bug - in_drive_list() from icside.c used
!strstr() instead of strstr() so it was missing two entries from a blacklist.

Let's play spot the bug here...

Amazing that it survives in modular case...

It was broken by IDE modular fixes in 2.6.2.

From: Matthew Wilcox <willy@debian.org>

From: Christoph Hellwig <hch@lst.de>

Instead of using the old MOD_INC_USE_COUNT and getting warnings all the
time preventing module unload can be much easier achived by just not
implementing a module_exit handler.

into intel.com:/home/lenb/src/linux-acpi-test-2.6.3

ARM added a definition for PT_SINGLESTEP which conflicted with
our definition.  So define PT_SINGLESTEP_BIT, PT_BLOCKSTEP_BIT and
PT_BLOCKSTEP to prevent similar problems in the future.  (James Bottomley)

Update harmony driver to the latest in the PA-RISC tree (Helge Deller)

PA-RISC also uses the 64-bit version of the IPC structs.

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

into home.osdl.org:/home/torvalds/v2.5/linux

into redhat.com:/spare/repo/net-drivers-2.5