From: Joe Thornber <thornber@sistina.com>

Split the dm_create() function into two variants, depending on whether you
want the device to have a specific minor number.  This avoids the broken
overloading of the minor argument to the old dm_create().

From: Hugh Dickins <hugh@veritas.com>, Christoph Rohland <cr@sap.com>

Switches the long GPL blurb which Hugh took from the 2.4-ac version of
mm/shmem.c, back to the original oneliner (as in 2.4.current).

Christoph spotted that the long blurb says "either version 2 of the License,
or (at your option) any later version", whereas he intended it be as COPYING
says, version 2 period.  Hugh and Christoph are happier with known version 2
than undefined future variants.

From: James Morris <jmorris@intercode.com.au>

This patch against current bk removes anon_hash_chain from fs/inode.c, as
all inodes in the 2.5 kernel must be associated with a superblock, and the
anon hash is no longer needed.

(akpm: Well there's one inode with a null i_sb: swapper_inode.  But it
shouldn't get here and if it does, we'll fix it by other means).

From: Dipankar Sarma <dipankar@in.ibm.com>

Forward-port of a fix Alan merged into 2.4.21-pre.  It fixes arcnet oopses
with raw sockets.

From: george anzinger <george@mvista.com>

This patch addresses issues of roundoff error in the time keeping and NTP
code as follows:

The conversion of "actual jiffies" to TICK_USEC and then to TICK_NSEC
introduced large errors if jiffies was not a power of 10 (e.g.  1024 for
the ia64).  Most of this is avoided by converting directly to TICK_NSEC.

The calculation of MAX_SEC_IN_JIFFIES (the largest timespec or timeval the
kernel will attempt) had overflow problems in the 64-bit machines.  We
introduce a different equation for those machines.

The NTP frequency update code was allowing a micro second of error to
accumulate before applying the correction.  We change FINEUSEC to FINENSEC
to do the correction as soon as a full nanosecond has accumulated.

The initial calculation of time_freq for NTP had severe roundoff errors for
HZ not a power of 10 (i.e.  1024).  A new equation fixes this.

clock_nanosleep is changed to round up to the next jiffie to cover starting
between jiffies.

From: george anzinger <george@mvista.com>

This patch does the following:

Pushs down the change from timeval to timespec in the settime routines.

Fixes two places where time was set without updating the monotonic clock
offset.  (Changes sys_stime() to call do_settimeofday() and changes
clock_warp to do the update directly.) These were bugs!

Changes the uptime code to use the posix_clock_monotonic notion of uptime
instead of the jiffies.  This time will track NTP changes and so should be
better than your standard wristwatch (if your using ntp).

Changes posix_clock_monotonic to start at 0 on boot (was set to start at
initial jiffies).

Fixes a bug (never experienced) in timer_create() in posix-timers.c where
we "could" have released timer_id 0 if "id resources" were low.

Adds a test in do_settimeofday() to error out (EINVAL) attempts to use
unnormalized times.  This is passed back up to both settimeofday and
posix_setclock().

Warning: Requires changes in .../arch/???/kernel/time.c to change
do_settimeofday() to return an error if time is not normalized and to use a
timespec instead of timeval for its input.

From: Vojtech Pavlik <vojtech@suse.cz>

- irq_return_t fixes

- remove obsolete pci_driver.driver bits

From: "Protasevich, Natalie" <Natalie.Protasevich@UNISYS.com>

Include ES7000 specific code for es7000 subarch.

From: "Protasevich, Natalie" <Natalie.Protasevich@unisys.com>

The patch set adds support for the Unisys ES7000 series architecture.
It covers different ES7000 platform variations with their corresponding APIC
architectures, such as Fosters-Gallatins based (XAPIC) ES7000s as well as
older Cascades based (logical cluster) ES7000 machines.
The patch was discussed among platform developers on LK and later resided in
the Alan's tree for 4-5 weeks for testing. Alan indicated that there are
"Zero complaints of any kind, zero problems from the changed code". He
recommended submit the patch. The code was tested by IBM as well as others
and did not have any impact on existing subarchs and on standard systems.


Here is the "generic" patch #1.  It deals with the IRQ overrides (most of
changes in mpparse.c and PCI range adjustment in io_apic.c), insufficient
width of IO-APIC ID, and a corner case for defining quantities of things like
APICs, IRQ sources, etc.  (the attempt to make easier defining these items in
subarchs resulted in separation of MP definition portion in
include/asm-i386/mpparse.h from its enumeration portion).  It has also
Venkatesh Palladi's fix that he suggested earlier for the correct APIC
version; I had to have it for this platform since the correct APIC version
will be needed later for dynamic platform discovery when es7000 will be
incorporated into genapic.

From: Anton Blanchard <anton@samba.org>

- Add BLKRASET/BLKFRASET as compatible ioctls

- Remove ifdefs around RTC ioctls - ppc64 implements its own rtc driver
  and doesnt define CONFIG_RTC.

- remove SG_IO which is not compatible

- extra loop, autofs, NCP and bluetooth ioctls

From: Anton Blanchard <anton@samba.org>


Anton has been testing odd setups:

/* node 0 - no cpus, no memory */
/* node 1 - 1 cpu, no memory */
/* node 2 - 0 cpus, 1GB memory */
/* node 3 - 3 cpus, 3GB memory */

Two things tripped so far.  Firstly the ppc64 debug check for invalid cpus
in cpu_to_node().  Fix that in kernel/sched.c:node_nr_running_init().

The other problem concerned nodes with memory but no cpus.  kswapd tries to
set_cpus_allowed(0) and bad things happen.  So we only set cpu affinity
for kswapd if there are cpus in the node.

compat_sys_getrusage was returning the result of a logical or. Make
it return the result of sys_getrusage if it fails or -EFAULT if we
cant write it out to userspace.

Last night, Manfred and I found an interesting bug with kmalloc on
ppc32, where the kmalloc in alloc_super() (fs/super.c) was requesting
432 bytes but only getting 256 bytes.  The reason was that PAGE_SIZE
wasn't defined at the point where the kmalloc() inline function
occurs.  Thus the CACHE(32) entry got omitted from the list in
kmalloc_sizes.h, and kmalloc therefore used the entry in
malloc_sizes[] before the correct entry.

This patch fixes it by including asm/page.h and asm/cache.h in
linux/slab.h.  The list in kmalloc_sizes.h depends on L1_CACHE_BYTES
as well as PAGE_SIZE, which is why I added asm/cache.h.

This patch cleans up a few leftovers after Mochel's recent
driver model changes. Tested. Please apply.

- arch/i386/kernel/apic.c: <linux/module.h> no longer needed
- include/asm-i386/apic.h: device_lapic no longer exported

This patch removes the x86 local APIC blacklist rules for the
Microstar 6163 and Intel AL440LX mainboards. These boards do
work with local APIC enabled. Heavily tested. Please apply.

Long story: When I worked on the initial UP_APIC code, these
two mainboard (MS-6163 and AL440LX) were the only ones that
caused hangs when the local APIC was enabled. At the time,
I attributed these hangs to broken BIOSen and implemented DMI
scan blacklist rules to prevent enabling the local APIC on them.

However, the last year I've observed hangs on other mainboards
that initially were believed to be safe for local APIC. The
hangs turned out to be caused by APM's DISPLAY_BLANK option:
When this option is enabled, APM will invoke BIOS and graphics
card BIOS code without disabling the local APIC first. In many
cases, a local APIC timer interrupt while APM is trying to blank
the console will hang the system.

APM also has two other options that can hang the system: CPU_IDLE
and building APM as a module, both of which causes it to do BIOS
calls without disabling the local APIC first. On all my systems,
the only reliable configuration is to keep APM's DISPLAY_BLANK and
CPU_IDLE disabled, and to build APM non-modular.

Patch by Matthew Wilcox.

Allows all architectures to simply include the drivers/ide/Kconfig file.

Patch by Benjamin Herrenschmidt, minor fixes by me.

Unexpectedly there are some taskfile IO bugreports,
so bring back old code and add taskfile IO config option (default y).
This aids testing and lets people use safe(?) old code.

into home.transmeta.com:/home/torvalds/v2.5/linux

From: Manfred Spraul <manfred@colorfullife.com>

Martin found out what caused the deadlocks with 2.5.70 and SDET: I forgot
to initialize proc_lock during fork().  Usually this works, because the
parent lock is unlocked.  But a proc lookup during fork can cause it to be
locked, and then the system locks up.

1) Fix the comments for the migration_thread.  A while back Ingo
   agreed they were exactly wrong, IIRC. 8).

2) Changed spin_lock_irqsave to spin_lock_irq, since it's in a
   kernel thread.

3) Don't repeat if the task has moved off the original CPU, just finish.
   This is because we are simply trying to push the task off this CPU:
   if it's already moved, great.  Currently we might theoretically move
   a task which is actually running on another CPU, which is v. bad.

4) Replace the __ffs(p->cpus_allowed) with any_online_cpu(), since
   that's what it's for, and __ffs() can give the wrong answer, eg. if
   there's no CPU 0.

5) Move the core functionality of migrate_task into a separate function,
   move_task_away, which I want for the hotplug CPU patch.