de-HZ-ification of the granularity defaults unearthed a pre-existing
property of CFS: while it correctly converges to the granularity goal,
it does not prevent run-time fluctuations in the range of
[-gran ... 0 ... +gran].

With the increase of the granularity due to the removal of HZ
dependencies, this becomes visible in chew-max output (with 5 tasks
running):

 out:  28 . 27. 32 | flu:  0 .  0 | ran:    9 .   13 | per:   37 .   40
 out:  27 . 27. 32 | flu:  0 .  0 | ran:   17 .   13 | per:   44 .   40
 out:  27 . 27. 32 | flu:  0 .  0 | ran:    9 .   13 | per:   36 .   40
 out:  29 . 27. 32 | flu:  2 .  0 | ran:   17 .   13 | per:   46 .   40
 out:  28 . 27. 32 | flu:  0 .  0 | ran:    9 .   13 | per:   37 .   40
 out:  29 . 27. 32 | flu:  0 .  0 | ran:   18 .   13 | per:   47 .   40
 out:  28 . 27. 32 | flu:  0 .  0 | ran:    9 .   13 | per:   37 .   40

average slice is the ideal 13 msecs and the period is picture-perfect 40
msecs. But the 'ran' field fluctuates around 13.33 msecs and there's no
mechanism in CFS to keep that from happening: it's a perfectly valid
solution that CFS finds.

to fix this we add a granularity/preemption rule that knows about
the "target latency", which makes tasks that run longer than the ideal
latency run a bit less. The simplest approach is to simply decrease the
preemption granularity when a task overruns its ideal latency. For this
we have to track how much the task executed since its last preemption.

( this adds a new field to task_struct, but we can eliminate that
  overhead in 2.6.24 by putting all the scheduler timestamps into an
  anonymous union. )

with this change in place, chew-max output is fluctuation-less all
around:

 out:  28 . 27. 39 | flu:  0 .  2 | ran:   13 .   13 | per:   41 .   40
 out:  28 . 27. 39 | flu:  0 .  2 | ran:   13 .   13 | per:   41 .   40
 out:  28 . 27. 39 | flu:  0 .  2 | ran:   13 .   13 | per:   41 .   40
 out:  28 . 27. 39 | flu:  0 .  2 | ran:   13 .   13 | per:   41 .   40
 out:  28 . 27. 39 | flu:  0 .  1 | ran:   13 .   13 | per:   41 .   40
 out:  28 . 27. 39 | flu:  0 .  1 | ran:   13 .   13 | per:   41 .   40

this patch has no impact on any fastpath or on any globally observable
scheduling property. (unless you have sharp enough eyes to see
millisecond-level ruckles in glxgears smoothness :-)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mike Galbraith <efault@gmx.de>

There is an Amarok song switch time increase (regression) under
hefty load.

What is happening is that sleeper_bonus is never consumed, and only
rarely goes below runtime_limit, so for the most part, Amarok isn't
getting any bonus at all.  We're keeping sleeper_bonus right at
runtime_limit (sched_latency == sched_runtime_limit == 40ms) forever, ie
we don't consume if we're lower that that, and don't add if we're above
it.  One Amarok thread waking (or anybody else) will push us past the
threshold, so the next thread waking gets nada, but will reap pain from
the previous thread waking until we drop back to runtime_limit.  It
looks to me like under load, some random task gets a bonus, and
everybody else pays, whether deserving or not.

This diff fixed the regression for me at any load rate.
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>

Without this, we get qla2xxx complaining about "ISP System Error".

What's happening here is the firmware is detecting a Xfer-ready from the
storage when in fact the data-direction for a mode-select should be a
write (DATA_OUT).

The following patch fixes the problem (typo). Verified by Brian, as
well.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Verified-by: Brian De Wolf <bldewolf@csupomona.edu>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
  [SPARC32]: Make flush_tlb_kernel_range() an inline function.
  [SERIAL]: Fix 32-bit warnings in sunzilog.c and sunsu.c
  [SPARC32]: Kill unused vars and macros from prom/console.c
  [SPARC32]: Add __cmpdi2() libcall implementation ala. MIPS.
  [VIDEO]: Do not prom_halt() in cg3 and bw2 device probe.
  [SUNVDC]: Use slice 0xff on VD_DISK_TYPE_DISK.

* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [NET]: Mark Paul Moore as maintainer of labelled networking.
  [VLAN/BRIDGE]: Fix "skb_pull_rcsum - Fatal exception in interrupt"
  [ISDN]: Get rid of some pointless allocation casts in common and bsd comp.
  [NET]: Avoid pointless allocation casts in BSD compression module
  [IRDA]: Do not do pointless kmalloc return value cast in KingSun driver
  [NET]: Fix crash in dev_mc_sync()/dev_mc_unsync()
  [PPPOL2TP]: Fix endianness annotations.
  [IOAT]: ioatdma needs to to play nice in a multi-dma-client world
  [SLIP]: trivial sparse warning fix
  [EQL]: sparse warning fix
  [NET]: is_power_of_2 in net/core/neighbour.c
  [TCP]: Describe tcp_init_cwnd() thoroughly in a comment.
  [NET]: Fix IP_ADD/DROP_MEMBERSHIP to handle only connectionless
  [KBUILD]: Sanitize tc_ematch headers.
  [IPSEC] AH4: Update IPv4 options handling to conform to RFC 4302.

Fix bogus DEBUG_PREEMPT warning on x86_64, when cpu brought online after
bootup: current_is_keventd is right to note its use of smp_processor_id
is preempt-safe, but should use raw_smp_processor_id to avoid the warning.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Sigh.  Again an ACPI assault on the Thinkpad's Fn+F4 to suspend to RAM.
The default and text for CONFIG_THINKPAD_ACPI_INPUT_ENABLED were fixed
in -rc3, but now commit 14e04fb3 ("ACPI:
Schedule /proc/acpi/event for removal") introduces the ACPI_PROC_EVENT
config entry, and defaults it to 'n' to disable it again.

Change default to y, and add comment to make it clearer that n is for
future distros.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Len Brown <len.brown@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Commit 61ec7567 ('ACPI: boot correctly
with "nosmp" or "maxcpus=0"') broke 'maxcpus=' handling on x86[-64].

maxcpus=N is now having no effect on x86_64, and freezing bootup on i386
(because of inconsistency with the separate maxcpus parsing down in
arch/i386, I guess).  That's because early_param parsing is a little
different from __setup parsing, and needs the "=" omitted: then it seems
to work as the original commit intended (no mention of IO-APIC in
/proc/interrupts when maxcpus=0).
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Len Brown <len.brown@intel.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus: (60 commits)
  [MIPS] Fulong doesn't need ISA DMA.
  [MIPS] IP27: intr_sconnect_level: don't disable interrupts.
  [MIPS] IP27: startup_bridge_irq: connect interrupt.
  [MIPS] IP27: shutdown_bridge_irq: don't free irq.
  [MIPS] Sort out handling of ISA-less PCI systems.
  [MIPS] Add __cmpdi2
  [MIPS] HOTPLUG: Make register_pci_controller __devinit.
  [MIPS] PCI: Remove __devinit attribute from pcibios_fixup_bus.
  [MIPS] PCI: Remove __devinit attribute from pcibios_fixup_bus.
  [MIPS] Delete duplicate inclusion of <linux/delay.h>.
  [MIPS] Polish <asm/edac.h>.
  [MIPS] IP22: Export sgi_gfxaddr for use by the Newport console driver.
  [MIPS] Maintain si_code field properly for FP exceptions
  [MIPS] SMTC: Fix duplicate status dumps on NMI
  [MIPS] Unconditionally writeback and invalidate caches on kexec.
  [PATCH] rtc: Make rtc-rs5c348 driver hotplug-aware
  [MIPS] Fix gcc 3.3 warning.
  [MIPS] Fix invalid semicolon after if statement
  [MIPS] Update Cobalt defconfig
  [MIPS] Update workpad_defconfig
  ...

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc:
  [POWERPC] Fix SLB initialization at boot time
  [POWERPC] Fix undefined reference to device_power_up/resume
  [POWERPC] cell: Update cell_defconfig for 2.6.23
  [POWERPC] axonram: Do not delete gendisks queue in error path
  [POWERPC] axonram: Module modification for latest firmware API changes
  [POWERPC] cell: Support pinhole-reset on IBM cell blades
  [POWERPC] spu_manage: Use newer physical-id attribute
  [POWERPC] pasemi: Another IOMMU bugfix for 64K PAGE_SIZE

* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kyle/parisc-2.6:
  [PARISC] Add NOTES section
  [PARISC] Use compat_sys_getdents
  [PARISC] Do not allow STI_CONSOLE to be modular
  [PARISC] Clean up sti_flush
  [PARISC] Add dummy isa_(bus|virt)_to_(virt|bus) inlines
  [PARISC] Add empty <asm-parisc/vga.h>

Bisected bizarre kernel-space nullptr dereference in udev to commit
18991197, adding the NOTES section fixes
it.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Switch to using the generic compat_sys_getdents instead of a
homebrew one.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

It doesn't really make much sense, anyways, and would need a pile of
symbols exported.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

sti_flush is supposed to flush the caches so we can execute the STI rom
we copied to memory. Anything more than flush_icache_range is overkill.

Fixes a missing symbol when built as a module.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Less painful than fixing up the Kconfig for a pile of drivers to only build
on X86 && ARM && MIPS...

Just make them BUG(), as defining them to be 1:1 with physical memory will
likely HPMC the box anyways.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>

This avoids unused variable warnings in places like mm/vmalloc.c:

mm/vmalloc.c: In function ‘unmap_kernel_range’:
mm/vmalloc.c:75: warning: unused variable ‘start’

caused by it previously being a macro.
Signed-off-by: David S. Miller <davem@davemloft.net>

resource_size_t can be either a u64 or a u32, and we can't
really know for sure, so when printing such a value out
always use long-long printf formatting and cast the argument
to that type.
Signed-off-by: David S. Miller <davem@davemloft.net>

This should have been removed during the of_console_device
changes.
Signed-off-by: David S. Miller <davem@davemloft.net>

Device mapper generates calls to this with recent versions
of gcc.
Signed-off-by: David S. Miller <davem@davemloft.net>

Just give a normal kernel log message of the problem and
return failure.

Based upon a patch from Mark Fortescue.
Signed-off-by: David S. Miller <davem@davemloft.net>

While debugging issues with the VDS server I made the
driver use partition 2 to get at the whole disk since
this is the "whole disk" partition in the Sun disk
label.

We really should use slice 0xff which really means
the whole physical disk in the VIO disk protocol.
Otherwise things won't work well on a disk image
that doesn't have a proper disk label on it.
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: James Morris <jmorris@namei.org>
Acked-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

I tried to preserve bridging code as it was before, but logic is quite
strange - I think we should free skb on error, since it is already
unshared and thus will just leak.

Herbert Xu states:

> +	if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL)
> +		goto out;

If this happens it'll be a double-free on skb since we'll
return NF_DROP which makes the caller free it too.

We could return NF_STOLEN to prevent that but I'm not sure
whether that's correct netfilter semantics.  Patrick, could
you please make a call on this?

Patrick McHardy states:

NF_STOLEN should work fine here.
Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

vmalloc() returns a void pointer - no need to cast the return value.
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The general kernel memory allocation functions return void pointers
and there is no need to cast their return values.
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

kmalloc() returns a void pointer, so there is no need to cast it in
 drivers/net/irda/kingsun-sir.c::kingsun_probe().
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch fixes a crash that may occur when the routine dev_mc_sync()
deletes an address from the list it is currently going through. It
saves the pointer to the next element before deleting the current one.
The problem may also exist in dev_mc_unsync().
Signed-off-by: Benjamin Thery <benjamin.thery@bull.net>
Acked-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

{s,d}_{session,tunnel} in pppol2tp_addr are actually host-endian
everywhere.  We might switch them to net-endian, of course, but
that structure is exposed to userland via getname...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Now that the DMA engine has a multi-client interface, fix the ioatdma
driver to play along.  At the same time, remove a couple of unnecessary
reads and writes.
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Function declared static in forward declaration, but not in actual code.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

More noodlin on long flights, patch bin. Sparse warning fix for eql.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Replacing n & (n - 1) for power of 2 check by is_power_of_2(n)
Signed-off-by: vignesh babu <vignesh.babu@wipro.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

People often get tripped up by this function and think that
it does not implemented the prescribed algorithms from
RFC2414 and RFC3390, even though it does.

So add a comment to head off such misunderstandings in the
future.
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix IP[V6]_ADD_MEMBERSHIP and IP[V6]_DROP_MEMBERSHIP to
return -EPROTO for connection oriented sockets.
Signed-off-by: Flavio Leitner <fleitner@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The headers in tc_ematch are used by iproute2, so these headers should
be processed.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

In testing our ESP/AH offload hardware, I discovered an issue with how
AH handles mutable fields in IPv4.  RFC 4302 (AH) states the following
on the subject:

        For IPv4, the entire option is viewed as a unit; so even
        though the type and length fields within most options are immutable
        in transit, if an option is classified as mutable, the entire option
        is zeroed for ICV computation purposes.

The current implementation does not zero the type and length fields,
resulting in authentication failures when communicating with hosts
that do (i.e. FreeBSD).

I have tested record route and timestamp options (ping -R and ping -T)
on a small network involving Windows XP, FreeBSD 6.2, and Linux hosts,
with one router.  In the presence of these options, the FreeBSD and
Linux hosts (with the patch or with the hardware) can communicate.
The Windows XP host simply fails to accept these packets with or
without the patch.

I have also been trying to test source routing options (using
traceroute -g), but haven't had much luck getting this option to work
*without* AH, let alone with.
Signed-off-by: Nick Bowler <nbowler@ellipticsemi.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>