Convert the PDA code to use %fs rather than %gs as the segment for
per-processor data.  This is because some processors show a small but
measurable performance gain for reloading a NULL segment selector (as %fs
generally is in user-space) versus a non-NULL one (as %gs generally is).

On modern processors the difference is very small, perhaps undetectable.
Some old AMD "K6 3D+" processors are noticably slower when %fs is used
rather than %gs; I have no idea why this might be, but I think they're
sufficiently rare that it doesn't matter much.

This patch also fixes the math emulator, which had not been adjusted to
match the changed struct pt_regs.

[frederik.deweerdt@gmail.com: fixit with gdb]
[mingo@elte.hu: Fix KVM too]
Signed-off-by: NJeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: NAndi Kleen <ak@suse.de>
Cc: Ian Campbell <Ian.Campbell@XenSource.com>
Acked-by: NIngo Molnar <mingo@elte.hu>
Acked-by: NZachary Amsden <zach@vmware.com>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: NFrederik Deweerdt <frederik.deweerdt@gmail.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>

- Removed an extraneous debug message from allocate_cachealigned_map

- Changed extract_lsb_from_nodes to return 63 for the case where there was
  only one memory node.  The prevents the creation of the dynamic hashmap.

- Changed extract_lsb_from_nodes to use only the starting memory address of
  a node.  On an ES7000, our nodes overlap the starting and ending address,
  meaning, that we see nodes like

	00000 - 10000
	10000 - 20000

  But other systems have nodes whose start and end addresses do not overlap.
   For example:

	00000 - 0FFFF
	10000 - 1FFFF

  In this case, using the ending address will result in an LSB much lower
  than what is possible.  In this case an LSB of 1 when in reality it should
  be 16.

Cc: Andi Kleen <ak@suse.de>
Cc: Rohit Seth <rohitseth@google.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NAndi Kleen <ak@suse.de>

Remove the statically allocated memory to NUMA node hash map in favor of a
dynamically allocated memory to node hash map (it is cache aligned).

This patch has the nice side effect in that it allows the hash map to grow
for systems with large amounts of memory (256GB - 1TB), but suffer from
having small PCI space tacked onto the boot node (which is somewhere
between 192MB to 512MB on the ES7000).
Signed-off-by: NAmul Shah <amul.shah@unisys.com>
Signed-off-by: NAndi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Rohit Seth <rohitseth@google.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>

This does user copies in fs write() into the page cache with write combining.
This pushes the destination out of the CPU's cache, but allows higher bandwidth
in some case.

The theory is that the page cache data is usually not touched by the
CPU again and it's better to not pollute the cache with it. Also it is a little
faster.
Signed-off-by: NAndi Kleen <ak@suse.de>

Signed-off-by: NAndi Kleen <ak@suse.de>

Signed-off-by: NAndi Kleen <ak@suse.de>

* master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
  [SPARC]: Re-export saved_command_line to modules.
  [SPARC64]: Increase command line size to 2048 like other arches.
  [SPARC64]: We do not need ZONE_DMA.

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6: (25 commits)
  [XFRM]: Fix OOPSes in xfrm_audit_log().
  [TCP]: cleanup of htcp (resend)
  [TCP]: Use read mostly for CUBIC parameters.
  [NETFILTER]: nf_conntrack_tcp: make sysctl variables static
  [NETFILTER]: ip6t_mh: drop piggyback payload packet on MH packets
  [NETFILTER]: Fix whitespace errors
  [NETFILTER]: Kconfig: improve dependency handling
  [NETFILTER]: xt_mac/xt_CLASSIFY: use IPv6 hook names for IPv6 registration
  [NETFILTER]: nf_conntrack: change nf_conntrack_l[34]proto_unregister to void
  [NETFILTER]: nf_conntrack: properly use RCU for nf_conntrack_destroyed callback
  [NETFILTER]: ip_conntrack: properly use RCU for ip_conntrack_destroyed callback
  [NETFILTER]: nf_conntrack: fix invalid conntrack statistics RCU assumption
  [NETFILTER]: ip_conntrack: fix invalid conntrack statistics RCU assumption
  [NETFILTER]: nf_conntrack: properly use RCU API for nf_ct_protos/nf_ct_l3protos arrays
  [NETFILTER]: ip_conntrack: properly use RCU API for ip_ct_protos array
  [NETFILTER]: nf_nat: properly use RCU API for nf_nat_protos array
  [NETFILTER]: ip_nat: properly use RCU API for ip_nat_protos array
  [NETFILTER]: nf_log: minor cleanups
  [NETFILTER]: nf_log: switch logger registration/unregistration to mutex
  [NETFILTER]: nf_log: make nf_log_unregister_pf return void
  ...

This reverts some bogosity from the dynamic command-line
changes made on sparc32 and sparc64.

Drivers such as drivers/sbus/char/openprom.c reference
saved_command_line, and can be modular.

The boot_command_line is __initdata, yet the dynamic command-line
changes add modular exports of that symbol, obviously wrong.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since we look in both source and object directories for localversion*
files, we accidentally ended up getting them twice.  Use 'sort -u' to
avoid that.
Reported-by: NTony Luck <tony.luck@intel.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Make sure that this function is called correctly, and
add BUG() checking to ensure the arguments are sane.

Based upon a patch by Joy Latten.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Minor non-invasive cleanups:
 * white space around operators and line wrapping
 * use const
 * use __read_mostly
Signed-off-by: NStephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

These module parameters should be in the read mostly area to avoid
cache pollution.
Signed-off-by: NStephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

sysctls are registered by the protocol module itself since 2.6.19, no need
to have them visible to others.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Regarding RFC3775, MH payload proto field should be IPPROTO_NONE. Otherwise
it must be discarded (and the receiver should send ICMP error).

We assume filter should drop such piggyback everytime to disallow slipping
through firewall rules, even the final receiver will discard it.
Signed-off-by: NMasahide NAKAMURA <nakam@linux-ipv6.org>
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NYOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of depending on internally needed options and letting users
figure out what is needed, select them when needed:

- IP_NF_IPTABLES, IP_NF_ARPTABLES and IP6_NF_IPTABLES select
  NETFILTER_XTABLES

- NETFILTER_XT_TARGET_CONNMARK, NETFILTER_XT_MATCH_CONNMARK and
  IP_NF_TARGET_CLUSTERIP select NF_CONNTRACK_MARK

- NETFILTER_XT_MATCH_CONNBYTES selects NF_CT_ACCT
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use NF_IP6_ instead of NF_IP_. The values are identical, this is merely
cleanup.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

No caller checks the return value, and since its usually called within the
module unload path there's nothing a module could do about errors anyway,
so BUG on invalid conditions and return void.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

NF_CT_STAT_INC assumes rcu_read_lock in nf_hook_slow disables
preemption as well, making it legal to use __get_cpu_var without
disabling preemption manually. The assumption is not correct anymore
with preemptable RCU, additionally we need to protect against softirqs
when not holding nf_conntrack_lock.

Add NF_CT_STAT_INC_ATOMIC macro, which disables local softirqs,
and use where necessary.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

CONNTRACK_STAT_INC assumes rcu_read_lock in nf_hook_slow disables
preemption as well, making it legal to use __get_cpu_var without
disabling preemption manually. The assumption is not correct anymore
with preemptable RCU, additionally we need to protect against softirqs
when not holding ip_conntrack_lock.

Add CONNTRACK_STAT_INC_ATOMIC macro, which disables local softirqs,
and use where necessary.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace preempt_{enable,disable} based RCU by proper use of the
RCU API and add missing rcu_read_lock/rcu_read_unlock calls in
all paths not obviously only used within packet process context
(nfnetlink_conntrack).
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace preempt_{enable,disable} based RCU by proper use of the
RCU API and add missing rcu_read_lock/rcu_read_unlock calls in
all paths not obviously only used within packet process context
(nfnetlink_conntrack).
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace preempt_{enable,disable} based RCU by proper use of the
RCU API and add missing rcu_read_lock/rcu_read_unlock calls in
paths used outside of packet processing context (nfnetlink_conntrack).
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace preempt_{enable,disable} based RCU by proper use of the
RCU API and add missing rcu_read_lock/rcu_read_unlock calls in
paths used outside of packet processing context (nfnetlink_conntrack).
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

- rename nf_logging to nf_loggers since its an array of registered loggers

- rename nf_log_unregister_logger() to nf_log_unregister() to make it
  symetrical to nf_log_register() and convert all users
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The spinlock is only used in process context (register/unregister),
switch to a mutex.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since the only user of nf_log_unregister_pf (nfnetlink_log) doesn't
check the return value, change it to void and bail out silently when
a non-existant address family is supplied.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Also replace synchronize_net() calls by synchronize_rcu() since the
RCU protected data is also used for sysfs.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The spinlock is only used in process context (register/unregister)
since RCU is used for the nf_hook lists, switch to a mutex.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The spinlock is only used in process context (register/unregister),
switch to a mutex.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We're only adding to the list, no need to synchronize.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use rcu_assign_pointer/rcu_dereference for ip_ct_attach pointer instead
of self-made RCU and use rcu_read_lock to make sure the conntrack module
doesn't disappear below us while calling it, since this function can be
called from outside the netfilter hooks.
Signed-off-by: NPatrick McHardy <kaber@trash.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

* 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [S390] remove __io_virt and mmiowb.
  [S390] cio: use ARRAY_SIZE in device_id.c
  [S390] cio: Fixup interface for setting options on ccw devices.
  [S390] smp_call_function/smp_call_function_on locking.

This patch is inspired by Arjan's "Patch series to mark struct
file_operations and struct inode_operations const".

Compile tested with gcc & sparse.
Signed-off-by: NJosef 'Jeff' Sipek <jsipek@cs.sunysb.edu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Many struct inode_operations in the kernel can be "const".  Marking them const
moves these to the .rodata section, which avoids false sharing with potential
dirty data.  In addition it'll catch accidental writes at compile time to
these shared resources.
Signed-off-by: NArjan van de Ven <arjan@linux.intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>