pandora-kernel.git
16 years agomempolicy: update NUMA memory policy documentation
David Rientjes [Mon, 28 Apr 2008 09:12:31 +0000 (02:12 -0700)]
mempolicy: update NUMA memory policy documentation

Updates Documentation/vm/numa_memory_policy.txt and
Documentation/filesystems/tmpfs.txt to describe optional mempolicy mode flags.

Cc: Christoph Lameter <clameter@sgi.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomempolicy: add MPOL_F_RELATIVE_NODES flag
David Rientjes [Mon, 28 Apr 2008 09:12:30 +0000 (02:12 -0700)]
mempolicy: add MPOL_F_RELATIVE_NODES flag

Adds another optional mode flag, MPOL_F_RELATIVE_NODES, that specifies
nodemasks passed via set_mempolicy() or mbind() should be considered relative
to the current task's mems_allowed.

When the mempolicy is created, the passed nodemask is folded and mapped onto
the current task's mems_allowed.  For example, consider a task using
set_mempolicy() to pass MPOL_INTERLEAVE | MPOL_F_RELATIVE_NODES with a
nodemask of 1-3.  If current's mems_allowed is 4-7, the effected nodemask is
5-7 (the second, third, and fourth node of mems_allowed).

If the same task is attached to a cpuset, the mempolicy nodemask is rebound
each time the mems are changed.  Some possible rebinds and results are:

mems result
1-3 1-3
1-7 2-4
1,5-6 1,5-6
1,5-7 5-7

Likewise, the zonelist built for MPOL_BIND acts on the set of zones assigned
to the resultant nodemask from the relative remap.

In the MPOL_PREFERRED case, the preferred node is remapped from the currently
effected nodemask to the relative nodemask.

This mempolicy mode flag was conceived of by Paul Jackson <pj@sgi.com>.

Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomempolicy: add bitmap_onto() and bitmap_fold() operations
Paul Jackson [Mon, 28 Apr 2008 09:12:29 +0000 (02:12 -0700)]
mempolicy: add bitmap_onto() and bitmap_fold() operations

The following adds two more bitmap operators, bitmap_onto() and bitmap_fold(),
with the usual cpumask and nodemask wrappers.

The bitmap_onto() operator computes one bitmap relative to another.  If the
n-th bit in the origin mask is set, then the m-th bit of the destination mask
will be set, where m is the position of the n-th set bit in the relative mask.

The bitmap_fold() operator folds a bitmap into a second that has bit m set iff
the input bitmap has some bit n set, where m == n mod sz, for the specified sz
value.

There are two substantive changes between this patch and its
predecessor bitmap_relative:
 1) Renamed bitmap_relative() to be bitmap_onto().
 2) Added bitmap_fold().

The essential motivation for bitmap_onto() is to provide a mechanism for
converting a cpuset-relative CPU or Node mask to an absolute mask.  Cpuset
relative masks are written as if the current task were in a cpuset whose CPUs
or Nodes were just the consecutive ones numbered 0..N-1, for some N.  The
bitmap_onto() operator is provided in anticipation of adding support for the
first such cpuset relative mask, by the mbind() and set_mempolicy() system
calls, using a planned flag of MPOL_F_RELATIVE_NODES.  These bitmap operators
(and their nodemask wrappers, in particular) will be used in code that
converts the user specified cpuset relative memory policy to a specific system
node numbered policy, given the current mems_allowed of the tasks cpuset.

Such cpuset relative mempolicies will address two deficiencies
of the existing interface between cpusets and mempolicies:
 1) A task cannot at present reliably establish a cpuset
    relative mempolicy because there is an essential race
    condition, in that the tasks cpuset may be changed in
    between the time the task can query its cpuset placement,
    and the time the task can issue the applicable mbind or
    set_memplicy system call.
 2) A task cannot at present establish what cpuset relative
    mempolicy it would like to have, if it is in a smaller
    cpuset than it might have mempolicy preferences for,
    because the existing interface only allows specifying
    mempolicies for nodes currently allowed by the cpuset.

Cpuset relative mempolicies are useful for tasks that don't distinguish
particularly between one CPU or Node and another, but only between how many of
each are allowed, and the proper placement of threads and memory pages on the
various CPUs and Nodes available.

The motivation for the added bitmap_fold() can be seen in the following
example.

Let's say an application has specified some mempolicies that presume 16 memory
nodes, including say a mempolicy that specified MPOL_F_RELATIVE_NODES (cpuset
relative) nodes 12-15.  Then lets say that application is crammed into a
cpuset that only has 8 memory nodes, 0-7.  If one just uses bitmap_onto(),
this mempolicy, mapped to that cpuset, would ignore the requested relative
nodes above 7, leaving it empty of nodes.  That's not good; better to fold the
higher nodes down, so that some nodes are included in the resulting mapped
mempolicy.  In this case, the mempolicy nodes 12-15 are taken modulo 8 (the
weight of the mems_allowed of the confining cpuset), resulting in a mempolicy
specifying nodes 4-7.

Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: <kosaki.motohiro@jp.fujitsu.com>
Cc: <ray-lk@madrabbit.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomempolicy: add MPOL_F_STATIC_NODES flag
David Rientjes [Mon, 28 Apr 2008 09:12:27 +0000 (02:12 -0700)]
mempolicy: add MPOL_F_STATIC_NODES flag

Add an optional mempolicy mode flag, MPOL_F_STATIC_NODES, that suppresses the
node remap when the policy is rebound.

Adds another member to struct mempolicy, nodemask_t user_nodemask, as part of
a union with cpuset_mems_allowed:

struct mempolicy {
...
union {
nodemask_t cpuset_mems_allowed;
nodemask_t user_nodemask;
} w;
}

that stores the the nodemask that the user passed when he or she created the
mempolicy via set_mempolicy() or mbind().  When using MPOL_F_STATIC_NODES,
which is passed with any mempolicy mode, the user's passed nodemask
intersected with the VMA or task's allowed nodes is always used when
determining the preferred node, setting the MPOL_BIND zonelist, or creating
the interleave nodemask.  This happens whenever the policy is rebound,
including when a task's cpuset assignment changes or the cpuset's mems are
changed.

This creates an interesting side-effect in that it allows the mempolicy
"intent" to lie dormant and uneffected until it has access to the node(s) that
it desires.  For example, if you currently ask for an interleaved policy over
a set of nodes that you do not have access to, the mempolicy is not created
and the task continues to use the previous policy.  With this change, however,
it is possible to create the same mempolicy; it is only effected when access
to nodes in the nodemask is acquired.

It is also possible to mount tmpfs with the static nodemask behavior when
specifying a node or nodemask.  To do this, simply add "=static" immediately
following the mempolicy mode at mount time:

mount -o remount mpol=interleave=static:1-3

Also removes mpol_check_policy() and folds its logic into mpol_new() since it
is now obsoleted.  The unused vma_mpol_equal() is also removed.

Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomempolicy: support optional mode flags
David Rientjes [Mon, 28 Apr 2008 09:12:25 +0000 (02:12 -0700)]
mempolicy: support optional mode flags

With the evolution of mempolicies, it is necessary to support mempolicy mode
flags that specify how the policy shall behave in certain circumstances.  The
most immediate need for mode flag support is to suppress remapping the
nodemask of a policy at the time of rebind.

Both the mempolicy mode and flags are passed by the user in the 'int policy'
formal of either the set_mempolicy() or mbind() syscall.  A new constant,
MPOL_MODE_FLAGS, represents the union of legal optional flags that may be
passed as part of this int.  Mempolicies that include illegal flags as part of
their policy are rejected as invalid.

An additional member to struct mempolicy is added to support the mode flags:

struct mempolicy {
...
unsigned short policy;
unsigned short flags;
}

The splitting of the 'int' actual passed by the user is done in
sys_set_mempolicy() and sys_mbind() for their respective syscalls.  This is
done by intersecting the actual with MPOL_MODE_FLAGS, rejecting the syscall of
there are additional flags, and storing it in the new 'flags' member of struct
mempolicy.  The intersection of the actual with ~MPOL_MODE_FLAGS is stored in
the 'policy' member of the struct and all current users of pol->policy remain
unchanged.

The union of the policy mode and optional mode flags is passed back to the
user in get_mempolicy().

This combination of mode and flags within the same actual does not break
userspace code that relies on get_mempolicy(&policy, ...) and either

switch (policy) {
case MPOL_BIND:
...
case MPOL_INTERLEAVE:
...
};

statements or

if (policy == MPOL_INTERLEAVE) {
...
}

statements.  Such applications would need to use optional mode flags when
calling set_mempolicy() or mbind() for these previously implemented statements
to stop working.  If an application does start using optional mode flags, it
will need to mask the optional flags off the policy in switch and conditional
statements that only test mode.

An additional member is also added to struct shmem_sb_info to store the
optional mode flags.

[hugh@veritas.com: shmem mpol: fix build warning]
Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomempolicy: convert MPOL constants to enum
David Rientjes [Mon, 28 Apr 2008 09:12:23 +0000 (02:12 -0700)]
mempolicy: convert MPOL constants to enum

The mempolicy mode constants, MPOL_DEFAULT, MPOL_PREFERRED, MPOL_BIND, and
MPOL_INTERLEAVE, are better declared as part of an enum since they are
sequentially numbered and cannot be combined.

The policy member of struct mempolicy is also converted from type short to
type unsigned short.  A negative policy does not have any legitimate meaning,
so it is possible to change its type in preparation for adding optional mode
flags later.

The equivalent member of struct shmem_sb_info is also changed from int to
unsigned short.

For compatibility, the policy formal to get_mempolicy() remains as a pointer
to an int:

int get_mempolicy(int *policy, unsigned long *nmask,
  unsigned long maxnode, unsigned long addr,
  unsigned long flags);

although the only possible values is the range of type unsigned short.

Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: move cache_line_size() to <linux/cache.h>
Pekka Enberg [Mon, 28 Apr 2008 09:12:22 +0000 (02:12 -0700)]
mm: move cache_line_size() to <linux/cache.h>

Not all architectures define cache_line_size() so as suggested by Andrew move
the private implementations in mm/slab.c and mm/slob.c to <linux/cache.h>.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Reviewed-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agohugetlb: decrease hugetlb_lock cycling in gather_surplus_huge_pages
Adam Litke [Mon, 28 Apr 2008 09:12:20 +0000 (02:12 -0700)]
hugetlb: decrease hugetlb_lock cycling in gather_surplus_huge_pages

To reduce hugetlb_lock acquisitions and releases when freeing excess surplus
pages, scan the page list in two parts.  First, transfer the needed pages to
the hugetlb pool.  Then drop the lock and free the remaining pages back to the
buddy allocator.

In the common case there are zero excess pages and no lock operations are
required.

Thanks Mel Gorman for this improvement.

Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: try both endianess when checking for endianess
Chris Dearman [Mon, 28 Apr 2008 09:12:19 +0000 (02:12 -0700)]
mm: try both endianess when checking for endianess

When checking for the swap header try byteswapping the endianess dependent
fields to allow the swap partition to be shared between big & little endian
systems.

Signed-off-by: Chris Dearman <chris@mips.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: filter based on a nodemask as well as a gfp_mask
Mel Gorman [Mon, 28 Apr 2008 09:12:18 +0000 (02:12 -0700)]
mm: filter based on a nodemask as well as a gfp_mask

The MPOL_BIND policy creates a zonelist that is used for allocations
controlled by that mempolicy.  As the per-node zonelist is already being
filtered based on a zone id, this patch adds a version of __alloc_pages() that
takes a nodemask for further filtering.  This eliminates the need for
MPOL_BIND to create a custom zonelist.

A positive benefit of this is that allocations using MPOL_BIND now use the
local node's distance-ordered zonelist instead of a custom node-id-ordered
zonelist.  I.e., pages will be allocated from the closest allowed node with
available memory.

[Lee.Schermerhorn@hp.com: Mempolicy: update stale documentation and comments]
[Lee.Schermerhorn@hp.com: Mempolicy: make dequeue_huge_page_vma() obey MPOL_BIND nodemask]
[Lee.Schermerhorn@hp.com: Mempolicy: make dequeue_huge_page_vma() obey MPOL_BIND nodemask rework]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: have zonelist contains structs with both a zone pointer and zone_idx
Mel Gorman [Mon, 28 Apr 2008 09:12:17 +0000 (02:12 -0700)]
mm: have zonelist contains structs with both a zone pointer and zone_idx

Filtering zonelists requires very frequent use of zone_idx().  This is costly
as it involves a lookup of another structure and a substraction operation.  As
the zone_idx is often required, it should be quickly accessible.  The node idx
could also be stored here if it was found that accessing zone->node is
significant which may be the case on workloads where nodemasks are heavily
used.

This patch introduces a struct zoneref to store a zone pointer and a zone
index.  The zonelist then consists of an array of these struct zonerefs which
are looked up as necessary.  Helpers are given for accessing the zone index as
well as the node index.

[kamezawa.hiroyu@jp.fujitsu.com: Suggested struct zoneref instead of embedding information in pointers]
[hugh@veritas.com: mm-have-zonelist: fix memcg ooms]
[hugh@veritas.com: just return do_try_to_free_pages]
[hugh@veritas.com: do_try_to_free_pages gfp_mask redundant]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Christoph Lameter <clameter@sgi.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: use two zonelist that are filtered by GFP mask
Mel Gorman [Mon, 28 Apr 2008 09:12:16 +0000 (02:12 -0700)]
mm: use two zonelist that are filtered by GFP mask

Currently a node has two sets of zonelists, one for each zone type in the
system and a second set for GFP_THISNODE allocations.  Based on the zones
allowed by a gfp mask, one of these zonelists is selected.  All of these
zonelists consume memory and occupy cache lines.

This patch replaces the multiple zonelists per-node with two zonelists.  The
first contains all populated zones in the system, ordered by distance, for
fallback allocations when the target/preferred node has no free pages.  The
second contains all populated zones in the node suitable for GFP_THISNODE
allocations.

An iterator macro is introduced called for_each_zone_zonelist() that interates
through each zone allowed by the GFP flags in the selected zonelist.

Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: remember what the preferred zone is for zone_statistics
Mel Gorman [Mon, 28 Apr 2008 09:12:14 +0000 (02:12 -0700)]
mm: remember what the preferred zone is for zone_statistics

On NUMA, zone_statistics() is used to record events like numa hit, miss and
foreign.  It assumes that the first zone in a zonelist is the preferred zone.
When multiple zonelists are replaced by one that is filtered, this is no
longer the case.

This patch records what the preferred zone is rather than assuming the first
zone in the zonelist is it.  This simplifies the reading of later patches in
this set.

Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Christoph Lameter <clameter@sgi.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: introduce node_zonelist() for accessing the zonelist for a GFP mask
Mel Gorman [Mon, 28 Apr 2008 09:12:14 +0000 (02:12 -0700)]
mm: introduce node_zonelist() for accessing the zonelist for a GFP mask

Introduce a node_zonelist() helper function.  It is used to lookup the
appropriate zonelist given a node and a GFP mask.  The patch on its own is a
cleanup but it helps clarify parts of the two-zonelist-per-node patchset.  If
necessary, it can be merged with the next patch in this set without problems.

Reviewed-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: use zonelists instead of zones when direct reclaiming pages
Mel Gorman [Mon, 28 Apr 2008 09:12:12 +0000 (02:12 -0700)]
mm: use zonelists instead of zones when direct reclaiming pages

The following patches replace multiple zonelists per node with two zonelists
that are filtered based on the GFP flags.  The patches as a set fix a bug with
regard to the use of MPOL_BIND and ZONE_MOVABLE.  With this patchset, the
MPOL_BIND will apply to the two highest zones when the highest zone is
ZONE_MOVABLE.  This should be considered as an alternative fix for the
MPOL_BIND+ZONE_MOVABLE in 2.6.23 to the previously discussed hack that filters
only custom zonelists.

The first patch cleans up an inconsistency where direct reclaim uses
zonelist->zones where other places use zonelist.

The second patch introduces a helper function node_zonelist() for looking up
the appropriate zonelist for a GFP mask which simplifies patches later in the
set.

The third patch defines/remembers the "preferred zone" for numa statistics, as
it is no longer always the first zone in a zonelist.

The forth patch replaces multiple zonelists with two zonelists that are
filtered.  The two zonelists are due to the fact that the memoryless patchset
introduces a second set of zonelists for __GFP_THISNODE.

The fifth patch introduces helper macros for retrieving the zone and node
indices of entries in a zonelist.

The final patch introduces filtering of the zonelists based on a nodemask.
Two zonelists exist per node, one for normal allocations and one for
__GFP_THISNODE.

Performance results varied depending on the machine configuration.  In real
workloads the gain/loss will depend on how much the userspace portion of the
benchmark benefits from having more cache available due to reduced referencing
of zonelists.

These are the range of performance losses/gains when running against
2.6.24-rc4-mm1.  The set and these machines are a mix of i386, x86_64 and
ppc64 both NUMA and non-NUMA.
     loss   to  gain
Total CPU time on Kernbench: -0.86% to  1.13%
Elapsed   time on Kernbench: -0.79% to  0.76%
page_test from aim9:         -4.37% to  0.79%
brk_test  from aim9:         -0.71% to  4.07%
fork_test from aim9:         -1.84% to  4.60%
exec_test from aim9:         -0.71% to  1.08%

This patch:

The allocator deals with zonelists which indicate the order in which zones
should be targeted for an allocation.  Similarly, direct reclaim of pages
iterates over an array of zones.  For consistency, this patch converts direct
reclaim to use a zonelist.  No functionality is changed by this patch.  This
simplifies zonelist iterators in the next patch.

Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomake swap_pte_to_pagemap_entry() static
Adrian Bunk [Mon, 28 Apr 2008 09:12:11 +0000 (02:12 -0700)]
make swap_pte_to_pagemap_entry() static

Make the needlessly global swap_pte_to_pagemap_entry() static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: remove nopage
Nick Piggin [Mon, 28 Apr 2008 09:12:10 +0000 (02:12 -0700)]
mm: remove nopage

Nothing in the tree uses nopage any more.  Remove support for it in the
core mm code and documentation (and a few stray references to it in
comments).

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agommap_region: cleanup the final vma_merge() related code
Oleg Nesterov [Mon, 28 Apr 2008 09:12:10 +0000 (02:12 -0700)]
mmap_region: cleanup the final vma_merge() related code

It is not easy to actually understand the "if (!file || !vma_merge())"
code, turn it into "if (file && vma_merge())".  This makes immediately
obvious that the subsequent "if (file)" is superfluous.

As Hugh Dickins pointed out, we can also factor out the ->i_writecount
corrections, and add a small comment about that.

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agofix invalidate_inode_pages2_range() to not clear ret
Hisashi Hifumi [Mon, 28 Apr 2008 09:12:08 +0000 (02:12 -0700)]
fix invalidate_inode_pages2_range() to not clear ret

DIO invalidates page cache through invalidate_inode_pages2_range().
invalidate_inode_pages2_range() sets ret=-EIO when
invalidate_complete_page2() fails, but this ret is cleared if
do_launder_page() succeed on a page of next index.

In this case, dio is carried out even if invalidate_complete_page2() fails
on some pages.

This can cause inconsistency between memory and blocks on HDD because the
page cache still exists.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Hisashi Hifumi <hifumi.hisashi@oss.ntt.co.jp>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Chuck Lever <cel@citi.umich.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoremove sparse warning for mmzone.h
Harvey Harrison [Mon, 28 Apr 2008 09:12:07 +0000 (02:12 -0700)]
remove sparse warning for mmzone.h

include/linux/mmzone.h:640:22: warning: potentially expensive pointer subtraction

Calculate the offset into the node_zones array rather than the index
using casts to (char *) and comparing against the index * sizeof(struct zone).

On X86_32 this saves a sar, but code size increases by one byte per
is_highmem() use due to 32-bit cmps rather than 16 bit cmps.

Before:
 207:   2b 80 8c 07 00 00       sub    0x78c(%eax),%eax
 20d:   c1 f8 0b                sar    $0xb,%eax
 210:   83 f8 02                cmp    $0x2,%eax
 213:   74 16                   je     22b <kmap_atomic_prot+0x144>
 215:   83 f8 03                cmp    $0x3,%eax
 218:   0f 85 8f 00 00 00       jne    2ad <kmap_atomic_prot+0x1c6>
 21e:   83 3d 00 00 00 00 02    cmpl   $0x2,0x0
 225:   0f 85 82 00 00 00       jne    2ad <kmap_atomic_prot+0x1c6>
 22b:   64 a1 00 00 00 00       mov    %fs:0x0,%eax

After:
 207:   2b 80 8c 07 00 00       sub    0x78c(%eax),%eax
 20d:   3d 00 10 00 00          cmp    $0x1000,%eax
 212:   74 18                   je     22c <kmap_atomic_prot+0x145>
 214:   3d 00 18 00 00          cmp    $0x1800,%eax
 219:   0f 85 8f 00 00 00       jne    2ae <kmap_atomic_prot+0x1c7>
 21f:   83 3d 00 00 00 00 02    cmpl   $0x2,0x0
 226:   0f 85 82 00 00 00       jne    2ae <kmap_atomic_prot+0x1c7>
 22c:   64 a1 00 00 00 00       mov    %fs:0x0,%eax

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoRemove set_migrateflags()
Christoph Lameter [Mon, 28 Apr 2008 09:12:05 +0000 (02:12 -0700)]
Remove set_migrateflags()

Migrate flags must be set on slab creation as agreed upon when the antifrag
logic was reviewed.  Otherwise some slabs of a slabcache will end up in the
unmovable and others in the reclaimable section depending on which flag was
active when a new slab page was allocated.

This likely slid in somehow when antifrag was merged. Remove it.

The buffer_heads are always allocated with __GFP_RECLAIMABLE because the
SLAB_RECLAIM_ACCOUNT option is set.  The set_migrateflags() never had any
effect there.

Radix tree allocations are not directly reclaimable but they are allocated
with __GFP_RECLAIMABLE set on each allocation.  We now set
SLAB_RECLAIM_ACCOUNT on radix tree slab creation making sure that radix
tree slabs are consistently placed in the reclaimable section.  Radix tree
slabs will also be accounted as such.

There is then no user left of set_migratepages. So remove it.

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoaio: io_getevents() should return if io_destroy() is invoked
Jeff Moyer [Mon, 28 Apr 2008 09:12:04 +0000 (02:12 -0700)]
aio: io_getevents() should return if io_destroy() is invoked

This patch wakes up a thread waiting in io_getevents if another thread
destroys the context.  This was tested using a small program that spawns a
thread to wait in io_getevents while the parent thread destroys the io context
and then waits for the getevents thread to exit.  Without this patch, the
program hangs indefinitely.  With the patch, the program exits as expected.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: Christopher Smith <x@xman.org>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agohotplug-memory: make online_page() common
Jeremy Fitzhardinge [Mon, 28 Apr 2008 09:12:03 +0000 (02:12 -0700)]
hotplug-memory: make online_page() common

All architectures use an effectively identical definition of online_page(), so
just make it common code.  x86-64, ia64, powerpc and sh are actually
identical; x86-32 is slightly different.

x86-32's differences arise because it puts its hotplug pages in the highmem
zone.  We can handle this in the generic code by inspecting the page to see if
its in highmem, and update the totalhigh_pages count appropriately.  This
leaves init_32.c:free_new_highpage with a single caller, so I folded it into
add_one_highpage_init.

I also removed an incorrect comment referring to the NUMA case; any NUMA
details have already been dealt with by the time online_page() is called.

[akpm@linux-foundation.org: fix indenting]
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Acked-by: Dave Hansen <dave@linux.vnet.ibm.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamez.hiroyu@jp.fujitsu.com>
Tested-by: KAMEZAWA Hiroyuki <kamez.hiroyu@jp.fujitsu.com>
Cc: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: Christoph Lameter <clameter@sgi.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agohotplug memory remove: generic __remove_pages() support
Badari Pulavarty [Mon, 28 Apr 2008 09:12:01 +0000 (02:12 -0700)]
hotplug memory remove: generic __remove_pages() support

Generic helper function to remove section mappings and sysfs entries for the
section of the memory we are removing.  offline_pages() correctly adjusted
zone and marked the pages reserved.

TODO: Yasunori Goto is working on patches to free up allocations from bootmem.

Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com>
Acked-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc: replace remaining __FUNCTION__ occurrences
Harvey Harrison [Mon, 28 Apr 2008 09:12:00 +0000 (02:12 -0700)]
rtc: replace remaining __FUNCTION__ occurrences

__FUNCTION__ is gcc-specific, use __func__

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Cc: David Brownell <david-b@pacbell.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agodrivers/char/rtc.c: use time_before, time_before_eq, etc
Julia Lawall [Mon, 28 Apr 2008 09:11:59 +0000 (02:11 -0700)]
drivers/char/rtc.c: use time_before, time_before_eq, etc

The functions time_before, time_before_eq, time_after, and time_after_eq
are more robust for comparing jiffies against other values.

A simplified version of the semantic patch making this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@ change_compare_np @
expression E;
@@

(
- jiffies <= E
+ time_before_eq(jiffies,E)
|
- jiffies >= E
+ time_after_eq(jiffies,E)
|
- jiffies < E
+ time_before(jiffies,E)
|
- jiffies > E
+ time_after(jiffies,E)
)

@ include depends on change_compare_np @
@@

#include <linux/jiffies.h>

@ no_include depends on !include && change_compare_np @
@@

  #include <linux/...>
+ #include <linux/jiffies.h>
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc: add the support for alarm time relative to current time in sysfs
Zhao Yakui [Mon, 28 Apr 2008 09:11:58 +0000 (02:11 -0700)]
rtc: add the support for alarm time relative to current time in sysfs

In current kernel if we want to set the alarm time, the absolute time the
seconds relative to 1970-01-01 00:00:00) should be written into
/sys/class/rtc/rtc0/wakealarm.  It is not convenient.

It is more reasonable to add the support for the alarm time relative to
current RTC time.(the unit is second)

For example:
If the RTC is required to generate alarm after 2 minutes, the following
will be OK.
echo +120 > /sys/class/rtc/rtc0/wakealarm
or      echo +0x78 > /sys/class/rtc/rtc0/wakealarm

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc: rtc-rs5c372: fix up NULL name in transfer error path
Paul Mundt [Mon, 28 Apr 2008 09:11:57 +0000 (02:11 -0700)]
rtc: rtc-rs5c372: fix up NULL name in transfer error path

rs5c_get_regs() currently uses rs5c->rtc->name for its debug printk when
i2c_transfer() fails, though it is used several times before the rtc dev
has been registered. The earliest we can get at the symbolic name is via
the i2c client's struct device, which can be handled by moving the first
rs5c_get_regs() until after the client pointer is assigned.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Cc: David Brownell <david-b@pacbell.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agokerneldoc for <linux/clk.h>
David Brownell [Mon, 28 Apr 2008 09:11:56 +0000 (02:11 -0700)]
kerneldoc for <linux/clk.h>

Add <linux/clk.h> to the generated kerneldoc, with some overview
to go along with those per-function descriptions.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomake ds1511_rtc_{read,set}_time() static
Adrian Bunk [Mon, 28 Apr 2008 09:11:55 +0000 (02:11 -0700)]
make ds1511_rtc_{read,set}_time() static

Make the needlessly global ds1511_rtc_{read,set}_time() static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Cc: David Brownell <david-b@pacbell.net>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc: silence section mismatch warning in rtc-test
Sam Ravnborg [Mon, 28 Apr 2008 09:11:55 +0000 (02:11 -0700)]
rtc: silence section mismatch warning in rtc-test

Fix following warning:
WARNING: vmlinux.o(.data+0x253e28): Section mismatch in reference from the variable test_drv to the function .devexit.text:test_remove()

Fix by renaming the platfrom_driver variable from *_drv to *_driver
so modpost ignore the reference to an __devexit section.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc-x1205: new style conversion
Alessandro Zummo [Mon, 28 Apr 2008 09:11:54 +0000 (02:11 -0700)]
rtc-x1205: new style conversion

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc-pcf8563: new style conversion
Alessandro Zummo [Mon, 28 Apr 2008 09:11:54 +0000 (02:11 -0700)]
rtc-pcf8563: new style conversion

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc-isl1208: new style conversion and minor bug fixes
Alessandro Zummo [Mon, 28 Apr 2008 09:11:53 +0000 (02:11 -0700)]
rtc-isl1208: new style conversion and minor bug fixes

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: Herbert Valerio Riedel <hvr@gnu.org>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc: avoid legacy drivers with generic framework
David Brownell [Mon, 28 Apr 2008 09:11:52 +0000 (02:11 -0700)]
rtc: avoid legacy drivers with generic framework

Kconfig tweaks to help reduce RTC configuration bugs, by avoiding
legacy RTC drivers when the generic RTC framework is enabled:

 - If rtc-cmos is selected, disable the legacy rtc driver;

 - When using generic RTC on x86, enable rtc-cmos by default;

 - In the old "chardev RTC" section of Kconfig, add a comment
   warning people off these (seven) legacy RTC drivers when
   the generic framework is in use.

People can still use the legacy drivers if they want (or need) to.

This doesn't fix the broken dependencies for the legacy "CMOS" RTC driver.
Ideally it would be a full list of platforms where it works, not a partial
list of ones where it won't.  Or better yet, it would depend on a
"HAVE_CMOS_RTC" flag defined by various platforms ...  surely there's a
Kconfig style guideline lurking there.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agortc-pcf8583 build fix
David Brownell [Mon, 28 Apr 2008 09:11:51 +0000 (02:11 -0700)]
rtc-pcf8583 build fix

Fix bogus #include in rtc-pcf8583, so it compiles on platforms that
don't support PC clone RTCs.  (Original issue noted by Adrian Bunk.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Cc: Adrian Bunk <bunk@kernel.org>
Acked-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agodz: test after postfix decrement fails in dz_console_putchar()
Roel Kluin [Mon, 28 Apr 2008 09:11:50 +0000 (02:11 -0700)]
dz: test after postfix decrement fails in dz_console_putchar()

When loops reaches 0 the postfix decrement still subtracts, so the subsequent
test fails.

Signed-off-by: Roel Kluin <12o3l@tiscali.nl>
Acked-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: Johannes Weiner <hannes@saeurebad.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agomm: fix possible off-by-one in walk_pte_range()
Johannes Weiner [Mon, 28 Apr 2008 09:11:47 +0000 (02:11 -0700)]
mm: fix possible off-by-one in walk_pte_range()

After the loop in walk_pte_range() pte might point to the first address after
the pmd it walks.  The pte_unmap() is then applied to something bad.

Spotted by Roel Kluin and Andreas Schwab.

Signed-off-by: Johannes Weiner <hannes@saeurebad.de>
Cc: Roel Kluin <12o3l@tiscali.nl>
Cc: Andreas Schwab <schwab@suse.de>
Acked-by: Matt Mackall <mpm@selenic.com>
Acked-by: Mikael Pettersson <mikpe@it.uu.se>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agox86: PAT fix
Ingo Molnar [Fri, 21 Mar 2008 14:42:28 +0000 (15:42 +0100)]
x86: PAT fix

Adrian Bunk noticed the following Coverity report:

> Commit e7f260a276f2c9184fe753732d834b1f6fbe9f17
> (x86: PAT use reserve free memtype in mmap of /dev/mem)
> added the following gem to arch/x86/mm/pat.c:
>
> <--  snip  -->
>
> ...
> int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
>                                 unsigned long size, pgprot_t *vma_prot)
> {
>         u64 offset = ((u64) pfn) << PAGE_SHIFT;
>         unsigned long flags = _PAGE_CACHE_UC_MINUS;
>         unsigned long ret_flags;
> ...
> ...  (nothing that touches ret_flags)
> ...
>         if (flags != _PAGE_CACHE_UC_MINUS) {
>                 retval = reserve_memtype(offset, offset + size, flags, NULL);
>         } else {
>                 retval = reserve_memtype(offset, offset + size, -1, &ret_flags);
>         }
>
>         if (retval < 0)
>                 return 0;
>
>         flags = ret_flags;
>
>         if (pfn <= max_pfn_mapped &&
>             ioremap_change_attr((unsigned long)__va(offset), size, flags) < 0) {
>                 free_memtype(offset, offset + size);
>                 printk(KERN_INFO
>                 "%s:%d /dev/mem ioremap_change_attr failed %s for %Lx-%Lx\n",
>                         current->comm, current->pid,
>                         cattr_name(flags),
>                         offset, offset + size);
>                 return 0;
>         }
>
>         *vma_prot = __pgprot((pgprot_val(*vma_prot) & ~_PAGE_CACHE_MASK) |
>                              flags);
>         return 1;
> }
>
> <--  snip  -->
>
> If (flags != _PAGE_CACHE_UC_MINUS) we pass garbage from the stack to
> ioremap_change_attr() and/or __pgprot().
>
> Spotted by the Coverity checker.

the fix simplifies the code as we get rid of the 'ret_flags'
complication.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agox86 PAT: tone down debugging messages some more
Linus Torvalds [Sun, 27 Apr 2008 18:59:30 +0000 (11:59 -0700)]
x86 PAT: tone down debugging messages some more

Ingo already fixed one of these at my request (in "x86 PAT: tone down
debugging messages", commit 1ebcc654f010d4a63f3ebf8ddd2cab5a709b1824),
but there was another one he missed.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6
Linus Torvalds [Sun, 27 Apr 2008 18:25:00 +0000 (11:25 -0700)]
Merge git://git./linux/kernel/git/jejb/scsi-misc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6: (40 commits)
  [SCSI] jazz_esp, sgiwd93, sni_53c710, sun3x_esp: fix platform driver hotplug/coldplug
  [SCSI] aic7xxx: add const
  [SCSI] aic7xxx: add static
  [SCSI] aic7xxx: Update _shipped files
  [SCSI] aic7xxx: teach aicasm to not emit unused debug code/data
  [SCSI] qla2xxx: Update version number to 8.02.01-k2.
  [SCSI] qla2xxx: Correct regression in relogin code.
  [SCSI] qla2xxx: Correct misc. endian and byte-ordering issues.
  [SCSI] qla2xxx: make qla2x00_issue_iocb_timeout() static
  [SCSI] qla2xxx: qla_os.c, make 2 functions static
  [SCSI] qla2xxx: Re-register FDMI information after a LIP.
  [SCSI] qla2xxx: Correct SRB usage-after-completion/free issues.
  [SCSI] qla2xxx: Correct ISP84XX verify-chip response handling.
  [SCSI] qla2xxx: Wakeup DPC thread to process any deferred-work requests.
  [SCSI] qla2xxx: Collapse RISC-RAM retrieval code during a firmware-dump.
  [SCSI] m68k: new mac_esp scsi driver
  [SCSI] zfcp: Add some statistics provided by the FCP adapter to the sysfs
  [SCSI] zfcp: Print some messages only during ERP
  [SCSI] zfcp: Wait for free SBAL during exchange config
  [SCSI] scsi_transport_fc: fc_user_scan correction
  ...

16 years ago[SCSI] jazz_esp, sgiwd93, sni_53c710, sun3x_esp: fix platform driver hotplug/coldplug
Kay Sievers [Fri, 18 Apr 2008 20:57:19 +0000 (13:57 -0700)]
[SCSI] jazz_esp, sgiwd93, sni_53c710, sun3x_esp: fix platform driver hotplug/coldplug

Since

commit 43cc71eed1250755986da4c0f9898f9a635cb3bf
Author: Kay Sievers <kay.sievers@vrfy.org>
Date:   Sat Aug 18 04:40:39 2007 +0200

    platform: prefix MODALIAS with "platform:"

the platform modalias is prefixed with "platform:". Add MODULE_ALIAS()
to the hotpluggable SCSI platform drivers, to re-enable auto loading.

[dbrownell@users.sourceforge.net: more drivers, registration fixes]
[akpm@linux-foundation.org: fix sgiwd93.c]
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] aic7xxx: add const
Denys Vlasenko [Fri, 25 Apr 2008 02:36:01 +0000 (04:36 +0200)]
[SCSI] aic7xxx: add const

This patch adds more const keywords where appropriate.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] aic7xxx: add static
Denys Vlasenko [Fri, 25 Apr 2008 02:34:49 +0000 (04:34 +0200)]
[SCSI] aic7xxx: add static

This patch adds static (and sometimes const) keywords where appropriate.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] aic7xxx: Update _shipped files
Hannes Reinecke [Fri, 25 Apr 2008 13:03:05 +0000 (15:03 +0200)]
[SCSI] aic7xxx: Update _shipped files

Update the precompiled sequencer code to match the latest
aicasm changes.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] aic7xxx: teach aicasm to not emit unused debug code/data
Hannes Reinecke [Fri, 25 Apr 2008 13:01:41 +0000 (15:01 +0200)]
[SCSI] aic7xxx: teach aicasm to not emit unused debug code/data

Add a 'count' variable to each symbol which gets increased every time
the symbol is referenced.  And then modify the register definition to
include counts for symbols which are referenced from the source code
only and not from the sequencer code.

This will give us an automatic usage count for the symbols with only
minimal hand-crafting.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Update version number to 8.02.01-k2.
Andrew Vasquez [Thu, 24 Apr 2008 22:21:31 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Update version number to 8.02.01-k2.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Correct regression in relogin code.
Andrew Vasquez [Thu, 24 Apr 2008 22:21:30 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Correct regression in relogin code.

Commit 63a8651f2548c6bb5132c0b4e7dad4f57a9274db ([SCSI] qla2xxx:
Correct infinite-login-retry issue.) introduced a small
regression where a successful relogin would result in an fcport's
loop_id to be incorrectly reset to FC_NO_LOOP_ID.  Only clear-out
loopid, if retries have been 'truly' exhausted.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Correct misc. endian and byte-ordering issues.
Seokmann Ju [Thu, 24 Apr 2008 22:21:29 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Correct misc. endian and byte-ordering issues.

There were several places in the driver which could cause byte
ordering problem as provided by Al Viro
<viro@ZenIV.linux.org.uk>.

Signed-off-by: Seokmann Ju <seokmann.ju@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: make qla2x00_issue_iocb_timeout() static
Adrian Bunk [Thu, 24 Apr 2008 22:21:28 +0000 (15:21 -0700)]
[SCSI] qla2xxx: make qla2x00_issue_iocb_timeout() static

This patch makes the needlessly global qla2x00_issue_iocb_timeout()
static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: qla_os.c, make 2 functions static
Adrian Bunk [Thu, 24 Apr 2008 22:21:27 +0000 (15:21 -0700)]
[SCSI] qla2xxx: qla_os.c, make 2 functions static

This patch makes the following needlessly global functions static:
- qla2x00_alloc_work()
- qla2x00_post_work()

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Re-register FDMI information after a LIP.
Andrew Vasquez [Thu, 24 Apr 2008 22:21:26 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Re-register FDMI information after a LIP.

Original code would (incorrectly) only re-register after a
loop-down condition.  Also, FDMI registration should be enabled
by default.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Correct SRB usage-after-completion/free issues.
Andrew Vasquez [Thu, 24 Apr 2008 22:21:25 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Correct SRB usage-after-completion/free issues.

The driver is incorrectly assuming that the 'sp' reference held
in qla2[x00|4xx]_abort_command() is valid after the mailbox
command is issued to abort the exchange.  It is *not*, as the
command may be completed during interrupt context before control
is returned to the mailbox caller.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Correct ISP84XX verify-chip response handling.
Andrew Vasquez [Thu, 24 Apr 2008 22:21:24 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Correct ISP84XX verify-chip response handling.

Earlier code could trigger an infinite-retry if 1st invocation
returned a non-CS_COMPLETE status.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Wakeup DPC thread to process any deferred-work requests.
Andrew Vasquez [Thu, 24 Apr 2008 22:21:23 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Wakeup DPC thread to process any deferred-work requests.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] qla2xxx: Collapse RISC-RAM retrieval code during a firmware-dump.
Andrew Vasquez [Thu, 24 Apr 2008 22:21:22 +0000 (15:21 -0700)]
[SCSI] qla2xxx: Collapse RISC-RAM retrieval code during a firmware-dump.

Use the more efficient read-DMA'ble-buffer mailbox commands
rather than reading a single word/dword at a time.  We also
remove a bulk of the duplicate mailbox command-handling codes in
favor of more generic read-memory() routines (qla2xxx_dump_ram()
and qla24xx_dump_ram()).

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] m68k: new mac_esp scsi driver
Finn Thain [Fri, 25 Apr 2008 15:06:05 +0000 (10:06 -0500)]
[SCSI] m68k: new mac_esp scsi driver

Replace the mac_esp driver with a new one based on the esp_scsi core.

For esp_scsi: add support for sync transfers for the PIO mode, add a new
esp_driver_ops method to get the maximum dma transfer size (like the old
NCR53C9x driver), and some cleanups.

Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] zfcp: Add some statistics provided by the FCP adapter to the sysfs
Swen Schillig [Thu, 24 Apr 2008 17:35:54 +0000 (19:35 +0200)]
[SCSI] zfcp: Add some statistics provided by the FCP adapter to the sysfs

The new FCP adapter statistics provide a variety of information about
the virtual adapter (subchannel). In order to collect this information
the zfcp driver is extended to query this information.

The information provided by the new FCP adapter statistics can be
fetched by reading from the following files in the sysfs filesystem

/sys/class/scsi_host/host<n>/seconds_active
/sys/class/scsi_host/host<n>/requests
/sys/class/scsi_host/host<n>/megabytes
/sys/class/scsi_host/host<n>/utilization

These are the statistics on a virtual adapter (subchannel) level.

The information provided is raw and not modified or interpreted by any
means.  No interpretation or modification of the values is done by the
zfcp driver.

Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] zfcp: Print some messages only during ERP
Swen Schillig [Thu, 24 Apr 2008 17:35:53 +0000 (19:35 +0200)]
[SCSI] zfcp: Print some messages only during ERP

When statistics are polled from sysfs, the statistics use the same
commands as the adapter initialization. Change the messages printed
here, so they are only printed during initialization and not for each
poll of adapter data.

Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] zfcp: Wait for free SBAL during exchange config
Swen Schillig [Thu, 24 Apr 2008 17:35:52 +0000 (19:35 +0200)]
[SCSI] zfcp: Wait for free SBAL during exchange config

When sending a exchange config data command, wait for a free SBAL.
This does not matter during adapter initialization, but this is
required for pulling adapter statistics during high I/O load.

Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] scsi_transport_fc: fc_user_scan correction
James Smart [Thu, 24 Apr 2008 16:12:46 +0000 (12:12 -0400)]
[SCSI] scsi_transport_fc: fc_user_scan correction

Way back when, when the fc_user_scan routine was created, it kept some
of its original logic that walked the rport list and kicked off a scan.
Unfortunately, it didn't keep any of the locking around the rport list,
nor did it consider the synchronous nature of the scan invoked. The result,
there are some scan requests where the rport list changes, thus a subsequent
scan is called on a bogus rport structure and the system NMI's.

Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] aha1542: minor irq handler cleanups
Jeff Garzik [Thu, 24 Apr 2008 23:45:32 +0000 (19:45 -0400)]
[SCSI] aha1542: minor irq handler cleanups

- where the 'irq' function argument is known never to be used, rename
  it to 'dummy' to make this more obvious

- replace per-irq lookup functions and tables with a direct reference
  to data object obtained via 'dev_id' function argument, passed from
  request_irq()

Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] scsi_transport_spi: include sysfs.h
Randy Dunlap [Wed, 23 Apr 2008 16:56:14 +0000 (09:56 -0700)]
[SCSI] scsi_transport_spi: include sysfs.h

scsi_transport_spi.c needs to #include <linux/sysfs.h>:

next-20080423/drivers/scsi/scsi_transport_spi.c:1467: error: implicit declaration of function 'sysfs_update_group'
make[3]: *** [drivers/scsi/scsi_transport_spi.o] Error 1

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] FlashPoint: fix off-by-one errors
Adrian Bunk [Wed, 23 Apr 2008 09:51:10 +0000 (12:51 +0300)]
[SCSI] FlashPoint: fix off-by-one errors

This patch fixes off-by-one errors in error checks (the variables are
used as array indexes for arrays with MAX_SCSI_TAR resp. MAX_LUN
elements) spotted by the Coverity checker.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] aic7xxx: Update type check in aicasm grammar
Hannes Reinecke [Wed, 23 Apr 2008 09:39:49 +0000 (11:39 +0200)]
[SCSI] aic7xxx: Update type check in aicasm grammar

The function type_check() in aicasm grammar code was
never used properly due to a bug.
This patch fixes it up and ensures it's only called if appropriate.

In addition the unused 16bit instruction are disabled, but left in
the code for reference.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years ago[SCSI] fix SLUB WARN_ON
James Bottomley [Mon, 21 Apr 2008 15:57:20 +0000 (10:57 -0500)]
[SCSI] fix SLUB WARN_ON

We're getting a WARN_ON from SLUB indicating that we're trying to free
caches with in-use objects.  The root cause is a new dependency in the
command/sense free on unchecked_isa_dma.  The WARN_ON is caused by
drivers which change this in their setup after the command/sense cache
is allocated.

The fix is to move the allocation of this cache into scsi_add_host()
so things like gdth have an opportunity to modify it between alloc and
add (but *not* after).

The true fix would be to move unchecked_isa_dma into the template and
out of the host, so it because a truly read only variable.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years agoMerge branch 'kvm-updates-2.6.26' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 27 Apr 2008 17:13:52 +0000 (10:13 -0700)]
Merge branch 'kvm-updates-2.6.26' of git://git./linux/kernel/git/avi/kvm

* 'kvm-updates-2.6.26' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm: (147 commits)
  KVM: kill file->f_count abuse in kvm
  KVM: MMU: kvm_pv_mmu_op should not take mmap_sem
  KVM: SVM: remove selective CR0 comment
  KVM: SVM: remove now obsolete FIXME comment
  KVM: SVM: disable CR8 intercept when tpr is not masking interrupts
  KVM: SVM: sync V_TPR with LAPIC.TPR if CR8 write intercept is disabled
  KVM: export kvm_lapic_set_tpr() to modules
  KVM: SVM: sync TPR value to V_TPR field in the VMCB
  KVM: ppc: PowerPC 440 KVM implementation
  KVM: Add MAINTAINERS entry for PowerPC KVM
  KVM: ppc: Add DCR access information to struct kvm_run
  ppc: Export tlb_44x_hwater for KVM
  KVM: Rename debugfs_dir to kvm_debugfs_dir
  KVM: x86 emulator: fix lea to really get the effective address
  KVM: x86 emulator: fix smsw and lmsw with a memory operand
  KVM: x86 emulator: initialize src.val and dst.val for register operands
  KVM: SVM: force a new asid when initializing the vmcb
  KVM: fix kvm_vcpu_kick vs __vcpu_run race
  KVM: add ioctls to save/store mpstate
  KVM: Rename VCPU_MP_STATE_* to KVM_MP_STATE_*
  ...

16 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
Linus Torvalds [Sun, 27 Apr 2008 17:13:06 +0000 (10:13 -0700)]
Merge git://git./linux/kernel/git/bart/ide-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6: (49 commits)
  ide-tape: remove tape->merge_stage
  ide-tape: mv tape->merge_stage_size tape->merge_bh_size
  ide-tape: mv idetape_empty_write_pipeline ide_tape_flush_merge_buffer
  ide-tape: mv idetape_discard_read_pipeline ide_tape_discard_merge_buffer
  ide-tape: make __idetape_discard_read_pipeline() of type void
  ide: remove now unused ide_pci_create_host_proc()
  ide: remove /proc/ide/ali
  ide-tape: improve buffer pages freeing strategy
  ide-tape: mv tape->pages_per_stage tape->pages_per_buffer
  ide-tape: mv tape->stage_size tape->buffer_size
  ide-tape: improve buffer allocation strategy
  ide: add struct ide_io_ports (take 3)
  ide: make ide_unregister() take 'ide_hwif_t *' as an argument (take 2)
  ide: sanitize ide_unregister() usage
  mpc8xx-ide: use ide_find_port()
  ide: add "noacpi" / "acpigtf" / "acpionboot" parameters
  gayle: add "doubler" parameter
  ide: add "cdrom=" and "chs=" parameters
  ide: add "nodma|noflush|noprobe|nowerr=" parameters
  ide: remove obsoleted "hdx=autotune" kernel parameter
  ...

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux...
Linus Torvalds [Sun, 27 Apr 2008 17:10:54 +0000 (10:10 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/x86/linux-2.6-idle-fix

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-idle-fix:
  fix idle (arch, acpi and apm) and lockdep

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sun, 27 Apr 2008 17:10:37 +0000 (10:10 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: xpad - fix build failure

16 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Sun, 27 Apr 2008 17:10:14 +0000 (10:10 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  mlx4_core: Add helper to move QP to ready-to-send
  mlx4_core: Add HW queues allocation helpers
  RDMA/nes: Remove volatile qualifier from struct nes_hw_cq.cq_vbase
  mlx4_core: CQ resizing should pass a 0 opcode modifier to MODIFY_CQ
  mlx4_core: Move kernel doorbell management into core
  IB/ehca: Bump version number to 0026
  IB/ehca: Make some module parameters bool, update descriptions
  IB/ehca: Remove mr_largepage parameter
  IB/ehca: Move high-volume debug output to higher debug levels
  IB/ehca: Prevent posting of SQ WQEs if QP not in RTS
  IPoIB: Handle 4K IB MTU for UD (datagram) mode
  RDMA/nes: Fix adapter reset after PXE boot
  RDMA/nes: Print IPv4 addresses in a readable format
  RDMA/nes: Use print_mac() to format ethernet addresses for printing

16 years agoasm/unaligned.h doesn't work well as the very first include
Al Viro [Sun, 27 Apr 2008 05:19:18 +0000 (06:19 +0100)]
asm/unaligned.h doesn't work well as the very first include

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoe1000e triggers sparc32 gcc bug
Al Viro [Sun, 27 Apr 2008 05:15:42 +0000 (06:15 +0100)]
e1000e triggers sparc32 gcc bug

... and isn't possible on sparc32 boxen anyway, unless somebody
had done JavaStation with PCIE lately.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
16 years agoKVM: kill file->f_count abuse in kvm
Al Viro [Sat, 19 Apr 2008 19:33:56 +0000 (20:33 +0100)]
KVM: kill file->f_count abuse in kvm

Use kvm own refcounting instead of playing with ->filp->f_count.
That will allow to get rid of a lot of crap in anon_inode_getfd() and
kill a race in kvm_dev_ioctl_create_vm() (file might have been closed
immediately by another thread, so ->filp might point to already freed
struct file when we get around to setting it).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: MMU: kvm_pv_mmu_op should not take mmap_sem
Marcelo Tosatti [Wed, 16 Apr 2008 20:19:06 +0000 (17:19 -0300)]
KVM: MMU: kvm_pv_mmu_op should not take mmap_sem

kvm_pv_mmu_op should not take mmap_sem. All gfn_to_page() callers down
in the MMU processing will take it if necessary, so as it is it can
deadlock.

Apparently a leftover from the days before slots_lock.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: remove selective CR0 comment
Joerg Roedel [Wed, 16 Apr 2008 15:01:05 +0000 (17:01 +0200)]
KVM: SVM: remove selective CR0 comment

There is not selective cr0 intercept bug. The code in the comment sets the
CR0.PG bit. But KVM sets the CR4.PG bit for SVM always to implement the paged
real mode. So the 'mov %eax,%cr0' instruction does not change the CR0.PG bit.
Selective CR0 intercepts only occur when a bit is actually changed. So its the
right behavior that there is no intercept on this instruction.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: remove now obsolete FIXME comment
Joerg Roedel [Wed, 16 Apr 2008 14:51:19 +0000 (16:51 +0200)]
KVM: SVM: remove now obsolete FIXME comment

With the usage of the V_TPR field this comment is now obsolete.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: disable CR8 intercept when tpr is not masking interrupts
Joerg Roedel [Wed, 16 Apr 2008 14:51:18 +0000 (16:51 +0200)]
KVM: SVM: disable CR8 intercept when tpr is not masking interrupts

This patch disables the intercept of CR8 writes if the TPR is not masking
interrupts. This reduces the total number CR8 intercepts to below 1 percent of
what we have without this patch using Windows 64 bit guests.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: sync V_TPR with LAPIC.TPR if CR8 write intercept is disabled
Joerg Roedel [Wed, 16 Apr 2008 14:51:17 +0000 (16:51 +0200)]
KVM: SVM: sync V_TPR with LAPIC.TPR if CR8 write intercept is disabled

If the CR8 write intercept is disabled the V_TPR field of the VMCB needs to be
synced with the TPR field in the local apic.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: export kvm_lapic_set_tpr() to modules
Joerg Roedel [Wed, 16 Apr 2008 14:51:16 +0000 (16:51 +0200)]
KVM: export kvm_lapic_set_tpr() to modules

This patch exports the kvm_lapic_set_tpr() function from the lapic code to
modules. It is required in the kvm-amd module to optimize CR8 intercepts.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: sync TPR value to V_TPR field in the VMCB
Joerg Roedel [Wed, 16 Apr 2008 14:51:15 +0000 (16:51 +0200)]
KVM: SVM: sync TPR value to V_TPR field in the VMCB

This patch adds syncing of the lapic.tpr field to the V_TPR field of the VMCB.
With this change we can safely remove the CR8 read intercept.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: ppc: PowerPC 440 KVM implementation
Hollis Blanchard [Thu, 17 Apr 2008 04:28:09 +0000 (23:28 -0500)]
KVM: ppc: PowerPC 440 KVM implementation

This functionality is definitely experimental, but is capable of running
unmodified PowerPC 440 Linux kernels as guests on a PowerPC 440 host. (Only
tested with 440EP "Bamboo" guests so far, but with appropriate userspace
support other SoC/board combinations should work.)

See Documentation/powerpc/kvm_440.txt for technical details.

[stephen: build fix]

Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: Add MAINTAINERS entry for PowerPC KVM
Hollis Blanchard [Thu, 17 Apr 2008 04:28:08 +0000 (23:28 -0500)]
KVM: Add MAINTAINERS entry for PowerPC KVM

Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: ppc: Add DCR access information to struct kvm_run
Hollis Blanchard [Thu, 17 Apr 2008 04:28:07 +0000 (23:28 -0500)]
KVM: ppc: Add DCR access information to struct kvm_run

Device Control Registers are essentially another address space found on PowerPC
4xx processors, analogous to PIO on x86. DCRs are always 32 bits, and can be
identified by a 32-bit number. We forward most DCR accesses to userspace for
emulation (with the exception of CPR0 registers, which can be read directly
for simplicity in timebase frequency determination).

Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoppc: Export tlb_44x_hwater for KVM
Hollis Blanchard [Thu, 17 Apr 2008 04:28:06 +0000 (23:28 -0500)]
ppc: Export tlb_44x_hwater for KVM

PowerPC 440 KVM needs to know how many TLB entries are used for the host kernel
linear mapping (it does not modify these mappings when switching between guest
and host execution).

Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>
Acked-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: Rename debugfs_dir to kvm_debugfs_dir
Hollis Blanchard [Tue, 15 Apr 2008 21:05:42 +0000 (16:05 -0500)]
KVM: Rename debugfs_dir to kvm_debugfs_dir

It's a globally exported symbol now.

Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: x86 emulator: fix lea to really get the effective address
Avi Kivity [Mon, 14 Apr 2008 20:46:37 +0000 (23:46 +0300)]
KVM: x86 emulator: fix lea to really get the effective address

We never hit this, since there is currently no reason to emulate lea.

Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: x86 emulator: fix smsw and lmsw with a memory operand
Avi Kivity [Mon, 14 Apr 2008 11:40:50 +0000 (14:40 +0300)]
KVM: x86 emulator: fix smsw and lmsw with a memory operand

lmsw and smsw were implemented only with a register operand.  Extend them
to support a memory operand as well.  Fixes Windows running some display
compatibility test on AMD hosts.

Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: x86 emulator: initialize src.val and dst.val for register operands
Avi Kivity [Mon, 14 Apr 2008 20:27:07 +0000 (23:27 +0300)]
KVM: x86 emulator: initialize src.val and dst.val for register operands

This lets us treat the case where mod == 3 in the same manner as other cases.

Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: SVM: force a new asid when initializing the vmcb
Avi Kivity [Mon, 14 Apr 2008 10:10:21 +0000 (13:10 +0300)]
KVM: SVM: force a new asid when initializing the vmcb

Shutdown interception clears the vmcb, leaving the asid at zero (which is
illegal.  so force a new asid on vmcb initialization.

Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: fix kvm_vcpu_kick vs __vcpu_run race
Marcelo Tosatti [Fri, 11 Apr 2008 18:01:22 +0000 (15:01 -0300)]
KVM: fix kvm_vcpu_kick vs __vcpu_run race

There is a window open between testing of pending IRQ's
and assignment of guest_mode in __vcpu_run.

Injection of IRQ's can race with __vcpu_run as follows:

CPU0                                CPU1
kvm_x86_ops->run()
vcpu->guest_mode = 0                SET_IRQ_LINE ioctl
..
kvm_x86_ops->inject_pending_irq
kvm_cpu_has_interrupt()

                                    apic_test_and_set_irr()
                                    kvm_vcpu_kick
                                    if (vcpu->guest_mode)
                                        send_ipi()

vcpu->guest_mode = 1

So move guest_mode=1 assignment before ->inject_pending_irq, and make
sure that it won't reorder after it.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: add ioctls to save/store mpstate
Marcelo Tosatti [Fri, 11 Apr 2008 16:24:45 +0000 (13:24 -0300)]
KVM: add ioctls to save/store mpstate

So userspace can save/restore the mpstate during migration.

[avi: export the #define constants describing the value]
[christian: add s390 stubs]
[avi: ditto for ia64]

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years ago[SCSI] esp_scsi: Make cur_residue and tot_residue signed.
David S. Miller [Sat, 19 Apr 2008 14:16:38 +0000 (09:16 -0500)]
[SCSI] esp_scsi: Make cur_residue and tot_residue signed.

Many of the overflow checks test whether the value has
gone negative, and we want to retain such checks.

Reported by Julia Lawall.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
16 years agoide-tape: remove tape->merge_stage
Borislav Petkov [Sun, 27 Apr 2008 13:38:34 +0000 (15:38 +0200)]
ide-tape: remove tape->merge_stage

Get rid of the pipeline merge stage but retain the chrdev req caching
functionality by using a merge buffer tape->merge_bh which is flushed in chunks
of several blocks at a time. Also, remove last references to pipelining, e.g.
typedef idetape_stage_s.

Signed-off-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoide-tape: mv tape->merge_stage_size tape->merge_bh_size
Borislav Petkov [Sun, 27 Apr 2008 13:38:34 +0000 (15:38 +0200)]
ide-tape: mv tape->merge_stage_size tape->merge_bh_size

This is the size of the merge buffer.

Signed-off-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoide-tape: mv idetape_empty_write_pipeline ide_tape_flush_merge_buffer
Borislav Petkov [Sun, 27 Apr 2008 13:38:34 +0000 (15:38 +0200)]
ide-tape: mv idetape_empty_write_pipeline ide_tape_flush_merge_buffer

Signed-off-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoide-tape: mv idetape_discard_read_pipeline ide_tape_discard_merge_buffer
Borislav Petkov [Sun, 27 Apr 2008 13:38:34 +0000 (15:38 +0200)]
ide-tape: mv idetape_discard_read_pipeline ide_tape_discard_merge_buffer

Also, rename its __-low level helper too.

Signed-off-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoide-tape: make __idetape_discard_read_pipeline() of type void
Borislav Petkov [Sun, 27 Apr 2008 13:38:34 +0000 (15:38 +0200)]
ide-tape: make __idetape_discard_read_pipeline() of type void

It always returns 0 which has no effect on tape positioning calculation so
simplify it by converting its type to void, bringing no functional change to the
driver.

Signed-off-by: Borislav Petkov <petkovbb@gmail.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoide: remove now unused ide_pci_create_host_proc()
Alexey Dobriyan [Sun, 27 Apr 2008 13:38:34 +0000 (15:38 +0200)]
ide: remove now unused ide_pci_create_host_proc()

It creates files in proc with obsoleted ->get_info interface.

Signed-off-by: Alexey Dobriyan <adobriyan@openvz.org>
Cc: Andrew Morton <akpm@osdl.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
16 years agoide: remove /proc/ide/ali
Alexey Dobriyan [Sun, 27 Apr 2008 13:38:33 +0000 (15:38 +0200)]
ide: remove /proc/ide/ali

Bart says: "can be done from user-space and is not especially interesting
even when debugging problems (raw PCI config space dump is far more useful)."

Signed-off-by: Alexey Dobriyan <adobriyan@openvz.org>
Cc: Andrew Morton <akpm@osdl.org>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>