sfrench/cifs-2.6.git
13 years ago[NETFILTER]: xt_sctp: fix mistake to pass a pointer where array is required
Li Zefan [Thu, 18 Oct 2007 12:12:21 +0000 (05:12 -0700)]
[NETFILTER]: xt_sctp: fix mistake to pass a pointer where array is required

Macros like SCTP_CHUNKMAP_XXX(chukmap) require chukmap to be an array,
but match_packet() passes a pointer to these macros. Also remove the
ELEMCOUNT macro and fix a bug in SCTP_CHUNKMAP_COPY.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[NET]: Fix OOPS due to missing check in dev_parse_header().
Patrick McHardy [Thu, 18 Oct 2007 12:09:28 +0000 (05:09 -0700)]
[NET]: Fix OOPS due to missing check in dev_parse_header().

[ This is kernel bugzilla 9174 "linux-2.6.23-git11 kernel panic" ]

The device in question is an IPv6-over-IPv4 tunnel, which doesn't have
any header_ops, so the crash happens in dev_parse_header when
dereferencing them.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[TCP]: Remove lost_retrans zero seqno special cases
Ilpo Järvinen [Thu, 18 Oct 2007 12:07:57 +0000 (05:07 -0700)]
[TCP]: Remove lost_retrans zero seqno special cases

Both high-sack detection and new lowest seq variables have
unnecessary zero special case which are now removed by setting
safe initial seqnos.

This also fixes problem which caused zero received_upto being
passed to tcp_mark_lost_retrans which confused after relations
within the marker loop causing incorrect TCPCB_SACKED_RETRANS
clearing. The problem was noticed because of a performance
report from TAKANO Ryousei <takano@axe-inc.co.jp>.

Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Acked-by: Ryousei Takano <takano-ryousei@aist.go.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[NET]: fix carrier-on bug?
Jeff Garzik [Thu, 18 Oct 2007 06:26:43 +0000 (23:26 -0700)]
[NET]: fix carrier-on bug?

While looking at a net driver with the following construct,

if (!netif_carrier_ok(dev))
netif_carrier_on(dev);

it stuck me that the netif_carrier_ok() check was redundant, since
netif_carrier_on() checks bit __LINK_STATE_NOCARRIER anyway.  This is
the same reason why netif_queue_stopped() need not be called prior to
netif_wake_queue().

This is true, but there is however an unwanted side effect from assuming
that netif_carrier_on() can be called multiple times:  it touches the
watchdog, regardless of pre-existing carrier state.

The fix:  move watchdog-up inside the bit-cleared code path.

Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[NET]: Fix uninitialised variable in ip_frag_reasm()
David Howells [Thu, 18 Oct 2007 04:37:22 +0000 (21:37 -0700)]
[NET]: Fix uninitialised variable in ip_frag_reasm()

Fix uninitialised variable in ip_frag_reasm().  err should be set to
-ENOMEM if the initial call of skb_clone() fails.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Rename mode to outer_mode and add inner_mode
Herbert Xu [Thu, 18 Oct 2007 04:35:51 +0000 (21:35 -0700)]
[IPSEC]: Rename mode to outer_mode and add inner_mode

This patch adds a new field to xfrm states called inner_mode.  The existing
mode object is renamed to outer_mode.

This is the first part of an attempt to fix inter-family transforms.  As it
is we always use the outer family when determining which mode to use.  As a
result we may end up shoving IPv4 packets into netfilter6 and vice versa.

What we really want is to use the inner family for the first part of outbound
processing and the outer family for the second part.  For inbound processing
we'd use the opposite pairing.

I've also added a check to prevent silly combinations such as transport mode
with inter-family transforms.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Disallow combinations of RO and AH/ESP/IPCOMP
Herbert Xu [Thu, 18 Oct 2007 04:35:15 +0000 (21:35 -0700)]
[IPSEC]: Disallow combinations of RO and AH/ESP/IPCOMP

Combining RO and AH/ESP/IPCOMP does not make sense.  So this patch adds a
check in the state initialisation function to prevent this.

This allows us to safely remove the mode input function of RO since it
can never be called anymore.  Indeed, if somehow it does get called we'll
know about it through an OOPS instead of it slipping past silently.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Use the top IPv4 route's peer instead of the bottom
Herbert Xu [Thu, 18 Oct 2007 04:34:46 +0000 (21:34 -0700)]
[IPSEC]: Use the top IPv4 route's peer instead of the bottom

For IPv4 we were using the bottom route's peer instead of the top one.
This is wrong because the peer is only used by TCP to keep track of
information about the TCP destination address which certainly does not
live in the bottom route.

This patch fixes that which allows us to get rid of the family check
since the bottom route could be IPv6 while the top one must always
be IPv4.

I've also changed the other fields which are IPv4-specific to get the
info from the top route instead of potentially bogus data from the
bottom route.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Store afinfo pointer in xfrm_mode
Herbert Xu [Thu, 18 Oct 2007 04:33:12 +0000 (21:33 -0700)]
[IPSEC]: Store afinfo pointer in xfrm_mode

It is convenient to have a pointer from xfrm_state to address-specific
functions such as the output function for a family.  Currently the
address-specific policy code calls out to the xfrm state code to get
those pointers when we could get it in an easier way via the state
itself.

This patch adds an xfrm_state_afinfo to xfrm_mode (since they're
address-specific) and changes the policy code to use it.  I've also
added an owner field to do reference counting on the module providing
the afinfo even though it isn't strictly necessary today since IPv6
can't be unloaded yet.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Add missing BEET checks
Herbert Xu [Thu, 18 Oct 2007 04:31:50 +0000 (21:31 -0700)]
[IPSEC]: Add missing BEET checks

Currently BEET mode does not reinject the packet back into the stack
like tunnel mode does.  Since BEET should behave just like tunnel mode
this is incorrect.

This patch fixes this by introducing a flags field to xfrm_mode that
tells the IPsec code whether it should terminate and reinject the packet
back into the stack.

It then sets the flag for BEET and tunnel mode.

I've also added a number of missing BEET checks elsewhere where we check
whether a given mode is a tunnel or not.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Move type and mode map into xfrm_state.c
Herbert Xu [Thu, 18 Oct 2007 04:31:12 +0000 (21:31 -0700)]
[IPSEC]: Move type and mode map into xfrm_state.c

The type and mode maps are only used by SAs, not policies.  So it makes
sense to move them from xfrm_policy.c into xfrm_state.c.  This also allows
us to mark xfrm_get_type/xfrm_put_type/xfrm_get_mode/xfrm_put_mode as
static.

The only other change I've made in the move is to get rid of the casts
on the request_module call for types.  They're unnecessary because C
will promote them to ints anyway.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Fix length check in xfrm_parse_spi
Herbert Xu [Thu, 18 Oct 2007 04:30:34 +0000 (21:30 -0700)]
[IPSEC]: Fix length check in xfrm_parse_spi

Currently xfrm_parse_spi requires there to be 16 bytes for AH and ESP.
In contrived cases there may not actually be 16 bytes there since the
respective header sizes are less than that (8 and 12 currently).

This patch changes the test to use the actual header length instead of 16.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Move ip_summed zapping out of xfrm6_rcv_spi
Herbert Xu [Thu, 18 Oct 2007 04:30:07 +0000 (21:30 -0700)]
[IPSEC]: Move ip_summed zapping out of xfrm6_rcv_spi

Not every transform needs to zap ip_summed.  For example, a pure tunnel
mode encapsulation does not affect the hardware checksum at all.  In fact,
every algorithm (that needs this) other than AH6 already does its own
ip_summed zapping.

This patch moves the zapping into AH6 which is in line with what IPv4 does.

Possible future optimisation: Checksum the data as we copy them in IPComp.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Get nexthdr from caller in xfrm6_rcv_spi
Herbert Xu [Thu, 18 Oct 2007 04:29:25 +0000 (21:29 -0700)]
[IPSEC]: Get nexthdr from caller in xfrm6_rcv_spi

Currently xfrm6_rcv_spi gets the nexthdr value itself from the packet.
This means that we need to fix up the value in case we have a 4-on-6
tunnel.  Moving this logic into the caller simplifies things and allows
us to merge the code with IPv4.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Move tunnel parsing for IPv4 out of xfrm4_input
Herbert Xu [Thu, 18 Oct 2007 04:28:53 +0000 (21:28 -0700)]
[IPSEC]: Move tunnel parsing for IPv4 out of xfrm4_input

This patch moves the tunnel parsing for IPv4 out of xfrm4_input and into
xfrm4_tunnel.  This change is in line with what IPv6 does and will allow
us to merge the two input functions.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPSEC]: Fix pure tunnel modes involving IPv6
Herbert Xu [Thu, 18 Oct 2007 04:28:06 +0000 (21:28 -0700)]
[IPSEC]: Fix pure tunnel modes involving IPv6

I noticed that my recent patch broke 6-on-4 pure IPsec tunnels (the ones
that are only used for incompressible IPsec packets).  Subsequent reviews
show that I broke 6-on-6 pure tunnels more than three years ago and nobody
ever noticed. I suppose every must be testing 6-on-6 IPComp with large
pings which are very compressible :)

This patch fixes both cases.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPV6]: Cleanup snmp6_alloc_dev()
Pavel Emelyanov [Thu, 18 Oct 2007 04:25:32 +0000 (21:25 -0700)]
[IPV6]: Cleanup snmp6_alloc_dev()

This functions is never called with NULL or not setup argument,
so the checks inside are redundant.

Also, the return value is always -ENOMEM, so no need in
additional variable for this.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPV6]: Fix return type for snmp6_free_dev()
Pavel Emelyanov [Thu, 18 Oct 2007 04:23:43 +0000 (21:23 -0700)]
[IPV6]: Fix return type for snmp6_free_dev()

This call is essentially void.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[NET]: Fix the race between sk_filter_(de|at)tach and sk_clone()
Pavel Emelyanov [Thu, 18 Oct 2007 04:22:42 +0000 (21:22 -0700)]
[NET]: Fix the race between sk_filter_(de|at)tach and sk_clone()

The proposed fix is to delay the reference counter decrement
until the quiescent state pass. This will give sk_clone() a
chance to get the reference on the cloned filter.

Regular sk_filter_uncharge can happen from the sk_free() only
and there's no need in delaying the put - the socket is dead
anyway and is to be release itself.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[NET]: Cleanup the error path in sk_attach_filter
Pavel Emelyanov [Thu, 18 Oct 2007 04:22:17 +0000 (21:22 -0700)]
[NET]: Cleanup the error path in sk_attach_filter

The sk_filter_uncharge is called for error handling and
for releasing the former filter, but this will have to
be done in a bit different manner, so cleanup the error
path a bit.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[NET]: Move the filter releasing into a separate call
Pavel Emelyanov [Thu, 18 Oct 2007 04:21:51 +0000 (21:21 -0700)]
[NET]: Move the filter releasing into a separate call

This is done merely as a preparation for the fix.

The sk_filter_uncharge() unaccounts the filter memory and calls
the sk_filter_release(), which in turn decrements the refcount
anf frees the filter.

The latter function will be required separately.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[NET]: Introduce the sk_detach_filter() call
Pavel Emelyanov [Thu, 18 Oct 2007 04:21:26 +0000 (21:21 -0700)]
[NET]: Introduce the sk_detach_filter() call

Filter is attached in a separate function, so do the
same for filter detaching.

This also removes one variable sock_setsockopt().

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[MAC80211]: only honor IW_SCAN_THIS_ESSID in STA, IBSS, and AP modes
John W. Linville [Thu, 18 Oct 2007 04:16:16 +0000 (21:16 -0700)]
[MAC80211]: only honor IW_SCAN_THIS_ESSID in STA, IBSS, and AP modes

The previous IW_SCAN_THIS_ESSID patch left a hole allowing scan
requests on interfaces in inappropriate modes.

Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'fixes-davem' of master.kernel.org:/pub/scm/linux/kernel/git/linville...
David S. Miller [Thu, 18 Oct 2007 04:14:35 +0000 (21:14 -0700)]
Merge branch 'fixes-davem' of /linux/kernel/git/linville/wireless-2.6

13 years ago[INET]: Consolidate frag queues freeing
Pavel Emelyanov [Thu, 18 Oct 2007 02:48:26 +0000 (19:48 -0700)]
[INET]: Consolidate frag queues freeing

Since we now allocate the queues in inet_fragment.c, we
can safely free it in the same place. The ->destructor
callback thus becomes optional for inet_frags.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[INET]: Remove no longer needed ->equal callback
Pavel Emelyanov [Thu, 18 Oct 2007 02:47:56 +0000 (19:47 -0700)]
[INET]: Remove no longer needed ->equal callback

Since this callback is used to check for conflicts in
hashtable when inserting a newly created frag queue, we can
do the same by checking for matching the queue with the
argument, used to create one.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[INET]: Consolidate xxx_find() in fragment management
Pavel Emelyanov [Thu, 18 Oct 2007 02:47:21 +0000 (19:47 -0700)]
[INET]: Consolidate xxx_find() in fragment management

Here we need another callback ->match to check whether the
entry found in hash matches the key passed. The key used
is the same as the creation argument for inet_frag_create.

Yet again, this ->match is the same for netfilter and ipv6.
Running a frew steps forward - this callback will later
replace the ->equal one.

Since the inet_frag_find() uses the already consolidated
inet_frag_create() remove the xxx_frag_create from protocol
codes.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[INET]: Consolidate xxx_frag_create()
Pavel Emelyanov [Thu, 18 Oct 2007 02:46:47 +0000 (19:46 -0700)]
[INET]: Consolidate xxx_frag_create()

This one uses the xxx_frag_intern() and xxx_frag_alloc()
routines, which are already consolidated, so remove them
from protocol code (as promised).

The ->constructor callback is used to init the rest of
the frag queue and it is the same for netfilter and ipv6.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[INET]: Consolidate xxx_frag_alloc()
Pavel Emelyanov [Thu, 18 Oct 2007 02:45:23 +0000 (19:45 -0700)]
[INET]: Consolidate xxx_frag_alloc()

Just perform the kzalloc() allocation and setup common
fields in the inet_frag_queue(). Then return the result
to the caller to initialize the rest.

The inet_frag_alloc() may return NULL, so check the
return value before doing the container_of(). This looks
ugly, but the xxx_frag_alloc() will be removed soon.

The xxx_expire() timer callbacks are patches,
because the argument is now the inet_frag_queue, not
the protocol specific queue.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[INET]: Consolidate xxx_frag_intern
Pavel Emelyanov [Thu, 18 Oct 2007 02:44:34 +0000 (19:44 -0700)]
[INET]: Consolidate xxx_frag_intern

This routine checks for the existence of a given entry
in the hash table and inserts the new one if needed.

The ->equal callback is used to compare two frag_queue-s
together, but this one is temporary and will be removed
later. The netfilter code and the ipv6 one use the same
routine to compare frags.

The inet_frag_intern() always returns non-NULL pointer,
so convert the inet_frag_queue into protocol specific
one (with the container_of) without any checks.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[INET]: Omit double hash calculations in xxx_frag_intern
Pavel Emelyanov [Thu, 18 Oct 2007 02:43:37 +0000 (19:43 -0700)]
[INET]: Omit double hash calculations in xxx_frag_intern

Since the hash value is already calculated in xxx_find, we can
simply use it later. This is already done in netfilter code,
so make the same in ipv4 and ipv6.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[BR2684]: get rid of broken header code.
Stephen Hemminger [Thu, 18 Oct 2007 02:39:22 +0000 (19:39 -0700)]
[BR2684]: get rid of broken header code.

Recent header_ops change would break the following dead
code in br2684. Maintaining conditonal code in mainline is wrong.

"Do, or do not. There is no 'try.'"

Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[UMP]: header_ops conversion needed for non-ethernet drivers
Jeff Dike [Thu, 18 Oct 2007 02:35:04 +0000 (19:35 -0700)]
[UMP]: header_ops conversion needed for non-ethernet drivers

UML's two non-ethernet drivers need some header_ops conversion.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IRDA]: IrCOMM discovery indication simplification
Ryan Reading [Thu, 18 Oct 2007 02:34:11 +0000 (19:34 -0700)]
[IRDA]: IrCOMM discovery indication simplification

From: Ryan Reading <ryanr23@gmail.com>

Every IrCOMM socket is registered with the discovery subsystem, so we don't
need to loop over all of them for every discovery event. We just need to
do it for the registered IrCOMM socket.

Signed-off-by: Samuel Ortiz <samuel@sortiz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[DCCP]: fix link error with !CONFIG_SYSCTL
Ingo Molnar [Thu, 18 Oct 2007 02:33:06 +0000 (19:33 -0700)]
[DCCP]: fix link error with !CONFIG_SYSCTL

Do not define the sysctl_dccp_sync_ratelimit sysctl variable in the
CONFIG_SYSCTL dependent sysctl.c module - move it to input.c instead.

This fixes the following build bug:

 net/built-in.o: In function `dccp_check_seqno':
 input.c:(.text+0xbd859): undefined reference to `sysctl_dccp_sync_ratelimit'
 distcc[29953] ERROR: compile (null) on localhost failed
 make: *** [vmlinux] Error 1

Found via 'make randconfig' build testing.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[IPV6]: Fix memory leak in cleanup_ipv6_mibs()
Pavel Emelyanov [Thu, 18 Oct 2007 02:30:40 +0000 (19:30 -0700)]
[IPV6]: Fix memory leak in cleanup_ipv6_mibs()

The icmpv6msg mib statistics is not freed.

This is almost not critical for current kernel, since ipv6
module is unloadable, but this can happen on load error and
will happen every time we stop the network namespace (when
we have one, of course).

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: David L Stevens <dlstevens@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[BNX2]: Update version to 1.6.8.
Michael Chan [Thu, 18 Oct 2007 02:26:28 +0000 (19:26 -0700)]
[BNX2]: Update version to 1.6.8.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[BNX2]: Fix Serdes WoL bug.
Michael Chan [Thu, 18 Oct 2007 02:26:15 +0000 (19:26 -0700)]
[BNX2]: Fix Serdes WoL bug.

The bug is in the code in bnx2_set_power_state() that assumes copper
devices when setting up WoL.  This is no longer true after adding WoL
support for Serdes devices.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years ago[BNX2]: Update 5709 firmware to 3.7.1.
Michael Chan [Thu, 18 Oct 2007 02:25:27 +0000 (19:25 -0700)]
[BNX2]: Update 5709 firmware to 3.7.1.

This firmware update fixes a problem running with IPMI management
firmware.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Wed, 17 Oct 2007 22:13:42 +0000 (15:13 -0700)]
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] fix non-numa build

13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh...
Linus Torvalds [Wed, 17 Oct 2007 22:05:58 +0000 (15:05 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/ericvh/v9fs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
  9p: remove sysctl
  9p: fix bad kconfig cross-dependency
  9p: soften invalidation in loose_mode
  9p: attach-per-user
  9p: rename uid and gid parameters
  9p: define session flags
  9p: Make transports dynamic

13 years ago[IA64] fix non-numa build
Andrew Morton [Wed, 17 Oct 2007 21:28:38 +0000 (14:28 -0700)]
[IA64] fix non-numa build

arch/ia64/kernel/machine_kexec.c: In function `arch_crash_save_vmcoreinfo':
arch/ia64/kernel/machine_kexec.c:131: error: `pgdat_list' undeclared (first use in this function)
arch/ia64/kernel/machine_kexec.c:131: error: (Each undeclared identifier is reported only once
arch/ia64/kernel/machine_kexec.c:131: error: for each function it appears in.)
arch/ia64/kernel/machine_kexec.c:134: error: `node_memblk' undeclared (first use in this function)
arch/ia64/kernel/machine_kexec.c:135: error: `NR_NODE_MEMBLKS' undeclared (first use in this function)
arch/ia64/kernel/machine_kexec.c:136: error: invalid application of `sizeof' to incomplete type `node_memblk_s'
arch/ia64/kernel/machine_kexec.c:137: error: dereferencing pointer to incomplete type
arch/ia64/kernel/machine_kexec.c:138: error: dereferencing pointer to incomplete type
make[1]: *** [arch/ia64/kernel/machine_kexec.o] Error 1

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>
13 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
Linus Torvalds [Wed, 17 Oct 2007 21:12:44 +0000 (14:12 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/drzeus/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
  net: libertas sdio driver
  mmc: at91_mci: cleanup: use MCI_ERRORS
  mmc: possible leak in mmc_read_ext_csd

13 years agonet: libertas sdio driver
Pierre Ossman [Wed, 17 Oct 2007 20:24:24 +0000 (22:24 +0200)]
net: libertas sdio driver

Add driver for Marvell's Libertas 8385 and 8686 wifi chips.

Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Acked-by: Dan Williams <dcbw@redhat.com>
13 years agoMerge ssh://master.kernel.org/pub/scm/linux/kernel/git/tglx/linux-2.6-x86
Linus Torvalds [Wed, 17 Oct 2007 20:13:16 +0000 (13:13 -0700)]
Merge ssh:///linux/kernel/git/tglx/linux-2.6-x86

* ssh://master.kernel.org/pub/scm/linux/kernel/git/tglx/linux-2.6-x86: (114 commits)
  x86: delete vsyscall files during make clean
  kbuild: fix typo SRCARCH in find_sources
  x86: fix kernel rebuild due to vsyscall fallout
  .gitignore update for x86 arch
  x86: unify include/asm/debugreg_32/64.h
  x86: unify include/asm/unwind_32/64.h
  x86: unify include/asm/types_32/64.h
  x86: unify include/asm/tlb_32/64.h
  x86: unify include/asm/siginfo_32/64.h
  x86: unify include/asm/bug_32/64.h
  x86: unify include/asm/mman_32/64.h
  x86: unify include/asm/agp_32/64.h
  x86: unify include/asm/kdebug_32/64.h
  x86: unify include/asm/ioctls_32/64.h
  x86: unify include/asm/floppy_32/64.h
  x86: apply missing DMA/OOM prevention to floppy_32.h
  x86: unify include/asm/cache_32/64.h
  x86: unify include/asm/cache_32/64.h
  x86: unify include/asm/dmi_32/64.h
  x86: unify include/asm/delay_32/64.h
  ...

13 years agox86: delete vsyscall files during make clean
Sam Ravnborg [Wed, 17 Oct 2007 20:06:30 +0000 (22:06 +0200)]
x86: delete vsyscall files during make clean

make clean failed to delete a few files in
x86/kernel. This is because kbuild does not
see the correct/full kernel/Makefile.

As a workaround until the Makefiles are merged specify
the files to be deleted in the common Makefile.

Reported by Mike Galbraith <efault@gmx.de>

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agokbuild: fix typo SRCARCH in find_sources
Yinghai Lu [Tue, 16 Oct 2007 07:28:06 +0000 (00:28 -0700)]
kbuild: fix typo SRCARCH in find_sources

otherwise get the two copy file list in SRCARCH

for cscope:
C symbol: start_kernel

  File           Function            Line
0 proto.h        <global>              11 extern void start_kernel(void );
1 start_kernel.h <global>              10 extern asmlinkage void __init start_kernel(void );
2 head32.c       i386_start_kernel     37 start_kernel();
3 head32.c       i386_start_kernel     37 start_kernel();
4 head64.c       x86_64_start_kernel   85 start_kernel();
5 head64.c       x86_64_start_kernel   85 start_kernel();
6 head_32.S      options              199 cmpb $0,%cl #the first CPU calls start_kernel
7 head_32.S      options              199 cmpb $0,%cl #the first CPU calls start_kernel
8 enlighten.c    xen_start_kernel    1145 start_kernel();
9 enlighten.c    xen_start_kernel    1145 start_kernel();
a lguest.c       lguest_init         1095 start_kernel();
b main.c         start_kernel         513 asmlinkage void __init start_kernel(void )

after the patch:
C symbol: start_kernel

  File           Function            Line
0 proto.h        <global>              11 extern void start_kernel(void );
1 start_kernel.h <global>              10 extern asmlinkage void __init start_kernel(void );
2 head32.c       i386_start_kernel     37 start_kernel();
3 head64.c       x86_64_start_kernel   85 start_kernel();
4 head_32.S      options              199 cmpb $0,%cl #the first CPU calls start_kernel
5 enlighten.c    xen_start_kernel    1145 start_kernel();
6 lguest.c       lguest_init         1095 start_kernel();
7 main.c         start_kernel         513 asmlinkage void __init start_kernel(void )

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years ago9p: remove sysctl
Eric Van Hensbergen [Wed, 17 Oct 2007 19:35:15 +0000 (14:35 -0500)]
9p: remove sysctl

A sysctl method was added to enable and disable debugging levels.  After
further review, it was decided that there are better approaches to doing this
and the sysctl methodology isn't really desirable.  This patch removes the
sysctl code from 9p.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years agommc: at91_mci: cleanup: use MCI_ERRORS
Andrew Victor [Wed, 17 Oct 2007 09:53:40 +0000 (11:53 +0200)]
mmc: at91_mci: cleanup: use MCI_ERRORS

A small MMC driver cleanup.

Use the defined AT91_MCI_ERRORS in at91_mci_completed_command() instead
of specifying all the error bits individually.

Signed-off-by: Andrew Victor <andrew@sanpeople.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
13 years agommc: possible leak in mmc_read_ext_csd
Florin Malita [Sat, 13 Oct 2007 16:27:20 +0000 (12:27 -0400)]
mmc: possible leak in mmc_read_ext_csd

The exception path associated with an invalid ext_csd_struct returns
without freeing ext_csd.

Coverity CID 1909.

Signed-off-by: Florin Malita
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
13 years ago9p: fix bad kconfig cross-dependency
Eric Van Hensbergen [Wed, 17 Oct 2007 19:31:07 +0000 (14:31 -0500)]
9p: fix bad kconfig cross-dependency

This patch moves transport dynamic registration and matching to the net
module to prevent a bad Kconfig dependency between the net and fs 9p modules.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years ago9p: soften invalidation in loose_mode
Eric Van Hensbergen [Wed, 17 Oct 2007 19:31:07 +0000 (14:31 -0500)]
9p: soften invalidation in loose_mode

Loose mode in 9p utilizes the page cache without respecting coherency with
the server.  Any writes previously invaldiated the entire mapping for a file.
This patch softens the behavior to only invalidate the region of the actual
write.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years ago9p: attach-per-user
Latchesar Ionkov [Wed, 17 Oct 2007 19:31:07 +0000 (14:31 -0500)]
9p: attach-per-user

The 9P2000 protocol requires the authentication and permission checks to be
done in the file server. For that reason every user that accesses the file
server tree has to authenticate and attach to the server separately.
Multiple users can share the same connection to the server.

Currently v9fs does a single attach and executes all I/O operations as a
single user. This makes using v9fs in multiuser environment unsafe as it
depends on the client doing the permission checking.

This patch improves the 9P2000 support by allowing every user to attach
separately. The patch defines three modes of access (new mount option
'access'):

- attach-per-user (access=user) (default mode for 9P2000.u)
 If a user tries to access a file served by v9fs for the first time, v9fs
 sends an attach command to the server (Tattach) specifying the user. If
 the attach succeeds, the user can access the v9fs tree.
 As there is no uname->uid (string->integer) mapping yet, this mode works
 only with the 9P2000.u dialect.

- allow only one user to access the tree (access=<uid>)
 Only the user with uid can access the v9fs tree. Other users that attempt
 to access it will get EPERM error.

- do all operations as a single user (access=any) (default for 9P2000)
 V9fs does a single attach and all operations are done as a single user.
 If this mode is selected, the v9fs behavior is identical with the current
 one.

Signed-off-by: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years ago9p: rename uid and gid parameters
Latchesar Ionkov [Wed, 17 Oct 2007 19:31:07 +0000 (14:31 -0500)]
9p: rename uid and gid parameters

Change the names of 'uid' and 'gid' parameters to the more appropriate
'dfltuid' and 'dfltgid'.  This also sets the default uid/gid to -2
(aka nfsnobody)

Signed-off-by: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years ago9p: define session flags
Latchesar Ionkov [Wed, 17 Oct 2007 19:31:07 +0000 (14:31 -0500)]
9p: define session flags

Create more general flags field in the v9fs_session_info struct and move the
'extended' flag as a bit in the flags.

Signed-off-by: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years ago9p: Make transports dynamic
Eric Van Hensbergen [Wed, 17 Oct 2007 19:31:07 +0000 (14:31 -0500)]
9p: Make transports dynamic

This patch abstracts out the interfaces to underlying transports so that
new transports can be added as modules.  This should also allow kernel
configuration of transports without ifdef-hell.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
13 years agox86: fix kernel rebuild due to vsyscall fallout
Sam Ravnborg [Wed, 17 Oct 2007 19:16:33 +0000 (21:16 +0200)]
x86: fix kernel rebuild due to vsyscall fallout

Fix rebuild of kernel when there is no changes.
This happened for i386.
Using make V=2 hinted that the output files were
not assigned to targets - fixed by this patch.

Reported by: Boaz Harrosh <bharrosh@panasas.com>

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years ago.gitignore update for x86 arch
Denis V. Lunev [Tue, 16 Oct 2007 07:22:21 +0000 (11:22 +0400)]
.gitignore update for x86 arch

This patch:
- makes .gitignore files visible to git
- makes arch/x86/kernel/vsyscall_32.lds and arch/i386/boot invisible

Signed-off-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agoMerge branch 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block
Linus Torvalds [Wed, 17 Oct 2007 18:56:06 +0000 (11:56 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block:
  [BLOCK] blk_rq_map_sg() next_sg fixup
  Revert "[SCSI] Remove full sg table memset()"

13 years agox86: unify include/asm/debugreg_32/64.h
Thomas Gleixner [Wed, 17 Oct 2007 18:35:37 +0000 (20:35 +0200)]
x86: unify include/asm/debugreg_32/64.h

Almost identical except for the extra DR_LEN_8 and the different
DR_CONTROL_RESERVED defines.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Conflicts:

include/asm-x86/Kbuild

13 years agox86: unify include/asm/unwind_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:21 +0000 (23:28 +0200)]
x86: unify include/asm/unwind_32/64.h

32bit has an extra UNW_FP define, which does not hurt.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/types_32/64.h
Thomas Gleixner [Wed, 17 Oct 2007 18:32:07 +0000 (20:32 +0200)]
x86: unify include/asm/types_32/64.h

Mostly the same. Make the few exceptions conditional.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Conflicts:

include/asm-x86/types_32.h

13 years agox86: unify include/asm/tlb_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:21 +0000 (23:28 +0200)]
x86: unify include/asm/tlb_32/64.h

Same file, except for whitespace, comment formatting.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/siginfo_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:21 +0000 (23:28 +0200)]
x86: unify include/asm/siginfo_32/64.h

Same file, except for the 64bit PREAMBLE_SIZE define.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/bug_32/64.h
Thomas Gleixner [Wed, 17 Oct 2007 15:19:30 +0000 (17:19 +0200)]
x86: unify include/asm/bug_32/64.h

Same file, except for whitespace, comment formatting and the
.long/.quad delta which can be solved by a define.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/mman_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:21 +0000 (23:28 +0200)]
x86: unify include/asm/mman_32/64.h

Same file, except for the extra 64bit MAP_32BIT define, which does not hurt
for 32 bit compiles.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/agp_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/agp_32/64.h

The 32bit D(n) debug addon can be made exclusive for 32 bit compiles.
Otherwise all the same.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/kdebug_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/kdebug_32/64.h

The 64 bit variant has additional function prototypes which do no harm
for 32 bit.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/ioctls_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/ioctls_32/64.h

Same file, except for whitespace and comment formatting.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/floppy_32/64.h
Thomas Gleixner [Wed, 17 Oct 2007 18:24:56 +0000 (20:24 +0200)]
x86: unify include/asm/floppy_32/64.h

Same file, except for whitespace, comment formatting and:

32-bit: if((unsigned int) addr >= (unsigned int) high_memory)
64-bit: if((unsigned long) addr >= (unsigned long) high_memory)

where the latter can be used safely for both.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Conflicts:
include/asm-x86/floppy_32.h
include/asm-x86/floppy_64.h

13 years agox86: apply missing DMA/OOM prevention to floppy_32.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: apply missing DMA/OOM prevention to floppy_32.h

commit 554d284ba90bc2306c31e5363789f05c320969c3 added _GPF_NORETRY
to floppy_64.h to prevent OOM killer on floppy DMA allocations.

Apply the same to the 32 bit variant.

Found during the attempt to unify the _32/_64 variants. Seperate commit
to document the resulting code change.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/cache_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/cache_32/64.h

Same file, except for whitespace, comment formatting and the two variants
of fb_is_primary_device()

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/cache_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/cache_32/64.h

Same file, except for whitespace, comment formatting and:

32-bit: unsigned long *virt_addr = va;
64-bit: unsigned int *virt_addr = va;

Both can be safely replaced by:
u32 i, *virt_addr = va;

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/dmi_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/dmi_32/64.h

Unification, so we have these things in one file.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/delay_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/delay_32/64.h

Same file, except for whitespace, comment formatting and the extra
function prototype usc_tsc_delay() in _32.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/cache_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/cache_32/64.h

Same file, except for whitespace, comment formatting and the extra
defines in _64, which are conditional on VSMP anyway.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/cacheflush_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:20 +0000 (23:28 +0200)]
x86: unify include/asm/cacheflush_32/64.h

Same file, except for whitespace, comment formatting and the extra
DEBUG_PAGE_ALLOC function in _32.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/auxvec_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:19 +0000 (23:28 +0200)]
x86: unify include/asm/auxvec_32/64.h

Same file, except for whitespace, comment formatting and the
AT_SYSINFO define for 32bit

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify include/asm/agp_32/64.h
Thomas Gleixner [Mon, 15 Oct 2007 21:28:19 +0000 (23:28 +0200)]
x86: unify include/asm/agp_32/64.h

Same file, except for whitespace, comment formatting and the
usage of wbinvd() instead of asm volatile("wbinvd":::"memory"), which is
the same.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: unify some more trivial include/asm-x86/ 32/64 variants
Thomas Gleixner [Mon, 15 Oct 2007 11:57:47 +0000 (13:57 +0200)]
x86: unify some more trivial include/asm-x86/ 32/64 variants

Scripted unification.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: merge some trivially mergeable headers
Roland Dreier [Mon, 15 Oct 2007 11:57:46 +0000 (13:57 +0200)]
x86: merge some trivially mergeable headers

Merge errno.h, resource.h, rtc.h, sections.h, serial.h and sockios.h,
where i386 and x86_64 have no or only trivial comment/include guard
differences.

Build tested on both 32-bit and 64-bit, and booted on 64-bit.

[tglx: fixup Kbuild as well]

Signed-off-by: Roland Dreier <roland@digitalvampire.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: trivial header merges
Brian Gerst [Mon, 15 Oct 2007 11:57:46 +0000 (13:57 +0200)]
x86: trivial header merges

Merge 32/64-bit headers that simply redirect to asm-generic

[tglx: fixup Kbuild as well]

Signed-off-by: Brian Gerst <bgerst@didntduck.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: vdso linker script cleanup
Roland McGrath [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
x86: vdso linker script cleanup

I can't see the reason ".  = VDSO_PRELINK + 0x900;" was ever there in
the linker script for the x86_64 vDSO.  I can't find anything that
depends on this magic offset, or that should care at all about the
particular location of of the .data section (all from vvar.c) in the
vDSO image.  If it is really desireable to place .data at 0x900, then it
should be after all the other sections so they fill in the space up to
0x900.

This removes the 0x900 magic and cleans up the output sections generally
in the vDSO linker script.  This saves a few hundred bytes in the size
of the vDSO file, bringing it back well under 4kb total so that its vma
only needs one page.

Signed-off-by: Roland McGrath <roland@redhat.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: convert mm_context_t semaphore to a mutex
Luiz Fernando N. Capitulino [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
x86: convert mm_context_t semaphore to a mutex

convert mm_context_t semaphore to a mutex.

Signed-off-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: multi-byte single instruction NOPs
Jan Beulich [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
x86: multi-byte single instruction NOPs

Add support for and use the multi-byte NOPs recently documented to be
available on all PentiumPro and later processors.

This patch only applies cleanly on top of the "x86: misc.
constifications" patch sent earlier.

[ tglx: arch/x86 adaptation ]

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
 arch/x86/kernel/alternative.c  |   23 ++++++++++++++++++++++-
 include/asm-x86/processor_32.h |   22 ++++++++++++++++++++++
 include/asm-x86/processor_64.h |   22 ++++++++++++++++++++++
 3 files changed, 66 insertions(+), 1 deletion(-)

13 years agox86: remove duplicated vsyscall nsec update
Andi Kleen [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
x86: remove duplicated vsyscall nsec update

Spotted by Chuck Ebbert

[ tglx: arch/x86 adaptation ]

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agoi386: remove unnecessary code
Ingo Molnar [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
i386: remove unnecessary code

Oleg Nesterov pointed out that the set_fs() calls in setup_frame()
and setup_rt_frame() were superfluous.

[ tglx: arch/x86 adaptation ]

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Chuck Ebbert <76306.1226@compuserve.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: convert mm_context_t semaphore to a mutex
Luiz Fernando N. Capitulino [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
x86: convert mm_context_t semaphore to a mutex

[ tglx: arch/x86 adaptation ]

Signed-off-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: Use string instruction memcpy on AMD Fam11h
Andi Kleen [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
x86: Use string instruction memcpy on AMD Fam11h

[ tglx: arch/x86 adaptation ]

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: make io_apic not connected pin print complete
Yinghai Lu [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
x86: make io_apic not connected pin print complete

Normally we will have two segment not connected pin pin0, and pin after
15...

So we need to print out "not connected\n" for previous segment, before
printing out connected pins info...

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
13 years agoi386: print better early fault info
Ingo Molnar [Wed, 17 Oct 2007 16:04:41 +0000 (18:04 +0200)]
i386: print better early fault info

improve early fault output.

old format:

 Int 14: CR2 010001e3  err 00000002  EIP c011f2f9  CS 00000060  flags 00010046
 Stack: c073695e c0791c10 00000000 ffffffff 00000000 01000000 00001000 c0791c10

new format:

 BUG: Int 14: CR2 010001e3
      EDI c1000000  ESI c0693c10  EBP c0637f9c  ESP c0637f08
      EBX 00000000  EDX 0000000e  ECX 00000000  EAX 010001e3
      err 00000002  EIP c0123119   CS 00000060  flg 00010046
 Stack: c064d589 c0693000 00000000 c0637f60 00c001e3 01000000 00038000 00000163
        00000000 00000163 00000000 ffffffff 00038000 00000000 00000000 00001000
        00001000 00000000 c0637f88 c06509be c0a2ae60 00001000 00001000 00000000

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: also show non-zero IRQ counts for vectors that currently don't have a handler
Jan Beulich [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: also show non-zero IRQ counts for vectors that currently don't have a handler

It doesn't seem to make sense to hide these, even if their counts
can't change at the point in time they're being displayed.

[ tglx: arch/x86 adaptation ]

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: expand /proc/interrupts to include missing vectors, v2
Joe Korty [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: expand /proc/interrupts to include missing vectors, v2

Add missing IRQs and IRQ descriptions to /proc/interrupts.

/proc/interrupts is most useful when it displays every IRQ vector in use by
the system, not just those somebody thought would be interesting.

This patch inserts the following vector displays to the i386 and x86_64
platforms, as appropriate:

rescheduling interrupts
TLB flush interrupts
function call interrupts
thermal event interrupts
threshold interrupts
spurious interrupts

A threshold interrupt occurs when ECC memory correction is occuring at too
high a frequency.  Thresholds are used by the ECC hardware as occasional
ECC failures are part of normal operation, but long sequences of ECC
failures usually indicate a memory chip that is about to fail.

Thermal event interrupts occur when a temperature threshold has been
exceeded for some CPU chip.  IIRC, a thermal interrupt is also generated
when the temperature drops back to a normal level.

A spurious interrupt is an interrupt that was raised then lowered by the
device before it could be fully processed by the APIC.  Hence the apic sees
the interrupt but does not know what device it came from.  For this case
the APIC hardware will assume a vector of 0xff.

Rescheduling, call, and TLB flush interrupts are sent from one CPU to
another per the needs of the OS.  Typically, their statistics would be used
to discover if an interrupt flood of the given type has been occuring.

AK: merged v2 and v4 which had some more tweaks
AK: replace Local interrupts with Local timer interrupts
AK: Fixed description of interrupt types.

[ tglx: arch/x86 adaptation ]
[ mingo: small cleanup ]

Signed-off-by: Joe Korty <joe.korty@ccur.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Tim Hockin <thockin@hockin.org>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agoi386: clean up oops/bug reports
Pavel Emelyanov [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
i386: clean up oops/bug reports

Typically the oops first lines look like this:

BUG: unable to handle kernel NULL pointer dereference at virtual address 00000000
 printing eip:
c049dfbd
*pde = 00000000
Oops: 0002 [#1]
PREEMPT SMP
...

Such output is gained with some ugly if (!nl) printk("\n"); code and
besides being a waste of lines, this is also annoying to read. The
following output looks better (and it is how it looks on x86_64):

BUG: unable to handle kernel NULL pointer dereference at virtual address 00000000
printing eip: c049dfbd *pde = 00000000
Oops: 0002 [#1] PREEMPT SMP
...

[ tglx: arch/x86 adaptation ]

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: print info about late C1E detection on 32bit as well
Thomas Gleixner [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: print info about late C1E detection on 32bit as well

Some BIOSes set the C1E flag only on the second core. Print a warning so
the Firmware Toolkit can check for it.

mingo: fix C1E build bug on 32-bit

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: call cache_add_dev() from cache_sysfs_init() explicitly
Satyam Sharma [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: call cache_add_dev() from cache_sysfs_init() explicitly

Call cache_add_dev() from cache_sysfs_init() explicitly, instead of
referencing the CPU notifier callback directly from generic startup
code. Looks cleaner (to me at least) this way, and also makes it
possible to use other tricks to replace __cpuinit{data} annotations, as
recently discussed on this list.

Signed-off-by: Satyam Sharma <satyam@infradead.org>
Cc: Andi Kleen <ak@suse.de>
Cc: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
13 years agox86: fix dmi const-ify fallout
Thomas Gleixner [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: fix dmi const-ify fallout

The dmi const-ification missed acer_cpu_freq_pst. Fix it.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: unify include/asm/ldt_32/64.h
Thomas Gleixner [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: unify include/asm/ldt_32/64.h

The additional struct member of user_desc can be made conditional for
64 bit compiles.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: unify include/asm/ptrace-abi_32/64.h
Thomas Gleixner [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: unify include/asm/ptrace-abi_32/64.h

Aside of the register defines the content can be shared.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: unify include/asm/mce_32/64.h
Thomas Gleixner [Wed, 17 Oct 2007 16:04:40 +0000 (18:04 +0200)]
x86: unify include/asm/mce_32/64.h

Merge the files together.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>