profile/ivi/kernel-x86-ivi.git
12 years agoMerge branch 'master' of git://1984.lsi.us.es/net-next
David S. Miller [Thu, 8 Mar 2012 06:27:56 +0000 (22:27 -0800)]
Merge branch 'master' of git://1984.lsi.us.es/net-next

12 years agotcp: md5: correct a RCU lockdep splat
Eric Dumazet [Wed, 7 Mar 2012 04:45:43 +0000 (04:45 +0000)]
tcp: md5: correct a RCU lockdep splat

commit a8afca0329 (tcp: md5: protects md5sig_info with RCU) added a
lockdep splat in tcp_md5_do_lookup() in case a timer fires a tcp
retransmit.

At this point, socket lock is owned by the sofirq handler, not the user,
so we should adjust a bit the lockdep condition, as we dont hold
rcu_read_lock().

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Reported-by: Valdis Kletnieks <valdis.kletnieks@vt.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetfilter: xt_LOG: add __printf() to sb_add()
Eric Dumazet [Thu, 1 Mar 2012 02:56:59 +0000 (02:56 +0000)]
netfilter: xt_LOG: add __printf() to sb_add()

Helps to find format mismatches at compile time

Suggested-by: David Laight <David.Laight@ACULAB.COM>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
12 years agonetfilter: xt_CT: allow to attach timeout policy + glue code
Pablo Neira Ayuso [Wed, 29 Feb 2012 01:19:19 +0000 (02:19 +0100)]
netfilter: xt_CT: allow to attach timeout policy + glue code

This patch allows you to attach the timeout policy via the
CT target, it adds a new revision of the target to ensure
backward compatibility. Moreover, it also contains the glue
code to stick the timeout object defined via nfnetlink_cttimeout
to the given flow.

Example usage (it requires installing the nfct tool and
libnetfilter_cttimeout):

1) create the timeout policy:

 nfct timeout add tcp-policy0 inet tcp \
established 1000 close 10 time_wait 10 last_ack 10

2) attach the timeout policy to the packet:

 iptables -I PREROUTING -t raw -p tcp -j CT --timeout tcp-policy0

You have to install the following user-space software:

a) libnetfilter_cttimeout:
   git://git.netfilter.org/libnetfilter_cttimeout

b) nfct:
   git://git.netfilter.org/nfct

You also have to get iptables with -j CT --timeout support.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_ct_ext: add timeout extension
Pablo Neira Ayuso [Tue, 28 Feb 2012 22:36:48 +0000 (23:36 +0100)]
netfilter: nf_ct_ext: add timeout extension

This patch adds the timeout extension, which allows you to attach
specific timeout policies to flows.

This extension is only used by the template conntrack.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: add cttimeout infrastructure for fine timeout tuning
Pablo Neira Ayuso [Tue, 28 Feb 2012 18:13:48 +0000 (19:13 +0100)]
netfilter: add cttimeout infrastructure for fine timeout tuning

This patch adds the infrastructure to add fine timeout tuning
over nfnetlink. Now you can use the NFNL_SUBSYS_CTNETLINK_TIMEOUT
subsystem to create/delete/dump timeout objects that contain some
specific timeout policy for one flow.

The follow up patches will allow you attach timeout policy object
to conntrack via the CT target and the conntrack extension
infrastructure.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_conntrack: pass timeout array to l4->new and l4->packet
Pablo Neira Ayuso [Tue, 28 Feb 2012 17:23:31 +0000 (18:23 +0100)]
netfilter: nf_conntrack: pass timeout array to l4->new and l4->packet

This patch defines a new interface for l4 protocol trackers:

unsigned int *(*get_timeouts)(struct net *net);

that is used to return the array of unsigned int that contains
the timeouts that will be applied for this flow. This is passed
to the l4proto->new(...) and l4proto->packet(...) functions to
specify the timeout policy.

This interface allows per-net global timeout configuration
(although only DCCP supports this by now) and it will allow
custom custom timeout configuration by means of follow-up
patches.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_ct_gre: add unsigned int array to define timeouts
Pablo Neira Ayuso [Tue, 28 Feb 2012 17:56:39 +0000 (18:56 +0100)]
netfilter: nf_ct_gre: add unsigned int array to define timeouts

This patch adds an array to define the default GRE timeouts.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_ct_tcp: move retransmission and unacknowledged timeout to array
Pablo Neira Ayuso [Tue, 28 Feb 2012 15:56:31 +0000 (16:56 +0100)]
netfilter: nf_ct_tcp: move retransmission and unacknowledged timeout to array

This patch moves the retransmission and unacknowledged timeouts
to the tcp_timeouts array. This change is required by follow-up
patches.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_ct_udp[lite]: convert UDP[lite] timeouts to array
Pablo Neira Ayuso [Tue, 28 Feb 2012 01:23:28 +0000 (02:23 +0100)]
netfilter: nf_ct_udp[lite]: convert UDP[lite] timeouts to array

Use one array to store the UDP timeouts instead of two variables.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ctnetlink: fix lockep splats
Hans Schillstrom [Mon, 5 Mar 2012 02:24:29 +0000 (02:24 +0000)]
netfilter: ctnetlink: fix lockep splats

net/netfilter/nf_conntrack_proto.c:70 suspicious rcu_dereference_check() usage!

other info that might help us debug this:

rcu_scheduler_active = 1, debug_locks = 0
3 locks held by conntrack/3235:
nfnl_lock+0x17/0x20
netlink_dump+0x32/0x240
ctnetlink_dump_table+0x3e/0x170 [nf_conntrack_netlink]

stack backtrace:
Pid: 3235, comm: conntrack Tainted: G W  3.2.0+ #511
Call Trace:
[<ffffffff8108ce45>] lockdep_rcu_suspicious+0xe5/0x100
[<ffffffffa00ec6e1>] __nf_ct_l4proto_find+0x81/0xb0 [nf_conntrack]
[<ffffffffa0115675>] ctnetlink_fill_info+0x215/0x5f0 [nf_conntrack_netlink]
[<ffffffffa0115dc1>] ctnetlink_dump_table+0xd1/0x170 [nf_conntrack_netlink]
[<ffffffff815fbdbf>] netlink_dump+0x7f/0x240
[<ffffffff81090f9d>] ? trace_hardirqs_on+0xd/0x10
[<ffffffff815fd34f>] netlink_dump_start+0xdf/0x190
[<ffffffffa0111490>] ? ctnetlink_change_nat_seq_adj+0x160/0x160 [nf_conntrack_netlink]
[<ffffffffa0115cf0>] ? ctnetlink_get_conntrack+0x2a0/0x2a0 [nf_conntrack_netlink]
[<ffffffffa0115ad9>] ctnetlink_get_conntrack+0x89/0x2a0 [nf_conntrack_netlink]
[<ffffffff81603a47>] nfnetlink_rcv_msg+0x467/0x5f0
[<ffffffff81603a7c>] ? nfnetlink_rcv_msg+0x49c/0x5f0
[<ffffffff81603922>] ? nfnetlink_rcv_msg+0x342/0x5f0
[<ffffffff81071b21>] ? get_parent_ip+0x11/0x50
[<ffffffff816035e0>] ? nfnetlink_subsys_register+0x60/0x60
[<ffffffff815fed49>] netlink_rcv_skb+0xa9/0xd0
[<ffffffff81603475>] nfnetlink_rcv+0x15/0x20
[<ffffffff815fe70e>] netlink_unicast+0x1ae/0x1f0
[<ffffffff815fea16>] netlink_sendmsg+0x2c6/0x320
[<ffffffff815b2a87>] sock_sendmsg+0x117/0x130
[<ffffffff81125093>] ? might_fault+0x53/0xb0
[<ffffffff811250dc>] ? might_fault+0x9c/0xb0
[<ffffffff81125093>] ? might_fault+0x53/0xb0
[<ffffffff815b5991>] ? move_addr_to_kernel+0x71/0x80
[<ffffffff815b644e>] sys_sendto+0xfe/0x130
[<ffffffff815b5c94>] ? sys_bind+0xb4/0xd0
[<ffffffff817a8a0e>] ? retint_swapgs+0xe/0x13
[<ffffffff817afcd2>] system_call_fastpath+0x16/0x1b

Reported-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Hans Schillstrom <hans.schillstrom@ericsson.com>
12 years agonetfilter: xt_LOG: fix bogus extra layer-4 logging information
Richard Weinberger [Thu, 1 Mar 2012 11:39:15 +0000 (11:39 +0000)]
netfilter: xt_LOG: fix bogus extra layer-4 logging information

In 16059b5 netfilter: merge ipt_LOG and ip6_LOG into xt_LOG, we have
merged ipt_LOG and ip6t_LOG.

However:

IN=wlan0 OUT= MAC=xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx:xx
SRC=213.150.61.61 DST=192.168.1.133 LEN=40 TOS=0x00 PREC=0x00 TTL=117
ID=10539 DF PROTO=TCP SPT=80 DPT=49013 WINDOW=0 RES=0x00 ACK RST
URGP=0 PROTO=UDPLITE SPT=80 DPT=49013 LEN=45843 PROTO=ICMP TYPE=0
       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Several missing break in the code led to including bogus layer-4
information. This patch fixes this problem.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: remove ipt_SAME.h and ipt_realm.h
WANG Cong [Wed, 29 Feb 2012 03:05:41 +0000 (04:05 +0100)]
netfilter: remove ipt_SAME.h and ipt_realm.h

These two headers are not required anymore, they have been
replaced by xt_SAME.h and xt_realm.h.

Florian Westphal pointed out this.

Cc: "David S. Miller" <davem@davemloft.net>
Cc: Florian Westphal <fw@strlen.de>
Signed-off-by: WANG Cong <xiyou.wangcong@gmail.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_ct_ecache: refactor nf_ct_deliver_cached_events
Tony Zelenoff [Wed, 22 Feb 2012 06:48:01 +0000 (10:48 +0400)]
netfilter: nf_ct_ecache: refactor nf_ct_deliver_cached_events

* identation lowered
* some CPU cycles saved at delayed item variable initialization

Signed-off-by: Tony Zelenoff <antonz@parallels.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: nf_ct_ecache: trailing whitespace removed
Tony Zelenoff [Wed, 22 Feb 2012 06:47:59 +0000 (10:47 +0400)]
netfilter: nf_ct_ecache: trailing whitespace removed

Signed-off-by: Tony Zelenoff <antonz@parallels.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: merge ipt_LOG and ip6_LOG into xt_LOG
Richard Weinberger [Fri, 10 Feb 2012 22:10:52 +0000 (23:10 +0100)]
netfilter: merge ipt_LOG and ip6_LOG into xt_LOG

ipt_LOG and ip6_LOG have a lot of common code, merge them
to reduce duplicate code.

Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ctnetlink: allow to set expectfn for expectations
Pablo Neira Ayuso [Sun, 5 Feb 2012 02:44:51 +0000 (03:44 +0100)]
netfilter: ctnetlink: allow to set expectfn for expectations

This patch allows you to set expectfn which is specifically used
by the NAT side of most of the existing conntrack helpers.

I have added a symbol map that uses a string as key to look up for
the function that is attached to the expectation object. This is
the best solution I came out with to solve this issue.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ctnetlink: add NAT support for expectations
Pablo Neira Ayuso [Sun, 5 Feb 2012 02:41:52 +0000 (03:41 +0100)]
netfilter: ctnetlink: add NAT support for expectations

This patch adds the missing bits to create expectations that
are created in NAT setups.

12 years agonetfilter: ctnetlink: allow to set expectation class
Pablo Neira Ayuso [Sun, 5 Feb 2012 02:21:12 +0000 (03:21 +0100)]
netfilter: ctnetlink: allow to set expectation class

This patch allows you to set the expectation class.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ctnetlink: allow to set helper for new expectations
Pablo Neira Ayuso [Sun, 5 Feb 2012 01:34:16 +0000 (02:34 +0100)]
netfilter: ctnetlink: allow to set helper for new expectations

This patch allow you to set the helper for newly created
expectations based of the CTA_EXPECT_HELP_NAME attribute.
Before this, the helper set was NULL.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ipset: hash:net,iface timeout bug fixed
Jozsef Kadlecsik [Fri, 13 Jan 2012 21:55:54 +0000 (22:55 +0100)]
netfilter: ipset: hash:net,iface timeout bug fixed

Timed out entries were still matched till the garbage collector
purged them out. The fix is verified in the testsuite.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ipset: Exceptions support added to hash:*net* types
Jozsef Kadlecsik [Sat, 14 Jan 2012 16:16:36 +0000 (17:16 +0100)]
netfilter: ipset: Exceptions support added to hash:*net* types

The "nomatch" keyword and option is added to the hash:*net* types,
by which one can add exception entries to sets. Example:

        ipset create test hash:net
        ipset add test 192.168.0/24
        ipset add test 192.168.0/30 nomatch

In this case the IP addresses from 192.168.0/24 except 192.168.0/30
match the elements of the set.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ipset: Log warning when a hash type of set gets full
Jozsef Kadlecsik [Tue, 10 Jan 2012 16:04:32 +0000 (17:04 +0100)]
netfilter: ipset: Log warning when a hash type of set gets full

If the set is full, the SET target cannot add more elements.
Log warning so that the admin got notified about it.

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ipset: expose userspace-relevant parts in ip_set.h
Jan Engelhardt [Sat, 14 Jan 2012 15:26:20 +0000 (16:26 +0100)]
netfilter: ipset: expose userspace-relevant parts in ip_set.h

iptables's libxt_SET.c depends on these.

Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agonetfilter: ipset: use NFPROTO_ constants
Jan Engelhardt [Mon, 13 Feb 2012 23:24:10 +0000 (00:24 +0100)]
netfilter: ipset: use NFPROTO_ constants

ipset is actually using NFPROTO values rather than AF (xt_set passes
that along).

Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
12 years agopowerpc/44x: Add more changes for APM821XX EMAC driver
Duc Dang [Mon, 5 Mar 2012 00:58:22 +0000 (00:58 +0000)]
powerpc/44x: Add more changes for APM821XX EMAC driver

This patch includes:

  Configure EMAC PHY clock source (clock from PHY or internal clock).

  Do not advertise PHY half duplex capability as APM821XX EMAC does not support half duplex mode.

  Add changes to support configuring jumbo frame for APM821XX EMAC.

[ Fix coding style -DaveM ]

Signed-off-by: Duc Dang <dhdang@apm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agopowerpc/44x: Add new compatible value for EMAC node of APM821XX dts file.
Duc Dang [Mon, 5 Mar 2012 00:57:42 +0000 (00:57 +0000)]
powerpc/44x: Add new compatible value for EMAC node of APM821XX dts file.

This compatible value will be used to distinguish some special features of APM821XX EMAC: no half duplex mode support, configuring jumbo frame.

Signed-off-by: Duc Dang <dhdang@apm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet: Use bool for return value of dev_valid_name().
David S. Miller [Tue, 6 Mar 2012 21:12:15 +0000 (16:12 -0500)]
net: Use bool for return value of dev_valid_name().

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4: defining functions as static
Yevgeny Petrilin [Tue, 6 Mar 2012 04:05:02 +0000 (04:05 +0000)]
net/mlx4: defining functions as static

Fixing sparse warnings, the 2 functions are only used in same
file. Defining them as static and not exporting them.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4: remove unused functions
Yevgeny Petrilin [Tue, 6 Mar 2012 04:04:54 +0000 (04:04 +0000)]
net/mlx4: remove unused functions

Removing functions that are no longer in use, but still exist

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4: fixing sparse warnings for not declared, functions
Yevgeny Petrilin [Tue, 6 Mar 2012 04:04:47 +0000 (04:04 +0000)]
net/mlx4: fixing sparse warnings for not declared, functions

The SET_PORT functions are implemented in port.c, which is part
of mlx4_core, these functions are exported. The functions are in use by
the mlx4_en module (were originally part of mlx4_en).
Their declaration remained in mlx4_en module, moving the declaration to the right location.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4: fixing sparse warnings when copying mac, address to gid entry
Yevgeny Petrilin [Tue, 6 Mar 2012 04:04:26 +0000 (04:04 +0000)]
net/mlx4: fixing sparse warnings when copying mac, address to gid entry

The mac should be written as __be64 the gid. The warning was because
we changed the mac parameter, which is u64, by writing result of cpu_to_be64
into it. Fixing by using new variable of type __be64.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4: fix sparse warnings on wrong type for RSS keys
Yevgeny Petrilin [Tue, 6 Mar 2012 04:04:07 +0000 (04:04 +0000)]
net/mlx4: fix sparse warnings on wrong type for RSS keys

The keys used for the hardware RSS topelitz hash are of type __be32
where the values provided by the driver are from array of u32,
this triggered sparse warning on incorrect type in assignment as of different base types.
Since these values are picked randomly,
the fix is to transform the key to __be32 by executing cpu_to_be_32()

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4: fix sparse warnings on TX blue flame buffer
Or Gerlitz [Tue, 6 Mar 2012 04:04:02 +0000 (04:04 +0000)]
net/mlx4: fix sparse warnings on TX blue flame buffer

The blue flame buffer is defined to be of type void __iomem *
but was passed to mlx4_bf_copy which gets unsigned long * .
This triggered sparse warning on different address spaces,
fix that by changing mlx4_bf_copy first param to be of type void __iomem * .

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4: fix sparse warnings on TX control flags, endianess
Or Gerlitz [Tue, 6 Mar 2012 04:03:41 +0000 (04:03 +0000)]
net/mlx4: fix sparse warnings on TX control flags, endianess

Fix sparse warnings on incompatibility between the endianess of the ctrl_flags
field of struct mlx4_en_priv to the srcrb_flags field of struct
mlx4_wqe_ctrl_seg by changing the former to be __be32 instead of u32.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/mlx4_en: Saving mem access on data path
Yevgeny Petrilin [Tue, 6 Mar 2012 04:03:34 +0000 (04:03 +0000)]
net/mlx4_en: Saving mem access on data path

Localized the pdev->dev, and using dma_map instead of pci_map
There are multiple map/unmap operations on data path,
optimizing those by saving redundant pointer access.
Those places were identified as hot-spots when running kernel profiling
during some benchmarks.
The fixes had most impact when testing packet rate with small packets,
reducing several % from CPU load, and in some case being the difference
between reaching wire speed or being CPU bound.

Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqla3xxx: ethernet: Silence static checker warning.
Santosh Nayak [Fri, 2 Mar 2012 05:09:39 +0000 (05:09 +0000)]
qla3xxx: ethernet: Silence static checker warning.

Silence the following warning:
"warn: returning -1 instead of -ENOMEM is sloppy".

Signed-off-by: Santosh Nayak <santoshprasadnayak@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years ago8139too: Add 64bit statistics
Junchang Wang [Mon, 5 Mar 2012 17:13:05 +0000 (17:13 +0000)]
8139too: Add 64bit statistics

Switch to use ndo_get_stats64 to get 64bit statistics.
Two sync entries are used (one for Rx and one for Tx).

Signed-off-by: Junchang Wang <junchangwang@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
David S. Miller [Tue, 6 Mar 2012 02:16:26 +0000 (21:16 -0500)]
Merge git://git./linux/kernel/git/davem/net

Conflicts:
drivers/net/vmxnet3/vmxnet3_drv.c

Small vmxnet3 conflict with header size bug fix in 'net'.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 6 Mar 2012 00:23:12 +0000 (16:23 -0800)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "It contains three cherry-picked fixes from perf/core, which turned out
  to be more urgent than we originally thought."

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf tools: Handle kernels that don't support attr.exclude_{guest,host}
  perf tools: Change perf_guest default back to false
  perf record: No build id option fails

12 years agoMerge tag 'usb-3.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Tue, 6 Mar 2012 00:10:44 +0000 (16:10 -0800)]
Merge tag 'usb-3.3-rc6' of git://git./linux/kernel/git/gregkh/usb

USB: revert a powerpc EHCI patch

There is just one patch in here, a revert of a powerpc EHCI driver
patch that was reported to cause problems.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
* tag 'usb-3.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  Revert "powerpc/usb: fix issue of CPU halt when missing USB PHY clock"

12 years agoMerge tag 'tty-3.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Tue, 6 Mar 2012 00:10:27 +0000 (16:10 -0800)]
Merge tag 'tty-3.3-rc6' of git://git./linux/kernel/git/gregkh/tty

tty: build fix for 3.3-rc6

This contains one build fix for the powerpc udbg driver that was reported.

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
* tag 'tty-3.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty/powerpc: early udbg consoles can't be modules

12 years agoMerge tag 'md-3.3-fixes' of git://neil.brown.name/md
Linus Torvalds [Tue, 6 Mar 2012 00:01:25 +0000 (16:01 -0800)]
Merge tag 'md-3.3-fixes' of git://neil.brown.name/md

Pull md fixes from Neil Brown:
 "Three fixes for md in 3.3-rc: Two relate to the recently added drive
  replacement.  One fixes the problem where a read error in RAID10 would
  sometimes be retried indefinitely."

* tag 'md-3.3-fixes' of git://neil.brown.name/md:
  md/raid10: fix assembling of arrays with replacement devices.
  md/raid10: fix handling of error on last working device in array.
  md/raid1: fix buglet in md_raid1_contested.

12 years agoMerge branch 'akpm' (Andrew's patch bomb)
Linus Torvalds [Mon, 5 Mar 2012 23:50:25 +0000 (15:50 -0800)]
Merge branch 'akpm' (Andrew's patch bomb)

Merge the emailed seties of 19 patches from Andrew Morton

* akpm:
  rapidio/tsi721: fix queue wrapping bug in inbound doorbell handler
  memcg: fix mapcount check in move charge code for anonymous page
  mm: thp: fix BUG on mm->nr_ptes
  alpha: fix 32/64-bit bug in futex support
  memcg: fix GPF when cgroup removal races with last exit
  debugobjects: Fix selftest for static warnings
  floppy/scsi: fix setting of BIO flags
  memcg: fix deadlock by inverting lrucare nesting
  drivers/rtc/rtc-r9701.c: fix crash in r9701_remove()
  c2port: class_create() returns an ERR_PTR
  pps: class_create() returns an ERR_PTR, not NULL
  hung_task: fix the broken rcu_lock_break() logic
  vfork: kill PF_STARTING
  coredump_wait: don't call complete_vfork_done()
  vfork: make it killable
  vfork: introduce complete_vfork_done()
  aio: wake up waiters when freeing unused kiocbs
  kprobes: return proper error code from register_kprobe()
  kmsg_dump: don't run on non-error paths by default

12 years agorapidio/tsi721: fix queue wrapping bug in inbound doorbell handler
Alexandre Bounine [Mon, 5 Mar 2012 22:59:21 +0000 (14:59 -0800)]
rapidio/tsi721: fix queue wrapping bug in inbound doorbell handler

Fix a bug that causes a kernel panic when the number of received doorbells
is larger than number of entries in the inbound doorbell queue (current
default value = 512).

Another possible indication for this bug is large number of spurious
doorbells reported by tsi721 driver after reaching the queue size maximum.

Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Chul Kim <chul.kim@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: <stable@vger.kernel.org> [3.2.x+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomemcg: fix mapcount check in move charge code for anonymous page
Naoya Horiguchi [Mon, 5 Mar 2012 22:59:20 +0000 (14:59 -0800)]
memcg: fix mapcount check in move charge code for anonymous page

Currently the charge on shared anonyous pages is supposed not to moved in
task migration.  To implement this, we need to check that mapcount > 1,
instread of > 2.  So this patch fixes it.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: thp: fix BUG on mm->nr_ptes
Andrea Arcangeli [Mon, 5 Mar 2012 22:59:20 +0000 (14:59 -0800)]
mm: thp: fix BUG on mm->nr_ptes

Dave Jones reports a few Fedora users hitting the BUG_ON(mm->nr_ptes...)
in exit_mmap() recently.

Quoting Hugh's discovery and explanation of the SMP race condition:

  "mm->nr_ptes had unusual locking: down_read mmap_sem plus
   page_table_lock when incrementing, down_write mmap_sem (or mm_users
   0) when decrementing; whereas THP is careful to increment and
   decrement it under page_table_lock.

   Now most of those paths in THP also hold mmap_sem for read or write
   (with appropriate checks on mm_users), but two do not: when
   split_huge_page() is called by hwpoison_user_mappings(), and when
   called by add_to_swap().

   It's conceivable that the latter case is responsible for the
   exit_mmap() BUG_ON mm->nr_ptes that has been reported on Fedora."

The simplest way to fix it without having to alter the locking is to make
split_huge_page() a noop in nr_ptes terms, so by counting the preallocated
pagetables that exists for every mapped hugepage.  It was an arbitrary
choice not to count them and either way is not wrong or right, because
they are not used but they're still allocated.

Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Josh Boyer <jwboyer@redhat.com>
Cc: <stable@vger.kernel.org> [3.0.x, 3.1.x, 3.2.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoalpha: fix 32/64-bit bug in futex support
Andrew Morton [Mon, 5 Mar 2012 22:59:19 +0000 (14:59 -0800)]
alpha: fix 32/64-bit bug in futex support

Michael Cree said:

: : I have noticed some user space problems (pulseaudio crashes in pthread
: : code, glibc/nptl test suite failures, java compiler freezes on SMP alpha
: : systems) that arise when using a 2.6.39 or later kernel on Alpha.
: : Bisecting between 2.6.38 and 2.6.39 (using glibc/nptl test suite as
: : criterion for good/bad kernel) eventually leads to:
: :
: : 8d7718aa082aaf30a0b4989e1f04858952f941bc is the first bad commit
: : commit 8d7718aa082aaf30a0b4989e1f04858952f941bc
: : Author: Michel Lespinasse <walken@google.com>
: : Date:   Thu Mar 10 18:50:58 2011 -0800
: :
: :     futex: Sanitize futex ops argument types
: :
: :     Change futex_atomic_op_inuser and futex_atomic_cmpxchg_inatomic
: :     prototypes to use u32 types for the futex as this is the data type the
: :     futex core code uses all over the place.
: :
: : Looking at the commit I see there is a change of the uaddr argument in
: : the Alpha architecture specific code for futexes from int to u32, but I
: : don't see why this should cause a problem.

Richard Henderson said:

: futex_atomic_cmpxchg_inatomic(u32 *uval, u32 __user *uaddr,
:                               u32 oldval, u32 newval)
: ...
:         :       "r"(uaddr), "r"((long)oldval), "r"(newval)
:
:
: There is no 32-bit compare instruction.  These are implemented by
: consistently extending the values to a 64-bit type.  Since the
: load instruction sign-extends, we want to sign-extend the other
: quantity as well (despite the fact it's logically unsigned).
:
: So:
:
: -        :       "r"(uaddr), "r"((long)oldval), "r"(newval)
: +        :       "r"(uaddr), "r"((long)(int)oldval), "r"(newval)
:
: should do the trick.

Michael said:

: This fixes the glibc test suite failures and the pulseaudio related
: crashes, but it does not fix the java compiiler lockups that I was (and
: are still) observing.  That is some other problem.

Reported-by: Michael Cree <mcree@orcon.net.nz>
Tested-by: Michael Cree <mcree@orcon.net.nz>
Acked-by: Phil Carmody <ext-phil.2.carmody@nokia.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Michel Lespinasse <walken@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomemcg: fix GPF when cgroup removal races with last exit
Hugh Dickins [Mon, 5 Mar 2012 22:59:18 +0000 (14:59 -0800)]
memcg: fix GPF when cgroup removal races with last exit

When moving tasks from old memcg (with move_charge_at_immigrate on new
memcg), followed by removal of old memcg, hit General Protection Fault in
mem_cgroup_lru_del_list() (called from release_pages called from
free_pages_and_swap_cache from tlb_flush_mmu from tlb_finish_mmu from
exit_mmap from mmput from exit_mm from do_exit).

Somewhat reproducible, takes a few hours: the old struct mem_cgroup has
been freed and poisoned by SLAB_DEBUG, but mem_cgroup_lru_del_list() is
still trying to update its stats, and take page off lru before freeing.

A task, or a charge, or a page on lru: each secures a memcg against
removal.  In this case, the last task has been moved out of the old memcg,
and it is exiting: anonymous pages are uncharged one by one from the
memcg, as they are zapped from its pagetables, so the charge gets down to
0; but the pages themselves are queued in an mmu_gather for freeing.

Most of those pages will be on lru (and force_empty is careful to
lru_add_drain_all, to add pages from pagevec to lru first), but not
necessarily all: perhaps some have been isolated for page reclaim, perhaps
some isolated for other reasons.  So, force_empty may find no task, no
charge and no page on lru, and let the removal proceed.

There would still be no problem if these pages were immediately freed; but
typically (and the put_page_testzero protocol demands it) they have to be
added back to lru before they are found freeable, then removed from lru
and freed.  We don't see the issue when adding, because the
mem_cgroup_iter() loops keep their own reference to the memcg being
scanned; but when it comes to mem_cgroup_lru_del_list().

I believe this was not an issue in v3.2: there, PageCgroupAcctLRU and
PageCgroupUsed flags were used (like a trick with mirrors) to deflect view
of pc->mem_cgroup to the stable root_mem_cgroup when neither set.
38c5d72f3ebe ("memcg: simplify LRU handling by new rule") mercifully
removed those convolutions, but left this General Protection Fault.

But it's surprisingly easy to restore the old behaviour: just check
PageCgroupUsed in mem_cgroup_lru_add_list() (which decides on which lruvec
to add), and reset pc to root_mem_cgroup if page is uncharged.  A risky
change?  just going back to how it worked before; testing, and an audit of
uses of pc->mem_cgroup, show no problem.

And there's a nice bonus: with mem_cgroup_lru_add_list() itself making
sure that an uncharged page goes to root lru, mem_cgroup_reset_owner() no
longer has any purpose, and we can safely revert 4e5f01c2b9b9 ("memcg:
clear pc->mem_cgroup if necessary").

Calling update_page_reclaim_stat() after add_page_to_lru_list() in swap.c
is not strictly necessary: the lru_lock there, with RCU before memcg
structures are freed, makes mem_cgroup_get_reclaim_stat_from_page safe
without that; but it seems cleaner to rely on one dependency less.

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodebugobjects: Fix selftest for static warnings
Stephen Boyd [Mon, 5 Mar 2012 22:59:17 +0000 (14:59 -0800)]
debugobjects: Fix selftest for static warnings

debugobjects is now printing a warning when a fixup for a NOTAVAILABLE
object is run.  This causes the selftest to fail like:

ODEBUG: selftest warnings failed 4 != 5

We could just increase the number of warnings that the selftest is
expecting to see because that is actually what has changed.  But, it turns
out that fixup_activate() was written with inverted logic and thus a fixup
for a static object returned 1 indicating the object had been fixed, and 0
otherwise.  Fix the logic to be correct and update the counts to reflect
that nothing needed fixing for a static object.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agofloppy/scsi: fix setting of BIO flags
Muthu Kumar [Mon, 5 Mar 2012 22:59:16 +0000 (14:59 -0800)]
floppy/scsi: fix setting of BIO flags

Fix setting bio flags in drivers (sd_dif/floppy).

Signed-off-by: Muthukumar R <muthur@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomemcg: fix deadlock by inverting lrucare nesting
Hugh Dickins [Mon, 5 Mar 2012 22:59:16 +0000 (14:59 -0800)]
memcg: fix deadlock by inverting lrucare nesting

We have forgotten the rules of lock nesting: the irq-safe ones must be
taken inside the non-irq-safe ones, otherwise we are open to deadlock:

CPU0                          CPU1
----                          ----
lock(&(&pc->lock)->rlock);
                              local_irq_disable();
                              lock(&(&zone->lru_lock)->rlock);
                              lock(&(&pc->lock)->rlock);
<Interrupt>
lock(&(&zone->lru_lock)->rlock);

To check a different locking issue, I happened to add a spin_lock to
memcg's bit_spin_lock in lock_page_cgroup(), and lockdep very quickly
complained about __mem_cgroup_commit_charge_lrucare() (on CPU1 above).

So delete __mem_cgroup_commit_charge_lrucare(), passing a bool lrucare to
__mem_cgroup_commit_charge() instead, taking zone->lru_lock under
lock_page_cgroup() in the lrucare case.

The original was using spin_lock_irqsave, but we'd be in more trouble if
it were ever called at interrupt time: unconditional _irq is enough.  And
ClearPageLRU before del from lru, SetPageLRU before add to lru: no strong
reason, but that is the ordering used consistently elsewhere.

Fixes 36b62ad539498d00c2d280a151a ("memcg: simplify corner case handling
of LRU").

Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-r9701.c: fix crash in r9701_remove()
Anatolij Gustschin [Mon, 5 Mar 2012 22:59:16 +0000 (14:59 -0800)]
drivers/rtc/rtc-r9701.c: fix crash in r9701_remove()

If probing the RTC didn't succeed due to failed RTC register access, the
RTC device will be unregistered.  Then, when removing the module
r9701_remove() causes a kernel crash while trying to unregister a not
registered RTC device.  Fix this by doing RTC register access test before
RTC device registration.

Signed-off-by: Anatolij Gustschin <agust@denx.de>
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>
12 years agoc2port: class_create() returns an ERR_PTR
Dan Carpenter [Mon, 5 Mar 2012 22:59:15 +0000 (14:59 -0800)]
c2port: class_create() returns an ERR_PTR

class_create() doesn't return a NULL, it only returns ERR_PTRs.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agopps: class_create() returns an ERR_PTR, not NULL
Dan Carpenter [Mon, 5 Mar 2012 22:59:15 +0000 (14:59 -0800)]
pps: class_create() returns an ERR_PTR, not NULL

class_create() never returns NULLs only ERR_PTRs.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Rodolfo Giometti <giometti@enneenne.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohung_task: fix the broken rcu_lock_break() logic
Oleg Nesterov [Mon, 5 Mar 2012 22:59:14 +0000 (14:59 -0800)]
hung_task: fix the broken rcu_lock_break() logic

check_hung_uninterruptible_tasks()->rcu_lock_break() introduced by
"softlockup: check all tasks in hung_task" commit ce9dbe24 looks
absolutely wrong.

- rcu_lock_break() does put_task_struct(). If the task has exited
  it is not safe to even read its ->state, nothing protects this
  task_struct.

- The TASK_DEAD checks are wrong too. Contrary to the comment, we
  can't use it to check if the task was unhashed. It can be unhashed
  without TASK_DEAD, or it can be valid with TASK_DEAD.

  For example, an autoreaping task can do release_task(current)
  long before it sets TASK_DEAD in do_exit().

  Or, a zombie task can have ->state == TASK_DEAD but release_task()
  was not called, and in this case we must not break the loop.

Change this code to check pid_alive() instead, and do this before we drop
the reference to the task_struct.

Note: while_each_thread() under rcu_read_lock() is not really safe, it can
livelock.  This will be fixed later, but fortunately in this case the
"max_count" logic saves us anyway.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Mandeep Singh Baines <msb@google.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agovfork: kill PF_STARTING
Oleg Nesterov [Mon, 5 Mar 2012 22:59:14 +0000 (14:59 -0800)]
vfork: kill PF_STARTING

Previously it was (ab)used by utrace.  Then it was wrongly used by the
scheduler code.

Currently it is not used, kill it before it finds the new erroneous user.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocoredump_wait: don't call complete_vfork_done()
Oleg Nesterov [Mon, 5 Mar 2012 22:59:13 +0000 (14:59 -0800)]
coredump_wait: don't call complete_vfork_done()

Now that CLONE_VFORK is killable, coredump_wait() no longer needs
complete_vfork_done().  zap_threads() should find and kill all tasks with
the same ->mm, this includes our parent if ->vfork_done is set.

mm_release() becomes the only caller, unexport complete_vfork_done().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agovfork: make it killable
Oleg Nesterov [Mon, 5 Mar 2012 22:59:13 +0000 (14:59 -0800)]
vfork: make it killable

Make vfork() killable.

Change do_fork(CLONE_VFORK) to do wait_for_completion_killable().  If it
fails we do not return to the user-mode and never touch the memory shared
with our child.

However, in this case we should clear child->vfork_done before return, we
use task_lock() in do_fork()->wait_for_vfork_done() and
complete_vfork_done() to serialize with each other.

Note: now that we use task_lock() we don't really need completion, we
could turn task->vfork_done into "task_struct *wake_up_me" but this needs
some complications.

NOTE: this and the next patches do not affect in-kernel users of
CLONE_VFORK, kernel threads run with all signals ignored including
SIGKILL/SIGSTOP.

However this is obviously the user-visible change.  Not only a fatal
signal can kill the vforking parent, a sub-thread can do execve or
exit_group() and kill the thread sleeping in vfork().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agovfork: introduce complete_vfork_done()
Oleg Nesterov [Mon, 5 Mar 2012 22:59:13 +0000 (14:59 -0800)]
vfork: introduce complete_vfork_done()

No functional changes.

Move the clear-and-complete-vfork_done code into the new trivial helper,
complete_vfork_done().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoaio: wake up waiters when freeing unused kiocbs
Jeff Moyer [Mon, 5 Mar 2012 22:59:12 +0000 (14:59 -0800)]
aio: wake up waiters when freeing unused kiocbs

Bart Van Assche reported a hung fio process when either hot-removing
storage or when interrupting the fio process itself.  The (pruned) call
trace for the latter looks like so:

  fio             D 0000000000000001     0  6849   6848 0x00000004
   ffff880092541b88 0000000000000046 ffff880000000000 ffff88012fa11dc0
   ffff88012404be70 ffff880092541fd8 ffff880092541fd8 ffff880092541fd8
   ffff880128b894d0 ffff88012404be70 ffff880092541b88 000000018106f24d
  Call Trace:
    schedule+0x3f/0x60
    io_schedule+0x8f/0xd0
    wait_for_all_aios+0xc0/0x100
    exit_aio+0x55/0xc0
    mmput+0x2d/0x110
    exit_mm+0x10d/0x130
    do_exit+0x671/0x860
    do_group_exit+0x44/0xb0
    get_signal_to_deliver+0x218/0x5a0
    do_signal+0x65/0x700
    do_notify_resume+0x65/0x80
    int_signal+0x12/0x17

The problem lies with the allocation batching code.  It will
opportunistically allocate kiocbs, and then trim back the list of iocbs
when there is not enough room in the completion ring to hold all of the
events.

In the case above, what happens is that the pruning back of events ends
up freeing up the last active request and the context is marked as dead,
so it is thus responsible for waking up waiters.  Unfortunately, the
code does not check for this condition, so we end up with a hung task.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Reported-by: Bart Van Assche <bvanassche@acm.org>
Tested-by: Bart Van Assche <bvanassche@acm.org>
Cc: <stable@kernel.org> [3.2.x only]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agokprobes: return proper error code from register_kprobe()
Prashanth Nageshappa [Mon, 5 Mar 2012 22:59:12 +0000 (14:59 -0800)]
kprobes: return proper error code from register_kprobe()

register_kprobe() aborts if the address of the new request falls in a
prohibited area (such as ftrace pouch, __kprobes annotated functions,
non-kernel text addresses, jump label text).  We however don't return the
right error on this abort, resulting in a silent failure - incorrect
adding/reporting of kprobes ('perf probe do_fork+18' or 'perf probe
mcount' for instance).

In V2 we are incorporating Masami Hiramatsu's  feedback.

This patch fixes it by returning -EINVAL upon failure.

While we are here, rename the label used for exit to be more appropriate.

Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Signed-off-by: Prashanth K Nageshappa <prashanth@linux.vnet.ibm.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agokmsg_dump: don't run on non-error paths by default
Matthew Garrett [Mon, 5 Mar 2012 22:59:10 +0000 (14:59 -0800)]
kmsg_dump: don't run on non-error paths by default

Since commit 04c6862c055f ("kmsg_dump: add kmsg_dump() calls to the
reboot, halt, poweroff and emergency_restart paths"), kmsg_dump() gets
run on normal paths including poweroff and reboot.

This is less than ideal given pstore implementations that can only
represent single backtraces, since a reboot may overwrite a stored oops
before it's been picked up by userspace.  In addition, some pstore
backends may have low performance and provide a significant delay in
reboot as a result.

This patch adds a printk.always_kmsg_dump kernel parameter (which can also
be changed from userspace).  Without it, the code will only be run on
failure paths rather than on normal paths.  The option can be enabled in
environments where there's a desire to attempt to audit whether or not a
reboot was cleanly requested or not.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Acked-by: Seiji Aguchi <seiji.aguchi@hds.com>
Cc: Seiji Aguchi <seiji.aguchi@hds.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Marco Stornelli <marco.stornelli@gmail.com>
Cc: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Cc: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agor8169: add byte queue limit support.
Igor Maravic [Sun, 4 Mar 2012 23:01:25 +0000 (00:01 +0100)]
r8169: add byte queue limit support.

Nothing fancy:
- sent bytes count is notified in the start_xmit path right before
  updating the owner bit in the hardware Tx descriptor (E. Dumazet)
- avoid useless tp->dev dereferencing in start_xmit (E. Dumazet)

Use of netdev_reset_queue is favored over proper accounting in
rtl8169_tx_clear_range since the latter would need more work for the
same result (nb: said accounting degenerates to nothing in xmit_frags).

Signed-off-by: Igor Maravic <igorm@etf.rs>
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
12 years agor8169: add 64bit statistics.
Junchang Wang [Sun, 4 Mar 2012 22:30:32 +0000 (23:30 +0100)]
r8169: add 64bit statistics.

Switch to use ndo_get_stats64 to get 64bit statistics.
Two sync entries are used (one for Rx and one for Tx).

Signed-off-by: Junchang Wang <junchangwang@gmail.com>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
12 years agomd/raid10: fix assembling of arrays with replacement devices.
NeilBrown [Mon, 5 Mar 2012 06:48:12 +0000 (17:48 +1100)]
md/raid10: fix assembling of arrays with replacement devices.

commit 56a2559bb654a (md/raid10: recognise replacements ...)
changed 'run' to set ->replacement or ->rdev depending on the
'Replacement' status if the device, but it didn't remove the
old unconditional setting of 'rdev'.  So it was largely ineffective.

So remove that now.

Signed-off-by: NeilBrown <neilb@suse.de>
12 years agodrm, gma500: Fix Cedarview boot failures in 3.3-rc
Alan Cox [Mon, 5 Mar 2012 14:22:16 +0000 (14:22 +0000)]
drm, gma500: Fix Cedarview boot failures in 3.3-rc

Production GMA3600/3650 hardware turns out to be subtly different to the
development platforms.  This combined with a minor driver bug is causing
the kernel to hang on these platforms.

This patch does the following

 - turn down a couple of messages that were meant to be debug and are
   causing much confusion

 - ensure the hotplug interrupt is disabled on Cedartrail systems.

 - fix a bug where gtt roll mode called psbfb_sync, which tries to sync
   the 2D engine. On other devices it is harmless as the 2D engine is
   present but not in use when in gtt roll mode, on Cedartrail it causes
   a hang

Without these changes 3.3-rc hangs on boot on Cedartrail based systems.

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Mon, 5 Mar 2012 22:30:54 +0000 (14:30 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

1) TCP SACK processing can calculate an incorrect reordering value in
   some cases, fix from Neal Cardwell.

2) tcp_mark_head_lost() can split SKBs in situations where it should
   not, violating send queue invariants expected by other pieces of
   code and thus resulting (eventually) in corrupted retransmit state
   counters.  Also from Neal Cardwell.

3) qla3xxx erroneously calls spin_lock_irqrestore() with constant
   hw_flags of zero.  Fix from Santosh Nayak.

4) Fix NULL deref in rt2x00, from Gabor Juhos.

5) pch_gbe passes address of wrong typed object to pch_gbe_validate_option
   thus corrupting part of the value.  From Dan Carpenter.

6) We must check the return value of nlmsg_parse() before trying to use
   the results.  From Eric Dumazet.

7) Bridging code fails to check return value of ipv6_dev_get_saddr()
   thus potentially leaving uninitialized garbage in the outgoing ipv6
   header.  From Ulrich Weber.

8) Due to rounding and a reversed operation on jiffies, bridge message
   ages can go backwards instead of forwards, thus breaking STP.  Fixes
   from Joakim Tjernlund.

9) r8169 modifies Config* registers without properly holding the
   Config9346 lock, resulting in corrupted IP fragments on some chips.
   Fix from Francois Romieu.

10) NET_PACKET_ENGINE default wan't set properly during the network
   driver mega-move.  Fix from Stephen Hemminger.

11) vmxnet3 uses TCP header size where it actually should use the UDP
   header size, fix from Shreyas Bhatewara.

12) Netfilter bridge module autoload is busted in the compat case, fix
   from Florian Westphal.

13) Wireless Key removal was not setting multicast bits correctly thus
   accidently killing the unicast key 0 and thus all traffic stops.
   Fix from Johannes Berg.

14) Fix endless retries of A-MPDU transmissions in brcm80211 driver.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (22 commits)
  qla3xxx: ethernet: Fix bogus interrupt state flag.
  bridge: check return value of ipv6_dev_get_saddr()
  rtnetlink: fix rtnl_calcit() and rtnl_dump_ifinfo()
  bridge: message age needs to increase, not decrease.
  bridge: Adjust min age inc for HZ > 256
  tcp: don't fragment SACKed skbs in tcp_mark_head_lost()
  r8169: corrupted IP fragments fix for large mtu.
  packetengines: fix config default
  vmxnet3: Fix transport header size
  enic: fix an endian bug in enic_probe()
  pch_gbe: memory corruption calling pch_gbe_validate_option()
  tg3: Fix tg3_get_stats64 for 5700 / 5701 devs
  tcp: fix false reordering signal in tcp_shifted_skb
  tcp: fix comment for tp->highest_sack
  netfilter: bridge: fix module autoload in compat case
  brcm80211: smac: only print block-ack timeout message at trace level
  brcm80211: smac: fix endless retry of A-MPDU transmissions
  mac80211: Fix a warning on changing to monitor mode from STA
  mac80211: zero initialize count field in ieee80211_tx_rate
  iwlwifi: fix key removal
  ...

12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci
Linus Torvalds [Mon, 5 Mar 2012 22:30:12 +0000 (14:30 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci

Pull PCI fixes from Jesse Barnes:
 "A couple of fixes for booting specific machines, and one for a minor
  memory leak on pre-_CRS platforms."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci:
  x86/PCI: do not tie MSI MS-7253 use_crs quirk to BIOS version
  x86/PCI: use host bridge _CRS info on MSI MS-7253
  PCI: fix memleak when ACPI _CRS is not used.

12 years agoMerge branch 'for-3.3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
Linus Torvalds [Mon, 5 Mar 2012 22:28:36 +0000 (14:28 -0800)]
Merge branch 'for-3.3-fixes' of git://git./linux/kernel/git/tj/percpu

Pull per-cpu patches from Tejun Heo:
 "This pull request contains four patches.  One replaces manual clearing
  with bitmap_clear(), two fix generic definition of __this_cpu ops so
  that they don't choose unnecessarily strict arch version.  One makes
  _this_cpu definition use raw_local_irq_*() so that it doesn't end up
  wrecking irq on/off state tracking when used from inside lockdep.

  Of the four patches, the raw_local_irq_*() update is the most
  important, so please feel free to cherry pick only that one patch and
  ignore the rest if you want to - commit e920d5971d 'percpu: use
  raw_local_irq_* in _this_cpu op'."

* 'for-3.3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu:
  percpu: fix __this_cpu_{sub,inc,dec}_return() definition
  percpu: use raw_local_irq_* in _this_cpu op
  percpu: fix generic definition of __this_cpu_add_and_return()
  percpu: use bitmap_clear

12 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Mon, 5 Mar 2012 22:27:34 +0000 (14:27 -0800)]
Merge branch 'upstream' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fixes from Ralf Baechle:
 "What's in there: a number of MIPS fixes and touchups.  The most
  important change in this pull request is Kautuk Consul's port of
  changes to do_page_fault which fix a hang that affects some
  configurations.  Still not quite ready for a release, there are
  problems with 64-bit platforms."

* 'upstream' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: traps.c: Fix typo
  MIPS: PowerTV: Fix defconfigs for coverage builds
  MIPS: Netlogic: Fix defconfigs for coverage builds
  MIPS: ATH79: Avoid a kernel bug on AR913X
  MIPS: PCI: use list_for_each_entry() for bus->devices traversal
  MIPS: fault.c: Port OOM changes to do_page_fault
  MIPS: vmlinux.lds.S: remove duplicate _sdata symbol
  MIPS: Alchemy: Increase minimum timeout for 32kHz timer.
  MIPS: txx9 7segled fix struct device has no member
  MIPS: Alchemy: Update Au1300 inlined GPIO macros
  MIPS: Remove temporary kludge from <asm/page.h>
  MIPS: BMIPS: smp-bmips.c does not need to include version.h

12 years agoflush_tlb_range() needs ->page_table_lock when ->mmap_sem is not held
Al Viro [Mon, 5 Mar 2012 06:40:29 +0000 (06:40 +0000)]
flush_tlb_range() needs ->page_table_lock when ->mmap_sem is not held

All other callers already hold either ->mmap_sem (exclusive) or
->page_table_lock.  And we need it because some page table flushing
instanced do work explicitly with ge tables.

See e.g.  arch/powerpc/mm/tlb_hash32.c, flush_tlb_range() and
flush_range() in there.  The same goes for uml, with a lot more
extensive playing with page tables.

Almost all callers are actually fine - flush_tlb_range() may have no
need to bother playing with page tables, but it can do so safely; again,
this caller is the sole exception - everything else either has exclusive
->mmap_sem on the mm in question, or mm->page_table_lock is held.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoVM_GROWS{UP,DOWN} shouldn't be set on shmem VMAs
Al Viro [Mon, 5 Mar 2012 06:39:47 +0000 (06:39 +0000)]
VM_GROWS{UP,DOWN} shouldn't be set on shmem VMAs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoaout: move setup_arg_pages() prior to reading/mapping the binary
Al Viro [Mon, 5 Mar 2012 06:38:42 +0000 (06:38 +0000)]
aout: move setup_arg_pages() prior to reading/mapping the binary

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoqla3xxx: ethernet: Fix bogus interrupt state flag.
Santosh Nayak [Fri, 2 Mar 2012 05:09:05 +0000 (05:09 +0000)]
qla3xxx: ethernet: Fix bogus interrupt state flag.

In 'ql_adapter_initialize'
the first call for 'spin_unlock_irqrestore()' is with hw_flags = 0,
which is as good as 'spin_unlock_irq()' (unconditional interrupt
enabling). If this is intended, then for better performance
'spin_unlock_irqrestore()' can be replaced with 'spin_unlock_irq()'
and 'spin_lock_irqsave()' can be replaced by 'spin_lock_irq()

Signed-off-by: Santosh Nayak <santoshprasadnayak@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agobridge: check return value of ipv6_dev_get_saddr()
Ulrich Weber [Mon, 5 Mar 2012 04:52:44 +0000 (04:52 +0000)]
bridge: check return value of ipv6_dev_get_saddr()

otherwise source IPv6 address of ICMPV6_MGM_QUERY packet
might be random junk if IPv6 is disabled on interface or
link-local address is not yet ready (DAD).

Signed-off-by: Ulrich Weber <ulrich.weber@sophos.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agomlx4_en: remove sparse errors
Eric Dumazet [Mon, 5 Mar 2012 05:01:14 +0000 (05:01 +0000)]
mlx4_en: remove sparse errors

Fix new sparse errors introduced in commit 6221217199 (mlx4_en: dont
change mac_header on xmit)

Reported-by: Or Gerlitz <or.gerlitz@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'for-davem' of git://git.kernel.org/pub/scm/linux/kernel/git/linville...
David S. Miller [Mon, 5 Mar 2012 21:39:02 +0000 (16:39 -0500)]
Merge branch 'for-davem' of git://git./linux/kernel/git/linville/wireless-next

12 years agonet: export netdev_stats_to_stats64
Eric Dumazet [Mon, 5 Mar 2012 04:50:09 +0000 (04:50 +0000)]
net: export netdev_stats_to_stats64

Some drivers use internal netdev stats member to store part of their
stats, yet advertize ndo_get_stats64() to implement some 64bit fields.

Allow them to use netdev_stats_to_stats64() helper to make the copy of
netdev stats before they compute their 64bit counters.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Mon, 5 Mar 2012 20:05:54 +0000 (15:05 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless-next into for-davem

12 years agoMerge tag 'mmc-fixes-for-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc
Linus Torvalds [Mon, 5 Mar 2012 16:51:10 +0000 (08:51 -0800)]
Merge tag 'mmc-fixes-for-3.3' of git://git./linux/kernel/git/cjb/mmc

MMC fixes from Chris Ball for 3.3:
 - atmel-mci: oops fix against regression introduced in 3.2
 - core: power saving regression fix against 3.3-rc1
 - core: suspend/resume fix for UHS-I cards
 - esdhc-imx: MMC card regression fix against 3.0
 - mmci: oops fix for ARM systems with large (64k) pages
 - MAINTAINERS update for atmel-mci.

* tag 'mmc-fixes-for-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc:
  mmc: core: Fixup suspend/resume issues for UHS-I cards
  mmc: mmci: reduce max_blk_count to avoid overflowing max_req_size
  mmc: sdhci-esdhc-imx: fix for mmc cards on i.MX5
  mmc: core: fix regression: set default clock gating delay to 0
  MAINTAINERS: hand over atmel-mci (sd/mmc interface)
  mmc: atmel-mci: don't use dma features when using DMA with no chan available

12 years agoMerge branch 'upstream-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Mon, 5 Mar 2012 16:48:24 +0000 (08:48 -0800)]
Merge branch 'upstream-fixes' of git://git./linux/kernel/git/jikos/hid

Pull from Jiri Kosina:
 "Please pull to receive updates for HID layer.  Nikolai's patch is
  rather important and should still go in for 3.3, as it's a regression
  fix for commit b4b583d."

* 'upstream-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: hid-input: allow array fields out of range
  HID: usbhid: Add NOGET quirk for the AIREN Slim+ keyboard

12 years agoHID: hid-input: allow array fields out of range
Nikolai Kondrashov [Fri, 2 Mar 2012 20:13:58 +0000 (22:13 +0200)]
HID: hid-input: allow array fields out of range

Allow array field values out of range as per HID 1.11 specification,
section 6.2.25:

Rather than returning a single bit for each button in the group, an
array returns an index in each field that corresponds to the pressed
button (like keyboard scan codes). An out-of range value in and array
field is considered no controls asserted.

Apparently, "and" above is a typo and should be "an".

This fixes at least Waltop tablet pen clicks - otherwise BTN_TOUCH is never
released.

The relevant part of Waltop tablet report descriptors is this:

0x09, 0x42,         /*          Usage (Tip Switch),         */
0x09, 0x44,         /*          Usage (Barrel Switch),      */
0x09, 0x46,         /*          Usage (Tablet Pick),        */
0x15, 0x01,         /*          Logical Minimum (1),        */
0x25, 0x03,         /*          Logical Maximum (3),        */
0x75, 0x04,         /*          Report Size (4),            */
0x95, 0x01,         /*          Report Count (1),           */
0x80,               /*          Input,                      */

This is a regression fix for commit b4b583d ("HID: be more strict when
ignoring out-of-range fields").

Signed-off-by: Nikolai Kondrashov <spbnick@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
12 years agoMerge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Mon, 5 Mar 2012 08:05:44 +0000 (09:05 +0100)]
Merge tag 'perf-urgent-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/urgent

Cherry picked fixes from perf/core, together with the kernel fix (1018faa),
the sampling tools (top, record) are back working on AMD systems.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
12 years agortnetlink: fix rtnl_calcit() and rtnl_dump_ifinfo()
Eric Dumazet [Sun, 4 Mar 2012 12:32:10 +0000 (12:32 +0000)]
rtnetlink: fix rtnl_calcit() and rtnl_dump_ifinfo()

nlmsg_parse() might return an error, so test its return value before
potential random memory accesses.

Errors introduced in commit 115c9b81928 (rtnetlink: Fix problem with
buffer allocation)

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'master' of git://gitorious.org/linux-can/linux-can-next
David S. Miller [Mon, 5 Mar 2012 02:59:39 +0000 (21:59 -0500)]
Merge branch 'master' of git://gitorious.org/linux-can/linux-can-next

12 years agobridge: message age needs to increase, not decrease.
Joakim Tjernlund [Thu, 1 Mar 2012 08:12:19 +0000 (08:12 +0000)]
bridge: message age needs to increase, not decrease.

commit bridge: send proper message_age in config BPDU
added this gem:
  bpdu.message_age = (jiffies - root->designated_age)
  p->designated_age = jiffies + bpdu->message_age;
Notice how bpdu->message_age is negated when reassigned to
bpdu.message_age. This causes message age to decrease breaking the
STP protocol.

Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agobridge: Adjust min age inc for HZ > 256
Joakim Tjernlund [Thu, 1 Mar 2012 08:12:18 +0000 (08:12 +0000)]
bridge: Adjust min age inc for HZ > 256

min age increment needs to round up its min age tick for all
HZ values to guarantee message age is increasing.

Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agovmxnet3: Fix log messages and corrects some typos
Shreyas Bhatewara [Tue, 28 Feb 2012 22:08:39 +0000 (22:08 +0000)]
vmxnet3: Fix log messages and corrects some typos

Fix log messages and corrects some typos

Change logging of failure to enable MSI/MSI-X to display device's PCI
address instead of eth%d. Rectify a typo.

Signed-off-by: Shreyas N Bhatewara <sbhatewara@vmware.com>
Reviewed-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Recode PCI MRRS adjustment as a PCI quirk
Matt Carlson [Mon, 27 Feb 2012 09:44:48 +0000 (09:44 +0000)]
tg3: Recode PCI MRRS adjustment as a PCI quirk

This patch recodes the MRRS cap for 5719 A0 devices as a PCI quirk.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoucc_geth: separate out rx/tx ring alloc and free operations
Paul Gortmaker [Mon, 27 Feb 2012 02:36:29 +0000 (02:36 +0000)]
ucc_geth: separate out rx/tx ring alloc and free operations

Factor out the the existing allocation and free operations
so that they can be used individually.

This is to improve code readability, and also to prepare for
possible future changes like better error recovery and more
dynamic configuration (e.g on-the-fly resizing of the rings).

This change represents a straight up relocation of the existing
code into separate routines without changing any of the contained
code itself.  Local variables are relocated as necessary.

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoppp: Replace uses of <linux/if_ppp.h> with <linux/ppp-ioctl.h>
Paul Mackerras [Sun, 4 Mar 2012 12:56:55 +0000 (12:56 +0000)]
ppp: Replace uses of <linux/if_ppp.h> with <linux/ppp-ioctl.h>

Since all that include/linux/if_ppp.h does is #include <linux/ppp-ioctl.h>,
this replaces the occurrences of #include <linux/if_ppp.h> with
#include <linux/ppp-ioctl.h>.

It also corrects an error in Documentation/networking/l2tp.txt, where
it referenced include/linux/if_ppp.h as the source of some definitions
that are actually now defined in include/linux/if_pppol2tp.h.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoppp: Move ioctl definitions from if_ppp.h to new ppp-ioctl.h
Paul Mackerras [Sun, 4 Mar 2012 12:56:04 +0000 (12:56 +0000)]
ppp: Move ioctl definitions from if_ppp.h to new ppp-ioctl.h

This moves the definitions of the ioctls, constants and structures
relating to the ppp_generic interface to userspace out from if_ppp.h
to a new file, ppp-ioctl.h.  The new file has my copyright since I
designed and implemented the ppp_generic interface in the late 1990s.
None of the contents of this file comes from the original if_ppp.h
published by Carnegie Mellon University.

Of the remainder of if_ppp.h, only the PPP_MTU definition was being
used, and this replaces the uses of it with PPP_MRU (which is identical).
Therefore, this replaces the entire file with the single line

#include <linux/ppp-ioctl.h>

which clearly doesn't contain any CMU code.  Thus I have removed the
CMU copyright notice with its problematic advertising clause, and in
fact since it's only one trivial line I have not added any other
copyright notice.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoppp: Change copyright notices from ANU to me
Paul Mackerras [Sun, 4 Mar 2012 12:54:54 +0000 (12:54 +0000)]
ppp: Change copyright notices from ANU to me

This changes the copyright notices on the PPP code that I developed
in the late 1990s from being copyright The Australian National
University to copyright Paul Mackerras.  I can do this as I have an
acknowledgement in writing from the Head of the Computer Science
Department at ANU (where I worked then) that ANU does not claim any
intellectual property in this code.

While I'm at it, change the copyright notice from BSD-style to
GNU GPL like the rest of the kernel.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Fix poor tx performance on 57766 after MTU change
Michael Chan [Sun, 4 Mar 2012 14:48:15 +0000 (14:48 +0000)]
tg3: Fix poor tx performance on 57766 after MTU change

GRC reset causes the read DMA engine to go into a mode that breaks up
requests into 256 bytes.  A PHY reset is required to bring it back to
the normal mode.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Add memory barriers to sync BD data
Michael Chan [Sun, 4 Mar 2012 14:48:14 +0000 (14:48 +0000)]
tg3: Add memory barriers to sync BD data

for weak memory model architectures to ensure that the chip will DMA
valid BD data.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Reviewed-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotg3: Fix jumbo loopback test on 5719
Michael Chan [Sun, 4 Mar 2012 14:48:13 +0000 (14:48 +0000)]
tg3: Fix jumbo loopback test on 5719

Loopback on 9K packet fails because the chip has a DMA limit of 4K.  The
loopback test logic uses a single BD for simplicity.  Fix it by reducing
the jumbo packet size to the DMA limit.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Benjamin Li <benli@broadcom.com>
Reviewed-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6
Linus Torvalds [Mon, 5 Mar 2012 01:10:06 +0000 (17:10 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sameo/mfd-2.6

MFD fixes from Samuel Ortiz:
 "This is the pull request for the MFD fixes for 3.3.  We have a few
  NULL pointer dereferences fixes, an ACPI conflict check fix, and a
  couple of wm8994 fixes."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6:
  mfd: Correct readability of WM8994 DC servo 4E register
  mfd: Initialize tps65912 irq platform data properly
  mfd: Fix ACPI conflict check
  mfd: Fix ab8500 error path bug
  mfd: Test for jack detection when deciding if wm8994 should suspend
  mfd: Initialize tps65910 irq platform data properly
  mfd: Fix possible s5m null pointer dereference
  mfd: wm8350 variable dereferenced before check

12 years agovfs: move dentry_cmp from <linux/dcache.h> to fs/dcache.c
Linus Torvalds [Sun, 4 Mar 2012 23:51:42 +0000 (15:51 -0800)]
vfs: move dentry_cmp from <linux/dcache.h> to fs/dcache.c

It's only used inside fs/dcache.c, and we're going to play games with it
for the word-at-a-time patches.  This time we really don't even want to
export it, because it really is an internal function to fs/dcache.c, and
has been since it was introduced.

Having it in that extremely hot header file (it's included in pretty
much everything, thanks to <linux/fs.h>) is a disaster for testing
different versions, and is utterly pointless.

We really should have some kind of header file diet thing, where we
figure out which parts of header files are really better off private and
only result in more expensive compiles.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agopercpu: fix __this_cpu_{sub,inc,dec}_return() definition
Konstantin Khlebnikov [Tue, 28 Feb 2012 20:41:12 +0000 (00:41 +0400)]
percpu: fix __this_cpu_{sub,inc,dec}_return() definition

This patch adds missed "__" prefixes, otherwise these functions
works as irq/preemption safe.

Reported-by: Torsten Kaiser <just.for.lkml@googlemail.com>
Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Tejun Heo <tj@kernel.org>