GitHub/MotorolaMobilityLLC/kernel-slsi.git
13 years agoperf tools: Fix tracepoint id to string perf.data header table
Arnaldo Carvalho de Melo [Mon, 17 Jan 2011 20:28:13 +0000 (18:28 -0200)]
perf tools: Fix tracepoint id to string perf.data header table

It was broken by f006d25 that passed just the event name, not the complete
sys:event that it expected to open the /sys/.../sys/sys:event/id file to get
the id.

Fix it by moving it to after parse_events in cmd_record, as at that point
we can just traverse the evsel_list and use evsel->attr.config +
event_name(evsel) instead of re-opening the /id file.

Reported-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Cc: Franck Bui-Huu <vagabon.xyz@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Han Pingtian <phan@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <20110117202801.GG2085@ghostprotocols.net>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Fix handling of wildcards in tracepoint event selectors
Arnaldo Carvalho de Melo [Mon, 17 Jan 2011 16:25:06 +0000 (14:25 -0200)]
perf tools: Fix handling of wildcards in tracepoint event selectors

It wasn't accounting the ':' when consuming bytes in the the event
selector string, so parse_events() would fail in this test:

                if (!(*str == 0 || *str == ',' || isspace(*str)))
                        return -1;

as *str would be pointing to '*', the last character in the '-e' arg in:

$ perf record -q -a -D -e sched:sched_* | perf script -i - -s perf-script.py

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agopowerpc: perf: Fix frequency calculation for overflowing counters
Anton Blanchard [Mon, 17 Jan 2011 05:17:42 +0000 (16:17 +1100)]
powerpc: perf: Fix frequency calculation for overflowing counters

When profiling a benchmark that is almost 100% userspace, I noticed some wildly
inaccurate profiles that showed almost all time spent in the kernel.

Closer examination shows we were programming a tiny number of cycles into the
PMU after each overflow (about ~200 away from the next overflow). This gets us
stuck in a loop which we eventually break out of by throttling the PMU (there
are regular throttle/unthrottle events in the log).

It looks like we aren't setting event->hw.last_period to something same and the
frequency to period calculations in perf are going haywire.

With the following patch we find the correct period after a few interrupts and
stay there. I also see no more throttle events.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: linuxppc-dev@lists.ozlabs.org
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
LKML-Reference: <20110117161742.5feb3761@kryten>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoMerge branch 'tip/perf/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/roste...
Ingo Molnar [Sat, 15 Jan 2011 14:23:27 +0000 (15:23 +0100)]
Merge branch 'tip/perf/urgent' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into perf/urgent

13 years agotracing: Remove syscall_exit_fields
Lai Jiangshan [Fri, 14 Jan 2011 09:49:02 +0000 (17:49 +0800)]
tracing: Remove syscall_exit_fields

There is no need for syscall_exit_fields as the syscall
exit event class can already host the fields in its structure,
like most other trace events do by default. Use that
default behavior instead.

Following this scheme, we don't need anymore to override the
get_fields() callback of the syscall exit event class either.

Hence both syscall_exit_fields and syscall_get_exit_fields() can
be removed.

Also changed some indentation to keep the following under 80
characters:

".fields = LIST_HEAD_INIT(event_class_syscall_exit.fields),"

Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
LKML-Reference: <4D301C0E.8090408@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing: Only process module tracepoints once
Steven Rostedt [Fri, 14 Jan 2011 16:25:58 +0000 (11:25 -0500)]
tracing: Only process module tracepoints once

The commit:

 9f987b3141f086de27832514aad9f50a53f754
 tracing: Include module.h in define_trace.h

only solved half the problem. If the trace/events/module.h header is
included at the time of define_trace.h (or in ftrace.h within it),
the module.h TRACE_SYSTEM will override the current TRACE_SYSTEM
macro.

Since define_trace.h is included when CREATE_TRACE_POINTS is set,
and the first thing it does is to #undef CREATE_TRACE_POINTS,
by placing the module.h TRACE_SYSTEM inside a
 #ifdef CREATE_TRACE_POINTS
we can prevent it from overriding the TRACE_SYSTEM that is
being processed, and still process the module.h tracepoints
when the module code defines CREATE_TRACE_POINTS and includes
the trace/events/module.h header.

As with commit 9f987b3141, this is only an issue if module.h
is not included before the trace/events/<event>.h file is
included, which (luckily) has not happened yet.

Reported-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agoperf record: Add "nodelay" mode, disabled by default
Kirill Smelkov [Wed, 12 Jan 2011 14:59:36 +0000 (17:59 +0300)]
perf record: Add "nodelay" mode, disabled by default

Sometimes there is a need to use perf in "live-log" mode. The problem
is, for seldom events, actual info output is largely delayed because
perf-record reads sample data in whole pages.

So for such scenarious, add flag for perf-record to go in "nodelay"
mode. To track e.g. what's going on in icmp_rcv while ping is running
Use it with something like this:

(1) $ perf probe -L icmp_rcv | grep -U8 '^ *43\>'
                                    goto error;
                    }
         38         if (!pskb_pull(skb, sizeof(*icmph)))
                            goto error;
                    icmph = icmp_hdr(skb);

         43         ICMPMSGIN_INC_STATS_BH(net, icmph->type);
                    /*
                     *      18 is the highest 'known' ICMP type. Anything else is a mystery
                     *
                     *      RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently
                     *                discarded.
                     */
         50         if (icmph->type > NR_ICMP_TYPES)
                            goto error;

    $ perf probe icmp_rcv:43 'type=icmph->type'

(2) $ cat trace-icmp.py
    [...]
    def trace_begin():
            print "in trace_begin"

    def trace_end():
            print "in trace_end"

    def probe__icmp_rcv(event_name, context, common_cpu,
            common_secs, common_nsecs, common_pid, common_comm,
            __probe_ip, type):
                    print_header(event_name, common_cpu, common_secs, common_nsecs,
                            common_pid, common_comm)

                    print "__probe_ip=%u, type=%u\n" % \
                    (__probe_ip, type),
    [...]

(3) $ perf record -a -D -e probe:icmp_rcv -o - | \
      perf script -i - -s trace-icmp.py

Thanks to Peter Zijlstra for pointing how to do it.

Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>, Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <20110112140613.GA11698@tugrik.mns.mnsspb.ru>
Signed-off-by: Kirill Smelkov <kirr@mns.spb.ru>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf sched: Fix list of events, dropping unsupported ':r' modifier
Stephane Eranian [Wed, 12 Jan 2011 09:29:05 +0000 (10:29 +0100)]
perf sched: Fix list of events, dropping unsupported ':r' modifier

Looks to me like the :r modifier is not supported anymore, so remove it from
the list of events.

Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
LKML-Reference: <AANLkTim=jawJyBj0iFd0r4-LCKzvjFW+NddzJMD5GUB9@mail.gmail.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoRevert "perf tools: Emit clearer message for sys_perf_event_open ENOENT return"
Arnaldo Carvalho de Melo [Tue, 11 Jan 2011 18:58:54 +0000 (16:58 -0200)]
Revert "perf tools: Emit clearer message for sys_perf_event_open ENOENT return"

This reverts commit aa7bc7ef73efc46d7c3a0e185eefaf85744aec98.

It removed the fallback from hardware profiling to software profiling.
.e.g., in a VM with no PMU.

Reported-by: David Ahern <daahern@cisco.com>
Cc: David Ahern <daahern@cisco.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf top: Fix annotate segv
Arnaldo Carvalho de Melo [Tue, 11 Jan 2011 17:16:52 +0000 (15:16 -0200)]
perf top: Fix annotate segv

Before we had sym_counter, it was initialized to zero and we used that
as an index in the global attrs variable, now we have a list of evsel
entries, and sym_counter became sym_evsel, that remained initialized to
zero (NULL): b00m.

Fix it by initializing it to the first entry in the evsel list.

Bug-introduced: 69aad6f
Reported-by: Kirill Smelkov <kirr@mns.spb.ru>
Tested-by: Kirill Smelkov <kirr@mns.spb.ru>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Kirill Smelkov <kirr@mns.spb.ru>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evsel: Fix order of event list deletion
Arnaldo Carvalho de Melo [Tue, 11 Jan 2011 14:42:00 +0000 (12:42 -0200)]
perf evsel: Fix order of event list deletion

We need to defer calling perf_evsel_list__delete() till after atexit
registered routines, because we need to traverse the events being
recorded at that time at least on 'perf record'.

This fixes the problem reported by Thomas Renninger where cmd_record
called by cmd_timechart would not write the tracing data to the perf.data
file header because the evsel_list at atexit (control+C on 'perf timechart
record') time would be empty, being already deleted by run_builtin(),
and thus 'perf timechart' when trying to process such perf.data file would
die with:

"no trace data in the file"

Problem introduced in 70d544d.

Reported-by: Thomas Renninger <trenn@suse.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Renninger <trenn@suse.de>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf session: Fix infinite loop in __perf_session__process_events
Arnaldo Carvalho de Melo [Mon, 10 Jan 2011 23:37:57 +0000 (21:37 -0200)]
perf session: Fix infinite loop in __perf_session__process_events

In this if statement:

        if (head + event->header.size >= mmap_size) {
                if (mmaps[map_idx]) {
                        munmap(mmaps[map_idx], mmap_size);
                        mmaps[map_idx] = NULL;
                }

                page_offset = page_size * (head / page_size);
                file_offset += page_offset;
                head -= page_offset;
                goto remap;
        }

With, for instance, these values:

head=2992
event->header.size=48
mmap_size=3040

We end up endlessly looping back to remap. Off by one.

Problem introduced in 55b4462.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Ingo Molnar <mingo@elte.hu>
Reported-by: David Ahern <daahern@cisco.com>
Bisected-by: David Ahern <daahern@cisco.com>
Tested-by: David Ahern <daahern@cisco.com>
Cc: David Ahern <daahern@cisco.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evsel: Support perf_evsel__open(cpus > 1 && threads > 1)
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 13:55:27 +0000 (11:55 -0200)]
perf evsel: Support perf_evsel__open(cpus > 1 && threads > 1)

And a test for it:

[acme@felicio linux]$ perf test
 1: vmlinux symtab matches kallsyms: Ok
 2: detect open syscall event: Ok
 3: detect open syscall event on all cpus: Ok
[acme@felicio linux]$

Translating C the test does:

1. generates different number of open syscalls on each CPU
   by using sched_setaffinity
2. Verifies that the expected number of events is generated
   on each CPU

It works as expected.

LKML-Reference: <new-submission>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf sched: Use PTHREAD_STACK_MIN to avoid pthread_attr_setstacksize() fail
Jiri Pirko [Mon, 10 Jan 2011 16:14:23 +0000 (14:14 -0200)]
perf sched: Use PTHREAD_STACK_MIN to avoid pthread_attr_setstacksize() fail

on ppc64:
/usr/include/bits/local_lim.h:#define PTHREAD_STACK_MIN 131072

therefore following set of commands:

gives:
perf.2.6.37test: builtin-sched.c:493: create_tasks: Assertion `!(err)' failed.

So make sure we do not set stack size lower than PTHREAD_STACK_MIN.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <20110110160417.GB2685@psychotron.brq.redhat.com>
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Emit clearer message for sys_perf_event_open ENOENT return
Arnaldo Carvalho de Melo [Mon, 10 Jan 2011 15:36:24 +0000 (13:36 -0200)]
perf tools: Emit clearer message for sys_perf_event_open ENOENT return

Improve sys_perf_event_open ENOENT return handling in top and record, just
like 5a3446b does for stat.

Cc: David Ahern <daahern@cisco.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf stat: better error message for unsupported events
David Ahern [Sun, 9 Jan 2011 18:21:12 +0000 (11:21 -0700)]
perf stat: better error message for unsupported events

For unsupported events (e.g., H/W events when running in a VM)
perf stat currently fails with the error message:

      Error: open_counter returned with 2 (No such file or directory).
    /bin/dmesg may provide additional information.

      Fatal: Not all events could be opened.

dmesg is of no help and it is not clear as to why it fails to
open the counter. This patch changes the error message to

      Error: cache-misses event is not supported.
      Fatal: Not all events could be opened.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: a.p.zijlstra@chello.nl
LPU-Reference: <1294597272-17335-1-git-send-email-daahern@cisco.com>
Signed-off-by: David Ahern <daahern@cisco.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf sched: Fix allocation result check
Arnaldo Carvalho de Melo [Mon, 10 Jan 2011 12:48:47 +0000 (10:48 -0200)]
perf sched: Fix allocation result check

Bug introduced in ce47dc56.

Reported-by: Mike Galbraith <efault@gmx.de>
Cc: Chris Samuel <chris@csamuel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoMerge branch 'tip/perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Ingo Molnar [Sun, 9 Jan 2011 09:42:21 +0000 (10:42 +0100)]
Merge branch 'tip/perf/core' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into perf/urgent

13 years agoperf, x86: P4 PMU - Fix unflagged overflows handling
Cyrill Gorcunov [Fri, 7 Jan 2011 18:42:06 +0000 (21:42 +0300)]
perf, x86: P4 PMU - Fix unflagged overflows handling

Don found that P4 PMU reads CCCR register instead of counter
itself (in attempt to catch unflagged event) this makes P4
NMI handler to consume all NMIs it observes. So the other
NMI users such as kgdb simply have no chance to get NMI
on their hands.

Side note: at moment there is no way to run nmi-watchdog
together with perf tool. This is because both 'perf top' and
nmi-watchdog use same event. So while nmi-watchdog reserves
one event/counter for own needs there is no room for perf tool
left (there is a way to disable nmi-watchdog on boot of course).

Ming has tested this patch with the following results

 | 1. watchdog disabled
 |
 | kgdb tests on boot OK
 | perf works OK
 |
 | 2. watchdog enabled, without patch perf-x86-p4-nmi-4
 |
 | kgdb tests on boot hang
 |
 | 3. watchdog enabled, without patch perf-x86-p4-nmi-4 and do not run kgdb
 | tests on boot
 |
 | "perf top" partialy works
 |   cpu-cycles            no
 |   instructions          yes
 |   cache-references      no
 |   cache-misses          no
 |   branch-instructions   no
 |   branch-misses         yes
 |   bus-cycles            no
 |
 | 4. watchdog enabled, with patch perf-x86-p4-nmi-4 applied
 |
 | kgdb tests on boot OK
 | perf does not work, NMI "Dazed and confused" messages show up
 |

Which means we still have problems with p4 box due to 'unknown'
nmi happens but at least it should fix kgdb test cases.

Reported-by: Jason Wessel <jason.wessel@windriver.com>
Reported-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Acked-by: Don Zickus <dzickus@redhat.com>
Acked-by: Lin Ming <ming.m.lin@intel.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <4D275E7E.3040903@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agodynamic debug: Fix build issue with older gcc
Jason Baron [Fri, 7 Jan 2011 18:36:58 +0000 (13:36 -0500)]
dynamic debug: Fix build issue with older gcc

On older gcc (3.3) dynamic debug fails to compile:

include/net/inet_connection_sock.h: In function `inet_csk_reset_xmit_timer':
include/net/inet_connection_sock.h:236: error: duplicate label declaration `do_printk'
include/net/inet_connection_sock.h:219: error: this is a previous declaration
include/net/inet_connection_sock.h:236: error: duplicate label declaration `out'
include/net/inet_connection_sock.h:219: error: this is a previous declaration
include/net/inet_connection_sock.h:236: error: duplicate label `do_printk'
include/net/inet_connection_sock.h:236: error: duplicate label `out'

Fix, by reverting the usage of JUMP_LABEL() in dynamic debug for now.

Cc: <stable@kernel.org>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Tested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Jason Baron <jbaron@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing: Fix TRACE_EVENT power tracepoint creation
Mathieu Desnoyers [Thu, 28 Oct 2010 15:31:17 +0000 (11:31 -0400)]
tracing: Fix TRACE_EVENT power tracepoint creation

DEFINE_TRACE should also exist when CONFIG_EVENT_TRACING=n. Otherwise, setting
only TRACEPOINTS=y is broken.

Acked-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
LKML-Reference: <20101028153117.GA4051@Krystal>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing: Fix preempt count leak
Li Zefan [Thu, 9 Dec 2010 07:47:56 +0000 (15:47 +0800)]
tracing: Fix preempt count leak

While running my ftrace stress test, this showed up:

BUG: sleeping function called from invalid context at mm/mmap.c:233
...
note: cat[3293] exited with preempt_count 1

The bug was introduced by commit 91e86e560d0b3ce4c5fc64fd2bbb99f856a30a4e
("tracing: Fix recursive user stack trace")

Cc: <stable@kernel.org>
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <4D0089AC.1020802@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracepoint: Add __rcu annotation
Lai Jiangshan [Tue, 4 Jan 2011 08:06:09 +0000 (16:06 +0800)]
tracepoint: Add __rcu annotation

Add __rcu annotation to :
(struct tracepoint)->funcs

Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
LKML-Reference: <4D22D4F1.50505@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing: remove duplicate null-pointer check in skb tracepoint
Mathieu Desnoyers [Thu, 6 Jan 2011 17:53:19 +0000 (12:53 -0500)]
tracing: remove duplicate null-pointer check in skb tracepoint

The check for NULL skb in the kfree_skb trace event is a duplicate from the
check already done in its only caller, kfree_skb(). Remove this duplicate check.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
LKML-Reference: <20110106175319.GA30610@Krystal>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: David S. Miller <davem@davemloft.net>
CC: Steven Rostedt <rostedt@goodmis.org>
CC: Frederic Weisbecker <fweisbec@gmail.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Zhaolei <zhaolei@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing/trivial: Add missing comma in TRACE_EVENT comment
Mathieu Desnoyers [Thu, 6 Jan 2011 18:45:32 +0000 (13:45 -0500)]
tracing/trivial: Add missing comma in TRACE_EVENT comment

Add missing comma within the TRACE_EVENT() example in tracepoint.h.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
LKML-Reference: <20110106184532.GA2526@Krystal>
CC: Frederic Weisbecker <fweisbec@gmail.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agotracing: Include module.h in define_trace.h
Steven Rostedt [Fri, 7 Jan 2011 20:40:10 +0000 (15:40 -0500)]
tracing: Include module.h in define_trace.h

While doing some developing, Peter Zijlstra and I have found
that if a CREATE_TRACE_POINTS include is done before module.h
is included, it can break the build.

We have been lucky so far that this has not broke the build
since module.h is included in almost everything.

Reported-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
13 years agox86: Save rbp in pt_regs on irq entry
Frederic Weisbecker [Thu, 6 Jan 2011 14:22:47 +0000 (15:22 +0100)]
x86: Save rbp in pt_regs on irq entry

From the x86_64 low level interrupt handlers, the frame pointer is
saved right after the partial pt_regs frame.

rbp is not supposed to be part of the irq partial saved registers,
but it only requires to extend the pt_regs frame by 8 bytes to
do so, plus a tiny stack offset fixup on irq exit.

This changes a bit the semantics or get_irq_entry() that is supposed
to provide only the value of caller saved registers and the cpu
saved frame. However it's a win for unwinders that can walk through
stack frames on top of get_irq_regs() snapshots.

A noticeable impact is that it makes perf events cpu-clock and
task-clock events based callchains working on x86_64.

Let's then save rbp into the irq pt_regs.

As a result with:

perf record -e cpu-clock perf bench sched messaging
perf report --stdio

Before:
    20.94%             perf  [kernel.kallsyms]        [k] lock_acquire
                       |
                       --- lock_acquire
                          |
                          |--44.01%-- __write_nocancel
                          |
                          |--43.18%-- __read
                          |
                          |--6.08%-- fork
                          |          create_worker
                          |
                          |--0.88%-- _dl_fixup
                          |
                          |--0.65%-- do_lookup_x
                          |
                          |--0.53%-- __GI___libc_read
                           --4.67%-- [...]

After:
    19.23%         perf  [kernel.kallsyms]    [k] __lock_acquire
                   |
                   --- __lock_acquire
                      |
                      |--97.74%-- lock_acquire
                      |          |
                      |          |--21.82%-- _raw_spin_lock
                      |          |          |
                      |          |          |--37.26%-- unix_stream_recvmsg
                      |          |          |          sock_aio_read
                      |          |          |          do_sync_read
                      |          |          |          vfs_read
                      |          |          |          sys_read
                      |          |          |          system_call
                      |          |          |          __read
                      |          |          |
                      |          |          |--24.09%-- unix_stream_sendmsg
                      |          |          |          sock_aio_write
                      |          |          |          do_sync_write
                      |          |          |          vfs_write
                      |          |          |          sys_write
                      |          |          |          system_call
                      |          |          |          __write_nocancel

v2: Fix cfi annotations.

Reported-by: Soeren Sandmann Pedersen <sandmann@redhat.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: H. Peter Anvin <hpa@zytor.com
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Jan Beulich <JBeulich@novell.com>
13 years agox86, dumpstack: Fix unused variable warning
Rakib Mullick [Fri, 7 Jan 2011 15:59:49 +0000 (16:59 +0100)]
x86, dumpstack: Fix unused variable warning

In dump_stack function, bp isn't used anymore, which is introduced by
commit 9c0729dc8062bed96189bd14ac6d4920f3958743. This patch removes bp
completely.

Signed-off-by: Rakib Mullick <rakib.mullick@gmail.com>
Cc: Soeren Sandmann <sandmann@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
LKML-Reference: <AANLkTik9U_Z0WSZ7YjrykER_pBUfPDdgUUmtYx=R74nL@mail.gmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
13 years agox86, NMI: Clean-up default_do_nmi()
Don Zickus [Thu, 6 Jan 2011 21:18:52 +0000 (16:18 -0500)]
x86, NMI: Clean-up default_do_nmi()

Just re-arrange the code a bit to make it easier to follow what is
going on.  Basically un-negating the if-statement and swapping the code
inside the if-statement with code outside.

No functional changes.

Originally-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1294348732-15030-7-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86, NMI: Allow NMI reason io port (0x61) to be processed on any CPU
Don Zickus [Thu, 6 Jan 2011 21:18:51 +0000 (16:18 -0500)]
x86, NMI: Allow NMI reason io port (0x61) to be processed on any CPU

In original NMI handler, NMI reason io port (0x61) is only processed
on BSP.  This makes it impossible to hot-remove BSP.  To solve the
issue, a raw spinlock is used to allow the port to be processed on any
CPU.

Originally-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1294348732-15030-6-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86, NMI: Remove DIE_NMI_IPI
Don Zickus [Thu, 6 Jan 2011 21:18:50 +0000 (16:18 -0500)]
x86, NMI: Remove DIE_NMI_IPI

With priorities in place and no one really understanding the difference between
DIE_NMI and DIE_NMI_IPI, just remove DIE_NMI_IPI and convert everyone to DIE_NMI.

This also simplifies default_do_nmi() a little bit.  Instead of calling the
die_notifier in both the if and else part, just pull it out and call it before
the if-statement.  This has the side benefit of avoiding a call to the ioport
to see if there is an external NMI sitting around until after the (more frequent)
internal NMIs are dealt with.

Patch-Inspired-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1294348732-15030-5-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86, NMI: Add priorities to handlers
Don Zickus [Thu, 6 Jan 2011 21:18:49 +0000 (16:18 -0500)]
x86, NMI: Add priorities to handlers

In order to consolidate the NMI die_chain events, we need to setup the priorities
for the die notifiers.

I started by defining a bunch of common priorities that can be used by the
notifier blocks.  Then I modified the notifier blocks to use the newly created
priorities.

Now that the priorities are straightened out, it should be easier to remove the
event DIE_NMI_IPI.

Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1294348732-15030-4-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: Convert some devices to use DIE_NMIUNKNOWN
Don Zickus [Thu, 6 Jan 2011 21:18:48 +0000 (16:18 -0500)]
x86: Convert some devices to use DIE_NMIUNKNOWN

They are a handful of places in the code that register a die_notifier
as a catch all in case no claims the NMI.  Unfortunately, they trigger
on events like DIE_NMI and DIE_NMI_IPI, which depending on when they
registered may collide with other handlers that have the ability to
determine if the NMI is theirs or not.

The function unknown_nmi_error() makes one last effort to walk the
die_chain when no one else has claimed the NMI before spitting out
messages that the NMI is unknown.

This is a better spot for these devices to execute any code without
colliding with the other handlers.

The two drivers modified are only compiled on x86 arches I believe, so
they shouldn't be affected by other arches that may not have
DIE_NMIUNKNOWN defined.

Signed-off-by: Don Zickus <dzickus@redhat.com>
Cc: Russ Anderson <rja@sgi.com>
Cc: Corey Minyard <minyard@acm.org>
Cc: openipmi-developer@lists.sourceforge.net
Cc: dann frazier <dannf@hp.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1294348732-15030-3-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86, NMI: Add NMI symbol constants and rename memory parity to PCI SERR
Huang Ying [Thu, 6 Jan 2011 21:18:47 +0000 (16:18 -0500)]
x86, NMI: Add NMI symbol constants and rename memory parity to PCI SERR

Replace the NMI related magic numbers with symbol constants.

Memory parity error is only valid for IBM PC-AT, newer machine use
bit 7 (0x80) of 0x61 port for PCI SERR. While memory error is usually
reported via MCE. So corresponding function name and kernel log string
is changed.

But on some machines, PCI SERR line is still used to report memory
errors. This is used by EDAC, so corresponding EDAC call is reserved.

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1294348732-15030-2-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf_events: Add perf_event_time()
Stephane Eranian [Mon, 3 Jan 2011 16:20:01 +0000 (18:20 +0200)]
perf_events: Add perf_event_time()

Adds perf_event_time() to try and centralize access to event
timing and in particular ctx->time. Prepares for cgroup support.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4d22059c.122ae30a.5e0e.ffff8b8b@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf_events: Generalize use of event_filter_match()
Stephane Eranian [Mon, 3 Jan 2011 16:20:01 +0000 (18:20 +0200)]
perf_events: Generalize use of event_filter_match()

Replace all occurrences of:
event->cpu != -1 && event->cpu == smp_processor_id()
by a call to:
event_filter_match(event)

This makes the code more consistent and will make the cgroup
patch smaller.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4d220593.2308e30a.48c5.ffff8ae9@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf_events: Move code around to prepare for cgroup
Stephane Eranian [Mon, 3 Jan 2011 16:20:01 +0000 (18:20 +0200)]
perf_events: Move code around to prepare for cgroup

In particular this patch move perf_event_exit_task() before
cgroup_exit() to allow for cgroup support. The cgroup_exit()
function detaches the cgroups attached to a task.

Other movements include hoisting some definitions and inlines
at the top of perf_event.c

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4d22058b.cdace30a.4657.ffff95b1@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf tools: Pass whole attr to event selectors
Lin Ming [Fri, 7 Jan 2011 03:11:09 +0000 (11:11 +0800)]
perf tools: Pass whole attr to event selectors

Since commit 69aad6f1(perf tools: Introduce event selectors), only
perf_event_attr::type and ::config are passed to event selector, which
makes perf tool not work correctly.

For example, PEBS does not work because perf_event_attr::precise_ip is
not passed to the syscall.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1294369869.20563.19.camel@minggr.sh.intel.com>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Build with frame pointer
Frederic Weisbecker [Thu, 6 Jan 2011 14:51:53 +0000 (15:51 +0100)]
perf tools: Build with frame pointer

It seems that some gcc versions build by default with frame pointers
and some others omit them.

Just build the tools with frame pointers as the callchains can be an
important part of the perf workflow.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1294325513-14276-3-git-send-email-fweisbec@gmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Fix buffer overflow error when specifying all tracepoints
Han Pingtian [Thu, 6 Jan 2011 09:39:22 +0000 (17:39 +0800)]
perf tools: Fix buffer overflow error when specifying all tracepoints

I found when specifying all tracepoints with -e to one of subcommand,
such as 'stat', the program will trigger a buffer overflow error, like
this:

*** buffer overflow detected ***: ./perf terminated
======= Backtrace: =========
/lib64/libc.so.6(__fortify_fail+0x37)[0x382cefb2c7]
....

The tracepoints are separated by comma, something like this:

$ perf stat -a -e `perf list |grep Tracepoint|awk -F'[' '{gsub(/[[:space:]]+/,"",$1);array[FNR]=$1}END{outputs=array[1];for (i=2;i<=FNR;i++){ outputs=outputs "," array[i];};print outputs}'`

The root reason of this problem is that store_event_type() is called for all
events, and will overflow the 'filename' at:

    strncat(filename, orgname, strlen(orgname));

This patch fixes it by calling store_event_type() only when the event name has
been found.

LKML-Reference: <20110106093922.GB6713@hpt.nay.redhat.com>
Signed-off-by: Han Pingtian <phan@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoMerge branch 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 19:12:17 +0000 (11:12 -0800)]
Merge branch 'x86-mm-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, mm: Initialize initial_page_table before paravirt jumps

13 years agoMerge branches 'x86-alternatives-for-linus', 'x86-fpu-for-linus', 'x86-hwmon-for...
Linus Torvalds [Thu, 6 Jan 2011 19:11:50 +0000 (11:11 -0800)]
Merge branches 'x86-alternatives-for-linus', 'x86-fpu-for-linus', 'x86-hwmon-for-linus', 'x86-paravirt-for-linus', 'core-locking-for-linus' and 'irq-core-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-alternatives-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, suspend: Avoid unnecessary smp alternatives switch during suspend/resume

* 'x86-fpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86-64, asm: Use fxsaveq/fxrestorq in more places

* 'x86-hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, hwmon: Add core threshold notification to therm_throt.c

* 'x86-paravirt-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, paravirt: Use native_halt on a halt, not native_safe_halt

* 'core-locking-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  locking, lockdep: Convert sprintf_symbol to %pS

* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  irq: Better struct irqaction layout

13 years agoMerge branch 'x86-uv-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 19:09:57 +0000 (11:09 -0800)]
Merge branch 'x86-uv-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-uv-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, UV, BAU: Extend for more than 16 cpus per socket
  x86, UV: Fix the effect of extra bits in the hub nodeid register
  x86, UV: Add common uv_early_read_mmr() function for reading MMRs

13 years agoMerge branch 'x86-tsc-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 19:08:14 +0000 (11:08 -0800)]
Merge branch 'x86-tsc-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-tsc-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Check tsc available/disabled in the delayed init function
  x86: Improve TSC calibration using a delayed workqueue
  x86: Make tsc=reliable override boot time stability checks

13 years agoMerge branch 'x86-security-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 6 Jan 2011 19:07:33 +0000 (11:07 -0800)]
Merge branch 'x86-security-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-security-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  module: Move RO/NX module protection to after ftrace module update
  x86: Resume trampoline must be executable
  x86: Add RO/NX protection for loadable kernel modules
  x86: Add NX protection for kernel data
  x86: Fix improper large page preservation

13 years agoMerge branch 'x86-platform-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 6 Jan 2011 19:06:31 +0000 (11:06 -0800)]
Merge branch 'x86-platform-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-platform-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, earlyprintk: Move mrst early console to platform/ and fix a typo
  x86, apbt: Setup affinity for apb timers acting as per-cpu timer
  ce4100: Add errata fixes for UART on CE4100
  x86: platform: Move iris to x86/platform where it belongs
  x86, mrst: Check platform_device_register() return code
  x86/platform: Add Eurobraille/Iris power off support
  x86, mrst: Add explanation for using 1960 as the year offset for vrtc
  x86, mrst: Fix dependencies of "select INTEL_SCU_IPC"
  x86, mrst: The shutdown for MRST requires the SCU IPC mechanism
  x86: Ce4100: Add reboot_fixup() for CE4100
  ce4100: Add PCI register emulation for CE4100
  x86: Add CE4100 platform support
  x86: mrst: Set vRTC's IRQ to level trigger type
  x86: mrst: Add audio driver bindings
  rtc: Add drivers/rtc/rtc-mrst.c
  x86: mrst: Add vrtc driver which serves as a wall clock device
  x86: mrst: Add Moorestown specific reboot/shutdown support
  x86: mrst: Parse SFI timer table for all timer configs
  x86/mrst: Add SFI platform device parsing code

13 years agoMerge branch 'x86-microcode-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 6 Jan 2011 19:06:09 +0000 (11:06 -0800)]
Merge branch 'x86-microcode-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-microcode-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, microcode, AMD: Cleanup code a bit
  x86, microcode, AMD: Replace vmalloc+memset with vzalloc

13 years agoMerge branch 'x86-mce-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 19:05:21 +0000 (11:05 -0800)]
Merge branch 'x86-mce-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-mce-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  apic, amd: Make firmware bug messages more meaningful
  mce, amd: Remove goto in threshold_create_device()
  mce, amd: Add helper functions to setup APIC
  mce, amd: Shorten local variables mci_misc_{hi,lo}
  mce, amd: Implement mce_threshold_block_init() helper function

13 years agoMerge branch 'x86-cpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 18:56:02 +0000 (10:56 -0800)]
Merge branch 'x86-cpu-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-cpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Fix included-by file reference comments
  x86, cpu: Only CPU features determine NX capabilities
  x86, cpu: Call verify_cpu during 32bit CPU startup
  x86, cpu: Clear XD_DISABLED flag on Intel to regain NX
  x86, cpu: Rename verify_cpu_64.S to verify_cpu.S

13 years agoMerge branch 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 18:51:36 +0000 (10:51 -0800)]
Merge branch 'x86-apic-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86: Fix APIC ID sizing bug on larger systems, clean up MAX_APICS confusion
  x86, acpi: Parse all SRAT cpu entries even above the cpu number limitation
  x86, acpi: Add MAX_LOCAL_APIC for 32bit
  x86: io_apic: Split setup_ioapic_ids_from_mpc()
  x86: io_apic: Fix CONFIG_X86_IO_APIC=n breakage
  x86: apic: Move probe_nr_irqs_gsi() into ioapic_init_mappings()
  x86: Allow platforms to force enable apic

13 years agoMerge branch 'x86-amd-nb-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 6 Jan 2011 18:50:28 +0000 (10:50 -0800)]
Merge branch 'x86-amd-nb-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-amd-nb-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, cacheinfo: Cleanup L3 cache index disable support
  x86, amd-nb: Cleanup AMD northbridge caching code
  x86, amd-nb: Complete the rename of AMD NB and related code

13 years agoMerge branch 'timers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 18:42:43 +0000 (10:42 -0800)]
Merge branch 'timers-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'timers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  MAINTAINERS: Update timer related entries
  timers: Use this_cpu_read
  timerqueue: Make timerqueue_getnext() static inline
  hrtimer: fix timerqueue conversion flub
  hrtimers: Convert hrtimers to use timerlist infrastructure
  timers: Fixup allmodconfig build issue
  timers: Rename timerlist infrastructure to timerqueue
  timers: Introduce timerlist infrastructure.
  hrtimer: Remove stale comment on curr_timer
  timer: Warn when del_timer_sync() is called in hardirq context
  timer: Del_timer_sync() can be used in softirq context
  timer: Make try_to_del_timer_sync() the same on SMP and UP
  posix-timers: Annotate lock_timer()
  timer: Permit statically-declared work with deferrable timers
  time: Use ARRAY_SIZE macro in timecompare.c
  timer: Initialize the field slack of timer_list
  timer_list: Remove alignment padding on 64 bit when CONFIG_TIMER_STATS
  time: Compensate for rounding on odd-frequency clocksources

Fix up trivial conflict in MAINTAINERS

13 years agoMerge branch 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 6 Jan 2011 18:23:33 +0000 (10:23 -0800)]
Merge branch 'sched-core-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (30 commits)
  sched: Change wait_for_completion_*_timeout() to return a signed long
  sched, autogroup: Fix reference leak
  sched, autogroup: Fix potential access to freed memory
  sched: Remove redundant CONFIG_CGROUP_SCHED ifdef
  sched: Fix interactivity bug by charging unaccounted run-time on entity re-weight
  sched: Move periodic share updates to entity_tick()
  printk: Use this_cpu_{read|write} api on printk_pending
  sched: Make pushable_tasks CONFIG_SMP dependant
  sched: Add 'autogroup' scheduling feature: automated per session task groups
  sched: Fix unregister_fair_sched_group()
  sched: Remove unused argument dest_cpu to migrate_task()
  mutexes, sched: Introduce arch_mutex_cpu_relax()
  sched: Add some clock info to sched_debug
  cpu: Remove incorrect BUG_ON
  cpu: Remove unused variable
  sched: Fix UP build breakage
  sched: Make task dump print all 15 chars of proc comm
  sched: Update tg->shares after cpu.shares write
  sched: Allow update_cfs_load() to update global load
  sched: Implement demand based update_cfs_load()
  ...

13 years agoMerge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 18:17:26 +0000 (10:17 -0800)]
Merge branch 'perf-core-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (146 commits)
  tools, perf: Documentation for the power events API
  perf: Add calls to suspend trace point
  perf script: Make some lists static
  perf script: Use the default lost event handler
  perf session: Warn about errors when processing pipe events too
  perf tools: Fix perf_event.h header usage
  perf test: Clarify some error reports in the open syscall test
  x86, NMI: Add touch_nmi_watchdog to io_check_error delay
  x86: Avoid calling arch_trigger_all_cpu_backtrace() at the same time
  x86: Only call smp_processor_id in non-preempt cases
  perf timechart: Adjust perf timechart to the new power events
  perf: Clean up power events by introducing new, more generic ones
  perf: Do not export power_frequency, but power_start event
  perf test: Add test for counting open syscalls
  perf evsel: Auto allocate resources needed for some methods
  perf evsel: Use {cpu,thread}_map to shorten list of parameters
  perf tools: Refactor all_tids to hold nr and the map
  perf tools: Refactor cpumap to hold nr and the map
  perf evsel: Introduce per cpu and per thread open helpers
  perf evsel: Steal the counter reading routines from stat
  ...

13 years agoMerge branch 'core-futexes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 6 Jan 2011 18:07:05 +0000 (10:07 -0800)]
Merge branch 'core-futexes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-futexes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  futex: Add futex_q static initializer
  futex: Replace fshared and clockrt with combined flags
  futex: Cleanup stale fshared flag interfaces

13 years agoMerge branch 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 6 Jan 2011 18:06:26 +0000 (10:06 -0800)]
Merge branch 'core-rcu-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  rcu: remove unused __list_for_each_rcu() macro
  rculist: fix borked __list_for_each_rcu() macro
  rcu: reduce __call_rcu()-induced contention on rcu_node structures
  rcu: limit rcu_node leaf-level fanout
  rcu: fine-tune grace-period begin/end checks
  rcu: Keep gpnum and completed fields synchronized
  rcu: Stop chasing QS if another CPU did it for us
  rcu: increase synchronize_sched_expedited() batching
  rcu: Make synchronize_srcu_expedited() fast if running readers
  rcu: fix race condition in synchronize_sched_expedited()
  rcu: update documentation/comments for Lai's adoption patch
  rcu,cleanup: simplify the code when cpu is dying
  rcu,cleanup: move synchronize_sched_expedited() out of sched.c
  rcu: get rid of obsolete "classic" names in TREE_RCU tracing
  rcu: Distinguish between boosting and boosted
  rcu: document TINY_RCU and TINY_PREEMPT_RCU tracing.
  rcu: add tracing for TINY_RCU and TINY_PREEMPT_RCU
  rcu: priority boosting for TINY_PREEMPT_RCU
  rcu: move TINY_RCU from softirq to kthread
  rcu: add priority-inversion testing to rcutorture

13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw
Linus Torvalds [Thu, 6 Jan 2011 18:01:23 +0000 (10:01 -0800)]
Merge git://git./linux/kernel/git/steve/gfs2-2.6-nmw

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw:
  GFS2: Don't flush delete workqueue when releasing the transaction lock
  GFS2: fsck.gfs2 reported statfs error after gfs2_grow
  GFS2: Merge glock state fields into a bitfield
  GFS2: Fix uninitialised error value in previous patch
  GFS2: fix recursive locking during rindex truncates
  GFS2: reread rindex when necessary to grow rindex
  GFS2: Remove duplicate #defines from glock.h
  GFS2: Clean up of gdlm_lock function
  GFS2: Allow gfs2 to update quota usage values through the quotactl interface
  GFS2: fs/gfs2/glock.h: Add __attribute__((format(printf,2,3)) to gfs2_print_dbg
  GFS2: fs/gfs2/glock.c: Use printf extension %pV
  GFS2: Clean up duplicated setattr code
  GFS2: Remove unreachable calls to vmtruncate
  GFS2: fs/gfs2/glock.c: Convert sprintf_symbol to %pS
  GFS2: Change two WQ_RESCUERs into WQ_MEM_RECLAIM

13 years agokernel-doc: code reorganization
Randy Dunlap [Thu, 6 Jan 2011 00:28:43 +0000 (16:28 -0800)]
kernel-doc: code reorganization

Move 'main' code vs. subroutines around so that they are not so
intermixed, for better readability/understanding (relative to Perl).
It was messy to follow the primary flow of code execution with the
code being mixed.  Now the code begins with data initialization,
followed by all subroutines, then ends with the main code execution.

This is almost totally source code movement, with a few changes as
needed for forward declarations.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoDocumentation: update kernel-docs.txt
Nicolas Kaiser [Thu, 6 Jan 2011 00:27:53 +0000 (16:27 -0800)]
Documentation: update kernel-docs.txt

Fixed typos, and removed duplicated entries.

Signed-off-by: Nicolas Kaiser <nikai@nikai.net>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoDocumentation/dontdiff: add further autogenerated files to ignore list
Michael Prokop [Thu, 6 Jan 2011 00:27:15 +0000 (16:27 -0800)]
Documentation/dontdiff: add further autogenerated files to ignore list

Mainly resulting from (but not limited to) autogenerated files of
lib/raid6 and drivers/gpu/drm/radeon. List generated as result of
a diff of a clean 2.6.36 tree against a built one.

Signed-off-by: Michael Prokop <mika@grml.org>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agotools, perf: Documentation for the power events API
Jean Pihet [Wed, 5 Jan 2011 18:49:02 +0000 (19:49 +0100)]
tools, perf: Documentation for the power events API

Provides documentation for the following:
- the new power trace API,
- the old (legacy) power trace API,
- the DEPRECATED Kconfig option usage.

Signed-off-by: Jean Pihet <j-pihet@ti.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: trenn@suse.de
Cc: Len Brown <len.brown@intel.com>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: linux-pm@lists.linux-foundation.org
LKML-Reference: <1294253342-29056-3-git-send-email-j-pihet@ti.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf: Add calls to suspend trace point
Jean Pihet [Wed, 5 Jan 2011 18:49:01 +0000 (19:49 +0100)]
perf: Add calls to suspend trace point

Uses the machine_suspend trace point, called from the
generic kernel suspend_devices_and_enter function.

Signed-off-by: Jean Pihet <j-pihet@ti.com>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Arjan van de Ven <arjan@linux.intel.com>
CC: Thomas Renninger <trenn@suse.de>
Cc: Len Brown <len.brown@intel.com>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: linux-pm@lists.linux-foundation.org
LKML-Reference: <1294253342-29056-2-git-send-email-j-pihet@ti.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf script: Make some lists static
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 18:32:52 +0000 (16:32 -0200)]
perf script: Make some lists static

Not accessed outside builtin-script, so make them static.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf script: Use the default lost event handler
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 18:27:30 +0000 (16:27 -0200)]
perf script: Use the default lost event handler

That already does what was being done here. The warning is now unconditionally
given by __perf_session__process_pipe_events, just like for non pipe processing.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf session: Warn about errors when processing pipe events too
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 18:25:15 +0000 (16:25 -0200)]
perf session: Warn about errors when processing pipe events too

Just like we do at __perf_session__process_events

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Fix perf_event.h header usage
Stephane Eranian [Tue, 4 Jan 2011 14:30:01 +0000 (16:30 +0200)]
perf tools: Fix perf_event.h header usage

This patch fixes the usage of the perf_event.h header file
between command modules and the supporting code in util.

It is necessary to ensure that ALL files use the SAME
perf_event.h header from the kernel source tree.

There were a couple of #include <linux/perf_event.h> mixed
with #include "../../perf_event.h".

This caused issues on some distros because of mismatch
in the layout of struct perf_event_attr. That eventually
led perf stat to segfault.

Cc: David S. Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@gmail.com>
LKML-Reference: <4d233cf0.2308e30a.7b00.ffffc187@mx.google.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf test: Clarify some error reports in the open syscall test
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 12:40:08 +0000 (10:40 -0200)]
perf test: Clarify some error reports in the open syscall test

Rebooted my devel machine, first thing I ran was perf test, that expects
debugfs to be mounted, test fails. Be more clear about it.

Also add missing newlines and add more informative message when
sys_perf_event_open fails.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agox86, NMI: Add touch_nmi_watchdog to io_check_error delay
Huang Ying [Wed, 5 Jan 2011 03:38:09 +0000 (22:38 -0500)]
x86, NMI: Add touch_nmi_watchdog to io_check_error delay

Prevent the long delay in io_check_error making NMI watchdog
timeout.

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Don Zickus <dzickus@redhat.com>
LKML-Reference: <1294198689-15447-3-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: Avoid calling arch_trigger_all_cpu_backtrace() at the same time
Dongdong Deng [Wed, 5 Jan 2011 03:38:08 +0000 (22:38 -0500)]
x86: Avoid calling arch_trigger_all_cpu_backtrace() at the same time

The spin_lock_debug/rcu_cpu_stall detector uses
trigger_all_cpu_backtrace() to dump cpu backtrace.
Therefore it is possible that trigger_all_cpu_backtrace()
could be called at the same time on different CPUs, which
triggers and 'unknown reason NMI' warning. The following case
illustrates the problem:

      CPU1                    CPU2                     ...   CPU N
                       trigger_all_cpu_backtrace()
                       set "backtrace_mask" to cpu mask
                               |
generate NMI interrupts  generate NMI interrupts       ...
    \                          |                               /
     \                         |                              /

The "backtrace_mask" will be cleaned by the first NMI interrupt
at nmi_watchdog_tick(), then the following NMI interrupts
generated by other cpus's arch_trigger_all_cpu_backtrace() will
be taken as unknown reason NMI interrupts.

This patch uses a test_and_set to avoid the problem, and stop
the arch_trigger_all_cpu_backtrace() from calling to avoid
dumping a double cpu backtrace info when there is already a
trigger_all_cpu_backtrace() in progress.

Signed-off-by: Dongdong Deng <dongdong.deng@windriver.com>
Reviewed-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Cc: fweisbec@gmail.com
LKML-Reference: <1294198689-15447-2-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Don Zickus <dzickus@redhat.com>
13 years agox86: Only call smp_processor_id in non-preempt cases
Don Zickus [Wed, 5 Jan 2011 03:38:07 +0000 (22:38 -0500)]
x86: Only call smp_processor_id in non-preempt cases

There are some paths that walk the die_chain with preemption on.
Make sure we are in an NMI call before we start doing anything.

This was triggered by do_general_protection calling notify_die
with DIE_GPF.

Reported-by: Jan Kiszka <jan.kiszka@web.de>
Signed-off-by: Don Zickus <dzickus@redhat.com>
LKML-Reference: <1294198689-15447-1-git-send-email-dzickus@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoMerge commit 'v2.6.37' into perf/core
Ingo Molnar [Wed, 5 Jan 2011 13:22:08 +0000 (14:22 +0100)]
Merge commit 'v2.6.37' into perf/core

Merge reason: Add the final .37 tree.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agosched: Change wait_for_completion_*_timeout() to return a signed long
NeilBrown [Wed, 5 Jan 2011 01:50:16 +0000 (12:50 +1100)]
sched: Change wait_for_completion_*_timeout() to return a signed long

wait_for_completion_*_timeout() can return:

   0: if the wait timed out
 -ve: if the wait was interrupted
 +ve: if the completion was completed.

As they currently return an 'unsigned long', the last two cases
are not easily distinguished which can easily result in buggy
code, as is the case for the recently added
wait_for_completion_interruptible_timeout() call in
net/sunrpc/cache.c

So change them both to return 'long'.  As MAX_SCHEDULE_TIMEOUT
is LONG_MAX, a large +ve return value should never overflow.

Signed-off-by: NeilBrown <neilb@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: J. Bruce Fields <bfields@fieldses.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
LKML-Reference: <20110105125016.64ccab0e@notabene.brown>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoMerge commit 'v2.6.37' into sched/core
Ingo Molnar [Wed, 5 Jan 2011 13:14:42 +0000 (14:14 +0100)]
Merge commit 'v2.6.37' into sched/core

Merge reason: Merge the final .37 tree.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86: Fix APIC ID sizing bug on larger systems, clean up MAX_APICS confusion
Yinghai Lu [Wed, 5 Jan 2011 00:38:52 +0000 (16:38 -0800)]
x86: Fix APIC ID sizing bug on larger systems, clean up MAX_APICS confusion

Found one x2apic pre-enabled system, x2apic_mode suddenly get
corrupted after register some cpus, when compiled
CONFIG_NR_CPUS=255 instead of 512.

It turns out that generic_processor_info() ==> phyid_set(apicid,
phys_cpu_present_map) causes the problem.

phys_cpu_present_map is sized by MAX_APICS bits, and pre-enabled
system some cpus have an apic id > 255.

The variable after phys_cpu_present_map may get corrupted
silently:

 ffffffff828e8420 B phys_cpu_present_map
 ffffffff828e8440 B apic_verbosity
 ffffffff828e8444 B local_apic_timer_c2_ok
 ffffffff828e8448 B disable_apic
 ffffffff828e844c B x2apic_mode
 ffffffff828e8450 B x2apic_disabled
 ffffffff828e8454 B num_processors
 ...

Actually phys_cpu_present_map is referenced via apic id, instead
index. We should use MAX_LOCAL_APIC instead MAX_APICS.

For 64-bit it will be 32768 in all cases. BSS will increase by 4k bytes
on 64-bit:

text data bss dec filename
21696943 4193748 12787712 38678403 vmlinux.before
21696943 4193748 12791808 38682499 vmlinux.after

No change on 32bit.

Finally we can remove MAX_APCIS that was rather confusing.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
LKML-Reference: <4D23BD9C.3070102@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoLinux 2.6.37
Linus Torvalds [Wed, 5 Jan 2011 00:50:19 +0000 (16:50 -0800)]
Linux 2.6.37

13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Tue, 4 Jan 2011 21:55:49 +0000 (13:55 -0800)]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  ipv4/route.c: respect prefsrc for local routes
  bridge: stp: ensure mac header is set
  bridge: fix br_multicast_ipv6_rcv for paged skbs
  atl1: fix oops when changing tx/rx ring params
  drivers/atm/atmtcp.c: add missing atm_dev_put
  starfire: Fix dma_addr_t size test for MIPS
  tg3: fix return value check in tg3_read_vpd()
  Broadcom CNIC core network driver: fix mem leak on allocation failures in cnic_alloc_uio_rings()
  ISDN, Gigaset: Fix memory leak in do_disconnect_req()
  CAN: Use inode instead of kernel address for /proc file
  skfp: testing the wrong variable in skfp_driver_init()
  ppp: allow disabling multilink protocol ID compression
  ehea: Avoid changing vlan flags
  ueagle-atm: fix PHY signal initialization race

13 years agoipv4/route.c: respect prefsrc for local routes
Joel Sing [Mon, 3 Jan 2011 20:24:20 +0000 (20:24 +0000)]
ipv4/route.c: respect prefsrc for local routes

The preferred source address is currently ignored for local routes,
which results in all local connections having a src address that is the
same as the local dst address. Fix this by respecting the preferred source
address when it is provided for local routes.

This bug can be demonstrated as follows:

 # ifconfig dummy0 192.168.0.1
 # ip route show table local | grep local.*dummy0
 local 192.168.0.1 dev dummy0  proto kernel  scope host  src 192.168.0.1
 # ip route change table local local 192.168.0.1 dev dummy0 \
     proto kernel scope host src 127.0.0.1
 # ip route show table local | grep local.*dummy0
 local 192.168.0.1 dev dummy0  proto kernel  scope host  src 127.0.0.1

We now establish a local connection and verify the source IP
address selection:

 # nc -l 192.168.0.1 3128 &
 # nc 192.168.0.1 3128 &
 # netstat -ant | grep 192.168.0.1:3128.*EST
 tcp        0      0 192.168.0.1:3128        192.168.0.1:33228 ESTABLISHED
 tcp        0      0 192.168.0.1:33228       192.168.0.1:3128  ESTABLISHED

Signed-off-by: Joel Sing <jsing@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
13 years agoremove trim_fs method from Documentation/filesystems/Locking
Christoph Hellwig [Tue, 4 Jan 2011 06:14:24 +0000 (07:14 +0100)]
remove trim_fs method from Documentation/filesystems/Locking

The ->trim_fs has been removed meanwhile, so remove it from the documentation
as well.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agosched, autogroup: Fix reference leak
Mike Galbraith [Fri, 31 Dec 2010 08:32:30 +0000 (09:32 +0100)]
sched, autogroup: Fix reference leak

The cgroup exit mess also uncovered a struct autogroup reference leak.
copy_process() was simply freeing vs putting the signal_struct,
stranding a reference.

Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Oleg Nesterov <oleg@redhat.com>
LKML-Reference: <1293784350.6839.2.camel@marge.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agosched, autogroup: Fix potential access to freed memory
Mike Galbraith [Thu, 16 Dec 2010 14:09:52 +0000 (15:09 +0100)]
sched, autogroup: Fix potential access to freed memory

Oleg pointed out that the /proc interface kref_get() useage may race with
the final put during autogroup_move_group().  A signal->autogroup assignment
may be in flight when the /proc interface dereference, leaving them taking
a reference to an already dead group.

Reported-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1292508592.5940.28.camel@maggy.simson.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agox86, mm: Initialize initial_page_table before paravirt jumps
Rusty Russell [Tue, 4 Jan 2011 06:50:54 +0000 (17:20 +1030)]
x86, mm: Initialize initial_page_table before paravirt jumps

v2.6.36-rc8-54-gb40827f (x86-32, mm: Add an initial page table
for core bootstrapping) made x86 boot using initial_page_table
and broke lguest.

For 2.6.37 we simply cut & paste the initialization code into
lguest (da32dac10126 "lguest: populate initial_page_table"), now
we fix it properly by doing that initialization before the
paravirt jump.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: lguest <lguest@ozlabs.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <201101041720.54535.rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoMerge commit 'v2.6.37-rc8' into x86/apic
Ingo Molnar [Tue, 4 Jan 2011 08:43:42 +0000 (09:43 +0100)]
Merge commit 'v2.6.37-rc8' into x86/apic

Conflicts:
arch/x86/include/asm/io_apic.h

Merge reason: move to a fresh -rc, resolve the conflict.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agosched: Remove redundant CONFIG_CGROUP_SCHED ifdef
Yong Zhang [Fri, 31 Dec 2010 13:58:58 +0000 (21:58 +0800)]
sched: Remove redundant CONFIG_CGROUP_SCHED ifdef

CONFIG_[FAIR|RT]_GROUP_SCHED always means CONFIG_CGROUP_SCHED

Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1293803938-8157-1-git-send-email-yong.zhang0@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf timechart: Adjust perf timechart to the new power events
Thomas Renninger [Mon, 3 Jan 2011 16:50:45 +0000 (17:50 +0100)]
perf timechart: Adjust perf timechart to the new power events

builtin-timechart must only pass -e power:xy events if they are supported by
the running kernel, otherwise try to fetch the old power:power{start,end}
events.

For this I added the tiny helper function:

   int is_valid_tracepoint(const char *event_string)

to parse-events.[hc], which could be more generic as an interface and support
hardware/software/... events, not only tracepoints, but someone else could
extend that if needed...

Signed-off-by: Thomas Renninger <trenn@suse.de>
Acked-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Jean Pihet <j-pihet@ti.com>
LKML-Reference: <1294073445-14812-4-git-send-email-trenn@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf: Clean up power events by introducing new, more generic ones
Thomas Renninger [Mon, 3 Jan 2011 16:50:44 +0000 (17:50 +0100)]
perf: Clean up power events by introducing new, more generic ones

Add these new power trace events:

 power:cpu_idle
 power:cpu_frequency
 power:machine_suspend

The old C-state/idle accounting events:
  power:power_start
  power:power_end

Have now a replacement (but we are still keeping the old
tracepoints for compatibility):

  power:cpu_idle

and
  power:power_frequency

is replaced with:
  power:cpu_frequency

power:machine_suspend is newly introduced.

Jean Pihet has a patch integrated into the generic layer
(kernel/power/suspend.c) which will make use of it.

the type= field got removed from both, it was never
used and the type is differed by the event type itself.

perf timechart userspace tool gets adjusted in a separate patch.

Signed-off-by: Thomas Renninger <trenn@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Jean Pihet <jean.pihet@newoldbits.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: rjw@sisk.pl
LKML-Reference: <1294073445-14812-3-git-send-email-trenn@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <1290072314-31155-2-git-send-email-trenn@suse.de>

13 years agoperf: Do not export power_frequency, but power_start event
Thomas Renninger [Mon, 3 Jan 2011 16:50:43 +0000 (17:50 +0100)]
perf: Do not export power_frequency, but power_start event

power_frequency moved to drivers/cpufreq/cpufreq.c which has
to be compiled in, no need to export it.

intel_idle can a be module though...

Signed-off-by: Thomas Renninger <trenn@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Jean Pihet <jean.pihet@newoldbits.com>
Cc: Jean Pihet <j-pihet@ti.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: rjw@sisk.pl
LKML-Reference: <1294073445-14812-2-git-send-email-trenn@suse.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <1290072314-31155-2-git-send-email-trenn@suse.de>

13 years agoMerge branch 'perf/test' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux...
Ingo Molnar [Tue, 4 Jan 2011 07:10:28 +0000 (08:10 +0100)]
Merge branch 'perf/test' of git://git./linux/kernel/git/acme/linux-2.6 into perf/core

13 years agoMerge commit 'v2.6.37-rc8' into perf/core
Ingo Molnar [Tue, 4 Jan 2011 07:08:51 +0000 (08:08 +0100)]
Merge commit 'v2.6.37-rc8' into perf/core

Merge reason: pick up latest -rc.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
13 years agoperf test: Add test for counting open syscalls
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 02:16:20 +0000 (00:16 -0200)]
perf test: Add test for counting open syscalls

To test the use of the perf_evsel class on something other than
the tools from where we refactored code to create it.

It calls open() N times and then checks if the event created to
monitor it returns N events.

[acme@felicio linux]$ perf test
 1: vmlinux symtab matches kallsyms: Ok
 2: detect open syscall event: Ok
[acme@felicio linux]$

It does.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Han Pingtian <phan@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evsel: Auto allocate resources needed for some methods
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 02:13:17 +0000 (00:13 -0200)]
perf evsel: Auto allocate resources needed for some methods

While writing the first user of the routines created from the ad-hoc
routines in the existing builtins I noticed that the resulting set of
calls was too long, reduce it by doing some best effort allocations.

Tools that need to operate on multiple threads and cpus should pre-allocate
enough resources by explicitely calling the perf_evsel__alloc_{fd,counters}
methods.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evsel: Use {cpu,thread}_map to shorten list of parameters
Arnaldo Carvalho de Melo [Tue, 4 Jan 2011 01:09:46 +0000 (23:09 -0200)]
perf evsel: Use {cpu,thread}_map to shorten list of parameters

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Refactor all_tids to hold nr and the map
Arnaldo Carvalho de Melo [Mon, 3 Jan 2011 19:53:33 +0000 (17:53 -0200)]
perf tools: Refactor all_tids to hold nr and the map

So that later, we can pass the thread_map instance instead of
(thread_num, thread_map) for things like perf_evsel__open and friends,
just like was done with cpu_map.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf tools: Refactor cpumap to hold nr and the map
Arnaldo Carvalho de Melo [Mon, 3 Jan 2011 19:49:48 +0000 (17:49 -0200)]
perf tools: Refactor cpumap to hold nr and the map

So that later, we can pass the cpu_map instance instead of (nr_cpus, cpu_map)
for things like perf_evsel__open and friends.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evsel: Introduce per cpu and per thread open helpers
Arnaldo Carvalho de Melo [Mon, 3 Jan 2011 19:48:12 +0000 (17:48 -0200)]
perf evsel: Introduce per cpu and per thread open helpers

Abstracting away the loops needed to create the various event fd handlers.

The users have to pass a confiruged perf->evsel.attr field, which is already
usable after perf_evsel__new (constructor) time, using defaults.

Comes out of the ad-hoc routines in builtin-stat, that now uses it.

Fixed a small silly bug where we were die()ing before killing our
children, dysfunctional family this one 8-)

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoperf evsel: Steal the counter reading routines from stat
Arnaldo Carvalho de Melo [Mon, 3 Jan 2011 19:45:52 +0000 (17:45 -0200)]
perf evsel: Steal the counter reading routines from stat

Making them hopefully generic enough to be used in 'perf test',
well see.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
13 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Tue, 4 Jan 2011 00:37:01 +0000 (16:37 -0800)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  ARM: pxa: fix page table corruption on resume
  ARM: it8152: add IT8152_LAST_IRQ definition to fix build error
  ARM: pxa: PXA_ESERIES depends on FB_W100.
  ARM: 6605/1: Add missing include "asm/memory.h"
  ARM: 6540/1: Stop irqsoff trace on return to user
  ARM: 6537/1: update Nomadik, U300 and Ux500 maintainers
  ARM: 6536/1: Add missing SZ_{32,64,128}
  ARM: fix cache-feroceon-l2 after stack based kmap_atomic()
  ARM: fix cache-xsc3l2 after stack based kmap_atomic()
  ARM: get rid of kmap_high_l1_vipt()
  ARM: smp: avoid incrementing mm_users on CPU startup
  ARM: pxa: PXA_ESERIES depends on FB_W100.

13 years agoarch/mn10300/kernel/irq.c: fix build
Andrew Morton [Mon, 3 Jan 2011 22:59:11 +0000 (14:59 -0800)]
arch/mn10300/kernel/irq.c: fix build

Addresses https://bugzilla.kernel.org/show_bug.cgi?id=25702

Reported-by: Martin Ettl <ettl.martin@gmx.de>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoima: fix add LSM rule bug
Mimi Zohar [Mon, 3 Jan 2011 22:59:10 +0000 (14:59 -0800)]
ima: fix add LSM rule bug

If security_filter_rule_init() doesn't return a rule, then not everything
is as fine as the return code implies.

This bug only occurs when the LSM (eg. SELinux) is disabled at runtime.

Adding an empty LSM rule causes ima_match_rules() to always succeed,
ignoring any remaining rules.

 default IMA TCB policy:
  # PROC_SUPER_MAGIC
  dont_measure fsmagic=0x9fa0
  # SYSFS_MAGIC
  dont_measure fsmagic=0x62656572
  # DEBUGFS_MAGIC
  dont_measure fsmagic=0x64626720
  # TMPFS_MAGIC
  dont_measure fsmagic=0x01021994
  # SECURITYFS_MAGIC
  dont_measure fsmagic=0x73636673

  < LSM specific rule >
  dont_measure obj_type=var_log_t

  measure func=BPRM_CHECK
  measure func=FILE_MMAP mask=MAY_EXEC
  measure func=FILE_CHECK mask=MAY_READ uid=0

Thus without the patch, with the boot parameters 'tcb selinux=0', adding
the above 'dont_measure obj_type=var_log_t' rule to the default IMA TCB
measurement policy, would result in nothing being measured.  The patch
prevents the default TCB policy from being replaced.

Signed-off-by: Mimi Zohar <zohar@us.ibm.com>
Cc: James Morris <jmorris@namei.org>
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Cc: David Safford <safford@watson.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
13 years agoMerge branch 'fix' of git://git.kernel.org/pub/scm/linux/kernel/git/ycmiao/pxa-linux-2.6
Russell King [Mon, 3 Jan 2011 22:55:21 +0000 (22:55 +0000)]
Merge branch 'fix' of git://git./linux/kernel/git/ycmiao/pxa-linux-2.6

13 years agobridge: stp: ensure mac header is set
Florian Westphal [Mon, 3 Jan 2011 04:16:28 +0000 (04:16 +0000)]
bridge: stp: ensure mac header is set

commit bf9ae5386bca8836c16e69ab8fdbe46767d7452a
(llc: use dev_hard_header) removed the
skb_reset_mac_header call from llc_mac_hdr_init.

This seems fine itself, but br_send_bpdu() invokes ebtables LOCAL_OUT.

We oops in ebt_basic_match() because it assumes eth_hdr(skb) returns
a meaningful result.

Cc: acme@ghostprotocols.net
References: https://bugzilla.kernel.org/show_bug.cgi?id=24532
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>