firefly-linux-kernel-4.4.55.git
15 years agotracing: Don't use tracing_record_cmdline() in workqueue tracer
KOSAKI Motohiro [Mon, 9 Mar 2009 09:15:34 +0000 (18:15 +0900)]
tracing: Don't use tracing_record_cmdline() in workqueue tracer

Impact: improve workqueue tracer output

Currently, /sys/kernel/debug/tracing/trace_stat/workqueues can display
wrong and strange thread names.

Why?

Currently, ftrace has tracing_record_cmdline()/trace_find_cmdline()
convenience function that implements a task->comm string cache.

This can avoid unnecessary memcpy overhead and the workqueue tracer
uses it.

However, in general, any trace statistics feature shouldn't use
tracing_record_cmdline() because trace statistics can display
very old process. Then comm cache can return wrong string because
recent process overrides the cache.

Fortunately, workqueue trace guarantees that displayed processes
are live. Thus we can search comm string from PID at display time.

<before>

% cat workqueues
 # CPU  INSERTED  EXECUTED   NAME
 # |      |         |          |

   7 431913     431913       kondemand/7
   7      0          0       tail
   7     21         21       git
   7      0          0       ls
   7      9          9       cat
   7 832632     832632       unix_chkpwd
   7 236292     236292       ls

Note: tail, git, ls, cat unix_chkpwd are obiously not workqueue thread.

<after>

% cat workqueues
 # CPU  INSERTED  EXECUTED   NAME
 # |      |         |          |

   7    510        510       kondemand/7
   7      0          0       kmpathd/7
   7     15         15       ata/7
   7      0          0       aio/7
   7     11         11       kblockd/7
   7   1063       1063       work_on_cpu/7
   7    167        167       events/7

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: Steven Rostedt <srostedt@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoftrace: tracing header should put '#' at the beginning of a line
KOSAKI Motohiro [Sun, 8 Mar 2009 04:12:43 +0000 (13:12 +0900)]
ftrace: tracing header should put '#' at the beginning of a line

In a recent discussion, Andrew Morton pointed out that tracing header
should put '#' at the beginning of a line.

Then, we can easily filtered the header by following grep usage:

  cat trace | grep -v '^#'

Wakeup trace also has the same header problem.

Comparison of headers displayed:

before this patch:

 # tracer: wakeup
 #
 wakeup latency trace v1.1.5 on 2.6.29-rc7-tip-tip
 --------------------------------------------------------------------
  latency: 19059 us, #21277/21277, CPU#1 | (M:desktop VP:0, KP:0, SP:0 HP:0 #P:4)
     -----------------
     | task: kondemand/1-1644 (uid:0 nice:-5 policy:0 rt_prio:0)
     -----------------

 #                  _------=> CPU#
 #                 / _-----=> irqs-off
 #                | / _----=> need-resched
 #                || / _---=> hardirq/softirq
 #                ||| / _--=> preempt-depth
 #                |||| /
 #                |||||     delay
 #  cmd     pid   ||||| time  |   caller
 #     \   /      |||||   \   |   /
 irqbalan-1887    1d.s.    0us :   1887:120:R   + [001]  1644:115:S kondemand/1
 irqbalan-1887    1d.s.    1us : default_wake_function <-autoremove_wake_function
 irqbalan-1887    1d.s.    2us : check_preempt_wakeup <-try_to_wake_up

after this patch:

 # tracer: wakeup
 #
 # wakeup latency trace v1.1.5 on 2.6.29-rc7-tip-tip
 # --------------------------------------------------------------------
 # latency: 529 us, #530/530, CPU#0 | (M:desktop VP:0, KP:0, SP:0 HP:0 #P:4)
 #    -----------------
 #    | task: kondemand/0-1641 (uid:0 nice:-5 policy:0 rt_prio:0)
 #    -----------------
 #
 #                  _------=> CPU#
 #                 / _-----=> irqs-off
 #                | / _----=> need-resched
 #                || / _---=> hardirq/softirq
 #                ||| / _--=> preempt-depth
 #                |||| /
 #                |||||     delay
 #  cmd     pid   ||||| time  |   caller
 #     \   /      |||||   \   |   /
     sshd-2496    0d.s.    0us :   2496:120:R   + [000]  1641:115:S kondemand/0
     sshd-2496    0d.s.    1us : default_wake_function <-autoremove_wake_function
     sshd-2496    0d.s.    1us : check_preempt_wakeup <-try_to_wake_up

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <20090308124421.23C3.A69D9226@jp.fujitsu.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branches 'tracing/doc', 'tracing/ftrace', 'tracing/printk' and 'tracing/textedi...
Ingo Molnar [Sun, 8 Mar 2009 15:48:51 +0000 (16:48 +0100)]
Merge branches 'tracing/doc', 'tracing/ftrace', 'tracing/printk' and 'tracing/textedit' into tracing/core

15 years agotracing: trace_bprintk() cleanups
Ingo Molnar [Fri, 6 Mar 2009 16:52:03 +0000 (17:52 +0100)]
tracing: trace_bprintk() cleanups

Impact: cleanup

Remove a few leftovers and clean up the code a bit.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1236356510-8381-5-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing/core: drop the old trace_printk() implementation in favour of trace_bprintk()
Frederic Weisbecker [Fri, 6 Mar 2009 16:21:49 +0000 (17:21 +0100)]
tracing/core: drop the old trace_printk() implementation in favour of trace_bprintk()

Impact: faster and lighter tracing

Now that we have trace_bprintk() which is faster and consume lesser
memory than trace_printk() and has the same purpose, we can now drop
the old implementation in favour of the binary one from trace_bprintk(),
which means we move all the implementation of trace_bprintk() to
trace_printk(), so the Api doesn't change except that we must now use
trace_seq_bprintk() to print the TRACE_PRINT entries.

Some changes result of this:

- Previously, trace_bprintk depended of a single tracer and couldn't
  work without. This tracer has been dropped and the whole implementation
  of trace_printk() (like the module formats management) is now integrated
  in the tracing core (comes with CONFIG_TRACING), though we keep the file
  trace_printk (previously trace_bprintk.c) where we can find the module
  management. Thus we don't overflow trace.c

- changes some parts to use trace_seq_bprintk() to print TRACE_PRINT entries.

- change a bit trace_printk/trace_vprintk macros to support non-builtin formats
  constants, and fix 'const' qualifiers warnings. But this is all transparent for
  developers.

- etc...

V2:

- Rebase against last changes
- Fix mispell on the changelog

V3:

- Rebase against last changes (moving trace_printk() to kernel.h)

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1236356510-8381-5-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing: add trace_bprintk()
Lai Jiangshan [Fri, 6 Mar 2009 16:21:48 +0000 (17:21 +0100)]
tracing: add trace_bprintk()

Impact: add a generic printk() for tracing, like trace_printk()

trace_bprintk() uses the infrastructure to record events on ring_buffer.

[ fweisbec@gmail.com: ported to latest -tip, made it work if
  !CONFIG_MODULES, never free the format strings from modules
  because we can't keep track of them and conditionnaly create
  the ftrace format strings section (reported by Steven Rostedt) ]

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1236356510-8381-4-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing: infrastructure for supporting binary record
Lai Jiangshan [Fri, 6 Mar 2009 16:21:47 +0000 (17:21 +0100)]
tracing: infrastructure for supporting binary record

Impact: save on memory for tracing

Current tracers are typically using a struct(like struct ftrace_entry,
struct ctx_switch_entry, struct special_entr etc...)to record a binary
event. These structs can only record a their own kind of events.
A new kind of tracer need a new struct and a lot of code too handle it.

So we need a generic binary record for events. This infrastructure
is for this purpose.

[fweisbec@gmail.com: rebase against latest -tip, make it safe while sched
tracing as reported by Steven Rostedt]

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1236356510-8381-3-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'core/printk' into tracing/ftrace
Ingo Molnar [Fri, 6 Mar 2009 16:45:42 +0000 (17:45 +0100)]
Merge branch 'core/printk' into tracing/ftrace

15 years agovsprintf: unify the format decoding layer for its 3 users
Frederic Weisbecker [Fri, 6 Mar 2009 16:21:50 +0000 (17:21 +0100)]
vsprintf: unify the format decoding layer for its 3 users

An new optimization is making its way to ftrace. Its purpose is to
make trace_printk() consuming less memory and be faster.

Written by Lai Jiangshan, the approach is to delay the formatting
job from tracing time to output time.

Currently, a call to trace_printk() will format the whole string and
insert it into the ring buffer. Then you can read it on /debug/tracing/trace
file.

The new implementation stores the address of the format string and
the binary parameters into the ring buffer, making the packet more compact
and faster to insert.
Later, when the user exports the traces, the format string is retrieved
with the binary parameters and the formatting job is eventually done.

The new implementation rewrites a lot of format decoding bits from
vsnprintf() function, making now 3 differents functions to maintain
in their duplicated parts of printf format decoding bits.

Suggested by Ingo Molnar, this patch tries to factorize the most
possible common bits from these functions.
The real common part between them is the format decoding. Although
they do somewhat similar jobs, their way to export or import the parameters
is very different. Thus, only the decoding layer is extracted, unless you see
other parts that could be worth factorized.

Changes in V2:

- Address a suggestion from Linus to group the format_decode() parameters inside
  a structure.

Changes in v3:

- Address other cleanups suggested by Ingo and Linus such as passing the
  printf_spec struct to the format helpers: pointer()/number()/string()
  Note that this struct is passed by copy and not by address. This is to
  avoid side effects because these functions often change these values and the
  changes shoudn't be persistant when a callee helper returns.
  It would be too risky.

- Various cleanups (code alignement, switch/case instead of if/else fountains).

- Fix a bug that printed the first format specifier following a %p

Changes in v4:

- drop unapropriate const qualifier loss while casting fmt to a char *
  (thanks to Vegard Nossum for having pointed this out).

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1236356510-8381-6-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agovsprintf: add binary printf
Lai Jiangshan [Fri, 6 Mar 2009 16:21:46 +0000 (17:21 +0100)]
vsprintf: add binary printf

Impact: add new APIs for binary trace printk infrastructure

vbin_printf(): write args to binary buffer, string is copied
when "%s" is occurred.

bstr_printf(): read from binary buffer for args and format a string

[fweisbec@gmail.com: rebase]

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
LKML-Reference: <1236356510-8381-2-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: implement atomic text_poke() via fixmap
Masami Hiramatsu [Fri, 6 Mar 2009 15:37:54 +0000 (10:37 -0500)]
x86: implement atomic text_poke() via fixmap

Use fixmaps instead of vmap/vunmap in text_poke() for avoiding
page allocation and delayed unmapping.

At the result of above change, text_poke() becomes atomic and can be called
from stop_machine() etc.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
LKML-Reference: <49B14352.2040705@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing, Text Edit Lock - SMP alternatives support
Masami Hiramatsu [Fri, 6 Mar 2009 15:37:22 +0000 (10:37 -0500)]
tracing, Text Edit Lock - SMP alternatives support

Use the mutual exclusion provided by the text edit lock in alternatives code.
Since alternative_smp_* will be called from module init code, etc,
we'd better protect it from other subsystems.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
LKML-Reference: <49B14332.9030109@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing, Text Edit Lock - kprobes architecture independent support
Mathieu Desnoyers [Fri, 6 Mar 2009 15:36:38 +0000 (10:36 -0500)]
tracing, Text Edit Lock - kprobes architecture independent support

Use the mutual exclusion provided by the text edit lock in the kprobes code. It
allows coherent manipulation of the kernel code by other subsystems.

Changelog:

Move the kernel_text_lock/unlock out of the for loops.
Use text_mutex directly instead of a function.
Remove whitespace modifications.

(note : kprobes_mutex is always taken outside of text_mutex)

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Acked-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Acked-by: Masami Hiramatsu <mhiramat@redhat.com>
LKML-Reference: <49B14306.2080202@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing, Text Edit Lock - Architecture Independent Code
Mathieu Desnoyers [Fri, 6 Mar 2009 15:35:52 +0000 (10:35 -0500)]
tracing, Text Edit Lock - Architecture Independent Code

This is an architecture independant synchronization around kernel text
modifications through use of a global mutex.

A mutex has been chosen so that kprobes, the main user of this, can sleep
during memory allocation between the memory read of the instructions it
must replace and the memory write of the breakpoint.

Other user of this interface: immediate values.

Paravirt and alternatives are always done when SMP is inactive, so there
is no need to use locks.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
LKML-Reference: <49B142D8.7020601@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'x86/core' into tracing/textedit
Ingo Molnar [Fri, 6 Mar 2009 15:44:14 +0000 (16:44 +0100)]
Merge branch 'x86/core' into tracing/textedit

Conflicts:
arch/x86/Kconfig
block/blktrace.c
kernel/irq/handle.c

Semantic conflict:
kernel/trace/blktrace.c

Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing, power-trace: make it build even if the power-tracer is turned off
Ingo Molnar [Fri, 6 Mar 2009 11:47:08 +0000 (12:47 +0100)]
tracing, power-trace: make it build even if the power-tracer is turned off

Impact: build fix

The 'struct power_trace' definition is needed (for the event tracer) even if
the power-tracer plugin is turned off in the .config.

Cc: Steven Rostedt <srostedt@redhat.com>
LKML-Reference: <20090306104106.GF31042@elte.hu>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing: fix deadlock when setting set_ftrace_pid
KOSAKI Motohiro [Fri, 6 Mar 2009 06:29:04 +0000 (15:29 +0900)]
tracing: fix deadlock when setting set_ftrace_pid

Impact: fix deadlock while using set_ftrace_pid

Reproducer:

# cd /sys/kernel/debug/tracing
# echo $$ > set_ftrace_pid

then, console becomes hung.

Details:

when writing set_ftracepid, kernel callstack is following

ftrace_pid_write()
mutex_lock(&ftrace_lock);
ftrace_update_pid_func()
mutex_lock(&ftrace_lock);
mutex_unlock(&ftrace_lock);
mutex_unlock(&ftrace_lock);

then, system always deadlocks when ftrace_pid_write() is called.

In past days, ftrace_pid_write() used ftrace_start_lock, but
commit e6ea44e9b4c12325337cd1c06103cd515a1c02b2 consolidated
ftrace_start_lock to ftrace_lock.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: Steven Rostedt <srostedt@redhat.com>
LKML-Reference: <20090306151155.0778.A69D9226@jp.fujitsu.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing: current tip/master can't enable ftrace
KOSAKI Motohiro [Fri, 6 Mar 2009 01:40:53 +0000 (10:40 +0900)]
tracing: current tip/master can't enable ftrace

After commit 40ada30f9621fbd831ac2437b9a2a399aad34b00,
"make menuconfig" doesn't display "Tracer" item.

Following modification restores it.

15 years agoMerge branch 'tip/tracing/ftrace' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Fri, 6 Mar 2009 10:40:37 +0000 (11:40 +0100)]
Merge branch 'tip/tracing/ftrace' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into tracing/ftrace

15 years agoMerge branches 'tracing/ftrace' and 'tracing/function-graph-tracer' into tracing...
Ingo Molnar [Fri, 6 Mar 2009 10:39:18 +0000 (11:39 +0100)]
Merge branches 'tracing/ftrace' and 'tracing/function-graph-tracer' into tracing/core

15 years agotracing: add format files for ftrace default entries
Steven Rostedt [Fri, 6 Mar 2009 02:35:29 +0000 (21:35 -0500)]
tracing: add format files for ftrace default entries

Impact: allow user apps to read binary format of basic ftrace entries

Currently, only defined raw events export their formats so a binary
reader can parse them. There's no reason that the default ftrace entries
can't export their formats.

This patch adds a subsystem called "ftrace" in the events directory
that includes the ftrace entries for basic ftrace recorded items.

These only have three files in the events directory:

 type             : printf
 available_types  : printf
 format           : format for the event entry

For example:

 # cat /debug/tracing/events/ftrace/wakeup/format
name: wakeup
ID: 3
format:
        field:unsigned char type;       offset:0;       size:1;
        field:unsigned char flags;      offset:1;       size:1;
        field:unsigned char preempt_count;      offset:2;       size:1;
        field:int pid;  offset:4;       size:4;
        field:int tgid; offset:8;       size:4;

        field:unsigned int prev_pid;    offset:12;      size:4;
        field:unsigned char prev_prio;  offset:16;      size:1;
        field:unsigned char prev_state; offset:17;      size:1;
        field:unsigned int next_pid;    offset:20;      size:4;
        field:unsigned char next_prio;  offset:24;      size:1;
        field:unsigned char next_state; offset:25;      size:1;
        field:unsigned int next_cpu;    offset:28;      size:4;

print fmt: "%u:%u:%u  ==+ %u:%u:%u [%03u]"

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: move print of event format to separate file
Steven Rostedt [Thu, 5 Mar 2009 16:45:43 +0000 (11:45 -0500)]
tracing: move print of event format to separate file

Impact: clean up

Move the macro that creates the event format file to a separate header.
This will allow the default ftrace events to use this same macro
to create the formats to read those events.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: make all file_operations const
Steven Rostedt [Fri, 6 Mar 2009 02:44:55 +0000 (21:44 -0500)]
tracing: make all file_operations const

Impact: cleanup

All file_operations structures should be constant. No one is going to
change them.

Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: clean up menu
Ingo Molnar [Thu, 5 Mar 2009 20:19:55 +0000 (21:19 +0100)]
tracing: clean up menu

Clean up menu structure, introduce TRACING_SUPPORT switch that signals
whether an architecture supports various instrumentation mechanisms.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'x86/uv' into x86/core
Ingo Molnar [Thu, 5 Mar 2009 20:49:47 +0000 (21:49 +0100)]
Merge branch 'x86/uv' into x86/core

15 years agoMerge branch 'x86/doc' into x86/core
Ingo Molnar [Thu, 5 Mar 2009 20:49:44 +0000 (21:49 +0100)]
Merge branch 'x86/doc' into x86/core

15 years agoMerge branch 'x86/mm' into x86/core
Ingo Molnar [Thu, 5 Mar 2009 20:49:35 +0000 (21:49 +0100)]
Merge branch 'x86/mm' into x86/core

15 years agoMerge branch 'x86/mce2' into x86/core
Ingo Molnar [Thu, 5 Mar 2009 20:49:25 +0000 (21:49 +0100)]
Merge branch 'x86/mce2' into x86/core

15 years agoMerge branch 'x86/urgent' into x86/core
Ingo Molnar [Thu, 5 Mar 2009 20:48:31 +0000 (21:48 +0100)]
Merge branch 'x86/urgent' into x86/core

Conflicts:
arch/x86/include/asm/fixmap_64.h
Semantic merge:
arch/x86/include/asm/fixmap.h

Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing: add tracing_on/tracing_off to kernel.h
Steven Rostedt [Thu, 5 Mar 2009 15:35:56 +0000 (10:35 -0500)]
tracing: add tracing_on/tracing_off to kernel.h

Impact: cleanup

The functions tracing_start/tracing_stop have been moved to kernel.h.
These are not the functions a developer most likely wants to use
when they want to insert a place to stop tracing and restart it from
user space.

tracing_start/tracing_stop was created to work with things like
suspend to ram, where even calling smp_processor_id() can crash the
system. The tracing_start/tracing_stop was used to stop the tracer from
doing anything. These are still light weight functions, but add a bit
more overhead to be able to stop the tracers. They also have no interface
back to userland. That is, if the kernel calls tracing_stop, userland
can not start tracing.

What a developer most likely wants to use is tracing_on/tracing_off.
These are very light weight functions (simply sets or clears a bit).
These functions just stop recording into the ring buffer. The tracers
don't even know that this happens except that they would receive NULL
from the ring_buffer_lock_reserve function.

Also, there's a way for the user land to enable or disable this bit.
In debugfs/tracing/tracing_on, a user may echo "0" (same as tracing_off())
or echo "1" (same as tracing_on()) into this file. This becomes handy when
a kernel developer is debugging and wants tracing to turn off when it
hits an anomaly. Then the developer can examine the trace, and restart
tracing if they want to try again (echo 1 > tracing_on).

This patch moves the prototypes for tracing_on/tracing_off to kernel.h
and comments their use, so that a kernel developer will know how
to use them.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agox86: UV, SGI RTC: add generic system vector, build fix on UP
Ingo Molnar [Thu, 5 Mar 2009 14:15:44 +0000 (15:15 +0100)]
x86: UV, SGI RTC: add generic system vector, build fix on UP

Make ack_APIC_irq() build on !SMP && !APIC too.

Cc: Dimitri Sivanich <sivanich@sgi.com>
LKML-Reference: <20090304185605.GA24419@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: pre-initialize boot_cpu_data.x86_phys_bits to avoid system_state tests
Jeremy Fitzhardinge [Thu, 5 Mar 2009 00:16:51 +0000 (16:16 -0800)]
x86: pre-initialize boot_cpu_data.x86_phys_bits to avoid system_state tests

Impact: cleanup, micro-optimization

Pre-initialize boot_cpu_data.x86_phys_bits to a reasonable default
to remove the use of system_state tests in __virt_addr_valid()
and __phys_addr().

Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86-32: use specific __vmalloc_start_set flag in __virt_addr_valid
Jeremy Fitzhardinge [Thu, 5 Mar 2009 00:10:44 +0000 (16:10 -0800)]
x86-32: use specific __vmalloc_start_set flag in __virt_addr_valid

Rather than relying on the ever-unreliable system_state,
add a specific __vmalloc_start_set flag to indicate whether
the vmalloc area has meaningful boundaries yet, and use that
in x86-32's __phys_addr and __virt_addr_valid.

Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: clean up old gcc warnings
Jeremy Fitzhardinge [Thu, 5 Mar 2009 01:14:30 +0000 (17:14 -0800)]
x86: clean up old gcc warnings

gcc 3.2.2 reports:

In file included from /usr/src/all/linux-next/arch/x86/include/asm/page.h:8,
                 from /usr/src/all/linux-next/arch/x86/include/asm/processor.h:18,
                 from /usr/src/all/linux-next/arch/x86/include/asm/atomic_32.h:6,
                 from /usr/src/all/linux-next/arch/x86/include/asm/atomic.h:2,
                 from include/linux/crypto.h:20,
                 from arch/x86/kernel/asm-offsets_32.c:7,
                 from arch/x86/kernel/asm-offsets.c:2:
/usr/src/all/linux-next/arch/x86/include/asm/page_types.h:54: warning: parameter has incomplete type
/usr/src/all/linux-next/arch/x86/include/asm/page_types.h:56: warning: parameter has incomplete type
In file included from /usr/src/all/linux-next/arch/x86/include/asm/page.h:8,
                 from /usr/src/all/linux-next/arch/x86/include/asm/processor.h:18,
                 from include/linux/prefetch.h:14,
                 from include/linux/list.h:6,
                 from include/linux/module.h:9,
                 from init/main.c:13:
/usr/src/all/linux-next/arch/x86/include/asm/page_types.h:54: warning: parameter has incomplete type
/usr/src/all/linux-next/arch/x86/include/asm/page_types.h:56: warning: parameter has incomplete type

This is a bogus warning, but moving the pat-related functions
into asm/pat.h and including asm/pgtable_types.h should fix it.

Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: move init_memory_mapping() to common mm/init.c, build fix on 32-bit PAE
Ingo Molnar [Thu, 5 Mar 2009 13:39:03 +0000 (14:39 +0100)]
x86: move init_memory_mapping() to common mm/init.c, build fix on 32-bit PAE

Impact: build fix

Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-14-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: move function and variable declarations to asm/init.h
Pekka Enberg [Thu, 5 Mar 2009 12:55:08 +0000 (14:55 +0200)]
x86: move function and variable declarations to asm/init.h

Impact: cleanup

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-17-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: unify kernel_physical_mapping_init() function signatures
Pekka Enberg [Thu, 5 Mar 2009 12:55:07 +0000 (14:55 +0200)]
x86: unify kernel_physical_mapping_init() function signatures

Impact: cleanup

In preparation for moving the function declaration to a header file,
unify 32-bit and 64-bit signatures.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-16-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: fix up some bad global variable names in mm/init.c
Pekka Enberg [Thu, 5 Mar 2009 12:55:06 +0000 (14:55 +0200)]
x86: fix up some bad global variable names in mm/init.c

Impact: cleanup

The table_start, table_end, and table_top are too generic for global
namespace so rename them to be more specific.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-15-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: move init_memory_mapping() to common mm/init.c
Pekka Enberg [Thu, 5 Mar 2009 12:55:05 +0000 (14:55 +0200)]
x86: move init_memory_mapping() to common mm/init.c

Impact: cleanup

This patch moves the init_memory_mapping() function to common mm/init.c.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-14-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: add stub init_gbpages() for 32-bit init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:55:04 +0000 (14:55 +0200)]
x86: add stub init_gbpages() for 32-bit init_memory_mapping()

Impact: cleanup

This patch adds an empty static inline init_gbpages() for the 32-bit
version of init_memory_mapping() making both versions identical.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-13-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: ifdef 32-bit and 64-bit NR_RANGE_MR for save_mr() unification
Pekka Enberg [Thu, 5 Mar 2009 12:55:03 +0000 (14:55 +0200)]
x86: ifdef 32-bit and 64-bit NR_RANGE_MR for save_mr() unification

Impact: cleanup

As a trivial preparation for moving common code to arc/x86/mm/init.c,
ifdef the 32-bit and 64-bit versions of NR_RANGE_MR.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-12-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: ifdef 32-bit and 64-bit pfn setup in init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:55:02 +0000 (14:55 +0200)]
x86: ifdef 32-bit and 64-bit pfn setup in init_memory_mapping()

Impact: cleanup

To reduce the diff between the 32-bit and 64-bit versions of
init_memory_mapping(), ifdef configuration specific pfn setup
code in the function.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-11-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: ifdef 32-bit and 64-bit setup in init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:55:01 +0000 (14:55 +0200)]
x86: ifdef 32-bit and 64-bit setup in init_memory_mapping()

Impact: cleanup

To reduce the diff between the 32-bit and 64-bit versions of
init_memory_mapping(), ifdef configuration specific setup code
in the function.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-10-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: add table start and end sanity checks to 32-bit init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:55:00 +0000 (14:55 +0200)]
x86: add table start and end sanity checks to 32-bit init_memory_mapping()

Impact: cleanup

This patch adds a sanity check to the 32-bit version of
init_memory_mapping() to reduce the diff to the 64-bit version.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-9-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: unify kernel_physical_mapping_init() call in init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:54:59 +0000 (14:54 +0200)]
x86: unify kernel_physical_mapping_init() call in init_memory_mapping()

Impact: cleanup

The 64-bit version of init_memory_mapping() uses the last mapped
address returned from kernel_physical_mapping_init() whereas the
32-bit version doesn't. This patch adds relevant ifdefs to both
versions of the function to reduce the diff between them.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-8-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: rename after_init_bootmem to after_bootmem in mm/init_32.c
Pekka Enberg [Thu, 5 Mar 2009 12:54:58 +0000 (14:54 +0200)]
x86: rename after_init_bootmem to after_bootmem in mm/init_32.c

Impact: cleanup

This patch renames after_init_bootmem to after_bootmem in
mm/init_32.c to reduce the diff to the 64-bit version of of
init_memory_mapping().

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-7-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: remove unnecessary save_mr() sanity check
Pekka Enberg [Thu, 5 Mar 2009 12:54:57 +0000 (14:54 +0200)]
x86: remove unnecessary save_mr() sanity check

Impact: cleanup

The save_mr() function already checks that start_pfn is less than
end_pfn so we can remove the unnecessary check which reduces the
diff between the 32-bit and the 64-bit versions of init_memory_mapping().

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-6-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: ifdef 32-bit specific setup in init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:54:56 +0000 (14:54 +0200)]
x86: ifdef 32-bit specific setup in init_memory_mapping()

Impact: cleanup

Enabling NX, PSE, and PGE are only required on 32-bit so ifdef them
in both versions of the function.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-5-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: move pgd_base out of init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:54:55 +0000 (14:54 +0200)]
x86: move pgd_base out of init_memory_mapping()

Impact: cleanup

This patch moves pgd_base out of init_memory_mapping() to reduce
the diff between the 32-bit version and the 64-bit version of the
function.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-4-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: find_early_table_space() unification
Pekka Enberg [Thu, 5 Mar 2009 12:54:54 +0000 (14:54 +0200)]
x86: find_early_table_space() unification

Impact: cleanup

There are some minor differences between the 32-bit and 64-bit
find_early_table_space() functions. This patch wraps those
differences under CONFIG_X86_32 to make the function identical
on both configurations.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-3-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: add gbpages support to 32-bit init_memory_mapping()
Pekka Enberg [Thu, 5 Mar 2009 12:54:53 +0000 (14:54 +0200)]
x86: add gbpages support to 32-bit init_memory_mapping()

Impact: cleanup

To reduce the diff between the 32-bit and 64-bit versions of
init_memory_mapping(), add gbpages support to the 32-bit version.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-2-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: init_memory_mapping() trivial cleanups
Pekka Enberg [Thu, 5 Mar 2009 12:54:52 +0000 (14:54 +0200)]
x86: init_memory_mapping() trivial cleanups

Impact: cleanup

To reduce the diff between the 32-bit and 64-bit versions of
init_memory_mapping(), fix up all trivial issues.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236257708-27269-1-git-send-email-penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing/function-graph-tracer: use the more lightweight local clock
Frederic Weisbecker [Thu, 5 Mar 2009 00:49:22 +0000 (01:49 +0100)]
tracing/function-graph-tracer: use the more lightweight local clock

Impact: decrease hangs risks with the graph tracer on slow systems

Since the function graph tracer can spend too much time on timer
interrupts, it's better now to use the more lightweight local
clock. Anyway, the function graph traces are more reliable on a
per cpu trace.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <49af243d.06e9300a.53ad.ffff840c@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: UV, SGI RTC: fix uv_time.c for UP
Dimitri Sivanich [Wed, 4 Mar 2009 22:02:46 +0000 (16:02 -0600)]
x86: UV, SGI RTC: fix uv_time.c for UP

Fix non-smp build of uv_time.c.

Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
LKML-Reference: <20090304220246.GC6288@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86/doc: mini-howto for using earlyprintk=dbgp
Yinghai Lu [Thu, 5 Mar 2009 00:11:35 +0000 (16:11 -0800)]
x86/doc: mini-howto for using earlyprintk=dbgp

[ mingo: small edits and extensions. ]

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Greg KH <gregkh@suse.de>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Sarah Sharp <sarah.a.sharp@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <49AF18B7.4050305@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing: move utility functions from ftrace.h to kernel.h
Ingo Molnar [Thu, 5 Mar 2009 09:28:45 +0000 (10:28 +0100)]
tracing: move utility functions from ftrace.h to kernel.h

Make common utility functions such as trace_printk() and
tracing_start()/tracing_stop() generally available to kernel
code.

Cc: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agotracing: rename ftrace_printk() => trace_printk()
Ingo Molnar [Thu, 5 Mar 2009 09:24:48 +0000 (10:24 +0100)]
tracing: rename ftrace_printk() => trace_printk()

Impact: cleanup

Use a more generic name - this also allows the prototype to move
to kernel.h and be generally available to kernel developers who
want to do some quick tracing.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'tip/tracing/ftrace' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Thu, 5 Mar 2009 09:21:49 +0000 (10:21 +0100)]
Merge branch 'tip/tracing/ftrace' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into tracing/ftrace

15 years agoMerge branch 'tracing/ftrace' into tracing/core
Ingo Molnar [Thu, 5 Mar 2009 09:20:47 +0000 (10:20 +0100)]
Merge branch 'tracing/ftrace' into tracing/core

15 years agotracing: have latency tracers set the latency format
Steven Rostedt [Thu, 5 Mar 2009 03:15:30 +0000 (22:15 -0500)]
tracing: have latency tracers set the latency format

The latency tracers (irqsoff, preemptoff, preemptirqsoff, and wakeup)
are pretty useless with the default output format. This patch makes them
automatically enable the latency format when they are selected. They
also record the state of the latency option, and if it was not enabled
when selected, they disable it on reset.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: consolidate print_lat_fmt and print_trace_fmt
Steven Rostedt [Thu, 5 Mar 2009 02:57:29 +0000 (21:57 -0500)]
tracing: consolidate print_lat_fmt and print_trace_fmt

Impact: clean up

Both print_lat_fmt and print_trace_fmt do pretty much the same thing
except for one different function call. This patch consolidates the
two functions and adds an if statement to perform the difference.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: remove extra latency_trace method from trace structure
Steven Rostedt [Thu, 5 Mar 2009 02:42:04 +0000 (21:42 -0500)]
tracing: remove extra latency_trace method from trace structure

Impact: clean up

The trace and latency_trace function pointers are identical for
every tracer but the function tracer. The differences in the function
tracer are trivial (latency output puts paranthesis around parent).

This patch removes the latency_trace pointer and all prints will
now just use the trace output function pointer.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: add latency output format option
Steven Rostedt [Thu, 5 Mar 2009 01:34:24 +0000 (20:34 -0500)]
tracing: add latency output format option

With the removal of the latency_trace file, we lost the ability
to see some of the finer details in a trace. Like the state of
interrupts enabled, the preempt count, need resched, and if we
are in an interrupt handler, softirq handler or not.

This patch simply creates an option to bring back the old format.
This also removes the warning about an unused variable that held
the latency_trace file operations.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: fix seq read from trace files
Steven Rostedt [Thu, 5 Mar 2009 01:31:11 +0000 (20:31 -0500)]
tracing: fix seq read from trace files

The buffer used by trace_seq was updated incorrectly. Instead
of consuming what was actually read, it consumed the rest of the
buffer on reads.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: do not return EFAULT if read copied anything
Steven Rostedt [Thu, 5 Mar 2009 00:10:05 +0000 (19:10 -0500)]
tracing: do not return EFAULT if read copied anything

Impact: fix trace read to conform to standards

Andrew Morton, Theodore Tso and H. Peter Anvin brought to my attention
that a userspace read should not return -EFAULT if it succeeded in
copying anything. It should only return -EFAULT if it failed to copy
at all.

This patch modifies the check of copy_from_user and updates the return
code appropriately.

I also used H. Peter Anvin's short cut rule to just test ret == count.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agoring-buffer: fix timestamp in partial ring_buffer_page_read
Steven Rostedt [Wed, 4 Mar 2009 04:52:42 +0000 (23:52 -0500)]
ring-buffer: fix timestamp in partial ring_buffer_page_read

If a partial ring_buffer_page_read happens, then some of the
incremental timestamps may be lost. This patch writes the
recent timestamp into the page that is passed back to the caller.

A partial ring_buffer_page_read is where the full page would not
be written back to the user, and instead, just part of the page
is copied to the user. A full page would be a page swap with the
ring buffer and the timestamps would be correct.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agotracing: add cpu_file intialization for ftrace_dump
Steven Rostedt [Wed, 4 Mar 2009 23:20:36 +0000 (18:20 -0500)]
tracing: add cpu_file intialization for ftrace_dump

Impact: fix to ftrace_dump output corruption

The commit: b04cc6b1f6398b0e0b60d37e27ce51b4899672ec
  tracing/core: introduce per cpu tracing files

added a new field to the iterator called cpu_file. This was a handle
to differentiate between the per cpu trace output files and the
all cpu "trace" file. The all cpu "trace" file required setting this
to TRACE_PIPE_ALL_CPU.

The problem is that the ftrace_dump sets up its own iterator but was
not updated to handle this change. The result was only CPU 0 printing
out on crash and a lot of "<0>"'s also being printed.

Reported-by: Thomas Gleixner <tglx@linuxtronix.de>
Tested-by: Darren Hart <dvhtc@us.ibm.com>
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agox86: fix bootmem cross node for 32bit numa, cleanup
Yinghai Lu [Wed, 4 Mar 2009 20:21:24 +0000 (12:21 -0800)]
x86: fix bootmem cross node for 32bit numa, cleanup

Impact: clean up

Simplify the code, reuse some lines.
Remove min_low_pfn reference, it is always 0

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <49AEE2C4.2030602@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: move free_initrd_mem() to common mm/init.c
Pekka Enberg [Wed, 4 Mar 2009 09:13:40 +0000 (11:13 +0200)]
x86: move free_initrd_mem() to common mm/init.c

Impact: cleanup

The function is identical on 32-bit and 64-bit configurations so move it to the
common mm/init.c file.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
LKML-Reference: <1236158020.29024.28.camel@penberg-laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: add Dell XPS710 reboot quirk
Leann Ogasawara [Wed, 4 Mar 2009 19:53:00 +0000 (11:53 -0800)]
x86: add Dell XPS710 reboot quirk

Dell XPS710 will hang on reboot.  This is resolved by adding a quirk to
set bios reboot.

Signed-off-by: Leann Ogasawara <leann.ogasawara@canonical.com>
Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
Cc: "manoj.iyer" <manoj.iyer@canonical.com>
Cc: <stable@kernel.org>
LKML-Reference: <1236196380.3231.89.camel@emiko>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: reserve exact size of mptable
Yinghai Lu [Wed, 4 Mar 2009 09:25:54 +0000 (01:25 -0800)]
x86: reserve exact size of mptable

Impact: save a bit of RAM

Get the exact size for the reserve_bootmem() call.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <49AE4922.605@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: ioremap mptable
Yinghai Lu [Wed, 4 Mar 2009 09:25:21 +0000 (01:25 -0800)]
x86: ioremap mptable

Impact: fix boot with mptable above max_low_mapped

Try to use early_ioremap() to map MPC to make sure it works even it is
at the end of ram.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <49AE4901.3090801@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Reported-and-tested-by: Kevin O'Connor <kevin@koconnor.net>
15 years agox86: make 32-bit init_memory_mapping range change more like 64-bit
Yinghai Lu [Wed, 4 Mar 2009 09:24:04 +0000 (01:24 -0800)]
x86: make 32-bit init_memory_mapping range change more like 64-bit

Impact: cleanup

make code more readable and more like 64-bit

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <49AE48B4.8010907@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: fix bootmem cross node for 32bit numa
Yinghai Lu [Wed, 4 Mar 2009 09:22:35 +0000 (01:22 -0800)]
x86: fix bootmem cross node for 32bit numa

Impact: fix panic on system 2g x4 sockets

Found one system with 4 sockets and every sockets has 2g can not boot
with numa32 because boot mem is crossing nodes.

So try to have numa version of setup_bootmem_allocator().

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <49AE485B.8000902@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86, math-emu: fix init_fpu for task != current
Daniel Glöckner [Wed, 4 Mar 2009 18:42:27 +0000 (19:42 +0100)]
x86, math-emu: fix init_fpu for task != current

Impact: fix math-emu related crash while using GDB/ptrace

init_fpu() calls finit to initialize a task's xstate, while finit always
works on the current task. If we use PTRACE_GETFPREGS on another
process and both processes did not already use floating point, we get
a null pointer exception in finit.

This patch creates a new function finit_task that takes a task_struct
parameter. finit becomes a wrapper that simply calls finit_task with
current. On the plus side this avoids many calls to get_current which
would each resolve to an inline assembler mov instruction.

An empty finit_task has been added to i387.h to avoid linker errors in
case the compiler still emits the call in init_fpu when
CONFIG_MATH_EMULATION is not defined.

The declaration of finit in i387.h has been removed as the remaining
code using this function gets its prototype from fpu_proto.h.

Signed-off-by: Daniel Glöckner <dg@emlix.com>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: "Pallipadi Venkatesh" <venkatesh.pallipadi@intel.com>
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: Bill Metzenthen <billm@melbpc.org.au>
LKML-Reference: <E1Lew31-0004il-Fg@mailer.emlix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: UV, SGI RTC: add UV RTC clocksource/clockevents
Dimitri Sivanich [Wed, 4 Mar 2009 18:59:18 +0000 (12:59 -0600)]
x86: UV, SGI RTC: add UV RTC clocksource/clockevents

This patch provides a high resolution clock/timer source using the
SGI UV system-wide synchronized RTC clock/timer hardware.

Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: john stultz <johnstul@us.ibm.com>
LKML-Reference: <20090304185918.GC24419@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: UV, SGI RTC: loop through installed UV blades
Dimitri Sivanich [Wed, 4 Mar 2009 18:57:19 +0000 (12:57 -0600)]
x86: UV, SGI RTC: loop through installed UV blades

Add macro to loop through each possible blade.

Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: john stultz <johnstul@us.ibm.com>
LKML-Reference: <20090304185719.GB24419@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: UV, SGI RTC: add generic system vector
Dimitri Sivanich [Wed, 4 Mar 2009 18:56:05 +0000 (12:56 -0600)]
x86: UV, SGI RTC: add generic system vector

This patch allocates a system interrupt vector for various platform
specific uses.

Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: john stultz <johnstul@us.ibm.com>
LKML-Reference: <20090304185605.GA24419@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'x86/urgent' into x86/mm
Ingo Molnar [Wed, 4 Mar 2009 19:20:10 +0000 (20:20 +0100)]
Merge branch 'x86/urgent' into x86/mm

Conflicts:
arch/x86/include/asm/fixmap_64.h
Semantic merge:
arch/x86/include/asm/fixmap.h

Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: EFI: Back efi_ioremap with init_memory_mapping instead of FIX_MAP
Huang Ying [Wed, 4 Mar 2009 02:58:33 +0000 (10:58 +0800)]
x86: EFI: Back efi_ioremap with init_memory_mapping instead of FIX_MAP

Impact: Fix boot failure on EFI system with large runtime memory range

Brian Maly reported that some EFI system with large runtime memory
range can not boot. Because the FIX_MAP used to map runtime memory
range is smaller than run time memory range.

This patch fixes this issue by re-implement efi_ioremap() with
init_memory_mapping().

Reported-and-tested-by: Brian Maly <bmaly@redhat.com>
Signed-off-by: Huang Ying <ying.huang@intel.com>
Cc: Brian Maly <bmaly@redhat.com>
Cc: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <1236135513.6204.306.camel@yhuang-dev.sh.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: set_highmem_pages_init() cleanup, #2
Pekka Enberg [Wed, 4 Mar 2009 08:16:07 +0000 (10:16 +0200)]
x86: set_highmem_pages_init() cleanup, #2

Impact: cleanup

The zones are set up at this stage so there's a highmem zone
available even for the UMA case.

The only difference there is that for machines that have
CONFIG_HIGHMEM enabled but don't have any highmem available,
->zone_start_pfn is zero whereas highstart_pfn is non-zero).

The field is left zeroed because of the !size test in
free_area_init_core() but shouldn't be a problem because
add_highpages_with_active_regions() handles empty ranges just
fine.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Mel Gorman <mel@csn.ul.ie>
LKML-Reference: <1236154567.29024.23.camel@penberg-laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: fix DMI on EFI
Brian Maly [Wed, 4 Mar 2009 02:55:31 +0000 (21:55 -0500)]
x86: fix DMI on EFI

Impact: reactivate DMI quirks on EFI hardware

DMI tables are loaded by EFI, so the dmi calls must happen after
efi_init() and not before.

Currently Apple hardware uses DMI to determine the framebuffer mappings
for efifb. Without DMI working you also have no video on MacBook Pro.

This patch resolves the DMI issue for EFI hardware (DMI is now properly
detected at boot), and additionally efifb now loads on Apple hardware
(i.e. video works).

Signed-off-by: Brian Maly <bmaly@redhat>
Acked-by: Yinghai Lu <yinghai@kernel.org>
Cc: ying.huang@intel.com
LKML-Reference: <49ADEDA3.1030406@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
 arch/x86/kernel/setup.c |    5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

15 years agotracing: add lockdep tracepoints for lock acquire/release
Peter Zijlstra [Wed, 4 Mar 2009 11:32:55 +0000 (12:32 +0100)]
tracing: add lockdep tracepoints for lock acquire/release

Augment the traces with lock names when lockdep is available:

 1)               |  down_read_trylock() {
 1)               |    _spin_lock_irqsave() {
 1)               |      /* lock_acquire: &sem->wait_lock */
 1)   4.201 us    |    }
 1)               |    _spin_unlock_irqrestore() {
 1)               |      /* lock_release: &sem->wait_lock */
 1)   3.523 us    |    }
 1)               |  /* lock_acquire: try read &mm->mmap_sem */
 1) + 13.386 us   |  }
 1)   1.635 us    |  find_vma();
 1)               |  handle_mm_fault() {
 1)               |    __do_fault() {
 1)               |      filemap_fault() {
 1)               |        find_lock_page() {
 1)               |          find_get_page() {
 1)               |            /* lock_acquire: read rcu_read_lock */
 1)               |            /* lock_release: rcu_read_lock */
 1)   5.697 us    |          }
 1)   8.158 us    |        }
 1) + 11.079 us   |      }
 1)               |      _spin_lock() {
 1)               |        /* lock_acquire: __pte_lockptr(page) */
 1)   3.949 us    |      }
 1)   1.460 us    |      page_add_file_rmap();
 1)               |      _spin_unlock() {
 1)               |        /* lock_release: __pte_lockptr(page) */
 1)   3.115 us    |      }
 1)               |      unlock_page() {
 1)   1.421 us    |        page_waitqueue();
 1)   1.220 us    |        __wake_up_bit();
 1)   6.519 us    |      }
 1) + 34.328 us   |    }
 1) + 37.452 us   |  }
 1)               |  up_read() {
 1)               |  /* lock_release: &mm->mmap_sem */
 1)               |    _spin_lock_irqsave() {
 1)               |      /* lock_acquire: &sem->wait_lock */
 1)   3.865 us    |    }
 1)               |    _spin_unlock_irqrestore() {
 1)               |      /* lock_release: &sem->wait_lock */
 1)   8.562 us    |    }
 1) + 17.370 us   |  }

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: =?ISO-8859-1?Q?T=F6r=F6k?= Edwin <edwintorok@gmail.com>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1236166375.5330.7209.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'core/locking' into tracing/ftrace
Ingo Molnar [Wed, 4 Mar 2009 17:49:19 +0000 (18:49 +0100)]
Merge branch 'core/locking' into tracing/ftrace

15 years agolockdep: require framepointers for x86
Peter Zijlstra [Wed, 4 Mar 2009 11:27:27 +0000 (12:27 +0100)]
lockdep: require framepointers for x86

Require framepointers for x86, because otherwise we'll be having
empty stack traces, which is useless.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1236167295.5330.7240.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agolockdep: remove extra "irq" string
Peter Zijlstra [Wed, 4 Mar 2009 13:53:24 +0000 (14:53 +0100)]
lockdep: remove extra "irq" string

Impact: clarify lockdep printk text

print_irq_inversion_bug() gets handed state strings of the form

  "HARDIRQ", "SOFTIRQ", "RECLAIM_FS"

and appends "-irq-{un,}safe" to them, which is either redudant for *IRQ or
confusing in the RECLAIM_FS case.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1236175192.5330.7585.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agolockdep: fix incorrect state name
Peter Zijlstra [Wed, 4 Mar 2009 12:51:13 +0000 (13:51 +0100)]
lockdep: fix incorrect state name

In the recent mark_lock_irq() rework a bug snuck in that would report the
state of write locks causing irq inversion under a read lock as a read
lock.

Fix this by masking the read bit of the state when validating write
dependencies.

Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1236172646.5330.7450.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Wed, 4 Mar 2009 15:49:07 +0000 (07:49 -0800)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc: Run sbc610 USB fixup code only on the appropriate platform.

15 years agodrm: fix double lock typo
Helge Bahmann [Wed, 4 Mar 2009 11:49:14 +0000 (21:49 +1000)]
drm: fix double lock typo

[airlied: you shall not retype patches from other trees half asleep]

Signed-of-by: Dave Airlie <airlied@redhat.com>
15 years agox86, mce: fix build failure in arch/x86/kernel/cpu/mcheck/threshold.c
Ingo Molnar [Wed, 4 Mar 2009 10:47:17 +0000 (11:47 +0100)]
x86, mce: fix build failure in arch/x86/kernel/cpu/mcheck/threshold.c

Impact: build fix

The APIC code rewrite in the x86 tree broke the x86/mce branch:

 arch/x86/kernel/cpu/mcheck/threshold.c: In function ‘mce_threshold_interrupt’:
 arch/x86/kernel/cpu/mcheck/threshold.c:24: error: implicit declaration of function ‘ack_APIC_irq’

Also tidy up the file a bit while at it.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: move devmem_is_allowed() to common mm/init.c
Pekka Enberg [Wed, 4 Mar 2009 09:46:40 +0000 (11:46 +0200)]
x86: move devmem_is_allowed() to common mm/init.c

Impact: cleanup

The function is identical on 32-bit and 64-bit configurations so move
it to the common mm/init.c file.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
LKML-Reference: <1236160001.29024.29.camel@penberg-laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'rfc/splice/tip/tracing/ftrace' of git://git.kernel.org/pub/scm/linux...
Ingo Molnar [Wed, 4 Mar 2009 10:15:42 +0000 (11:15 +0100)]
Merge branch 'rfc/splice/tip/tracing/ftrace' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into tracing/ftrace

15 years agoMerge branch 'tracing/ftrace'; commit 'v2.6.29-rc7' into tracing/core
Ingo Molnar [Wed, 4 Mar 2009 10:14:47 +0000 (11:14 +0100)]
Merge branch 'tracing/ftrace'; commit 'v2.6.29-rc7' into tracing/core

15 years agopowerpc: Run sbc610 USB fixup code only on the appropriate platform.
Tony Breeds [Tue, 3 Mar 2009 17:59:30 +0000 (17:59 +0000)]
powerpc: Run sbc610 USB fixup code only on the appropriate platform.

commit a969e76a7101bf5f3d369563df1ca1253dd6131b (powerpc: Correct USB
support for GE Fanuc SBC610) introduced a fixup for NEC usb controllers.
This fixup should only run on GEF SBC610 boards.

Fixes Fedora bug #486511.
(https://bugzilla.redhat.com/show_bug.cgi?id=486511)

Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agoMerge branch 'x86/core' into x86/mce2
H. Peter Anvin [Wed, 4 Mar 2009 05:05:42 +0000 (21:05 -0800)]
Merge branch 'x86/core' into x86/mce2

15 years agotracing: add binary buffer files for use with splice
Steven Rostedt [Tue, 2 Dec 2008 03:20:19 +0000 (22:20 -0500)]
tracing: add binary buffer files for use with splice

Impact: new feature

This patch creates a directory of files that correspond to the
per CPU ring buffers. These are binary files and are made to
be used with splice. This is the fastest way to extract data from
the ftrace ring buffers.

Thanks to Jiaying Zhang for pushing me to get this code fixed,
 and to Eduard - Gabriel Munteanu for his splice code that helped
 me debug my code.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agoring-buffer: make ring_buffer_read_page read from start on partial page
Steven Rostedt [Wed, 4 Mar 2009 00:51:40 +0000 (19:51 -0500)]
ring-buffer: make ring_buffer_read_page read from start on partial page

Impact: dont leave holes in read buffer page

The ring_buffer_read_page swaps a given page with the reader page
of the ring buffer, if certain conditions are set:

 1) requested length is big enough to hold entire page data

 2) a writer is not currently on the page

 3) the page is not partially consumed.

Instead of swapping with the supplied page. It copies the data to
the supplied page instead. But currently the data is copied in the
same offset as the source page. This causes a hole at the start
of the reader page. This complicates the use of this function.
Instead, it should copy the data at the beginning of the function
and update the index fields accordingly.

Other small clean ups are also done in this patch.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agoring-buffer: replace sizeof of event header with offsetof
Steven Rostedt [Tue, 3 Mar 2009 18:53:07 +0000 (13:53 -0500)]
ring-buffer: replace sizeof of event header with offsetof

Impact: fix to possible alignment problems on some archs.

Some arch compilers include an NULL char array in the sizeof field.
Since the ring_buffer_event type includes one of these, it is better
to use the "offsetof" instead, to avoid strange bugs on these archs.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agoring-buffer: fix ring_buffer_read_page
Steven Rostedt [Tue, 3 Mar 2009 05:27:49 +0000 (00:27 -0500)]
ring-buffer: fix ring_buffer_read_page

The ring_buffer_read_page was broken if it were to only copy part
of the page. This patch fixes that up as well as adds a parameter
to allow a length field, in order to only copy part of the buffer page.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
15 years agoring-buffer: reset write field for ring_buffer_read_page
Steven Rostedt [Tue, 3 Mar 2009 01:56:48 +0000 (20:56 -0500)]
ring-buffer: reset write field for ring_buffer_read_page

Impact: fix ring_buffer_read_page

After a page is swapped into the ring buffer, the write field must
also be reset.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>