The e1000e driver is a great user of the usleep_range() API,
and has nice ranges that in principle help power management.
However the ranges that are used only during system startup are
very long (and can add easily 100 msec to the boot time) while
the power savings of such long ranges is irrelevant due to the
one-off, boot only, nature of these functions.
This patch shrinks some of the longest ranges to be shorter
(while still using a power friendly 1 msec range); this saves
100msec+ of boot time on my BDW NUCs
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Recent kernels support asynchronous probing; most hyperv drivers
can be probed async easily so set the required flag for this.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The kthread has two tasks; handling timeouts (for which it runs once per
second), and submitting queued BIOs. If a BIO happens to be queued after
the thread has processed the queue but before it calls schedule_timeout(),
the thread will sleep for a second before submitting it, which can cause
performance problems in some rare cases (that will become more common in
a subsequent patch).
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Tested-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>
The nvme driver has a "once per second" event where the management kthread
wakes up the system and then reschedules itself for 1 second later.
For power efficiency reasons, I'd like this timer to happen together
with other wakeups in the system.
This patch makes the schedule_timeout() call in the kthread use
round_jiffies_relative(), causing the wakeup to at least align with other
"once per X seconds" events in the kernel.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Tested-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>
The ACPI clocksource takes quite some time to initialize,
and this increases the boot time of the kernel for a
double digit percentage. This while almost all modern
systems will be using the HPET already anyway.
This patch turns the clocksource loading into an asynchronous
operation; which means it won't hold up the boot while
still becoming available normally.
To make this work well, an udelay() had to be turned into an
usleep_range() so that on UP systems, we yield the CPU to
regular boot tasks instead of spinning.
CC: John Stultz <johnstul@us.ibm.com>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Len Brown <lenb@kernel.org>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
the tcp and udp code creates a set of struct file_operations at runtime
while it can also be done at compile time, with the added benefit of then
having these file operations be const.
the trickiest part was to get the "THIS_MODULE" reference right; the naive
method of declaring a struct in the place of registration would not work
for this reason.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This partially reverts commit e6e1e25935.
That commit changed the structure layout of the trace structure, which
in turn broke PowerTOP (1.9x generation) quite badly.
I appreciate not wanting to expose the variable in question, and
PowerTOP was not using it, so I've replaced the variable with just a
padding field - that way if in the future a new field is needed it can
just use this padding field.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The touch screen driver tries to find a range of free channels (which
are an array of bytes), by scanning for the "end of used channel" marker.
However it tries to be WAAAAY too smart and does 32 bit logic on 8 bit
quantities, and in the process completely gets it wrong
(repeatedly read the same register instead of incrementing in the loop,
assuming that if any of the 4 bytes in the 32 byte quantity is free,
all four are free, returning the channel number divided by 4 rather than
the actual first free channel number)
On the setting side, the same mistakes are made by and large; changed
this to just use the byte SCU write functions....
With these fixes we go from a completely non detected touchscreen to
something that appears to completely get detected.
(after also fixing the ordering issue that Jacobs patch should solve)
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
the touch screen driver tries to find a range of free channels (which
are an array of bytes), by scanning for the "end of used channel" marker.
however it tries to be WAAAAY too smart and does 32 bit logic on 8 bit
quantities, and in the process completely gets it wrong
(repeatedly read the same register instead of incrementing in the loop,
assuming that if any of the 4 bytes in the 32 byte quantity is free,
all four are free, returning the channel number divided by 4 rather than
the actual first free channel number)
On the setting side, the same mistakes are made by and large; changed
this to just use the byte SCU write functions....
with these fixes we go from a completely non detected touchscreen to
something that appears to completely get detected.
(after also fixing the ordering issue that Jacobs patch should solve)
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
With the introduction of the new unified work queue thread pools,
we lost one feature: It's no longer possible to know which worker
is causing the CPU to wake out of idle. The result is that PowerTOP
now reports a lot of "kworker/a:b" instead of more readable results.
This patch adds a pair of tracepoints to the new workqueue code,
similar in style to the timer/hrtimer tracepoints.
With this pair of tracepoints, the next PowerTOP can correctly
report which work item caused the wakeup (and how long it took):
Interrupt (43) i915 time 3.51ms wakeups 141
Work ieee80211_iface_work time 0.81ms wakeups 29
Work do_dbs_timer time 0.55ms wakeups 24
Process Xorg time 21.36ms wakeups 4
Timer sched_rt_period_timer time 0.01ms wakeups 1
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
PowerTOP would like to be able to trace timers.
Unfortunately, the current timer tracing is not very useful: the
actual timer function is not recorded in the trace at the start
of timer execution.
Although this is recorded for timer "start" time (when it gets
armed), this is not useful; most timers get started early, and a
tracer like PowerTOP will never see this event, but will only
see the actual running of the timer.
This patch just adds the function to the timer tracing; I've
verified with PowerTOP that now it can get useful information
about timers.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: xiaoguangrong@cn.fujitsu.com
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: <stable@kernel.org> # .35.x, .34.x, .33.x
LKML-Reference: <4C6C5FA9.3000405@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The mrst_max3110 driver had a set of unsafe wakeup sequences
along the following line:
if (!atomic_read(&foo)) {
atomic_set(&foo, 1);
wake_up(worker_thread);
}
and the worker thread would do
if (atomic_read(&foo)) {
do_work();
atomic_set(&foo, 0);
}
which can result in various missed wakups due to test-then-set races,
as well as due to clear-after-work instead of clear-before-work.
This patch fixes these races by using the proper bit test-and-set operations,
and by doing clear-before-work.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The mrst_max3110.c driver uses an open coded, non atomic variable
to create exclusion between two of its worker threads. More than that,
while the main thread does a proper set-work-clear sequence,
the other thread only does a test, with the result that no actual
exclusion is happening.
this patch replaces this open coded variable with a proper mutex
in addition, the 'lock' spinlock is removed from the per adapter structure,
the lock was only ever initialized but never used
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The arguments to intel_scu_ipc_command are "command, subcommand"
the battery driver got this the wrong way around.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Anton Vorontsov <cbouatmailru@gmail.com>
some messages take 4 bytes, but only fill 3 bytes....
this patch makes sure that whatever we send to the SCU is zeroed first
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Matthew Garrett <mjg@redhat.com>
The stack buffer for IPC messages was 16 bytes, limiting messages to a
size of 4 (each message is 32 bit).
However, the touch screen driver is trying to send messages of size 5....
(AC: Set to 20 bytes having checked the max size allowed)
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Matthew Garrett <mjg@redhat.com>
slab has a "once every 2 second" timer for its housekeeping.
As the number of logical processors is growing, its more and more
common that this 2 second timer becomes the primary wakeup source.
This patch turns this housekeeping timer into a deferable timer,
which means that the timer does not interrupt idle, but just runs
at the next event that wakes the cpu up.
The impact is that the timer likely runs a bit later, but during the
delay no code is running so there's not all that much reason for
a difference in housekeeping to occur because of this delay.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
In order for PowerTOP to be able to report how well the new runtime PM is
working for the various drivers, the kernel needs to export some basic
statistics in sysfs.
This patch adds two sysfs files in the runtime PM domain that expose the
total time a device has been active, and the time a device has been
suspended.
With this PowerTOP can compute the activity percentage
Active %age = 100 * (delta active) / (delta active + delta suspended)
and present the information to the user.
I've written the PowerTOP code (slated for version 1.12) already, and the
output looks like this:
Runtime Device Power Management statistics
Active Device name
10.0% 06:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8101E/RTL8102E PCI Express Fast Ethernet controller
[version 2: fix stat update bugs noticed by Alan Stern]
[version 3: rebase to -next and move the sysfs declaration]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Currently, the menu governor uses the (corrected) next timer as key item
for predicting the idle duration.
It turns out that there are specific cases where this breaks down: There
are cases where we have a very repetitive pattern of idle durations, where
the idle period is pretty much the same, for reasons completely unrelated
to the next timer event. Examples of such repeating patterns are network
loads with irq mitigation, the mouse moving but in theory also the wifi
beacons.
This patch adds a relatively simple detector for such repeating patterns,
where the standard deviation of the last 8 idle periods is compared to a
threshold.
With this extra predictor in place, measurements show that the DECAY
factor can now be increased (the decaying average will now decay slower)
to get an even more stable result.
[arjan@infradead.org: fix bug identified by Frank]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Corrado Zoccolo <czoccolo@gmail.com>
Cc: Frank Rowand <frank.rowand@am.sony.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
commit 672917dcc7 ("cpuidle: menu governor: reduce latency on exit")
added an optimization, where the analysis on the past idle period moved
from the end of idle, to the beginning of the new idle.
Unfortunately, this optimization had a bug where it zeroed one key
variable for new use, that is needed for the analysis. The fix is
simple, zero the variable after doing the work from the previous idle.
During the audit of the code that found this issue, another issue was
also found; the ->measured_us data structure member is never set, a
local variable is always used instead.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Corrado Zoccolo <czoccolo@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pavel Machek pointed out that not all CPUs have an efficient
idle at high frequency. Specifically, older Intel and various
AMD cpus would get a higher powerusage when copying files from
USB.
Mike Chan pointed out that the same is true for various ARM
chips as well.
Thomas Renninger suggested to make this a sysfs tunable with a
reasonable default.
This patch adds a sysfs tunable for the new behavior, and uses
a very simple function to determine a reasonable default,
depending on the CPU vendor/type.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: davej@redhat.com
LKML-Reference: <20100509082651.46914d04@infradead.org>
[ minor tidyup ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The ondemand cpufreq governor uses CPU busy time (e.g. not-idle
time) as a measure for scaling the CPU frequency up or down.
If the CPU is busy, the CPU frequency scales up, if it's idle,
the CPU frequency scales down. Effectively, it uses the CPU busy
time as proxy variable for the more nebulous "how critical is
performance right now" question.
This algorithm falls flat on its face in the light of workloads
where you're alternatingly disk and CPU bound, such as the ever
popular "git grep", but also things like startup of programs and
maildir using email clients... much to the chagarin of Andrew
Morton.
This patch changes the ondemand algorithm to count iowait time
as busy, not idle, time. As shown in the breakdown cases above,
iowait is performance critical often, and by counting iowait,
the proxy variable becomes a more accurate representation of the
"how critical is performance" question.
The problem and fix are both verified with the "perf timechar"
tool.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dave Jones <davej@redhat.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20100509082606.3d9f00d0@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
For the ondemand cpufreq governor, it is desired that the iowait
time is microaccounted in a similar way as idle time is.
This patch introduces the infrastructure to account and expose
this information via the get_cpu_iowait_time_us() function.
[akpm@linux-foundation.org: fix CONFIG_NO_HZ=n build]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: davej@redhat.com
LKML-Reference: <20100509082523.284feab6@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Now that the only user of ts->idle_lastupdate is
update_ts_time_stats(), the entire field can be eliminated.
In update_ts_time_stats(), idle_lastupdate is first set to
"now", and a few lines later, the only user is an if() statement
that assigns a variable either to "now" or to
ts->idle_lastupdate, which has the value of "now" at that point.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: davej@redhat.com
LKML-Reference: <20100509082439.2fab0b4f@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Right now, get_cpu_idle_time_us() only reports the idle
statistics upto the point the CPU entered last idle; not what is
valid right now.
This patch adds an update of the idle statistics to
get_cpu_idle_time_us(), so that calling this function always
returns statistics that are accurate at the point of the call.
This includes resetting the start of the idle time for
accounting purposes to avoid double accounting.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: davej@redhat.com
LKML-Reference: <20100509082323.2d2f1945@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
While HR timers have had the concept of timer slack for quite some time
now, the legacy timers lacked this concept, and had to make do with
round_jiffies() and friends.
Timer slack is important for power management; grouping timers reduces the
number of wakeups which in turn reduces power consumption.
This patch introduces timer slack to the legacy timers using the following
pieces:
* A slack field in the timer struct
* An api (set_timer_slack) that callers can use to set explicit timer slack
* A default slack of 0.4% of the requested delay for callers that do not set
any explicit slack
* Rounding code that is part of mod_timer() that tries to
group timers around jiffies values every 'power of two'
(so quick timers will group around every 2, but longer timers
will group around every 4, 8, 16, 32 etc)
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: johnstul@us.ibm.com
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Since the rewrite of the CPU idle governor in 2.6.32, two laptops have
surfaced where the BIOS advertises a C2 power state, but for some reason
this state is not functioning (as verified in both cases by powertop
before the patch in .32).
The old governor had the accidental behavior that if a non-working state
was chosen too many times, it would end up falling back to C1. The new
governor works differently and this accidental behavior is no longer
there; the result is a high temperature on these two machines.
This patch adds these 2 machines to the DMI table for C state anomalies;
by just not using C2 both these machines are better off (the TSC can be
used instead of the pm timer, giving a performance boost for example).
Addresses http://bugzilla.kernel.org/show_bug.cgi?id=14742
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Reported-by: <akwatts@ymail.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Len Brown <len.brown@intel.com>
The ipvs code has a nifty system for doing the size of ioctl command
copies; it defines an array with values into which it indexes the cmd
to find the right length.
Unfortunately, the ipvs code forgot to check if the cmd was in the
range that the array provides, allowing for an index outside of the
array, which then gives a "garbage" result into the length, which
then gets used for copying into a stack buffer.
Fix this by adding sanity checks on these as well as the copy size.
[ horms@verge.net.au: adjusted limit to IP_VS_SO_GET_MAX ]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Patrick McHardy <kaber@trash.net>
show_regs() is called as a mini BUG() equivalent in some places,
specifically for the "scheduling while atomic" case.
Unfortunately right now it does not print a Code: line unlike
a real bug/oops.
This patch changes the x86 implementation of show_regs() so that
it calls the same function as oopses do to print the registers
as well as the Code: line.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
LKML-Reference: <20091102165915.4a980fc0@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
proc.c and video.c are a bit sloppy around types and style,
confusing gcc for a new feature that'll be in 2.6.33 and will
cause a warning on the current code.
This patch changes
if (foo + 1 > sizeof bar)
into
if (foo >= sizeof(bar))
which is more kernel-style.
it also changes a variable in proc.c to unsigned; it gets assigned
a value from an unsigned type, and is then only compared for > not
for negative, so using unsigned is just outright the right type
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Commit b6b39e8f3f (tcp: Try to catch MSG_PEEK bug) added a printk()
to the WARN_ON() that's in tcp.c. This patch changes this combination
to WARN(); the advantage of WARN() is that the printk message shows up
inside the message, so that kerneloops.org will collect the message.
In addition, this gets rid of an extra if() statement.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ACPI /proc write() code takes an unsigned length argument like any write()
function, but then assigned it to a *signed* integer called "len".
Only after this is a sanity check for len done to make it not larger than 4.
Due to the type change a len < 0 is in principle also possible; this patch
adds a check for this.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
For doing work on the Linux power management components, I need to
make long (30+ seconds) traces. Currently, this then results in a
HUGE svg file, with mostly process data that isn't interesting.
This patch adds a --power-only mode to perf timechart that only
outputs the CPU power section of the SVG; this significantly
reduces the size of the SVG file, making even 30+ second traces
viewable with inkscape.
As a minor tweak for the same effect, the minimum text size is
decreased; current inkscape cannot zoom in deep enough to show text
this small, but it reduces inkscape compute time.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: peterz@infradead.org
LKML-Reference: <20090924154013.0675ab71@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The wireless extensions have a copy_from_user to a local stack
array "essid", but both me and gcc have failed to find where
the bounds for this copy are located in the code.
This patch adds some basic sanity checks for the copy length
to make sure that we don't overflow the stack buffer.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: linux-wireless@vger.kernel.org
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The sys_socketcall() function has a very clever system for the copy
size of its arguments. Unfortunately, gcc cannot deal with this in
terms of proving that the copy_from_user() is then always in bounds.
This is the last (well 9th of this series, but last in the kernel) such
case around.
With this patch, we can turn on code to make having the boundary provably
right for the whole kernel, and detect introduction of new security
accidents of this type early on.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tweak the output SVG to increase performance in SVG viewers by
limiting the different types of font sizes and by smarter
transformations on the text.
At least with Inkscape this gives a notable performance improvement
during zoom and scrolling.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090920181438.3a49cb93@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch adds a command line option for timechart that allows the
user to specify the width of the SVG file.
This patch also makes sure that each second of recording has at
least 200 units (pixels at 96 DPI) of width. This impacts
recordings longer than 5 seconds; recordings shorter than 5 second
will scale up to have a width of 1000 units for the whole recording
(as before).
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090920181416.69570c5d@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Given that scheduler latencies are the hot thing nowadays, show the
duration of said latencies in the SVG in text form.
In addition, if the latency is more than 10 msec, pick a brighter
yellow color as a way to point these long delays out.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090920181353.796f4509@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
timechart is a tool to visualize what is going on in the system.
The user makes a trace of what is going on with
> perf record --timechart /usr/bin/some_command
and then can turn the output of this into an svg file
> perf timechart
which then can be viewed with any SVG view; inkscape works well
enough for me.
The idea behind timechart is to create a "infinitely zoomable"
picture; something that has high level information on a 1:1 zoom
level, but which exposes more details every time you zoom into a
specific area.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090912130713.6a77bbc0@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch converts the existing power tracer into an event tracer,
so that power events (C states and frequency changes) can be
tracked via "perf".
This also removes the perl script that was used to demo the tracer;
its functionality is being replaced entirely with timechart.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090912130542.6d314860@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The trace event name<->id mapping is dynamic for each kernel
compile. In order for perf.data to be useable outside the actual
system, we thus need to store a table of this mapping for later
use.
This patch adds this table to perf.data, and provides helper
functions for lookup up fields from this table.
To avoid mistakes, lookup-from-table is kept completely seprate
from lookup-from-local-debugfs.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090912130405.6960d099@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
For counting how long an application has been waiting for
(disk) IO, there currently is only the HZ sample driven
information available, while for all other counters in this
class, a high resolution version is available via
CONFIG_SCHEDSTATS.
In order to make an improved bootchart tool possible, we also
need a higher resolution version of the iowait time.
This patch below adds this scheduler statistic to the kernel.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4A64B813.1080506@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
for some reason, this structure gets compiled as 36 bytes in some files
(the ones that alloacte it) but 40 bytes in others (the ones that use it).
The cause is an off_t type that gets a different size in different
compilation units for some yet-to-be-explained reason.
But the effect is disasterous; the size/offset members of the struct
are at different offsets, and result in mostly complete garbage.
The parser in perf is so robust that this all gets hidden, and after
skipping an certain amount of samples, it recovers.... so this bug
is not normally noticed.
.... except when you want every sample to be exact.
Fix this by just using an explicitly sized type.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4A655917.9080504@linux.intel.com>
the "reserved" field was not initialized to zero, resulting in 4 bytes
of stack data leaking to userspace....
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Use the newly introduced pci_ioremap_bar() function in drivers/serial.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
the "perf report" utility crashed in some circumstances
because the "sym" stack variable was not initialized before used
(as also proven by valgrind).
With this fix both the crash goes away and valgrind no longer complains.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
There is currently only one way for userspace to say "wait for my storage
device to get ready for the modules I just loaded": to load the
scsi_wait_scan module. Expectations of userspace are that once this
module is loaded, all the (storage) devices for which the drivers
were loaded before the module load are present.
Now, there are some issues with the implementation, and the async
stuff got caught in the middle of this: The existing code only
waits for the scsy async probing to finish, but it did not take
into account at all that probing might not have begun yet.
(Russell ran into this problem on his computer and the fix works for him)
This patch fixes this more thoroughly than the previous "fix", which
had some bad side effects (namely, for kernel code that wanted to wait for
the scsi scan it would also do an async sync, which would deadlock if you did
it from async context already.. there's a report about that on lkml):
The patch makes the module first wait for all device driver probes, and then it
will wait for the scsi parallel scan to finish.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Tested-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There are several DMI tables in the i8042 (keyboard) driver already, but
not one for the i8042.reset option. This patch adds such an option.
Two users for this table are added as well, the MSI Wind U-100 and the LG
X110. The MSI Wind also needs to be in the "don't trust the pnp data" for
the touchpad to work on my machine.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Some touchpads don't reset right the first time (MSI Wind U-100 for
example). This patch will retry the reset up to 5 times.
In addition, on x86, we don't fail entire i8042 initialization if
controller reset fails in hope that keyboard port will still be
functional and user will still get a working keyboard. This is
especially important on netbooks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
The battery driver tends to take quite some time to initialize
(100ms-300ms is quite typical).
This patch initializes the batter driver asynchronously, so that other
things in the kernel can initialize in parallel to this 300 msec.
As part of this, the battery driver had to move to the back
of the ACPI init order (hence the Makefile change).
Without this move, the next ACPI driver would just block
on the ACPI/devicee layer semaphores until the battery driver was
done anyway, not gaining any boot time.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
udelay() of 20 milliseconds really ought to just use mdelay(), that avoids
the various wrap scenarios and also is more readable
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
There seems to be a common pattern in the kernel where drivers want to
call request_module() from inside a module_init() function. Currently
this would deadlock.
As a result, several drivers go through hoops like scheduling things via
kevent, or creating custom work queues (because kevent can deadlock on them).
This patch changes this to use a request_module_nowait() function macro instead,
which just fires the modprobe off but doesn't wait for it, and thus avoids the
original deadlock entirely.
On my laptop this already results in one less kernel thread running..
(Includes Jiri's patch to use enum umh_wait)
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (bool-ified)
Cc: Jiri Slaby <jirislaby@gmail.com>
this patch flips the order in which sata and network drivers are initialized.
SATA probing takes quite a bit of time, and with the asynchronous infrastructure
other drivers that run after it can execute in parallel. Network drivers do tend
to take some real time talking to the hardware, so running these later is
a good thing (the sata probe then runs concurrent)
This saves about 15% of my kernels boot time.
Both Dave and Jeff acked this patch and suggested it should go via the async
tree.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Jeff Garzik <jeff@garzik.org>
Now that everyone has been able to test the async code (and it's being used
in the Moblin betas by default), we can enable it by default.
The various fixes needed have gone into 2.6.29 already.
[With an important bugfix from Stefan Richter]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Right now, the kobject_uevent code blocks for each uevent that's being
generated, due to using (for hystoric reasons) UHM_WAIT_EXEC as flag to
call_usermode_helper(). Specifically, the effect is that each uevent
that is being sent causes the code to wake up keventd, then block until
keventd has processed the work. Needless to say, this happens many times
during the system boot.
This patches changes that to UHN_NO_WAIT (brilliant name for a constant
btw) so that we only schedule the work to fire the uevent message, but
do not wait for keventd to process the work.
This removes one of the bottlenecks during boot; each one of them is
only a small effect, but the sum of them does add up.
[Note, distros that need this are broken, they should be setting
CONFIG_UEVENT_HELPER_PATH to "", that way this code path will never be
excuted at all -- gregkh]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some devices use devres_request_irq() for to install their interrupt
handler. Add support for threaded interrupts to devres as well.
[tglx - simplified and adapted to latest threadirq version]
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
the resume code does not currently wait for device probing to finish.
Even without async function calls this is dicey and not correct,
but with async function calls during the boot sequence this is going
to get hit more...
This patch adds the synchronization using the newly introduced helper.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Len Brown <lenb@kernel.org>
Acked-by: Greg KH <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
there's a few places that currently loop over driver_probe_done(), and
I'm about to add another one. This patch abstracts it into a helper
to reduce duplication.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Len Brown <lenb@kernel.org>
Acked-by: Greg KH <gregkh@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
An oops dump also contains the register values.
This patch parses these for (32 bit) x86, and then annotates the
disassembly with these values; this helps in analysis of the oops by the
developer, for example, NULL pointer or other pointer bugs show up clearly
this way.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
The SSS flag, which directs the OS to spin up one disk at a time
to not have the PSU blow out, sometimes gets set even when not needed.
The effect of this is a longer-than-needed boot time.
This patch adds a module parameter that makes the driver ignore SSS
at least as far as the parallel scan during boot is concerned...
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
At 37000 feet somewhere near Greenland I woke up from a half-sleep with the
realisation that __lowest_in_progress() is buggy. After landing I checked
and there were indeed 2 problems with it; this patch fixes both:
* The order of the list checks was wrong
* The locking was not correct.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There has been some light flamewar on lkml about decoding oopses
in modules (as part of the crashdump flamewar).
Now this isn't rocket science, just the markup_oops.pl script
cheaped out and didn't handle modules. But really; a flamewar
all about that?? What happened to C++ in the kernel or reading
files from inside the kernel?
This patch adds module support to markup_oops.pl; it's not the
most pretty perl but it works for my testcases...
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It is useful for diagnosing boot performance to see where async function
calls are waiting on serialization... this patch adds this
functionality to the bootgraph.pl script.
The waiting time is shown as a half transparent, gray bar through the
block that is waiting.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In a discussio with Jeff Garzik, he mentioned that the serialization
for the libata port probes only needs to be within the domain of a host.
This means that for the first port of each host (with ID 0), we don't
need to wait, so we can relax our serialization a little.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch adds a per host flag that allows drivers to opt in into
having its busses scanned in parallel.
Drivers that do not set this flag get their ports scanned in
the "original" sequence.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
turns out that there are real problems with allowing async
tasks that are scheduled from async tasks to run after
the async_synchronize_full() returns.
This patch makes the _full more strict and a complete
synchronization. Later I might need to add back a lighter
form of synchronization for other uses.. but not right now.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dave Jones, in his blog, had some feedback about the bootchart script:
Primarily his complaint was that shorter delays weren't visualized.
The reason for that was that too small delays will have their labels
mixed up in the graph in an unreadable mess.
This patch has a fix for this; for one, it makes the output wider,
so more will fit.
The second part is that smaller delays are now shown with a
much smaller font for the label; while this isn't per se
readable at a 1:1 zoom, at least you can zoom in with most SVG
viewing applications and see what it is you are looking at.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Device drivers that use pci_request_regions() (and similar APIs) have a
reasonable expectation that they are the only ones accessing their device.
As part of the e1000e hunt, we were afraid that some userland (X or some
bootsplash stuff) was mapping the MMIO region that the driver thought it
had exclusively via /dev/mem or via various sysfs resource mappings.
This patch adds the option for device drivers to cause their reserved
regions to the "banned from /dev/mem use" list, so now both kernel memory
and device-exclusive MMIO regions are banned.
NOTE: This is only active when CONFIG_STRICT_DEVMEM is set.
In addition to the config option, a kernel parameter iomem=relaxed is
provided for the cases where developers want to diagnose, in the field,
drivers issues from userspace.
Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Use the newly introduced pci_ioremap_bar() function in drivers/usb.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
while tracking the asynchronous calls during boot using the initcall_debug
convention is useful, doing it once the kernel is done is actually
bad now that we use asynchronous operations post boot as well...
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Dave Jones, in his blog, had some feedback about the bootchart script:
Primarily his complaint was that shorter delays weren't visualized.
The reason for that was that too small delays will have their labels
mixed up in the graph in an unreadable mess.
This patch has a fix for this; for one, it makes the output wider,
so more will fit.
The second part is that smaller delays are now shown with a
much smaller font for the label; while this isn't per se
readable at a 1:1 zoom, at least you can zoom in with most SVG
viewing applications and see what it is you are looking at.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
this makes "rm -rf" on a (names cached) kernel tree go from
11.6 to 8.6 seconds on an ext3 filesystem
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
As suggested by Linus: Don't do the libata init in 2 separate
steps with a global sync inbetween, but do it as one async step,
with a local sync before registering the device.
This cuts the boottime on my machine with 2 sata controllers down
significantly, and it seems to work. Would be nice if the libata
folks take a good look at this patch though..
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
This patch makes the libata port scanning asynchronous (per device).
There is a synchronization point before doing the actual disk scan
so that device ordering is not affected.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
This patch makes part of the scsi probe (which is mostly device spin up and the
partition scan) asynchronous. Only the part that runs after getting the device
number allocated is asynchronous, ensuring that device numbering remains stable.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Right now, most of the kernel boot is strictly synchronous, such that
various hardware delays are done sequentially.
In order to make the kernel boot faster, this patch introduces
infrastructure to allow doing some of the initialization steps
asynchronously, which will hide significant portions of the hardware delays
in practice.
In order to not change device order and other similar observables, this
patch does NOT do full parallel initialization.
Rather, it operates more in the way an out of order CPU does; the work may
be done out of order and asynchronous, but the observable effects
(instruction retiring for the CPU) are still done in the original sequence.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Use the newly introduced pci_ioremap_bar() function in drivers/edac.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the newly introduced pci_ioremap_bar() function in drivers/video.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the newly introduced pci_ioremap_bar() function in drivers/misc.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the newly introduced pci_ioremap_bar() function in drivers/uio.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Hans J. Koch <hjk@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Use the new pci_ioremap_bar() function in drivers/mmc.
pci_ioremap_bar() just takes a pci device and a bar number, with the goal
of making it really hard to get wrong, while also having a central place
to stick sanity checks.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Impact: reduce false positives in iomem_map_sanity_check()
Some drivers (vesafb) only map/reserve a portion of a resource.
If then some other driver comes in and maps the whole resource,
the current code WARN_ON's. This is not the intent of the checks
in iomem_map_sanity_check(); rather these checks want to
warn when crossing *hardware* resources only.
This patch skips BUSY resources as suggested by Linus.
Note: having two drivers talk to the same hardware at the same
time is obviously not optimal behavior, but that's a separate story.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Arjan van de Ven