| Age | Commit message (Collapse) | Author |
|---|
|
My FreeRunner's load average was leveling off at 2.00 as a result of
these two kernel threads:
SDIO Helper D c02b4500 0 255 2
[<c02b4298>] (schedule+0x0/0x2d4) from [<c02b4040>] (__down+0x100/0x158)
[<c02b3f40>] (__down+0x0/0x158) from [<c02b3e2c>] (__down_failed+0xc/0x20)
r7:00000000 r6:c01fbd64 r5:c7cb7134 r4:c7d58000
[<c01fa6fc>] (CardDetectHelperFunction+0x0/0x1ac) from [<c01fbd80>] (HelperLaunch+0x1c/0x28)
r5:c7cb7134 r4:c7cb7134
[<c01fbd64>] (HelperLaunch+0x0/0x28) from [<c005bbd0>] (kthread+0x60/0x94)
r4:c7d58000
[<c005bb70>] (kthread+0x0/0x94) from [<c0048f7c>] (do_exit+0x0/0x744)
r6:00000000 r5:00000000 r4:00000000
SDIO Helper D c02b4500 0 259 2
[<c02b4298>] (schedule+0x0/0x2d4) from [<c02b4040>] (__down+0x100/0x158)
[<c02b3f40>] (__down+0x0/0x158) from [<c02b3e2c>] (__down_failed+0xc/0x20)
r7:c0382bb4 r6:c0382b34 r5:c7d5a000 r4:00000001
[<c01f9104>] (SDIOIrqHelperFunction+0x0/0x29c) from [<c01fbd80>] (HelperLaunch+0x1c/0x28)
r8:00000000 r7:00000000 r6:c01fbd64 r5:c0382bb4 r4:c0382bb4
[<c01fbd64>] (HelperLaunch+0x0/0x28) from [<c005bbd0>] (kthread+0x60/0x94)
r4:c7d5a000
[<c005bb70>] (kthread+0x0/0x94) from [<c0048f7c>] (do_exit+0x0/0x744)
r6:00000000 r5:00000000 r4:00000000
This fixes them to use interruptible sleep primitives while waiting in
their main loops, as is conventional for kernel threads. They can't
actually be interrupted since kernel threads ignore all signals, but
by sleeping this way they get classified as long term waiters, and
don't get counted as running for purposes of load average calculation.
This is intended as a minimal fix. In the longer term, it'd probably
make sense to replace the semaphores with completions or something,
and to do away with some of these StudlyCapped wrapper functions.
Signed-off-by: Jason Uhlenkott <jasonuhl@jasonuhl.org>
--
This is untested, but what could possibly go wrong? ;)
|
|
A few questions have been flying around about how optimal
our waitstates are for various things including Glamo.
This patch introduces new sysfs nodes
/sys/devices/platform/neo1973-memconfig.0/BANKCON0
...
/sys/devices/platform/neo1973-memconfig.0/BANKCON7
If you cat them you get translated info about bus speed on
that chip select, eg,
# cat /sys/devices/platform/neo1973-memconfig.0/BANKCON1
BANKCON1 = 0x00000A40
Type = ROM / SRAM
PMC = normal (1 data)
Tacp = 2 clocks
Tcah = 0 clocks
Tcoh = 1 clock
Tacc = 3 clocks
Tcos = 1 clock
Tacs = 0 clocks
You can write them in hex too
# echo 0x200 > /sys/devices/platform/neo1973-memconfig.0/BANKCON1
The write format for BANKCON0 - 5 looks like this
b1..b0 PMC Page Mode Config
b3..b2 Tacp Page Mode Access Cycle
b5..b4 Tcah Address hold after CS deasserted
b7..b6 Tcoh CS hold after OE deasserted
b10..b8 Tacc Access Cycle Period
b12..b11 Tcos CS setup before OE asserted
b14..b13 Tacs Address setup before CS asserted
BANKCON 6 and 7 have two extra bits
b16..b15 MT Memory type (00=ROM/SRAM, 11=DRAM)
If it's ROM/SRAM, the rest of the bits are as described above.
For DRAM
b1..b0 SCAN Column address number
b3..b2 RAS to CAS delay
The patch is intended to let people experiement on their own. But
of course you will crash things for sure if the timing is wrong, and
you can also trash SD Card data if you make Glamo unstable, so remove
it or remount ro first. Other horrible things are possible, but
because the settings aren't sticky, you should always be able to
recover by either normal reboot usually or at worst NOR boot and then
dfu. Most likely you will just crash your session and have to reboot
if your settings are bad, but consider yourself warned bad things are
possible. :-)
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
We at least always enabled backlight on resume, this patch
changes us to set backlight back to last requested backlight
brightness level on resume. Note it means that you can
resume with screen blanked, but it should come back if that
happened with touchscreen action as usual.
/sys/class/backlight/pcf50633-bl/actual_brightness
and
/sys/class/backlight/pcf50633-bl/brightness
seem to agree after resume when reportedly they didn't before.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
During the suspend current reduction campaign on suspend I
forced the GPS UART to be GPIO and to drive 0 into the GPS
unit so we would not burn current there. On resume it lets
the pins act as UARTs again. But really, we should do this
all the time that the GPS unit is off, lest we leak it
enough power to hold internal state and make trouble.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
This patch adds a sysfs node:
/sys/class/i2c-adapter/i2c-0/0-0073/force_usb_limit_dangerous
it allows to force the charging limit regardless of the one chosen
by pcf50633 kernel driver. As such, if you write a charging limit
here that is not suitable for the power source, and the power source
is not current limited on its side, it could draw more current than
your power source can handle, burn down you house, etc.
If you're certain that your power supply can handle it, you can use
this on your own responsibility to make the amount drawn by the
PMU match what you believed your power supply could handle.
Example usage, in case where you have a dumb 500mA USB charger that
does not have the ID resistor:
# cat /sys/class/i2c-adapter/i2c-0/0-0073/charger_type
host/500mA usb mode 100mA <=== dumb charger does not ennumerate us
# echo 500 > /sys/class/i2c-adapter/i2c-0/0-0073/force_usb_limit_dangerous
# cat /sys/class/i2c-adapter/i2c-0/0-0073/charger_type
host/500mA usb mode 500mA
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Reported-by: Mickey Lauer <mickey@openmoko.org>
AUX level detection is inverted based on GTA01 or 02
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
We leave SECOND unmasked on resume, it's like the
situation at probe() time, but there it makes us
turn SECOND off after coldplug action. So we need
to act like after that has happened, not exactly
like what we do at probe / init time.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Subject: [PATCH] [neo1973 leds] Move from mutex to spinlock because we may not use mutexes
The led triggers may call set_brightness from atomic contexts. As
mutex_lock calls might_sleep and sleeping is not allowed in atomic contexts
we have to switch to spinlocks here.
Signed-Off-By: Holger Freyther <zecke@openmoko.org>
|
|
Subject: [PATCH] [pcf50633] Avoid ooops on start with inserted usb cable
The pcf50633_global might not be initialized when we get the first
usb interrupt. We would oops inside the dev_err because we made up
a struct device.
Signed-Off-By: Holger Freyther <zecke@openmoko.org>
|
|
Add missing initialization for the touchscreen driver for the
gta01 platform.
Signed-off-by: Mike Westerhof <mwester@dls.net>
|
|
Signed-off-by: Mike Westerhof <mwester@dls.net>
|
|
Signed-off-by: Matt Hsu <matt_hsu@openmoko.org>
- add an interrupt for ar6k wifi module
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Subject: [PATCH] [pcf50633] Report more events to userspace using the default callback
Signed-Off-By: Holger Freyther <zecke@openmoko.org>
|
|
Subject: [PATCH] [battery] Make the bq27000 send an uevent when the charging state possible changed
Remove the todo entries from the pcf50633, make the mach-gta02
call the bq27000 driver from the pmu callback.
|
|
Subject: [PATCH] [janitor] make checkpatch.pl happy
|
|
Subject: [PATCH] [janitor] Make checkpatch happy on the header files
|
|
Subject: [PATCH] [bq27000] Make the checkpatch.pl happy
|
|
Subject: [PATCH] Hacky CONFIG_NO_IDLE_HZ (dyn-tick) support for S3C24xx.
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
This patch introduces a new resume debugging concept: if we
get an OOPS inbetween starting suspend and finishing resume, it
uses a new "emergency spew" device similar to BUT NOT REQUIRING
CONFIG_DEBUG_LL to dump the syslog buffer and then the OOPS
on the debug device defined by the existing CONFIG_DEBUG_S3C_UART
index. But neither CONFIG_DEBUG_LL nor the S3C low level configs
are needed to use this feature.
Another difference between this feature and CONFIG_DEBUG_LL is that
it does not affect resume timing, ordering or UART traffic UNLESS
there is an OOPS during resume.
The patch adds three global exports, one to say if we are inside
suspend / resume, and two callbacks for printk() to use to init
and dump the emergency data. The callbacks are set in s3c serial
device init, but the whole structure is arch independent.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Reported-by: John Lee <john_lee@openmoko.com>
We don't reset the devices either at init or resume, where init
means use the BOOT bit to reload device calibration coefficients
from internal EEPROM. John Lee saw brain-damaged behaviour after
resume and sometimes after boot (since it may not have lost power
to force a BOOT itself that makes sense).
This patch
- adds a diagnostic dump feature down /sys
- forces BOOT action on init and resume, and waits for
completion
- makes sure XYZ capture is enabled on resume
- adds some constants in the .h and removes some magic numbers
in the code by using them
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Reported-by: Holger Freyther <zecke@openmoko.org>
length can be zero... blowing a divide by zero exception...
which somehow I don't get (?) Anyway the code is wrong and
this should fix it.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Touchscreen on GTA01-02 experiences noise on the channel that serves the
"tall axis" of the LCM. The sample quality of the other axis is good.
The bad samples have a characteristic of one shot excursions that can
reach +/- 20% or more of the sample average.
Previously, we had a simple averaging scheme going in the touchscreen
driver that summed up 32 x and ys and then divided it by 32. This patch
first tidies up the existing code for style, then adds a new "running
average" concept with a FIFO. The running average is separate from the
summing average mentioned above, and is accurate for the last n samples
sample-by-sample, where n is set by 1 << excursion_filter_len_bits in the
machine / platform stuff.
The heuristic the patch implements for the filtering is to accept all
samples, but tag the *previous* sample with a flag if it differed from
the running average by more than reject_threshold_vs_avg in either
axis. The next sample time, a beauty contest is held if the flag was
set to decide if we think the previous sample was a one-shot excursion
(detected by the new sample being closer to the average than to the
flagged previous sample), or if we believe we are moving (detected by
the new sample being closer to the flagged previous sample than the
average. In the case that we believe the previous sample was an
excursion, we simply overwrite it with the new data and adjust the
summing average to use the new data instead of the excursion data.
I only tested this by eyeballing the output of ts_print_raw, but it
seemed to be quite a bit better. Gross movement appeared to be
tracked fine too. If folks want to try different heuristics on top
of this patch, be my guest; either way feedback on what it looks like
with a graphical app would be good.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
A panic is silent on GTA02, it would be good if we got a little hint
if we are crashing (eg, in suspend / resume) from a panic instead of
a deadlock, etc. On a normal PC i8042 blinks the keyboard lights if
we panic, this patch causes AUX to flash at 5Hz in event of a panic.
Tested by giving kernel fake root= that didn't exist.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
|
|
make it happy
Call SET_NET_DEV to set a parent device. All other net drivers
are doing this and hald needs a parent to add the network device.
|
|
Reported-by: Graeme Gregory <graeme@openmoko.org>
Basically-solved-by: Graeme Gregory <graeme@openmoko.org>
Graeme found a while back that on resume, IISCON register in
s3c2442 does not show LRCK toggling in LRINDEX bit any more,
causing s3c24xx_snd_lrsync() to timeout and return an error,
aborting restart of any live stream that was playing at
suspend.
I confirmed it was true, meddled around for a bit looking
for some magic to restart LRCK or at least the reporting of
it, and in the end worked around it using the method noted
by Graeme: just ignore LRCK sync if it timed out. The worst
that could happen would be L and R swap for the duration of
stream that was suspended into but probably not even that.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Everywhere in the sources except the probe function the context
pointer is called "pcf"... in there it's called "data" for some
reason. This stops confusion by changing it to be "pcf" in there
as well.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
pcf50633.c shouldn't know GTAxx at all. Move to using a
platform callback to allow definition of platform devices
with pcf50633 as parent device (good for enforcing suspend /
resume ordering). Remove all code references to GTAxx from
the sources (one string left for compatability).
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Two issues... we never took care to take down engines in suspend
and bring them back in resume. This was part of the display
corruption that could be seen briefly on resume. The other issue
that made the "noise" corruption was bad ordering of resume steps.
This patch simplifies (removing needless re-init) resume actions
and makes explicit the suspend and resume steps. It also adds
code to track which engines are up and push them down in suspend
and bring them back in resume.
The result is no more corruption of display buffer in suspend, it
comes back completely clean.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Backlight wasn't off by default on resume, so it was never really
deferred (until LCM is initialized). This fixes that and so removes
the brief white screen between pcf50633 resume and LCM init.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Disable pcf interrupt (not for wake, just as interrupt) in
suspend, re-enable it again just before we force-call the
workqueue function at end of pcf resume, which leads to
pcf interrupt source registers getting cleared so it can
signal an interrupt normally again.
This change ends the uncontrolled appearance of pcf interrupts
during resume time which previously caused the work to attempt
to use the I2C stuff before i2c host device had itself resumed.
Now the isr work is only queued, and the isr work function called,
definitively after pcf resume completes.
In suspend time, the work function may have been queued some
time before and be pending, and it could still show up at a
bad time. Therefore if the work function sees that it is
coming since the start of pcf50633 suspend function, it
aborts without attempting to read the pcf interrupt regs,
leaving them for resume to take care of.
USB current limit and no battery work functions are also made
aware of suspend state and act accordingly.
Lastly I noticed that in early resume, i2c_get_clientdata(&pcf->client)
returns NULL, presumably because i2c device is still suspended. This
could easily make trouble for async events like interrupt work,
since pcf pointer is the client data. Disabling appearance of the
work until after pcf50633 resume will also avoid that.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Use an enum to define pcf50633 suspend / resume state.
Add PCF50633_SS_RESUMING_BUT_NOT_US_YET to be the state
early in resume: add platform driver resume function just
to set this state so we can differentiate between early
resume and late suspend.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Whoops left it up in suspend
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
mach-gta02 meddles with the regulator platform struct after
it is defined, leading to LCM power getting lost in suspend
despite I set it to be left up. Fixing this finally removes
the incredibly stubborn white LCM on suspend "flash".
This is also going to be implicated in Sean McNeil's
experience of monochromatic LCM after resume, which was
previously attacked by resetting and re-initing the LCM
from scratch.
In addition, I realized that we take down core_1v3 in
pcf50633 suspend action, this is happening near the
start of suspend, so we are in a meta-race to finish
suspend in a controlled way before the caps on core_1v3
run out (I only saw 23.3uF total). If it's true, this
is where the weirdo sensitivity to timing during
suspend is coming from.
Therefore in this patch we also remove sleeps and
dev_info() etc (which have to flush on serial console)
from the pc50633 isr workqueue if we are in pcf50633
driver suspend state 1, ie, suspending... because we
don't have time for it.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Sean McNeil reports that he doesn't get pcf50633 interrupts any
more after resume. This adds back the call to ISR work in
the resume, removal of which is probably to do with it.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
If you boot from SDCARD, this helper script for the build host
mounts SD card part 1, copies the new uImage.bin, umounts it and
then remounts SD card part 2 as ro, before doing a reboot all
in one step. Read the instructions inside the script for adding
your public key to the GTAxx rootfs for really really simple and
nice automatic update and reboot action.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
The LCM spins for 100ms during resume for not much reason. Leave it powered
(it is meant to pull uA when suspended) and get nice fast resume to video.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
pcf50633 needs to take responsibility for managing current limit
changes asycnhrnously, ie, from USB stack enumeration. It's a feature of
pcf50633 not mach-gta02.c, and we can do better with taking care about
keeping it from firing at a bad time in there too.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Simplify and speed up bulk sequential I2C actions in pcf50633
the time savings are pretty considerable and so is the simplification
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
More pcf50633 major time saving by using i2c bulk autoincrement. Code
reduction too by using array for time elements.
Signed-off-by: Andy Green <andy@openmoko.com>
|
|
Improve pcf50633 interrupt service scheduling to enforce only servicing
when resume action is completed
Signed-off-by: Andy Green <andy@openmoko.com>
|
