With change b610695b61, we fixed a problem
with the number of FP regs that were being saved on the stack. That change
decreased the required stack size for non-FP tasks by 64 bytes, but
increased the size needed for FP tasks (such as the lightbar).
The lightbar task was previously using within 64 bytes of its alloted stack,
so handling the task switching correctly meant that it now overflowed.
The hooks task had the same problem, but was hidden by the lightbar task.
This CL bumps the LARGER_TASK_STACK_SIZE up a bit, and switches the lightbar
task to use it instead of the default size.
BUG=chrome-os-partner:27971, chrome-os-partner:28407
BRANCH=ToT
TEST=Try it on both Link and Samus.
Before this change, the Samus lightbar was overflowing its stack every time
the AP booted (causing the lightbar to do things). With this change, it
doesn't. Here are typical stack sizes after this CL:
Task Ready Name Events Time (s) StkUsed
0 R << idle >> 00000000 28.394913 328/512
1 HOOKS 00000000 0.534085 640/768
2 R LIGHTBAR 10000000 5.359356 520/768
3 CHARGER 00000000 0.094674 384/512
4 CHIPSET 00000000 0.003353 320/512
5 KEYPROTO 00000000 0.002814 312/512
6 HOSTCMD 00000000 0.002942 244/512
7 R CONSOLE 00000000 0.193776 340/768
8 POWERBTN 00000000 0.000392 248/512
9 KEYSCAN 00000000 0.409337 332/512
Change-Id: Ica93608c8adb225410a62ec3a0a27944c479270a
Signed-off-by: Bill Richardson <wfrichar@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/197733
Reviewed-by: Alec Berg <alecaberg@chromium.org>
Reviewed-by: Randall Spangler <rspangler@chromium.org>
Vboot hash calculation takes ~350 ms during EC boot. Since the hash
task is higher priority than the hook task, this starves all the hooks
during boot.
We could, in theory, fix that simply by swapping the priority of the
hook and hash tasks. But then watchdog detection (in the hook task)
wouldn't detect hangs in the hash task.
A better fix (implemented here) is to convert the hashing operation to
a series of deferred function calls. This gets rid of the hash task
entirely, and allows all pending hooks and other deferred function
calls to take place between each chunk of hashing.
On STM32-based boards, we need to bump up the hook task stack size,
since hashing is called from several layers deep in the hook task
instead of at the top of its own task, but this is still a net win of
several hundred bytes of SRAM.
BUG=chrome-os-partner:24892
BRANCH=rambi
TEST=Boot EC; look for "hash start" and "hash done" debug output.
'taskinfo' shows at least 32 bytes of unused stack for HOOKS task.
'hash ro' runs properly from EC console.
Change-Id: I9e580dc10fc0bc8e44896d84451218ef67578bbe
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/181954
AP throttling in the thermal task ends up calling a pretty deep nested
set of calls, and in the worst case can overflow the stack. Bump up
the stack size for the hook task on x86 platforms to compensate.
BUG=chrome-os-partner:24536
BRANCH=peppy/falco
TEST=taskinfo shows hook task increased from 512 to 640 bytes stack
shmem shows at least 4000 bytes free
Change-Id: I63da7c47b993c935d895f91d787844655071da0d
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/180684
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Problems with existing thermal control loop:
* Not multi-board friendly. thermal.c only supports Link and needs
refactoring. Temp thresholds and fan speeds are hard-coded.
* Only the PECI temp is used to determine the fan speed. Other temp sensors
are ignored.
* Has confusing data structures. Values in the CPU temp thresholds array mix
ACPI thresholds with fan step values.
With this change, the thermal task monitors all temp sensors in order to
perform two completely independent functions:
Function one: Determine if the host needs to be throttled by or informed of
any thermal events.
For thermal events, each temp sensor will have three threshold levels.
TEMP_HOST_WARN
* When any sensor goes above this level, host_throttle_cpu(1) will be called
to ask the CPU to slow itself down.
* When all sensors drop below this level, host_throttle_cpu(0) will be called.
* Exactly AT this level, nothing happens (this provides hysteresis).
TEMP_HOST_HIGH
* When any sensor goes above this level, chipset_throttle_cpu(1) will be
called to slow the CPU down whether it wants to or not.
* When all sensors drop below this level, chipset_throttle_cpu(0) will be
called.
* Exactly AT this level, nothing happens (this provides hysteresis).
TEMP_HOST_SHUTDOWN
* When any sensor is above this level, chipset_force_shutdown() will be
called to halt the CPU.
* Nothing turns the CPU back on again - the user just has to wait for things
to cool off. Pressing the power button too soon will just trigger shutdown
again as soon as the EC can read the host temp.
Function two: Determine the amount of fan cooling needed
For fan cooling, each temp sensor will have two levels.
TEMP_FAN_OFF
* At or below this temperature, no active cooling is needed.
TEMP_FAN_MAX
* At or above this temperature, active cooling should be running at maximum.
The highest level of all temp sensors will be used to request the amount of
active cooling needed. The function pwm_fan_percent_to_rpm() is invoked to
convert the amount of cooling to the target fan RPM.
The default pwm_fan_percent_to_rpm() function converts smoothly between the
configured CONFIG_PWM_FAN_RPM_MIN and CONFIG_PWM_FAN_RPM_MAX for percentages
between 1 and 100. 0% means "off".
The default function probably provide the smoothest and quietest behavior,
but individual boards can provide their own pwm_fan_percent_to_rpm() to
implement whatever curves, hysteresis, feedback, or other hackery they wish.
BUG=chrome-os-partner:20805
BRANCH=none
TEST=manual
Compile-time test with
make BOARD=falco runtests
On the EC console, the existing fan commands should work correctly:
faninfo - display the fan state
fanduty NUM - force the fan PWM to the specified percentage (0-100)
fanset RPM - force the fan to the specified RPM
fanset NUM% - force the fan to the specified percentage (0-100) between
its configured minimum and maximum speeds from board.h
(CONFIG_PWM_FAN_RPM_MIN and CONFIG_PWM_FAN_RPM_MAX)
fanauto - let the EC control the fan automatically
You can test the default pwm_fan_percent_to_rpm() with
fanset 1%
faninfo
The fan should be turning at CONFIG_PWM_FAN_RPM_MIN. Let the EC control it
automatically again with
fanauto
Also on the EC console, the thermal settings can be examined or changed:
> temps
PECI : 327 K = 54 C
ECInternal : 320 K = 47 C
G781Internal : 319 K = 46 C
G781External : 318 K = 45 C
>
> thermalget
sensor warn high shutdown fan_off fan_max name
0 373 387 383 333 363 PECI
1 0 0 0 0 0 ECInternal
2 0 0 0 0 0 G781Internal
3 0 0 0 0 0 G781External
>
> help thermalset
Usage: thermalset sensor warn [high [shutdown [fan_off [fan_max]]]]
set thermal parameters (-1 to skip)
>
> thermalset 2 -1 -1 999
sensor warn high shutdown fan_off fan_max name
0 373 387 383 333 363 PECI
1 0 0 0 0 0 ECInternal
2 0 0 999 0 0 G781Internal
3 0 0 0 0 0 G781External
>
From the host, ectool can be used to get and set these parameters with
nearly identical commands:
ectool thermalget
ectool thermalset 2 -1 -1 999
Change-Id: Idb27977278f766826045fb7d41929953ec6b1cca
Signed-off-by: Bill Richardson <wfrichar@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/66688
Reviewed-by: Randall Spangler <rspangler@chromium.org>
Power button logic is common across all platforms and is not
LM4-specific, so move it to its own module. Switch.c will eventually
be moving to common/ and will common across all platforms (not just
x86), and splitting out the x86 power button logic is needed before
that too.
BUG=chrome-os-partner:18343
BRANCH=none
TEST=manual
1) power on system with both lid and power button.
2) power+refresh -> reboots
3) power+refresh+esc -> recovery mode
4) power+refresh+downarrow -> reboots, AP stays off
5) toggling recovery GPIO via servo should generate SW debug output
showing bit 0x10 toggling
Change-Id: I07714e2c035dceece66f90407983397d2697e7d5
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/61780
For most tests, we don't need to power the AP. Let's exclude chipset
task to save memory space.
BUG=chrome-os-partner:18598
TEST=Run pingpong test on Spring
BRANCH=none
Change-Id: I545c5b3e1c27b0067d4ffe09a7971d32b75d6039
Signed-off-by: Vic Yang <victoryang@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/47833
Reviewed-by: Randall Spangler <rspangler@chromium.org>
Reviewed-by: Vincent Palatin <vpalatin@chromium.org>
This changes current TASK() syntax to TASK_BASE() and TASK_NORMAL(),
where TASK_BASE is necessary for the EC to boot on a board and
TASK_NORMAL represents the task that can be removed in a test binary.
Tasks introduced by a test should be listed as TASK_TEST().
Note that this CL breaks current tests (many of them are broken anyway),
which will be fixed in up coming CLs.
BUG=chrome-os-partner:18598
TEST=Build link/bds/spring/snow/daisy/mccroskey. (mccroskey failed for
unrelated issue)
BRANCH=none
Change-Id: Ic645cdae0906ed21dc473553f1f43c2537ec4bb9
Signed-off-by: Vic Yang <victoryang@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/47531
This makes the charging task name consistent across platform.
No functional changes, just renaming.
BUG=chrome-os-partner:18343
BRANCH=none
TEST=build all EC boards
Change-Id: I348b31313f6604df2a05b474bdf6e0be7450c8c9
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/46891
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Since it's really the keyboard protocol task, not just handling i8042
commands. For consistency across keyboard protocols.
No functional changes, just renaming.
BUG=chrome-os-partner:18360
BRANCH=none
TEST=boot link and type on keyboard
Change-Id: I800a691a344f82bf582693cae865414b7d5d382a
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/46885
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Getting rid of a task saves on RAM requirements.
BUG=chrome-os-partner:18360
BRANCH=none
TEST=boot link and hold a key down; see it repeat as expected
Change-Id: I1ae4dc486f6669d5ad15899202abef85b8c1e7e8
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/46826
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Rename tasks
TICK -> HOOKS
The hooks task handles more than just the TICK hook now.
X86POWER -> CHIPSET
GAIAPOWER -> CHIPSET
Kinda kludgy that the name of the task controls which chipset source gets
included. Change this to a CONFIG_CHIPSET_{X86,GAIA} #define to make it
easier to support future chipsets. Also, rename the task function to
chipset_task() so ec.tasklist is chipset-agnostic.
No code changes, just renaming constants and functions.
BUG=none
BRANCH=none
TEST=build bds,link,daisy,snow,spring
Change-Id: I163ce1cd27b2d8d030d42bb1f7eb46b880c244fb
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/45805
This reduces memory / code size, and gets rid of ifdefs in temp_sensor.c.
BUG=chrome-os-partner:15714
BRANCH=none
TEST=boot system and run 'ectool temps all' every few seconds
- ectool temps all
The numbers should update over time.
Change-Id: Idaac7e6e4cbc1d6689f5d3b607c623a5cc536a4f
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/36940
Since it handles not just power button, but also lid switch, AC
detect, and other switches.
No functional changes; just renaming.
BUG=chrome-os-partner:15579
BRANCH=none
TEST=boot system, power on/off with power button
Change-Id: I51628a52293f7207715f5f6bf368a08fe6c3dbce
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/36821
Adds a new HOOK_TICK event which is called every 250ms (LM4) or 500ms
(STM32). This will be used to consolidate a number of tasks which do
small amounts of work infrequently, and previously needed their own
task functions.
This CL adds the tick task; subsequent CLs will consolidate watchdog
and other tasks into tick hooks.
BUG=chrome-os-partner:15714
BRANCH=none
TEST=taskinfo shows TICK task as lowest priority
Change-Id: I9068ee99d56a5bf5c12afd86ad51998c013f4954
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/36908
Reviewed-by: Simon Glass <sjg@chromium.org>
Since POWERSTATE is confusing whether it refers to battery power or
system power.
BUG=chrome-os-partner:15579
BRANCH=none
TEST=taskinfo; see CHARGER task
Change-Id: I5a237b1329cace4ce48ae39d8954c08a9912ed4b
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/36707
This works around a potential LM4 chip problem where edges on the FRMH
status bit don't always trigger interrupts. The workaround is to look
at FRMH for each channel in the interrupt handler rather than the
interrupt status, and to trigger the interrupt every 250ms to sweep up
any missed writes. We already do this for port 80 writes; this just
extends the workaround to all channels.
BUG=chrome-os-partner:13965
BRANCH=link
TEST=manual
- boot system
- EC console should show a number of HC lines for host command
- EC console should show a number of ACPI queries
- switch to root shell; keyboard should work
- ectool version should work
Change-Id: If02d685519c69ee88c055c8374a6c655a277e637
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/35871
Reviewed-by: Simon Glass <sjg@chromium.org>
With this CL, if CONFIG_FPU is defined (only for Link, ATM), the EC task
switcher will enable CONTROL.FPCA and expect all stack contexts to include
floating point state as well as normal state (an additional 18 words).
To support this, we need to increase the allocated stack space for each
task. The stack sizes are already chosen empirically, so I'm just rounding
them up a bit.
BUG=chrome-os-partner:14766
BRANCH=Link
TEST=manual
There should be no noticeable change. If you run the EC command "taskinfo"
you'll see the increased size each thread's stack, but everything that was
working before should continue to work just fine.
The additional overhead required to load and store another 18 words on each
context switch is not really measurable (I tried).
Change-Id: Ibaca7d7a2565285f049fda6906f32761e83207af
Signed-off-by: Bill Richardson <wfrichar@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/34391
Reviewed-by: Vincent Palatin <vpalatin@chromium.org>
Increase stack size slightly for vboot hash task since the vboot
SHA256 function allocates ~300 bytes of stack data. Reduce stack size
for watchdog, power LED, and a few other tasks with simple call trees
where we can be sure an error path isn't going to blow past the
reduced stack.
This frees up ~1KB of RAM on STM32.
BUG=chrome-os-partner:13814
BRANCH=all
TEST=boot system; shmem should show more unused RAM; taskinfo should show
tasks still have unused stack
Change-Id: I47d6b77564a0180d15d86667cc0566a8919b776e
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/32608
Reviewed-by: Vincent Palatin <vpalatin@chromium.org>
EC computes a SHA-256 hash of its RW code on boot. Also adds host and
console commands to tell the EC to recompute the hash, or hash a
different section of flash memory.
BUG=chrome-os-partner:10777
TEST=manual
1) ectool echash -> should match what the EC precomputed
2a) ectool echash recalc 0 0x10000 5
2b) on EC console, 'hash 0 0x10000 5'
2c) results should agree
3a) on ec console, 'hash 0 0x3e000' then quickly 'hash abort'
3b) ectool echash -> status should be unavailable
4) ectool echash start 0 0x3e000 6 && ectool echash && ectool echash abort && sleep 2 && ectool echash
status should be busy, then unavailable
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Change-Id: I6806d7b4d4dca3a74f476092551b4dba875d558e
Reviewed-on: https://gerrit.chromium.org/gerrit/26023
Charging state machine doesn't need to be able to preempt everybody.
Keyboard scanning and power button should preempt, because they need
to debounce/scan at a stable rate.
Signed-off-by: Randall Spangler <rspangler@chromium.org>
BUG=none
TEST=system still boots
Change-Id: Id57c680b9fa4652bc10d19270620d63788a7b269
Mainly add a typematic task that counts down the delay. Set the initial delay
in the keyboard_state_changed() when key pressed and clean it when released.
BUS=chrome-os-partner:8463
TEST=press on a particular key and screen shows that key is repeating.
Change-Id: Ic8432f8b38b514476588e0b7ad8fdc8a0b0c0b51
I need to clean up the console commands and provide the same functionality
via ectool, but this is a good starting point.
BUG=chrome-os-partner:7839
TEST=manual
Power up the CPU. The lights should blink.
Change-Id: Ic05a171d2b647551f1cfc7d6b2fd101088cac137
Signed-off-by: Bill Richardson <wfrichar@chromium.org>
This CL adds a charge state machine for SMB compliant battery pack.
Vendor specific charge constraints can be applied through function
call, defined in battery_pack.h .
BUG=chrome-os-partner:7526
TEST=Attach EC serial console
Unplug AC adapter: state ==> "discharge"
Plug AC adapter: state ==> "charge"
Battery full: state ==> "idle"
Unplug battery: state ==> "error"
Change-Id: Iabff0988a6067d37c17c11b060bbb7e66505c118
The thermal engine monitors the temperature readings from all sensors.
For each sensor, five threshold temperatures can be set:
1. Low fan speed.
2. High fan speed.
3. SMI warning.
4. Shutdown CPU.
5. Shutdown everything we can.
Each of these thresholds can be set to either a fixed value or disabled.
Currently the real implementation of SMI warning and shutting down is
left as TODO, as indicated in the comment.
Signed-off-by: Vic Yang <victoryang@chromium.org>
BUG=chrome-os-partner:8250
TEST=Manually change threshold value to test all actions can be triggered.
Change-Id: If168dcff78ef2d7a3203cb227e1739a08eca961e
Add a task to update fan speed in LPC mapped memory once per second.
Also added read_mapped_mem16 and read_mapped_mem32.
Signed-off-by: Vic Yang <victoryang@chromium.org>
BUG=chrome-os-partner:8183
TEST="ectool pwmgetfanrpm" shows same result as "faninfo" from ec
console.
Change-Id: Ibc536acd39f836ffcad0bfa7c9c14e730220bd49
A temperature polling task is added to achieve temporal correction and
also reduce the latency of reading temperature.
Factor out sensor specific part to keep code clean.
Signed-off-by: Vic Yang <victoryang@chromium.org>
BUG=chrome-os-partner:7801
TEST=On link, 'temps' shows all temperature readings.
Cover each sensor with hand and see object temperature rise.
Compilation succeeded on bds/adv/daisy/discovery.
Change-Id: I3c44c8b2e3ab2aa9ce640d3fc25e7fba56534b86
You can now enable/disable tasks more easily.
To conditionally compile a C file depending on the task FOO activation,
just write something like that in the build.mk file :
common-$(CONFIG_TASK_FOO)+=foo_source.o
Signed-off-by: Vincent Palatin <vpalatin@chromium.org>
BUG=None
TEST=make all BOARD=link && make qemu-tests
Change-Id: I760fb248e1599d13190ccd937a68ef47da17b510
For bringup, this powers on the x86 unconditionally.
Signed-off-by: Randall Spangler <rspangler@chromium.org>
BUG=chrome-os-partner:7528
TEST=none
Change-Id: Ib23e56d38ab42f8d8a4dbd1ba9dce12f0c3eeec9
Signed-off-by: Randall Spangler <rspangler@chromium.org>
BUG=chrome-os-partner:7499
TEST=press and release power button; should see debug messages
Change-Id: I8909ae4643afc98753edb690771618ad43135e3e
By default the scanning code would pull-low all column pins and listen to
any key press interrupt on input pins. This can save power compared to the
repeatly polling.
Once a key is pressed, the scanning code enters the busy loop and pull-low
column pins one by one. Then generate the scan code to host.
The code keeps polling after 1 second after no key is pressed. Then goes
back to interrupt mode.
BUG=none
TEST=Manual tested on S*y machine.
Change-Id: I0bf8877450dbd6ad1197a2fe1714ab755dc49a80
Implement EC lid switch interrupt handler and debouncing.
BUG=chrome-os-partner:7363
TEST=Manually test lid switch output signal is correct.
Use UART console to see debouncing is correct.
Change-Id: I74aad63330716da017fc4a57002349461c6a9b26
