Add graceful shutdown support #844
Conversation
jonasdn
requested review from
ataffanel,
evoggy,
knmcguire,
krichardsson,
tobbeanton and
whoenig
Currently to shut-down the Crazyflie, the nRF51 simply cut the power
to thae system. This was working fine until we started implementing
SD-card functionality: when shutting down the system that way we have
huge chances of corrupting the SD-card filesystem. This has lead to
implementation of very inefficient sd-car logging procedure to decrease
the probability of corruption.
A better solution for the shutdown problem would be to implement a
graceful shutdown in the system ...
... when the nRF51 wants to shutdown the system (ie. after a button
press), it sends a message to the stm32 requesting shutdown
... in the stm32, modules that wants to be notified of shutdown are
called in sequence and return when they are ready for shutdown
... when all the module have returned, the stm32 sends to the nRF51 a
"ready for shutdown" message
... the nRF51 cuts the power. If no "ready to shutdown" was received
after a timeout, the power is cut anyway
Closes #623
|
@whoenig @ataffanel I would like to add a commit to provide graceful shutdown for the usddeck, could you help me with that? Or point me towards a source than could? thanks! clarification: I do not think this PR should be merged without any subsystem using the functionality |
jonasdn
force-pushed
the
jonasdn/623
branch
4 times, most recently
from
78bb54a to
753aba1
Compare
|
I agree that this should only be merged after we have at least one callback registered and tested. I would sign up for the new callback after it is confirmed that the usd card can be mounted and the config be read, namely at . In the writer-task, you could check for a new flag, e.g.,not_shutdown that is initially false at
. In your callback you would switch the flag to true and wait until the task is finished. This sort of wait is the tricky part of the whole design.
|
|
@ataffanel @whoenig would something like the latest commit work? |
jonasdn
force-pushed
the
jonasdn/623
branch
2 times, most recently
from
afcf401 to
181afcc
Compare
src/deck/drivers/src/usddeck.c
Outdated
| @@ -92,6 +93,10 @@ | |||
| #define FIXED_FREQUENCY_EVENT_ID (0xFFFF) | |||
| #define FIXED_FREQUENCY_EVENT_NAME "fixedFrequency" | |||
|
|
|||
|
|
|||
| /* set to true when graceful shutdown is triggered */ | |||
| static volatile bool in_shutdown; | |||
There was a problem hiding this comment.
this is implicit false ... should I make it explicit?
static volatile bool in_shutdown = false;
jonasdn
force-pushed
the
jonasdn/623
branch
3 times, most recently
from
0e29f47 to
d0cb239
Compare
src/deck/drivers/src/usddeck.c
Outdated
| uint32_t timeout = 15; /* ms */ | ||
| in_shutdown = true; | ||
| vTaskResume(xHandleWriteTask); | ||
| ulTaskNotifyTake(pdTRUE, timeout / portTICK_PERIOD_MS); |
There was a problem hiding this comment.
Should this be ...
ulTaskNotifyTake(pdTRUE, pdMS_TO_TICKS(timeout));...?
There was a problem hiding this comment.
Yes, that would be cleaner.
I have now, and fixed some bugs together with @ataffanel |
Currently to shut-down the Crazyflie, the nRF51 simply cut the power to the system. This was working fine until we started implementing SD-card functionality: when shutting down the system that way we have huge chances of corrupting the SD-card filesystem. This has lead to implementation of very inefficient sd-car logging procedure to decrease the probability of corruption.
This PR suggests a better solution for the shutdown problem would be to implement a graceful shutdown in the system ...
... when the nRF51 wants to shutdown the system (ie. after a button press), it sends a message to the stm32 requesting shutdown
... in the stm32, modules that wants to be notified of shutdown are called in sequence and return when they are ready for shutdown
... when all the module have returned, the stm32 sends to the nRF51 a "ready for shutdown" message
... the nRF51 cuts the power. If no "ready to shutdown" was received after a timeout, the power is cut anyway
Closes #623