You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mynewt.apache.org by GitBox <gi...@apache.org> on 2020/05/27 15:52:39 UTC
[GitHub] [mynewt-nimble] haukepetersen opened a new issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
haukepetersen opened a new issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824
## Context
I am still trying to understand the behavior described in #818. To make sure this is not a RIOT or a porting issue, I ported the benchmark application to mynewt (-> https://github.com/haukepetersen/riotsandbox/tree/master/tests/mynewt_nimble_l2cap_coc_flooding)
This is the first time I actually did something with mynewt and I am still learning - so please excuse the project structure and code maturity :-)
## Problem
When flooding an L2CAP COC link between two nodes for a long enough amount of time (~some seconds), the transmitting node will crash (reboot).
By e.g. trying to send 10k chunks of 100byte (`flood 100 10000`), the `source` node will reboot after a random time, mostly something around ~15-20 seconds.
## Setup
Two `nrf52dk` nodes on my desk, one flashed with the `source` application (target `source`), one with the `sink` (target `sink`) application.
The nodes open a BLE connection automatically, and once connected also open a custom L2CAP connection oriented channel.
Then, once connected, simply call the `flood` shell command on the `source` node.
Is anyone able to reproduce this behavior? And does anyone have an idea where to look further? I am a little stuck here...
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] haukepetersen commented on issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
haukepetersen commented on issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824#issuecomment-673363503
No worries, same sorry here :)
> If I understand corretly you want to wait here
Yes, that is correct. The aim is to allow other, possibly lower prio tasks to work while the sending task is waiting for L2CAP to sent the previous packet. Using a semaphore/mutes here seemed the obvious choice, right? Or do you have an idea for something better suited for synchronizing the sender task?
Will also try to run my code using a semaphore and report back with my results.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] rymanluk edited a comment on issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
rymanluk edited a comment on issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824#issuecomment-654128701
@haukepetersen Sorry for late reply, I had no time to look into it till now.
I could reproduce it and this is watchdog restarting source. Solution for your issue is to use `os_sem` instead of `os_mutex`. If I understand corretly you want to wait here: https://github.com/haukepetersen/riotsandbox/blob/master/tests/mynewt_nimble_l2cap_coc_flooding/apps/source/src/main.c#L64 untli UNSTALLED will be received.
It works for me when I changed mutex -> semaphore.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] haukepetersen commented on issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
haukepetersen commented on issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824#issuecomment-679017137
Confirmed: using os_sem solves this issue. Interesting system design to have the watchdog triggering for waiting mutexes but not for semaphores :-) Maybe that could be added to the doc pages?!
Anyhow, everything is in order now, so I consider this issue solved.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] rymanluk commented on issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
rymanluk commented on issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824#issuecomment-654128701
Sorry for late reply, I had no time to look into it till now.
I could reproduce it and this is watchdog restarting source. Solution for your issue is to use `os_sem` instead of `os_mutex`. If I udenrstand corretly you want to wait here: https://github.com/haukepetersen/riotsandbox/blob/master/tests/mynewt_nimble_l2cap_coc_flooding/apps/source/src/main.c#L64 untli UNSTALLED will be received.
It works for me when I changed that.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] rymanluk edited a comment on issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
rymanluk edited a comment on issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824#issuecomment-654128701
Sorry for late reply, I had no time to look into it till now.
I could reproduce it and this is watchdog restarting source. Solution for your issue is to use `os_sem` instead of `os_mutex`. If I udenrstand corretly you want to wait here: https://github.com/haukepetersen/riotsandbox/blob/master/tests/mynewt_nimble_l2cap_coc_flooding/apps/source/src/main.c#L64 untli UNSTALLED will be received.
It works for me when I changed mutex -> semaphore.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] andrzej-kaczmarek commented on issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
andrzej-kaczmarek commented on issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824#issuecomment-679076119
> Confirmed: using os_sem solves this issue. Interesting system design to have the watchdog triggering for waiting mutexes but not for semaphores :-) Maybe that could be added to the doc pages?!
that's not the problem with mutex itself. sample code acquires mutex in on task (flooder) and releases it in another (main). this means an attempt to release mutex always returns an error since task which tries to release mutex does not own it, while an attempt to acquire mutex will not block since mutexes are recursive so the same task can acquire a mutex more than once. so effectively sample code spins in `_send` which triggers watchdog.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] haukepetersen closed issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
haukepetersen closed issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [mynewt-nimble] rymanluk edited a comment on issue #824: L2CAP COC: nimble@mynewt crashes when flooding a channel
Posted by GitBox <gi...@apache.org>.
rymanluk edited a comment on issue #824:
URL: https://github.com/apache/mynewt-nimble/issues/824#issuecomment-654128701
@haukepetersen Sorry for late reply, I had no time to look into it till now.
I could reproduce it and this is watchdog restarting source. Solution for your issue is to use `os_sem` instead of `os_mutex`. If I udenrstand corretly you want to wait here: https://github.com/haukepetersen/riotsandbox/blob/master/tests/mynewt_nimble_l2cap_coc_flooding/apps/source/src/main.c#L64 untli UNSTALLED will be received.
It works for me when I changed mutex -> semaphore.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org