While the current approach for garbage collection in the 6Lo reassembly
buffer is good for best-effort handling of
*fragmented* packets and nicely RAM saving, it has the problem that
incomplete, huge datagrams can basically DoS a node, if no further
fragmented datagram is received for a while (since the packet buffer is
full and GC is not triggered).
This change adds a asynchronous GC (utilizing the existing
functionality) to the reassembly buffer, so that even if there is no new
fragmented packet received, fragments older than `RBUF_TIMEOUT` will be
removed from the reassembly buffer, freeing up the otherwise wasted
packet buffer space.
Since IPHC also manipulates the total number of bytes of a received
datagram (by decompressing it), this also needs to be exposed. I guess
I was too focused on introducing a *generic* packet buffer for a future
virtual reassembly buffer (where it isn't needed, but so isn't `pkt` to
be honest), that I totally forgot about it in #9352.
This fixes an alignment issue I encountered in the static version of
the packet buffer.
The bug is caused by a race-condition where a certain order of
operations leads to a chunk being released according to the
byte-alignment of the platform, but overlapping potential space for
a future `_unused_t` struct e.g. (x mark allocated regions):
Future leak of size sizeof(_unused_t) Time
v |
+------------+-----+--------------------+ |
|xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx| +
+------------+-----+--------------------+ |
|
+------------+--+--+--------------------+ |
| |xxxxxxxxxxxxxxxxxxxxxxx| +
+------------+--+--+--------------------+ |
|
+-----+------+--+--+--------------------+ |
|xxxxx| |xxxxxxxxxxxxxxxxxxxxxxx| +
+-----+------+--+--+--------------------+ |
|
+-----+------+-----+---------+----------+ |
|xxxxx| |xxxxxxxxxx| +
+-----+------+-----+---------+----------+ |
|
+-----+------+-----+--------------------+ |
|xxxxx| |xxxxxxxxxxxxxxxxxxxxxxxxxx| +
+-----+------+-----+--------------------+ |
|
+------------+-----+--------------------+ |
|xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx| +
+------------+-----+--------------------+ |
|
+------------+-----+--------------------+ |
|xxxxxxxxxxxxxxxxxx| | +
+------------+-----+--------------------+ |
|
+------------+-----+--------------------+ |
| |xxxxx| | +
+------------+-----+--------------------+ |
v
Sadly, I wasn't able to create a reproducable unittest that show-cases
this corner-case, since I don't understand the order of operations that
cause this one 100%, but the bug is reproducable (but also not
reliably) by sending large (i.e. fragmented) packets to a 6Lo-enabled
host from more than 1 host simultaneously (use `gnrc_pktbuf_cmd` to
check).
By making the size of `_unused_t` the only condition for alignment,
this bug is fixed.
- Use RIOT's GPIO interface to access the sensor to increase portability
- Changed API to allow more than one sensor per board
- Added `sht1x_params.h` that specifies how the sensors is connected - each
board can overwrite default settings by #defining SHT1X_PARAM_CLK and
SHT1X_PARAM_DATA
- Changed arithmetic to use integer calculations only instead of floating point
arithmetic
- Added support for checking the CRC sum
- Allow optional skipping of the CRC check to speed up measuring
- Added support for advanced features like reducing the resolution and skipping
calibration to speed up measuring
- Allow specifying the supply voltage of sensor which heavily influences the
temperature result (and use that information to calculate the correct
temperature)
- Reset sensor on initialization to bring it in a well known state
- Support for the obscure heater feature. (Can be useful to check the
temperature sensor?)
- Updated old SHT11 shell commands to the new driver interface, thus allowing
more than one SHT10/11/15 sensor to be used
- Added new shell command to allow full configuration of all attached SHT1x
sensors
- Removed old command for setting the SHT11 temperature offset, as this feature
is implemented in the new configuration command
The sensor family SHT10, SHT11 and SHT15 only differ in their accuracy (as in
calibration, not as in resolution). Thus, the same driver can be used for all.
The new driver name better reflects this fact.
This refactors the `gnrc_sixlowpan_frag` module for the API proposed
in #8511.
The `ctx` for `gnrc_sixlowpan_frag_send()` is required to be a
`gnrc_sixlowpan_msg_frag_t` object, so IPHC can later on use it to
provide the *original* datagram size (otherwise, we would need to adapt
the API just for that, which seems to me as convoluted as this
proposal).
I also provide an expose function with a future possibility to provide
more than just one `gnrc_sixlowpan_msg_frag_t` object later on (plus
having cleaner module separation in general).
I'm using something like this command for a while now for debugging
GNRC. Usually, I just patch it into the application I'm using it with,
but I think there is a benefit to also provide it to RIOT upstream
properly.
