Important bugfix for Linux MQ #394

muellerr · 2021-04-08T15:29:35+02:00

muellerr commented

2021-04-08 15:29:35 +02:00

This bug was not an issue on hosted Linux or Rapsberry Pi (at least on the surface), but causes crashes on a Zynq SoC.

A buffer write was exceeded by 4 bytes, causing buffer overflow errors on that system.

The former implementation used the hardcoded value MAX_DATA_SIZE.
I replaced this with getMaximumMessageSize() but this is actually a problem because this will return the size with the header as well.

I suspect even a unit test would not have caught this because this was not an obvious issue on host systems. Maybe valgrind or a static code analyzer would have caught it..

This bug was not an issue on hosted Linux or Rapsberry Pi (at least on the surface), but causes crashes on a Zynq SoC. A buffer write was exceeded by 4 bytes, causing buffer overflow errors on that system. The former implementation used the hardcoded value `MAX_DATA_SIZE`. I replaced this with `getMaximumMessageSize()` but this is actually a problem because this will return the size with the header as well. I suspect even a unit test would not have caught this because this was not an obvious issue on host systems. Maybe `valgrind` or a static code analyzer would have caught it..

muellerr added the

bug

label 2021-04-08 15:29:35 +02:00

muellerr added 1 commit 2021-04-08 15:29:35 +02:00

6983ddc3e0 super evil bugfix

muellerr changed title from ~~super evil bugfix for Linux MQ~~ to Important bugfix for Linux MQ

2021-04-08 15:32:03 +02:00

muellerr added 1 commit 2021-04-08 15:39:33 +02:00

23e3f2f34f now fixed properly

gaisser commented

2021-04-08 15:58:04 +02:00

Bug is also in OSAL for RTEMS!