Closed andreasbaumann closed 4 years ago
These changes look good to me.
To be completely honest, I don't think we've ever tested NRPE on non-x86 processors (nor do I know of any that we keep in the office).
np. I might have triggered a rare usage case, running the moniroting station on a Raspberry Pi II, just to switch to an Alix IA-32 applicance, to hit the bug again. :-) But the array bounds write is also there when using valgrind on x86_64..
Code should be checked again against packet buffer operations, is it with terminating '\0' or not, what's it's exact size?
The tot_bytes case is not severe (fixed on-passant), as the result (being 10 in case of SSL) is never used afterwards.
Also calling SSL_read with 0 bytes to read is maybe not the best idea, better check for bytes_to_read >0 before calling the SSL_read function.
Experienced problems on IA-32 (Archlinux32) and Archlinux (armv7, RaspPi 2).
In armv7 I also got a linked list corruption in SSL_free when read_packet was called twice (in case of a protocol fallback).