Hi, El 14/1/20 a les 8:50, Ondřej Tůma ha escrit: > V Sun, 22 Dec 2019 09:45:05 +0100 (CET) > Joan Lledó <jlle...@mailfence.com> napsáno: > >> Hi, >> >> El 10/12/19 a les 16:29, Ondřej Tůma ha escrit: >>> Creating TCP server with existing value in library crashed by bad >>> memory using. Valgrind found the problem in low_level_input >>> function, which uses pbuf_alloc, that allocate too less memory. >>> This "original" value will work. >> >> low_level_info is usually implemented by the driver. Which driver are >> you using? Are you using IPv6? > > Hi, > > I'm using tap, you can see here: > > https://github.com/prusa3d/Prusa-Firmware-Buddy/blob/master/tests/integration/main.c > > lwsapi is LUA WSAPI or Python WSGI like implementation somewhere code, > which is linked to this directory after git clone. > > And as you can see, I'm not using IPv6 at this moment. >
I'm reluctant to change the TCP tuning values without a good reason, because any value will always fit for some cases and be inappropriate for others. Besides, I don't see how it's the MSS value related to this bug. And anyway, memory bugs in C are always tricky, the problem can be produced in any other place than the place where they manifest. Reading your code I assume you're using the tapif driver that is distributed with lwip, could the problem be the harcoded size of the input buffer? [1] ------------ [1] http://git.savannah.nongnu.org/cgit/lwip.git/tree/contrib/ports/unix/port/netif/tapif.c#n267
