Bug #37
closedstream pool error messages
Description
[29054] 1/1/2010 -- 23:07:30 - (stream-tcp-reassemble.c:1608) <Error> (StreamTcpGetSegment) -- [ERRCODE: SC_ERR_POOL_EMPTY(41)] - segment_pool7->empty_list_size 512, alloc 512
[29054] 1/1/2010 -- 23:07:30 - (stream-tcp-reassemble.c:1086) <Error> (StreamTcpReassembleHandleSegmentHandleData) -- [ERRCODE: SC_ERR_POOL_EMPTY(41)] - segment_pool7 is empty
[29054] 1/1/2010 -- 23:07:30 - (stream-tcp-reassemble.c:1485) <Error> (StreamTcpReassembleHandleSegment) -- [ERRCODE: SC_ERR_REASSEMBLY_FAILED(42)] - StreamTcpReassembleHandleSegmentHandleData error
Updated by Victor Julien almost 15 years ago
Seeing a lot of those after a while on my firewall box running nfq inline.
Updated by Will Metcalf almost 15 years ago
I'm requesting that we make segment pools a user configurable option. Perhaps we also need some payload sizes between 1448 and 65535. Might also be nice to just have an option to set a generic memcap for reassembly.
/* We define serveral pools with prealloced segments with fixed size
* payloads. We do this to prevent having to do an malloc call for every
* data segment we receive, which would be a large performance penalty.
* The cost is in memory of course. */
#define segment_pool_num 8
static uint16_t segment_pool_pktsizes[segment_pool_num] = {4, 16, 112, 248, 512,
768, 1448, 0xffff};
static uint16_t segment_pool_poolsizes[segment_pool_num] = {2048, 3072, 3072,
3072, 3072, 8192,
8192, 512};
Updated by Victor Julien over 14 years ago
- Status changed from New to Closed
This was fixed a few masters ago with the stream memcap and pool changes.