Hi, On 07/07/2020 17:07, Sylvain Beucler wrote: > On 06/07/2020 19:11, Sylvain Beucler wrote: >> Do we have definite info on what versions are affected? >> >> I cannot reproduce the issue in jessie/stretch/buster (5.7.x). >> >> Incidentally Salvatore's test now yields an error in bullseye >> (5.8dfsg-3), though I suspect the issue is at the client's level: >> # snmpbulkget -v3 -Cn1 -Cr1472 -l authPriv -u testuser -a SHA -A >> testpass -x AES -X testpass 127.0.0.1 1.3.6.1.2.1.1.5 1.3.6.1.2.1.1.7 >> Error in packet. >> Reason: (genError) A general failure occured > > Bisecting gives a range of ~20 commits where the server is buggy (either > goes 100% CPU, or rejects the request with "send response: Too long"). > > 1a0dbe19bf2787bb5bea913f210a9a5eb4c0c80c > "new snmp token sendMessageMaxSize" > works fine. > > 3eb4b473fed816108d1843dadee1ce877415b96b > "add debug_enable_token_logs debug_disable_token_logs to output_api.h" > triggers the double-free. > > Anything in-between is random, and includes 2 "getbulk enhancements". > The date varies greatly so this may be a series of cherry-picks. > > In any case, all of this happens between 5.7.3 and 5.8.pre1.
Restricting further (good..bad): $ git shortlog 1a0dbe19bf2787bb5bea913f210a9a5eb4c0c80c..e207b8113260fd7d84df0ebdb66925ab70da29b2 Robert Story (2): Add VMware copyright tweak sndMsgMaxSize handling VMwareDev Randy (4): getbulk enhancements: limit responses gathered reduce session msg max sizes to transport max getbulk enhancements: response size + fallback to forward encoding move v3 engineID probe into initial packet build Cheers! Sylvain Beucler Debian LTS Team