SEGV when running out of database connection handles
Long shot, but thought I would ping the list first. Unfortunately I dont have any debug information, the bt was mangled and logging (screen) was off. Anyone aware of a possible SEGV bug in rlm_sql under case of repeated auth/acct packets but no available connections to mysql? (the database server was stopped for an upgrade) Otherwise I guess I will just have to try and catch it again.
Joe Maimon <jmaimon@ttec.com> wrote:
Anyone aware of a possible SEGV bug in rlm_sql under case of repeated auth/acct packets but no available connections to mysql? (the database server was stopped for an upgrade)
I haven't seen that. The code *should* be OK. As time goes on, I'm starting to see more and more of a need for an automated source code analysis tool. None of the free ones I've seen do what I want, but there's at least one commercial one that looks close. An analysis tool could answer this question trivally. "if conditions X, Y, and Z are met, line #A deferences a NULL pointer". Alan DeKok.
Joe Maimon wrote:
Anyone aware of a possible SEGV bug in rlm_sql under case of repeated auth/acct packets but no available connections to mysql? (the database server was stopped for an upgrade)
I don't think there is any known issue with the module rlm_sql. Without more informations, it's not sure that the bug really is in rlm_sql, the server core, or the libmysqlclient.
Otherwise I guess I will just have to try and catch it again.
I'd suggest to run FreeRADIUS in valgrind. -- Nicolas Baradakis
participants (3)
-
Alan DeKok -
Joe Maimon -
Nicolas Baradakis