On Sat, Jan 27, 2007 at 04:05:40PM -0700, Vernon Schryver wrote: > > From: Gary Mills > > > When I checked the DCC log, it was a typical multipart/alternative > > MIME message with a text/plain and a text/html portion. Why would > > there be no Fuz1 and Fuz2 checksums in this case? Could they be > > using funny MIME headers? Here's how they look: > > This problem turned out to be a very old bug. The states of the MIME > boundary strings matchers were not being sufficiently cleared at the > starts of MIME parts. When the previous mail message handled by a dccm > or dccifd thread had broken MIME encapsulation, the MIME boundaries > of the next message would might not be matched.
I wonder if this bug is related to a problem I've been seeing recently? Normally, dccm only uses a small portion of the CPU, but when this problem strikes, the CPU usage gradually rises to become dominant. When I stop and restart dccm, the CPU usage goes back to normal, at least for a while. I'm assuming that a few copies of a malformed message have put the parser into a loop. This evening, I found dccm running with two constantly active threads, as shown by `pstack'. All of the other threads were in various wait states. The active threads were both in the dcc_ck_body() function. Here are the stack traces, taken a few minutes apart: ----------------- lwp# 3086 / thread# 3086 -------------------- 000294e4 dcc_ck_body (122f88, b31058, 39d2, 1a7c, b32aa1, 1231eb) + 184 0001981c dccm_body (affed8, b31058, 39d2, affed8, 197e8, 121db0) + 34 0002385c mi_engine (affed8, 9, ffffffff, b31058, 92f64, 92fb0) + 28c 00021d68 mi_handle_session (affed8, ff021634, 0, 0, 0, ffffffff) + 44 0002142c mi_thread_handle_wrapper (affed8, 0, 0, 0, 0, 0) + 4 ff0657b4 _lwp_start (0, 0, 0, 0, 0, 0) ----------------- lwp# 3197 / thread# 3197 -------------------- 000294dc dcc_ck_body (11a5f4, b34b68, dd2e, 1ce0, b36832, 11a857) + 17c 0001981c dccm_body (af4720, b34b68, dd2e, af4720, 197e8, 11941c) + 34 0002385c mi_engine (af4720, 9, ffffffff, b34b68, 92f64, 92fb0) + 28c 00021d68 mi_handle_session (af4720, ff171e34, 0, 0, 0, ffffffff) + 44 0002142c mi_thread_handle_wrapper (af4720, 0, 0, 0, 0, 0) + 4 ff0657b4 _lwp_start (0, 0, 0, 0, 0, 0) ----------------- lwp# 3086 / thread# 3086 -------------------- 00029818 dcc_ck_body (122f88, b31058, 39d2, 1a7c, b32aa1, 12324d) + 4b8 0001981c dccm_body (affed8, b31058, 39d2, affed8, 197e8, 121db0) + 34 0002385c mi_engine (affed8, 9, ffffffff, b31058, 92f64, 92fb0) + 28c 00021d68 mi_handle_session (affed8, ff021634, 0, 0, 0, ffffffff) + 44 0002142c mi_thread_handle_wrapper (affed8, 0, 0, 0, 0, 0) + 4 ff0657b4 _lwp_start (0, 0, 0, 0, 0, 0) ----------------- lwp# 3197 / thread# 3197 -------------------- 00029818 dcc_ck_body (11a5f4, b34b68, dd2e, 1ce0, b36832, 11a8b9) + 4b8 0001981c dccm_body (af4720, b34b68, dd2e, af4720, 197e8, 11941c) + 34 0002385c mi_engine (af4720, 9, ffffffff, b34b68, 92f64, 92fb0) + 28c 00021d68 mi_handle_session (af4720, ff171e34, 0, 0, 0, ffffffff) + 44 0002142c mi_thread_handle_wrapper (af4720, 0, 0, 0, 0, 0) + 4 ff0657b4 _lwp_start (0, 0, 0, 0, 0, 0) -- -Gary Mills- -Unix Support- -U of M Academic Computing and Networking- _______________________________________________ DCC mailing list [email protected] http://www.rhyolite.com/mailman/listinfo/dcc
