Hi, Joachim Worringen wrote:
>> Way back, msgbuf was a buffer of fixed size like 8K. These days >> it it the console streams logging device - see the source in >> uts/common/os/printf.c and uts/common/os/logsubr.c. > > O.k., thanks, but is there a way to increase the backlog of messages in > a kernel core dump? I actually do not care too much on how it's > implemented... That was an exercise for the reader :-) (Another way of saying I have never had to raise the limit - in fact I've seen dumps with acres of messages). It looks like the log queuesare created with a hiwater mark of LOG_HIWAT - 1MB. During normal operation if syslog is not running then additional log messages attempted while we're at or above the hiwater mark will be discarded. In panic (see log_sendmsg) we raise the console queue hiwater to 2 * LOG_HIWAT, or 2MB, to capture panic messages. In-flight log device messages are recognised in the panic flow so that on subsequent reboot we can replay them to syslog on startup (out of the dump device). Gavin _______________________________________________ opensolaris-code mailing list opensolaris-code@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
