• Bug#264929: slapd segfaults

    From Roman Hodek@1:229/2 to All on Wed Aug 11 01:30:12 2004
    From: [email protected]

    Package: slapd
    Version: 2.1.30-2
    Severity: important

    A few days ago I upgraded my server to latest testing. Since that, the
    LDAP server crashes with a segfault after a few hours (1h..12h).

    Unfortunately, I can't provide you a really good strace logs, as the
    threads after some time always receive a SIGRTMIN and an attached
    strace exits on that. But maybe I still got something from the
    faulting thread. This one is PID 25820 and seems to be the controller:

    getppid() = 25819
    poll([{fd=10, events=POLLIN}], 1, 2000) = 0
    getppid() = 25819
    poll([{fd=10, events=POLLIN}], 1, 2000) = 0
    getppid() = 25819
    poll([{fd=10, events=POLLIN}], 1, 2000) = 0
    getppid() = 25819
    poll([{fd=10, events=POLLIN}], 1, 2000) = -1 EINTR (Interrupted system call) --- SIGRT_1 (Unknown signal 33) @ 0 (0) ---
    sigreturn() = ? (mask now ~[TRAP KILL STOP]) getppid() = 25819
    waitpid(-1, [{WIFSIGNALED(s) && WTERMSIG(s) == SIGSEGV}], WNOHANG|__WCLONE) = 25821
    kill(25861, SIGSEGV) = 0
    --- SIGRT_1 (Unknown signal 33) @ 0 (0) ---
    sigreturn() = ? (mask now ~[TRAP KILL STOP]) kill(25823, SIGSEGV) = 0
    --- SIGRT_1 (Unknown signal 33) @ 0 (0) ---
    sigreturn() = ? (mask now ~[TRAP KILL STOP]) kill(25819, SIGSEGV) = 0
    exit_group(0) = ?

    This looks like as PID 25821 first signalled the SIGSEGV. And
    fortunately, the strace for 25821 survived :) It says:

    select(39, [4 6 7 8 9 12 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35], [], NULL, NULL) = 1 (in [35])
    time(NULL) = 1091987719
    read(35, "\27\3\1\0\340", 5) = 5
    read(35, "\272\216\23v
  • From Roland Bauerschmidt@1:229/2 to Roman Hodek on Wed Aug 11 22:40:09 2004
    From: [email protected]

    Roman Hodek wrote:
    A few days ago I upgraded my server to latest testing. Since that, the
    LDAP server crashes with a segfault after a few hours (1h..12h).

    Do you use TLS? If so, please update to 2.1.30-3, which fixes related segfaults. I assume that the segfaults occur on the machine you filed
    this bug from? Would it be possible for you to run slapd in gdb, if I
    sent you a package compiled with debugging symbols (i386)?

    Roland


    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)
  • From Hodek Roman@1:229/2 to All on Thu Aug 12 11:10:11 2004
    From: [email protected]

    Do you use TLS?

    Yep.

    If so, please update to 2.1.30-3, which fixes related segfaults.

    I can try, let's see how much other stuff from unstable is needed. A
    few libs don't hurt, but it I hope it won't force my postfix or apache
    to update, too. (There, I use the ipv6 versions.)

    I assume that the segfaults occur on the machine you filed this bug
    from?

    I prepared the mail on the server (grufti) with bug -p, but sent it
    from another machine. Shouldn't matter :)

    Would it be possible for you to run slapd in gdb, if I sent you a
    package compiled with debugging symbols (i386)?

    Basically yes, but I'd have to do that on a weekend where I'm at home
    and can watch it. If slapd doesn't run lots of things start to break,
    so I want to restart it as quickly as possible.

    Roman


    --
    To UNSUBSCRIBE, email to [email protected]
    with a subject of "unsubscribe". Trouble? Contact [email protected]

    --- SoupGate-Win32 v1.05
    * Origin: you cannot sedate... all the things you hate (1:229/2)