Bug#389126: gnome-session: looking through strace

[Steinar H. Gunderson]

On Tue, Sep 26, 2006 at 12:59:13AM -0400, Stephen Touset wrote:
> I've been going through the strace, and I see at line 102641, (pid?) 4962
> calls an ioctl to read fd 15, from which it never returns. Not sure where it's
> getting fd 15 from, though, or where (pid?) 4962 comes from.

fd 15 is a connection to dbus:

   63604 4970  socket(PF_FILE, SOCK_STREAM, 0 <unfinished ...>
   63609 4970  <... socket resumed> )            = 15
   63614 4970  connect(15, {sa_family=AF_FILE, path="/var/run/dbus/system_bus_socket"}, 33 <unfinished ...>

4962 is a thread, which I presume shares file descriptors with 4970:

   46290 4961  clone( <unfinished ...>
   46295 4961  <... clone resumed> child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0xa735c708) = 4962

Your analysis is wrong, though, the ioctl definitely finishes:

   102641 4962  ioctl(15, FIONREAD <unfinished ...> 
   102645 4962  <... ioctl resumed> , [112])      = 0

The problem, AFAICS, is an infinite loop of some sort, doing the same over
and over again:

  169761 4962  read(15,  <unfinished ...>
  169762 4960  <... gettimeofday resumed> {1159192120, 184513}, NULL) = 0
  169763 4962  <... read resumed> "\4\0\0\0\2\0\0\0\0\0\0\0\20\0\0\0gnome-session\0\0\0", 1024) = 32
  169764 4960  gettimeofday( <unfinished ...>

I don't really know why dbus has this burning desire to write the same bytes
over and over again, though.

/* Steinar */
-- 
Homepage: http://www.sesse.net/