On Fri, 2009-04-24 at 05:37 -0700, Thomas Maier-Komor wrote: > Hi, > > I was just playing around with fuse a little bit on an opensolaris box > running in a virtual box. Then I had a situation where I had an assertion > fault in my libfuse linked executable in the middle of running a rename. This > caused the executable to terminate. After that the mv command that triggered > the fuse rename execution hung, so I hit CTRL-C. A second later the machine > crashed. mdb revealed the following stack trace: > > e320e8a0::findstack > stack pointer for thread e320e8a0: e078fac0 > e078faec 0x4521() > e078fb0c cpu_grow+0x1e() > e078fb5c cpu_update_pct+0x123() > e078fbbc die+0x93() > e078fc6c trap+0x1422() > e078fc80 0xfe80037c() > e078fd28 mutex_owner_running+0x11() > e078fd68 cv_wait_sig+0x151() > e078fd84 fuse_queue_request_wait+0x5c() > e078fdc0 fuse_rename_i+0x157() > e078fdf4 fuse_rename+0x217() > e078fe44 fop_rename+0x86() > e078ff34 vn_renameat+0x2e7() > e078ff64 vn_rename+0x18() > e078ff84 rename+0x13() > e078ffac sys_call+0x10c() > > So this looks to me as if mutex_owner_running is somehow unable to deal with > this situation and caused a null pointer exception. The panic message was: > > ::status > debugging crash dump vmcore.0 (32-bit) from opensolaris > operating system: 5.11 snv_111 (i86pc) > panic message: > BAD TRAP: type=e (#pf Page fault) rp=e078fc80 addr=10 occurred in module > "unix" > due to a NULL pointer dereference > dump content: kernel pages only > > Is this a bug or should I be prepared that the machine might panic if I play > around with libfuse?
This is a bug fuse on Solaris. You should be prepared for crashes when running fuse on Solaris (we're working to get things stable ASAP). -M _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
