Netgraph/mpd5 stability issues

[Przemyslaw Frasunek]

> On 14.01.2011 18:46, Mike Tancsa wrote:
> I also have very loaded mpd/PPPoE servers that panic all the time:
> 
> http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/153255
> http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/153671

I've just got yet another panic on mpd5 box after about 30 days of uptime. This
time it seems to be unrelated with Netgraph:

(kgdb) bt
#0  doadump () at pcpu.h:196
#1  0xc0836ac7 in boot (howto=260) at ../../../kern/kern_shutdown.c:418
#2  0xc0836d99 in panic (fmt=Variable "fmt" is not available.
) at ../../../kern/kern_shutdown.c:574
#3  0xc0b5ef1c in trap_fatal (frame=0xc524fa2c, eva=56)
    at ../../../i386/i386/trap.c:950
#4  0xc0b5f1a0 in trap_pfault (frame=0xc524fa2c, usermode=0, eva=56)
    at ../../../i386/i386/trap.c:863
#5  0xc0b5fb95 in trap (frame=0xc524fa2c) at ../../../i386/i386/trap.c:541
#6  0xc0b42e7b in calltrap () at ../../../i386/i386/exception.s:166
#7  0xc0a683d0 in softdep_disk_io_initiation (bp=0xd96921e0)
    at ../../../ufs/ffs/ffs_softdep.c:3785
#8  0xc0a6e9ef in ffs_geom_strategy (bo=0xc58d10c0, bp=0xd96921e0) at buf.h:436
#9  0xc08a8830 in bufwrite (bp=0xd96921e0) at buf.h:429
#10 0xc0a6e3b8 in ffs_bufwrite (bp=0xd96921e0)
    at ../../../ufs/ffs/ffs_vfsops.c:1893
#11 0xc08a3af8 in vfs_bio_awrite (bp=0xd96921e0) at buf.h:417
#12 0xc08ad598 in vop_stdfsync (ap=0xc524fcd4)
    at ../../../kern/vfs_default.c:466
#13 0xc07bd3dc in devfs_fsync (ap=0xc524fcd4)
    at ../../../fs/devfs/devfs_vnops.c:499
#14 0xc0b743a2 in VOP_FSYNC_APV (vop=0xc0cd0e00, a=0xc524fcd4)
    at vnode_if.c:1007
#15 0xc08bdeb8 in sched_sync () at vnode_if.h:538
#16 0xc080e9f9 in fork_exit (callout=0xc08bd7b0 <sched_sync>, arg=0x0,
    frame=0xc524fd38) at ../../../kern/kern_fork.c:811
#17 0xc0b42ef0 in fork_trampoline () at ../../../i386/i386/exception.s:271
(kgdb) frame 7
#7  0xc0a683d0 in softdep_disk_io_initiation (bp=0xd96921e0)
    at ../../../ufs/ffs/ffs_softdep.c:3785
3785                    LIST_INSERT_AFTER(wk, &marker, wk_list);
(kgdb) print *wk
$4 = {wk_mp = 0x8, wk_list = {le_next = 0x30, le_prev = 0x3}, wk_type = 1,
  wk_state = 0}
(kgdb) print *bp
$5 = {b_bufobj = 0xc58d10c0, b_bcount = 16384, b_caller1 = 0x0,
  b_data = 0xde759000 "ýA\002", b_error = 0, b_iocmd = 2 '\002',
  b_ioflags = 2 '\002', b_iooffset = 16186277888, b_resid = 0, b_iodone = 0,
  b_blkno = 31613824, b_offset = 16186277888, b_bobufs = {
    tqe_next = 0xd96e4c5c, tqe_prev = 0xd9559d94}, b_left = 0x0,
  b_right = 0xd96e4c5c, b_vflags = 1, b_freelist = {tqe_next = 0xd955a2ec,
    tqe_prev = 0xd960ecf0}, b_qindex = 2, b_flags = 2684485668,
  b_xflags = 2 '\002', b_lock = {lk_object = {lo_name = 0xc0c114c5 "bufwait",
      lo_type = 0xc0c114c5 "bufwait", lo_flags = 70844416, lo_witness_data = {
        lod_list = {stqe_next = 0x0}, lod_witness = 0x0}},
    lk_interlock = 0xc0d3c030, lk_flags = 262144, lk_sharecount = 0,
    lk_waitcount = 0, lk_exclusivecount = 1, lk_prio = 80, lk_timo = 0,
    lk_lockholder = 0xfffffffe, lk_newlock = 0x0}, b_bufsize = 16384,
  b_runningbufspace = 16384, b_kvabase = 0xde759000 "ýA\002",
  b_kvasize = 16384, b_lblkno = 31613824, b_vp = 0xc58d1000, b_dirtyoff = 0,
  b_dirtyend = 0, b_rcred = 0x0, b_wcred = 0x0, b_saveaddr = 0xde759000,
  b_pager = {pg_reqpage = 0}, b_cluster = {cluster_head = {
      tqh_first = 0xd955d234, tqh_last = 0xd9509d5c}, cluster_entry = {
      tqe_next = 0xd955d234, tqe_prev = 0xd9509d5c}}, b_pages = {0xc3843bd8,
    0xc3842ab0, 0xc3843dd0, 0xc3843e18, 0x0 <repeats 28 times>}, b_npages = 4,
  b_dep = {lh_first = 0xc672ca80}, b_fsprivate1 = 0x0, b_fsprivate2 = 0x0,
  b_fsprivate3 = 0x0, b_pin_count = 0}
(kgdb) print *bp->b_dep->lh_first
$17 = {wk_mp = 0x8, wk_list = {le_next = 0x30, le_prev = 0x3}, wk_type = 1,
  wk_state = 0}
(kgdb) frame 12
#12 0xc08ad598 in vop_stdfsync (ap=0xc524fcd4)
    at ../../../kern/vfs_default.c:466
466                             vfs_bio_awrite(bp);
(kgdb) list
461                         ("bp %p wrong b_bufobj %p should be %p",
462                         bp, bp->b_bufobj, &vp->v_bufobj));
463                     if ((bp->b_flags & B_DELWRI) == 0)
464                             panic("fsync: not dirty");
465                     if ((vp->v_object != NULL) && (bp->b_flags & B_CLUSTEROK)) {
466                             vfs_bio_awrite(bp);
467                     } else {
468                             bremfree(bp);
469                             bawrite(bp);
470                     }
(kgdb) print *bp->b_dep->lh_first
$24 = {wk_mp = 0x8, wk_list = {le_next = 0x30, le_prev = 0x3}, wk_type = 1,
  wk_state = 0}