Current still seriously broken

[Martin Husemann <martin%duskware.de@localhost>]

Hi folks,

I know I will get a "bah, rump broken, news at 11" reaction - but just
wanted to point out that despite the ld.elf_so relocation fixes current
is still seriously broken on evbarm.

Doing a full atf test run on aarch64 leaves several rump server processes
around:

  776 ?     Rsl  105:32.28 rump_server -lrumpvfs -lrumpnet -lrumpnet_net -lrumpnet_netinet unix://commsock 
 4808 ?     Rsl   49:20.88 rump_server -lrumpvfs -lrumpnet -lrumpnet_net -lrumpnet_netinet unix://commsock 
 5719 ?     Osl  144:37.58 rump_server -lrumpvfs -lrumpdev -lrumpnet -lrumpnet_net -lrumpnet_netinet -lrumpnet_netinet6 -lrumpnet_local -lrumpnet_shmif -lrumpdev_disk -lrumpfs_ffs -lrumpfs_nfs -lrumpfs_nfsserver -d key unix://test 
17108 ?     Osl  242:50.00 rump_server unix://csock 
25496 ?     Osl   76:46.62 rump_server unix://commsock 

... where nearly all threads seem to be in __lwp_park, on in condwait and
the clock thread in a timedwait. Then a single thread is different:

#0  lwproc_proc_free (p=0xf1a36d469000)
    at /work/src/lib/librump/../../sys/rump/librump/rumpkern/lwproc.c:173
#1  lwproc_freelwp (l=<optimized out>)
    at /work/src/lib/librump/../../sys/rump/librump/rumpkern/lwproc.c:341
#2  rump_lwproc_switch (newlwp=<optimized out>)
    at /work/src/lib/librump/../../sys/rump/librump/rumpkern/lwproc.c:521
#3  0x0000f1a3770054b0 in lwproc_makelwp (p=p@entry=0xf1a364b6cc00, 
    doswitch=doswitch@entry=true, procmake=procmake@entry=false)
    at /work/src/lib/librump/../../sys/rump/librump/rumpkern/lwproc.c:385
#4  0x0000f1a3770055c0 in rump_lwproc_newlwp (pid=<optimized out>)
    at /work/src/lib/librump/../../sys/rump/librump/rumpkern/lwproc.c:439
#5  0x0000f1a376e66f20 in lwproc_newlwp (pid=311)
    at /work/src/lib/librumpuser/rumpuser_sp.c:212
#6  serv_handlesyscall (rhdr=0xf1a31eda8e48, rhdr=0xf1a31eda8e48, data=0x0, 
    spc=0xf1a376e7ce70) at /work/src/lib/librumpuser/rumpuser_sp.c:684
#7  serv_workbouncer (arg=<optimized out>)
    at /work/src/lib/librumpuser/rumpuser_sp.c:767
#8  0x0000f1a376e3d2f4 in pthread__create_tramp (cookie=0xf1a3623cbc00)
    at /work/src/lib/libpthread/pthread.c:560
#9  0x0000f1a376aa9f04 in __mknod50 () from /usr/lib/libc.so.12

and I see this backtrace (exact this one, starting at __mknod50) in all
of the hanging processes.


What is worse is that 32bit machines do not seem to be able to complete a
full atf test run (but maybe I have been unlucky).

Martin