Re: Bug in Xen 4.1.0: Xen leaks tapdisk2 processes

[Nathan March <nathan@gt.net>]

> in Xen 4.1.0 using "tap2:aio:/dev/sdaX" Xen leaks tapdisk2 processes.
> After running some VMs and terminating them, only dom0 running, "ps
> aux" shows:

I can definitely confirm this, on a machine with 2 active VMs:

xen1 ~ # ps auxf | grep -c tapdisk2
115

Although obviously many have been started/stopped/migrated away.
Stracing a few random ones shows them all waiting on fd8:

xen1 ~ # strace -p 11580
Process 11580 attached - interrupt to quit
select(8, [3 4 7], [], [], {425, 231813}^C <unfinished ...>
Process 11580 detached
xen1 ~ # strace -p 19110
Process 19110 attached - interrupt to quit
select(8, [3 4 7], [], [], {421, 228456}^C <unfinished ...>
Process 19110 detached
xen1 ~ # strace -p 20738
Process 20738 attached - interrupt to quit
select(8, [3 4 7], [], [], {415, 152712}^C <unfinished ...>
Process 20738 detached
xen1 ~ #

lsof:

tapdisk2 20738 root    3u  unix 0xffff880018585340        0t0 10490566
/var/run/blktap-control/ctl20738
tapdisk2 20738 root    4u  0000                0,8          0     1000
anon_inode
tapdisk2 20738 root    5u  unix 0xffff8800185858c0        0t0 10490568
socket
tapdisk2 20738 root    7u   CHR              251,3        0t0 10490549
/dev/xen/blktap-2/blktap3
tapdisk2 20738 root    8u   REG               0,22 2147483648 32367811
/mnt/xenDisks1/nathanxen2/swap (10.1.5.1:/vol/xenDisks1)

Let me know if I can grab any useful information for debugging.

- Nathan