[Qemu-devel] Migration issue with 4.0.x

[Andrey Korolyov]

Hello,

during tests against 4.0.5/4.0.6 for the problem described in
https://lists.gnu.org/archive/html/qemu-devel/2015-06/msg03117.html I
had noticed another weird issue: VM hangs after a couple of minutes
after being migrated if the hypervisor is running mentioned kernel
version. It does not matter for workload type during migration (idle
VM hangs too) or memory configuration of the VM - I can type
characters in the shell but dmesg listing takes about a minute to be
displayed (as well as other I/O operations lilke ls). The main visible
symptom is that I`ve not able to launch 'top' just after migration -
it 'hangs' after reading its own pid stats:

     0.000058 read(4, "2436 (bash) S 2434 2436 2436 34817 2514 4202752
4321 9165 0 3 0 13 0 6 20 0 1 0 55138 20561920 1045
18446744073709551615 4194304 5130324 140723572820800 140723572819456
139788860395630 0 65536 3686404 1266761467 0 0 0 17 0 0 0 0 0 0
7228872 7264912 17436672 140723572821400 140723572821406
140723572821406 140723572821998 0\n", 4095) = 324
     0.000096 close(4)                  = 0
     0.000061 open("/proc/2436/statm", O_RDONLY) = 4
     0.000061 read(4, "5020 1045 755 229 0 293 0\n", 4095) = 26
     0.000052 close(4)                  = 0
     0.000053 brk(0x15da000)            = 0x15da000
     0.000074 stat("/proc/2514", {st_mode=S_IFDIR|0555, st_size=0, ...}) = 0
     0.000083 open("/proc/2514/stat", O_RDONLY) = 4
     0.000065 read(4, "2514 (strace) S 2436 2514 2436 34817 2514
4202496 107 0 0 0 0 0 0 0 20 0 1 0 64452 4546560 124
18446744073709551615 4194304 4502860 140732305872800 140732305872152
139895671099674 0 0 3166215 0 0 0 0 17 4 0 0 0 0 0 6600016 6601324
9408512 140732305878369 140732305878401 140732305878401
140732305879016 0\n", 4095) = 305
     0.000094 close(4)                  = 0
     0.000062 open("/proc/2514/statm", O_RDONLY) = 4
     0.000067 read(4, "1110 124 106 76 0 97 0\n", 4095) = 23
     0.000051 close(4)                  = 0
     0.000056 stat("/proc/2515", {st_mode=S_IFDIR|0555, st_size=0, ...}) = 0
     0.000079 open("/proc/2515/stat", O_RDONLY) = 4
     0.000056 read(4, "2515 (top) R 2514 2514 2436 34817 2514 4202496
206 0 0 0 0 0 0 0 20 0 1 0 64452 23875584 698 18446744073709551615
4194304 4257604 140736479316592 140736479312856 140491962546544 0 0 0
2147221247 0 0 0 17 4 0 0 0 0 0 6356400 6361288 22503424
140736479321472 140736479321476 140736479321476 140736479322091 0\n",
4095) = 307
     0.000125 close(4)                  = 0
     0.000063 open("/proc/2515/statm", O_RDONLY) = 4
     0.000067 read(4, "5829 698 582 16 0 246 0\n", 4095) = 24
     0.000052 close(4)                  = 0
     0.000055 getdents(3, /* 0 entries */, 32768) = 0
     0.000056 close(3)                  = 0
     0.000055 nanosleep({0, 150000000}, NULL) = ?
ERESTART_RESTARTBLOCK (To be restarted)
    10.943440 --- SIGINT (Interrupt) @ 0 (0) ---



Please consider this to be a bug report, hosts are running 4.0.5, qemu
is a tagged 2.3.0 and the guest is debian jessie 3.16. Unfortunately I
do not have much time to bisect this one or go a bit more deeply right
now, hopefully it would be possible on a weekend with enough spare
time. Thanks!