Jenkins sends this files as a message body if something failed.
Cc: Sergey Bronnikov <sergeyb@openvz.org>
Fixes: 474f2dfc5d ("zdtm.py: add option --keep-going")
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Reported-by: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com>
Cc: Stanislav Kinsburskiy <skinsbursky@virtuozzo.com>
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We were luckely on built-in tests, because always
AIO pages were present, and, thus, they are dumped.
But AIO pages are not guaranteed to be present
in common case.
This patch makes AIO pages are dumped certanly.
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
nr_req always depends on ring size and we even check for that.
Kill parasite_aio::max_reqs,vma_nr_reqs and vma_area::aio_nr_req,
because they are not need and excess.
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
The first is a preparation to kill aio_nr_req. The second
and the third are to have more debug info at restore.
v2: Print mmap err code
Signed-off-by: Kirill Tkhai <ktkhai@vituozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
This should fix errors on build envs, where git tags haven't
been fetched. I.e, Travis-ci does not fetch tags before
building a project, wich results in:
---> Running in 0d63a3d28a7a
fatal: No names found, cannot describe anything.
fatal: No names found, cannot describe anything.
make[1]: Entering directory '/criu'
The other way is to fetch tags on `git describe` failure, but
it that way has two cons:
- it will change user's git (and I'm not sure it's right);
- we need not only fetch tags in that case, but also remotes,
as otherwise it will fail with "fatal: No tags can describe <hash>":
https://travis-ci.org/0x7f454c46/criu/jobs/133362045#L1748
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
to fix error:
./crit/crit decode -i test/dump/zdtm/static/maps04/118/1/stats-dump
Traceback (most recent call last):
File "./crit/crit", line 276, in <module>
main()
File "./crit/crit", line 273, in main
opts["func"](opts)
File "./crit/crit", line 28, in decode
img = pycriu.images.load(inf(opts), opts['pretty'], opts['nopl'])
KeyError: 'nopl'
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
In some images there can be quite a long "payload" -- some
raw data that is represented by base64 encoding. If we want
to explore huge images reading tons of base64 symbols can
be quite time consuming :) E.g. I a 1.5 gigs image with sysv
shmem was sent to me some time ago for investigation %)
So here is the --nopl argument for show action (decode should
produce encode-able image, so payload there is needed) that
just shows the amount of bytes in payload (if any).
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
This seems to be known problem in util-vdso.c on aarch64 [1].
Now restorer segfaults with the following log:
[ 8107.384817] criu[5135]: unhandled level 3 translation fault (11) at 0x00005b98, esr 0x90000007
[ 8107.385538] pgd = ffffffc038dbc000
[ 8107.386046] [00005b98] *pgd=0000000078d6c003, *pud=0000000078d6c003, *pmd=0000000073c31003, *pte=0000000000000000
[ 8107.391920]
[ 8107.392521] CPU: 0 PID: 5135 Comm: criu Not tainted 4.5.0 #29
[ 8107.392805] Hardware name: linux,dummy-virt (DT)
[ 8107.393140] task: ffffffc039a2a400 ti: ffffffc033c34000 task.ti: ffffffc033c34000
[ 8107.393782] PC is at 0x13514
[ 8107.406332] LR is at 0x1342c
[ 8107.406550] pc : [<0000000000013514>] lr : [<000000000001342c>] pstate: a0000000
This is because gcc for aarch64 adds vdso_symbols array to symbols
table and by that reason, it needs run-time relocations in place.
How it goes (with cut not interesting assembly):
0x14104: adrp x7, 0x14000 ; adrp+add loading of
0x14114: add x0, x7, #0x928 ; symbol table's address,
0x14134: ldp x2, x3, [x0] ; loading address of symbol from
; symbol table
0x1414c: stp x2, x3, [x29,#112] ; saving it on stack on
; function's begin
0x14188: ldr x2, [x29,#112] ; using symbol's address in code
The symbol for this in symbol table is:
[root@alarm cr]# readelf -s criu/pie/restorer.built-in.bin.o | grep 5b98
248: 0000000000005b98 0 NOTYPE LOCAL DEFAULT 1 $d
And also may be visible this way:
objdump -dS criu/pie/restorer.built-in.bin.o | less
...
0000000000004924 <cur_loglevel>:
4924: 00000002 00005b98 00000000 00005ed0 .....[.......^..
4934: 00000000 00005ee8 00000000 00005f00 .....^......._..
So, in a symbol table lies not relocated address of symbol.
The real address may be visible with added printing of vdso_symbols[0]:
pie: vdso: vdso_symbols[0] 0x15b98
(this way gcc by some reason does access symbol through
local adrp+add calculations, than through stack-saved pointer
of a symbol from symbol table).
While we don't handling properly relocs here, I suggest this
ugly workaround.
Temporary fix for: #150
[1]: https://lists.openvz.org/pipermail/criu/2015-October/022453.html
Cc: Wang Long <long.wanglong@huawei.com>
Cc: Christopher Covington <cov@codeaurora.org>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Tested-by: long.wanglong <long.wanglong@huawei.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
As for compatible 32-bit pie size of uint64_t is bigger than size of
unsigned long type, add proper casts, so we could build compatible pie.
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We are going to build everything from toplevel
for convenience (just like it was before criu 2)
so move one more snippet on top.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
They are already defined and exported.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We use in our PCS7 packaging and I think it's
suitable for everything else as well.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
$ git log v4.4-rc1-465-gd6df198
commit d6df198d924775e4751561cf60ef0294e95f74df
Author: Phil Sutter <phil@nwl.cc>
Date: Tue Dec 1 22:45:15 2015 +0100
net: ipv6: restrict hop_limit sysctl setting to range [1; 255]
Setting a value bigger than 255 resulted in using only the lower eight
bits of that value as it is assigned to the u8 header field. To avoid
this unexpected result, reject such values.
https://github.com/xemul/criu/issues/164
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
shmem.c:290:4: note: in expansion of macro 'pr_err'
pr_err("Shmem %lx already collected as SYSVIPC\n", vi->shmid);
^
v2: fix one more place
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
====================== Run zdtm/static/mnt_tracefs in uns ======================
Start test
./mnt_tracefs --pidfile=mnt_tracefs.pid --outfile=mnt_tracefs.out --dirname=mnt_tracefs.test
Run criu dump
Wait for zdtm/static/mnt_tracefs to die for 0.100000
Run criu restore
Run criu dump
Run criu restore
Run criu dump
Run criu restore
1: Old mounts lost: set([])
1: New mounts appeared: set([('/', '/debugfs/tracing')])
Wait for zdtm/static/mnt_tracefs to die for 0.100000
Cc: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Acked-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We have 3 structures and 2 ways to keep them while doing C/R.
Let's unify the vma_area is -- one struct with union and a
hash table.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
We now have two places where we if/elseif/elseif/else a vma
and take actions -- the prepare_mm_pid and open_vmas. I propose
to make this if/... onlce once (while preparing) and set up
a callback on vma to be called later (while opening).
This requires my patch that doesn't set -1 in alloc_vma or
Andrey's patch that does the same.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Image a process has done shmget(2 pages), then shmat() then
mprotect(1 page, ro). In this case criu will dump 1 shmem
segment 2 pages long and 2 vmas 1 page each.
But on restore time we'll call shmat() for _each_ vma and the
very first one will occupy the whole 2 pages space in vm
(there's no size argument for shmat, only for shmget) thus
blocking the 2nd vma from shmat()-in again.
The solution is:
1. check that each shmem segment is attached by
the sequence of vmas that cover one w/o holes
2. shmat() only the first one
3. mprotect() all of them if needed (there's no hunks
for this step in this path, mprotect is already
called in pie/restorer.c and does things well)
v2:
* List can contain anon shmems (caught by zdtm)
* There can be many attachments of a segment (caught by transition/ipc)
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
When testing for queued data present on
peers we setup non-blocking mode over them
to obtain -EAGAIN if there is no data
at all, but I forget to restore former
mode on return, ie blocking read.
Reported-by: Andrey Vagin <avagin@virtuozzo.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
For easier reading: we need it for pty queued data only,
but this way the code is simplier.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Usually unix stream sockets have consistent shutdown state
(if one end shutdown for read than other end will be shut down for write).
However sometimes it's possible to have stream sockets with mismatched
shutdown state. This happens if we connect() already shut down socket
(see upcoming sockets03.c test for example).
On read()/write() kernel always checks shutdown state on both ends, so
from userspace POV sockets with mismatched shutdown state behave the same
way as with matched state. So I think that we can just ignore shutdown
mismatch during dump, and just restore sockets with matching shutdown
state. Process should not see the differences.
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
The inital mount() call for a bind mount needs to have the MS_REC flag
present if the bind mount is recursive.
Signed-off-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Sting field sysctl_entry.sarg has wrong length in protobuf encoding in
netdev-9 image, according to
https://developers.google.com/protocol-buffers/docs/encoding :
In netdev-9.img binary representation of sarg field is: "1a 2c 32 36 30
37 3a 66 30 64 30 3a 31 30 30 32 3a 30 30 35 31 3a 30 30 30 30 3a 30 30
30 30 3a 30 30 30 30 3a 30 30 30 34 0a c0 f4 a7 01"
Field key is 0x1a - means type 2(Length-delimited field), field 3 - that
is sarg field.
Field len is 0x2c which is 44, first 40-byte
"2607:f0d0:1002:0051:0000:0000:0000:0004\0" and then last 4 - "c0 f4 a7
01" where 0xc0 is not utf-8, and that makes crit fail.
In sysctl_op we just read() from sysctl so no '\0' is added in the end
of the string. So we can zero-init arrays and that will fix the issue.
https://github.com/xemul/criu/issues/161
*Alternatively or additionaly we can put '\0' in the end of string in
sysctl_read_char to make it harder to make such a mistake in future.
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Acked-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
use native SYSCTL_TYPE__CTL_32 and SYSCTL_TYPE__CTL_STR
v2: add BUILD_BUG_ONS to check SysctlType constants are equal to
__CTL_STR and CTL_32, change __CTL_STR to CTL_STR in SysctlType enum.
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
For more convenient log parsing.
Reported-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
man of inet_ntop says that INET_ADDR_LEN should not be less than
INET_ADDRSTRLEN and INET6_ADDRSTRLEN
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Because stable_secret sysctl c/r produces a lot of spam error messages
in tests: https://github.com/xemul/criu/issues/160 But EIO for these
sysctl is valid behavior and no error need to be printed.
More general solution will also silence error messages for
sysctl_read_{u32,s32,u64}, but the flag CTL_FLAGS_READ_EIO_SKIP is
now used only for __CTL_STR and only for stable_secret so we can do
silence error only in sysctl_read_char.
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Acked-by: Andrew Vagin <avagin@virtuozzo.com>
Tested-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
It's required when a target path is overmounted by one of child mounts.
Signed-off-by: Andrew Vagin <avagin@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
When we're walking over controllers and their directories
we should update the directory at the very end because
otherwise we will not match the rest of them once variable
is updated.
So I desided to simply rewrite the helper. Hopefully it
is more clear now.
Signed-off-by: Cyrill Gorcunov <gorcunov@virtuozzo.com>
Acked-by: Tycho Andersen <tycho.andersen@canonical.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
