- 07 Apr, 2016 19 commits
-
-
Cyrill Gorcunov authored
In case of error inside collect_cgroups we need - @ctls list has been spliced into @cs->ctls so we cant just free it, but rather use put_ctls on @cs->ctls - delete @cs->l from global @cg_sets list before free it Signed-off-by:
Cyrill Gorcunov <gorcunov@virtuozzo.com> Reviewed-by:
Andrew Vagin <avagin@openvz.org> Acked-by:
Tycho Andersen <tycho.andersen@canonical.com> Signed-off-by:
Pavel Emelyanov <xemul@virtuozzo.com>
-
Cyrill Gorcunov authored
If add_cgroup_properties failed we have @ncd sitting in @match->children or @current_controller->heads list, and then we free @ncd entry leaving those lists carrying dead pointers. Add proper rollback action. Signed-off-by:
-
Cyrill Gorcunov authored
Since we're freeing list entries don't forget to initialize list then, otherwise it gets out with free entries and may hit use-after-free bug. Signed-off-by:
-
Dmitry Safonov authored
We will still able to distinguish if it's garbage in ax register, but this is definitely prettier: (00.036900) 687 was trapped (00.036903) 687 is going to execute the syscall 158 (00.036928) 687 was trapped (00.036931) 687 is going to execute the syscall 173 Than that: (00.024403) 687 was trapped (00.024407) 687 is going to execute the syscall 9e (00.024419) 687 was trapped (00.024421) 687 is going to execute the syscall ad Signed-off-by:
Dmitry Safonov <dsafonov@virtuozzo.com> Acked-by:
Cyrill Gorcunov <gorcunov@openvz.org> Acked-by:
Andrew Vagin <avagin@virtuozzo.com> Signed-off-by:
-
Cyrill Gorcunov authored
Previously when we've been on early stage we tried to workaround lack of prlimit libc call simply providing own implementation, but it cause problems on some libc configurations so simply use rlimit64 and __NR_prlimit64 syscall directly. The kernel must support __NR_prlimit64 syscall and provide rlimit64 structure as well. Signed-off-by:
-
Pavel Emelyanov authored
The option was in use when we didn't have the /proc/pid/ns links and thus ns ids. Later we used it at restore time to tell which namespaces to restore the pstree into. Right now ids are always generated and this option has no effect (and meaning). Signed-off-by: -
Pavel Emelyanov authored
We've been writing this file since 0.4, hopefully nobody's using such an oldie :) v2: Don't BUG_ON, but return error, as this can be triggered by corrupting images. Signed-off-by: -
Cyrill Gorcunov authored
This helper is used in both PIE and non-PIE code. For the second we can use normal pr_perror. Signed-off-by:
-
Andrew Vagin authored
Atomic operations doesn't work for non-aligned pointers. =[log]=> dump/zdtm/static/autofs/30/2/restore.log ------------------------ grep Error ------------------------ (00.185037) 5: Error (/root/git/criu/criu/include/lock.h:78): BUG at /root/git/criu/criu/include/lock.h:78 (00.207193) Error (cr-restore.c:1407): 100 killed by signal 9: Killed (00.219187) Error (cr-restore.c:2237): Restoring FAILED. ------------------------ ERROR OVER ------------------------ 4339 futex(0x7fa79c1561da, FUTEX_WAKE, 2147483647 <unfinished ...> 4339 <... futex resumed> ) = -1 EINVAL (Invalid argument) Signed-off-by:
-
Andrew Vagin authored
v2: copy aio00.desc Signed-off-by:
-
Kirill Tkhai authored
Since AIO ring is never pre-dumped, it never has a parent in pre-dump. Make generate_iovs() to know this. Signed-off-by:
Kirill Tkhai <ktkhai@virtuozzo.com> Signed-off-by:
-
Kirill Tkhai authored
Signed-off-by:
-
Kirill Tkhai authored
Description: 1)Create io context, submit several io operations and get events of some of them. 2)Sleep on signal. 3)Check tail, head and nr events restored correct. 4)Emit one more io operation. 5)Check tail has moved one number forward (i.e. synchronize in-kernel and userspace tails if they were different). Signed-off-by:
-
Kirill Tkhai authored
Check nr in aio header during dump and correctness of tail, head and nr during restore. Signed-off-by:
-
Kirill Tkhai authored
1)Dump/restore mmaped aio ring like any other private vma entry, with the only exception we do not predump it. 2)Create io context, set head and tail using write to /dev/null. 3)Copy aio ring restored in (1) to created in (2). 4)Remap (2) to address of (1). Signed-off-by:
-
Kirill Tkhai authored
Signed-off-by:
-
Kirill Tkhai authored
Signed-off-by:
-
Kirill Tkhai authored
Move the declaration from .c to header file. Also rename parasite_check_aios() since we have one more function with the same name. Signed-off-by:
-
Adrian Reber authored
The page server already knows how to create a TCP server and client. The code has been hidden behind static-s in page-xfer.c. This commit moves the common TCP server/client code to util.c to be able to also use it in the upcoming remote lazy-server/lazy-client userfaultfd enhanced criu. Signed-off-by:
Adrian Reber <areber@redhat.com> Signed-off-by:
-
- 04 Apr, 2016 21 commits
-
-
Andrew Vagin authored
Signed-off-by:
-
Dmitry Safonov authored
Impact: no functional changes, cleanup. Signed-off-by:
-
Dmitry Safonov authored
Impact: no functional changes, cleanup. Signed-off-by:
-
Dmitry Safonov authored
Impact: no functional changes, cleanup Signed-off-by:
-
Dmitry Safonov authored
Signed-off-by:
-
Dmitry Safonov authored
Add checks that structure end is inside memory area [mem, mem + size] before dereferencing structures. Otherwise, for example, vdso_fill_symtable will try to dereference ehdr members even with vma size 0. Signed-off-by:
-
Dmitry Safonov authored
Impact: drop additional casts all around. Cleanup. Signed-off-by:
-
Dmitry Safonov authored
Impact: refactor Signed-off-by:
-
Dmitry Safonov authored
As I am here to fix some things, I would like to remove UB-based pointer compare (6.5.8.5). We have -fno-strict-aliasing in pie CFLAGS, but it even looks nicer and more readable with uintptr_t. Impact: refactoring Signed-off-by:
-
Dmitry Safonov authored
Add Elf32 types and wrap them with Elf64 into some more generic types. This is for parsing vDSO from image file. Signed-off-by:
-
Pavel Emelyanov authored
Right now we get SOL_TCP.TCP_INFO option twice, but one time is enough. Signed-off-by: -
Pavel Emelyanov authored
This is prerequisite for the next patch and, actually, for the libsoccr :) Signed-off-by: -
Dmitry Safonov authored
Parasite should cleanup after fauilure and remove read-protection from VMAs. Check it with maps00 test & fault injection. Tested on travis with some addition: python test/zdtm.py run -t zdtm/static/maps00 --fault 3 --keep-going --report report -f h || false https://travis-ci.org/0x7f454c46/criu/jobs/119252291 Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by:
-
Pavel Emelyanov authored
It's just a cleanup to get rid of in-function {<code>} block. Signed-off-by: -
Pavel Emelyanov authored
By default criu tries to create parasite mem with memfd syscall, which is present in many systems. Let's check also the legacy way. Signed-off-by: -
Pavel Emelyanov authored
Signed-off-by: -
Pavel Emelyanov authored
Signed-off-by: -
Pavel Emelyanov authored
Signed-off-by: -
Pavel Emelyanov authored
The plan is to allow ->collect callbacks to read more stuff from image files. Signed-off-by: -
Pavel Emelyanov authored
Signed-off-by: -
Pavel Emelyanov authored
Signed-off-by:
-
