- 02 May, 2013 5 commits
-
-
Kir Kolyshkin authored
Signed-off-by:
Kir Kolyshkin <kir@openvz.org> Signed-off-by:
Pavel Emelyanov <xemul@parallels.com>
-
Kir Kolyshkin authored
$(E) is used for 'silent make' to inform about what is going on when we are not printing actual commands being executed, and do nothing in case of 'verbose make (i.e. make V=1). For 'make help' we want to output the text regardless of whether 'silent make' is enabled or not. Signed-off-by:
-
Kir Kolyshkin authored
Signed-off-by:
-
Kir Kolyshkin authored
This is developers documentation, I doubt any user of crui is interested in that. So let's not build it by default. It can still be built manually: make -C Documentation Makefile.build.1 Signed-off-by:
-
Kir Kolyshkin authored
...using "git archive" command line from Pavel. Signed-off-by:
-
- 30 Apr, 2013 10 commits
-
-
Pavel Emelyanov authored
This is lots-of-bug-fixes release -- many new people started playing with criu and reported a lot of issues with it. Other than this, we have several new features. First of all it's C/R of netlink sockets C/R (required for Fedora container dump) and pending signals. Next, we've changed the format of memory dumps. This allowed us to implement so called page server, that allows to send memory dumps directly on another box w/o intermediate dump file. Also this new format made it possible to implement the (not yet final) --snapshot option, that demonstrates the work of user memory changes tracker in the kernel. Signed-off-by: -
Kir Kolyshkin authored
"Tools" in the context of a binary name doesn't make any sense, because every binary is a tool (lstool, dftool), plus it's one tool. So let's rename the beast to criu, which is short, pretty unique and coincides with the name of the project. Backward-compatible symlink is added for the sake of tests, we can fix them later and remove the link. Packagers beware: our binary is no longer crtools but criu. Signed-off-by:
-
Cyrill Gorcunov authored
We don't use tabs in docs. Signed-off-by:
Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by:
-
Cyrill Gorcunov authored
Signed-off-by:
-
Cyrill Gorcunov authored
Signed-off-by:
-
Pavel Emelyanov authored
It has been merged, so do require one. Signed-off-by: -
Pavel Emelyanov authored
There are some proc files (ns ones in particular) that are fake files and on open they screw their path up. Don't dump them for now. Signed-off-by: -
Alexander Kartashov authored
Use the specifier %lu instead of %zu to print an unsinged long integer. Signed-off-by:
Alexander Kartashov <alekskartashov@parallels.com> Signed-off-by:
-
Alexander Kartashov authored
Use the format specifier PRIx64 instead of %lx to print an off_t integer. Signed-off-by:
-
Alexander Kartashov authored
Use the format specifier %zu instead of %lu to print a size_t integer. Signed-off-by:
-
- 29 Apr, 2013 15 commits
-
-
Andrey Vagin authored
tcore (per thread core) should be used here Signed-off-by:
Andrey Vagin <avagin@openvz.org> Signed-off-by:
-
Andrey Vagin authored
pie/restorer.c: In function ‘restore_thread_common’: pie/restorer.c:262:39: error: incompatible types when assigning to type ‘k_rtsigset_t’ from type ‘u64’ RT_SIGFRAME_UC(sigframe).uc_sigmask = args->blk_sigset; ^ Signed-off-by: -
Andrey Vagin authored
Signed-off-by:
-
Andrey Vagin authored
RT_SIGFRAME_UC(sigframe).uc_sigmask.sig = args->blk_sigset; blk_sigset is u64, but uc_sigmask.sig has type ulong [2], so only a part of mask is restore here. This patch reworks restoring of blocking masks symmetrically to dumping these masks. Signed-off-by:
-
Andrey Vagin authored
v2: handle errors from setXids and securebits manipulations handle errors of restoring creds after finishing CR_STATE_RESTORE_CREDS, because a sigchild handler is already restored in this moment. Only the current process is killed in a error case. Signed-off-by: -
Andrew Vagin authored
Signed-off-by:
-
Andrey Vagin authored
Signed-off-by:
-
Andrey Vagin authored
Signed-off-by:
-
Libo Chen authored
fix fdinfo leak. Signed-off-by:
Libo Chen <libo.chen@huawei.com> Signed-off-by:
-
Vasily Averin authored
"crtools check" crashes on ubuntu 12.10 (00.011275) Error (proc_parse.c:1049): No records of type 6 found in fdinfo file (00.011281) Error (proc_parse.c:1052): parse_fdinfo: error parsing [flags: 02 ] for 6 : Operation not permitted *** glibc detected *** /home/vvs/devel/criu/crtools/crtools: double free or corruption (top): 0x000000000068a5a0 ** Signed-off-by:
Vasily Averin <vvs@parallels.com> diff-double-fclose-in-parse_fdinfo Signed-off-by:
-
Libo Chen authored
break running./crtools check : *** buffer overflow detected ***: ./crtools terminated ======= Backtrace: ========= /lib/x86_64-linux-gnu/libc.so.6(__fortify_fail+0x5c)[0x7fdaacb3e82c] /lib/x86_64-linux-gnu/libc.so.6(+0x109700)[0x7fdaacb3d700] ./crtools[0x423d00] ./crtools[0x41d021] ./crtools[0x41d259] ./crtools[0x41e4b6] ./crtools[0x418f03] ./crtools[0x404988] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xed)[0x7fdaaca5576d] ./crtools[0x404ad9] ======= Memory map: ======== 00400000-00463000 r-xp 00000000 08:01 189158 /home/clb/crtools/crtools 00662000-00663000 r--p 00062000 08:01 189158 /home/clb/crtools/crtools 00663000-00671000 rw-p 00063000 08:01 189158 /home/clb/crtools/crtools 00671000-00677000 rw-p 00000000 00:00 0 01801000-01822000 rw-p 00000000 00:00 0 [heap] 7fdaac81e000-7fdaac833000 r-xp 00000000 08:01 410533 /lib/x86_64-linux-gnu/libgcc_s.so.1 7fdaac833000-7fdaaca32000 ---p 00015000 08:01 410533 /lib/x86_64-linux-gnu/libgcc_s.so.1 7fdaaca32000-7fdaaca33000 r--p 00014000 08:01 410533 /lib/x86_64-linux-gnu/libgcc_s.so.1 7fdaaca33000-7fdaaca34000 rw-p 00015000 08:01 410533 /lib/x86_64-linux-gnu/libgcc_s.so.1 7fdaaca34000-7fdaacbe9000 r-xp 00000000 08:01 393459 /lib/x86_64-linux-gnu/libc-2.15.so 7fdaacbe9000-7fdaacde8000 ---p 001b5000 08:01 393459 /lib/x86_64-linux-gnu/libc-2.15.so 7fdaacde8000-7fdaacdec000 r--p 001b4000 08:01 393459 /lib/x86_64-linux-gnu/libc-2.15.so 7fdaacdec000-7fdaacdee000 rw-p 001b8000 08:01 393459 /lib/x86_64-linux-gnu/libc-2.15.so 7fdaacdee000-7fdaacdf3000 rw-p 00000000 00:00 0 7fdaacdf3000-7fdaace03000 r-xp 00000000 08:01 937695 /usr/local/lib/libprotobuf-c.so.0.0.0 7fdaace03000-7fdaad002000 ---p 00010000 08:01 937695 /usr/local/lib/libprotobuf-c.so.0.0.0 7fdaad002000-7fdaad003000 r--p 0000f000 08:01 937695 /usr/local/lib/libprotobuf-c.so.0.0.0 7fdaad003000-7fdaad004000 rw-p 00010000 08:01 937695 /usr/local/lib/libprotobuf-c.so.0.0.0 7fdaad004000-7fdaad01c000 r-xp 00000000 08:01 393528 /lib/x86_64-linux-gnu/libpthread-2.15.so 7fdaad01c000-7fdaad21b000 ---p 00018000 08:01 393528 /lib/x86_64-linux-gnu/libpthread-2.15.so 7fdaad21b000-7fdaad21c000 r--p 00017000 08:01 393528 /lib/x86_64-linux-gnu/libpthread-2.15.so 7fdaad21c000-7fdaad21d000 rw-p 00018000 08:01 393528 /lib/x86_64-linux-gnu/libpthread-2.15.so 7fdaad21d000-7fdaad221000 rw-p 00000000 00:00 0 7fdaad221000-7fdaad243000 r-xp 00000000 08:01 393441 /lib/x86_64-linux-gnu/ld-2.15.so 7fdaad437000-7fdaad43b000 rw-p 00000000 00:00 0 7fdaad440000-7fdaad443000 rw-p 00000000 00:00 0 7fdaad443000-7fdaad444000 r--p 00022000 08:01 393441 /lib/x86_64-linux-gnu/ld-2.15.so 7fdaad444000-7fdaad446000 rw-p 00023000 08:01 393441 /lib/x86_64-linux-gnu/ld-2.15.so 7fff9eb3c000-7fff9eb5d000 rw-p 00000000 00:00 0 [stack] 7fff9ebcf000-7fff9ebd1000 r-xp 00000000 00:00 0 [vdso] ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0 [vsyscall] I find this overflow by the phase: memcpy(sd->groups, groups, sd->gsize); The reason is sd->gsize = 8 Byte, but sd->groups = xmalloc(sizeof(sd->gsize)) is exact 4 Byte. Signed-off-by:
Qiang Huang <h.huangqiang@huawei.com> Acked-by:
-
Kir Kolyshkin authored
1 The comment before MAKEFLAGS assignment (see commit 25e8cb5c) talks about -r (and possibly -R) flags and has nothing to do with --no-print-directory. Let's remove the last option to not confuse a reader. 2 The --no-print-directory flag to build and build-crtools shorthands is redundant, because scripts/Makefile.build has it already. Remove. Cc: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by:
-
Kir Kolyshkin authored
We don't use TAGS, AWK or ECHO anywhere in the code (at least according to git grep). Let's remove this. Cc: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by:
-
Kir Kolyshkin authored
This reverts part of commit a519e0c8 which broke the man page build: [kir@kirpad Documentation]$ make " GEN " crtools.1 /bin/sh: GEN : command not found make: *** [crtools.1] Error 127 So let's reintroduce Q and E in a separate Makefile.inc as it was before. Cc: Stanislav Kinsbursky <skinsbursky@openvz.org> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Pavel Emelyanov <xemul@parallels.com> Signed-off-by:
-
Libo Chen authored
Update to patch "check: Introduce --ms option" (commit: 379abca9) I find this error in my box: cc -g -O2 -Wall -Werror -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=0 -iquote ../../lib/arch/x86/include -I../../lib session03.c ../../lib/libzdtmtst.a -o session03 session03.c:112:7: error: variably modified ‘stack’ at file scope The reason is : "#define CLONE_STACK_SIZE PAGE_SIZE" is conflict with "#define PAGE_SIZE (sysconf(_SC_PAGESIZE))" defined in <sys/user.h>. I think CLONE_STACK_SIZE should be seted 4096 as session02.c. If this patch is appropriate, please let me know. Signed-off-by:
Andrew Vagin <avagin@parallels.com> Signed-off-by:
-
- 23 Apr, 2013 4 commits
-
-
Pavel Emelyanov authored
This option skips checking for features, that are known to be absent in mainstream kernel. Signed-off-by: -
Pavel Emelyanov authored
To know what exactly is missing. Signed-off-by: -
Pavel Emelyanov authored
It will appear not earlier than in 3.10. Signed-off-by: -
Andrey Vagin authored
According to the commit message in the kernel git, a new trap flag JOBCTL_TRAP_INTERRUPT is added, which is set on PTRACE_INTERRUPT and cleared when any trap happens. Currently we have a race window beetwen PTRACE_CONT and PTRACE_INTERRUPT. If a process stops before PTRACE_INTERRUPT, the flag JOBCTL_TRAP_INTERRUPT will be set once again. https://bugzilla.openvz.org/show_bug.cgi?id=2569Signed-off-by:
-
- 19 Apr, 2013 3 commits
-
-
Andrew Vagin authored
Signed-off-by:
-
Cyrill Gorcunov authored
This reverts commit 396a08eb. As Andrew requested he needs the symbols to be generated for all possible sources in the crtools code. Signed-off-by:
-
Andrey Vagin authored
For security reason processes can be resumed only when all credentials are restored. Otherwise someone can attach to a process, which are not restored credentials yet and execute some code. https://bugzilla.openvz.org/show_bug.cgi?id=2561Signed-off-by:
-
- 18 Apr, 2013 3 commits
-
-
Pavel Emelyanov authored
Just add new protocol command and call local_xfer->write_hole on server side. Also tune the mem-snap test to test memory xfer via page server. Signed-off-by: -
Pavel Emelyanov authored
It's not complete and is about to change significantly. Better not show it oficially. Signed-off-by: -
Pavel Emelyanov authored
Signed-off-by:
-
