- 16 Jun, 2015 2 commits
-
-
Tycho Andersen authored
o_O this is one of those "how has nobody ever hit this" bugs. According to: http://stackoverflow.com/questions/2535989/what-are-the-calling-conventions-for-unix-linux-system-calls-on-x86-64 syscall on x86_64 clobbers rcx and r11. We should mark these registers as clobbered in RUN_CLONE_RESTORE_FN so that gcc avoids using them, e.g.: ...<__export_restore_task> ... 1894: 49 89 db mov r11,rbx 1897: 4c 8d 4c 24 60 lea r9,[rsp+0x60] 189c: 48 89 eb mov rbx,rbp 189f: 48 89 44 24 10 mov QWORD PTR [rsp+0x10],rax 18a4: 48 8d 44 24 30 lea rax,[rsp+0x30] 18a9: 48 89 44 24 18 mov QWORD PTR [rsp+0x18],rax 18ae: eb 53 jmp 1903 <clone_end+0x15> 00000000000018b0 <clone_emul>: 18b0: 48 8b 74 24 08 mov rsi,QWORD PTR [rsp+0x8] 18b5: 48 83 ee 10 sub rsi,0x10 18b9: 48 89 ef mov rdi,rbp 18bc: 48 89 7e 08 mov QWORD PTR [rsi+0x8],rdi 18c0: 49 8b 7b 20 mov rdi,QWORD PTR [r11+0x20] 18c4: 48 89 3e mov QWORD PTR [rsi],rdi 18c7: 48 c7 c7 00 0d 05 00 mov rdi,0x50d00 18ce: 48 8b 54 24 18 mov rdx,QWORD PTR [rsp+0x18] 18d3: 4d 89 fa mov r10,r15 18d6: b8 38 00 00 00 mov eax,0x38 18db: 0f 05 syscall 18dd: 48 85 c0 test rax,rax 18e0: 74 05 je 18e7 <thread_run> 18e2: 48 89 c5 mov rbp,rax 18e5: eb 07 jmp 18ee <clone_end> 00000000000018e7 <thread_run>: 18e7: 48 31 ed xor rbp,rbp 18ea: 58 pop rax 18eb: 5f pop rdi 18ec: ff d0 call rax 00000000000018ee <clone_end>: 18ee: 41 83 c5 01 add r13d,0x1 18f2: 49 81 c7 00 8c 00 00 add r15,0x8c00 18f9: 45 39 6b 14 cmp DWORD PTR [r11+0x14],r13d will cause a seg fault because r11 is clobbered. Signed-off-by:
Tycho Andersen <tycho.andersen@canonical.com> Signed-off-by:
Pavel Emelyanov <xemul@parallels.com>
-
Cyrill Gorcunov authored
Otherwise getting | parasite-syscall.c: In function ‘parasite_infect_seized’: | parasite-syscall.c:1222:5: error: ‘elf_relocs’ undeclared (first use in this function) Simply wrap the @elf_relocs_apply with macros. Reported-by:
Andrey Vagin <avagin@openvz.org> Signed-off-by:
Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by:
-
- 15 Jun, 2015 8 commits
-
-
Cyrill Gorcunov authored
When been playing wich checkpoint/restore of container I found that we can't reuse existing controller if they were pre-created. For example currently in PCS7 we're bindmount cgroups which belong to a container in a form of /sys/fs/cgroup/<controller>/<container> ==> /sys/fs/cgroup/<controller> so that CRIU dumps such configuration fine but on restore it recreates controllers from the scratch which we would like to bindmount them and ask CRIU to restore subcgroups and their parameters. So I extended --manage-cgroups option to take <mode> arguments. Detailed description in docs. Signed-off-by:
-
Cyrill Gorcunov authored
Currently we always create temporary directory where we restore cgroups, but this won't work in case if mounting cgroups is forbidden from inside of a container for some reason (as in OpenVZ kernel). So one can pass --cgroup-yard option to specify an existing directory where cgroups are living. By default we assume it lays in /sys/fs/cgroup. Signed-off-by:
-
Cyrill Gorcunov authored
For example some linkers generate @__export_parasite_args as symbol which won't relocate. Handle such case properly. Signed-off-by:
-
Cyrill Gorcunov authored
Signed-off-by:
-
Cyrill Gorcunov authored
Signed-off-by:
-
Cyrill Gorcunov authored
Otherwise it's confusing since. Signed-off-by:
-
Tycho Andersen authored
The error I got was: CC pie/piegen/elf-x86-64.o In file included from pie/piegen/elf-x86-32.c:16:0: pie/piegen/elf.c: In function ‘handle_elf_x86_32’: pie/piegen/elf.c:476:3: error: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 6 has type ‘Elf32_Word’ [-Werror=format=] pr_debug("Copying section '%s'\n" \ ^ Signed-off-by: -
Laurent Dufour authored
On PPC64, the hard definition of TFD_IOC_SET_TICKS doesn't match the kernel one. We should use the _IOW based on to be more flexible here. Signed-off-by:
Laurent Dufour <ldufour@linux.vnet.ibm.com> Acked-by:
-
- 11 Jun, 2015 9 commits
-
-
Tycho Andersen authored
We'll need this for use in the restorer blob for restoring LSMs. It looks like arm already has openat, so I think it's just x86 and ppc that need it. In any case, please double check this, as I've only tested it on x86. Signed-off-by:
-
Andrew Vagin authored
Reported-by: Mr Jenkins Signed-off-by:
Ruslan Kuprieiev <rkuprieiev@cloudlinux.com> Signed-off-by:
-
Andrey Vagin authored
If the netns image is absent, the NetnsEntry entry will not be initialized. Currently restore from old images crashes: Core was generated by `criu swrk 3'. Program terminated with signal SIGSEGV, Segmentation fault. $0 0x0000000000427d80 in netns_entry.free_unpacked () (gdb) bt $0 0x0000000000427d80 in netns_entry.free_unpacked () $1 0x0000000000436d07 in prepare_net_ns () $2 0x0000000000457c78 in prepare_namespace () $3 0x0000000000432917 in restore_task_with_children () $4 0x00007fc86acfccfd in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111 v2: remove debugging code Cc: Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Signed-off-by:
Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Signed-off-by:
-
Laurent Dufour authored
The checkpoint and restore of the Power PC floating point registers is buggy. The issue is that the signal frame context is defined to store double value while the protocol buffer is handling unsigned 64bits integer value. A silent cast done by the compiler was modifying the restored value in our back. This fix changes the type used when manipulating the FP registers value to be consistent between checkpoint and restart. Signed-off-by:
-
Andrey Vagin authored
Otherwise the root yard can be propagated into the host mount namespace and remain there and criu will fail, because it will not be able to remove the roots yard. It occures if we give a shared mount as root to "criu restore" and criu converts it into a slave mount. Reported-by:
-
Laurent Dufour authored
Add prefix to the piegen's error and debug output to avoid confusion and fix few debug lines. Signed-off-by:
-
Laurent Dufour authored
Since opts is defined as extern in piegen.h, there is no need to pass it as argument. Signed-off-by:
-
Laurent Dufour authored
Introduce a new -o argument to piegen to specify generate file name. Send the debug stream to stdout and force it to /dev/null in the makefile if V=1 is not specify. Signed-off-by:
-
Laurent Dufour authored
handle elf should return 0 on success. Signed-off-by:
-
- 08 Jun, 2015 21 commits
-
-
Cyrill Gorcunov authored
Signed-off-by:
-
Cyrill Gorcunov authored
To hide @nr_gotpcrel Signed-off-by:
-
Laurent Dufour authored
When building the blob in the generated header file, we may shrink the output blobk and only copy the sections with the SHF_ALLOC bit set, the other ones are not needed at runtime. Signed-off-by:
-
Laurent Dufour authored
This cleans the assembly code, removing no more needed trick with the register 2 (TOC pointer). As a consequence, the __export_restore_task_trampoline() and __export_unmap_trampoline() are no more needed. Thus, the changes introduced by the commit de9df910 ("Per architecture restorer trampolines") in cr-restore.c are no more used but are not impacting runtime code anyway. Signed-off-by:
-
Laurent Dufour authored
PowerPC linking requires the TOC to be in its own section and to be aligned. Signed-off-by:
-
Cyrill Gorcunov authored
- Move relocs application into a separate file which get compiled as a regular C file in criu (pie/pie-relocs.[ch]) - Move types used by piegen into pie/piegen/uapi/types.h Signed-off-by:
-
Cyrill Gorcunov authored
At moment both parasite and restorer do not have any relocs because we support x86-64 only, but this will be changed soon so do a call and apply relocations. Signed-off-by:
-
Cyrill Gorcunov authored
We should use provided @nr_relocs instead of ARRAY_SIZE here. Didn't spot it earlier simply because at moment on x86-64 there is no relocs at all. Also when we apply relocation they are to be computed from virtual base of parasite address, not from local memory map address, so add @vbase parameter. And fix typo on addend in gotpcrel. Signed-off-by:
-
Cyrill Gorcunov authored
In case of @gotpcrel relocations we need additional space to carry pointers. Signed-off-by:
-
Cyrill Gorcunov authored
After this patch one can run ARCH="ia32" make to build 32bit version on CRIU on 64bit host. Note this is only build procedure which tuned up, the CRIU itself is not yet ready to make a checkpoint/restore cycle -- a lot of additional code is needed and here we rather put stubs simply to make build procedure run. Signed-off-by:
-
Cyrill Gorcunov authored
Signed-off-by:
-
Cyrill Gorcunov authored
There is no rip addressing in 32bit code but PIE code require GOT tables and friend which we better escape for performance sake. So lets use pc relocations it should do the trick. Signed-off-by:
-
Cyrill Gorcunov authored
Signed-off-by:
-
Cyrill Gorcunov authored
To support x86-32 mode we will need own syscall table. Here is it. Note the CRIU itself doesn't support such mode yet. Meanwhile put syscall table here just in case if someone is adding new syscall 32 bit variant should be updated as well. Signed-off-by:
-
Cyrill Gorcunov authored
Since at the moment we're running only x86-64 not 32 bit tasks, and our code is not carrying any big statically defined structures we can use relocatable files directly with all relocation applied during building. This gonna be changed soon once we start supporting 32 bit tasks. IOW even currently we need (which is not yet done but it's safe) - check for gotpcrel relocations - apply relocations with generated elf_apply_relocs helper Currently overall scheme looks this way - our object files are linked together into parasite.built-in.bin.o file - then pie/piegen/piegen tool is called which parses this file and generates C source code file with bytestream and all information needed to rellocate this bytestream into a new place (see elf_apply_relocs helper) Signed-off-by:
-
Cyrill Gorcunov authored
Here we simply build piegen tool which gonna be used to generate parasite code safe to rellocate. The tool is taking object file as an argument, parses it and generates C file with rellocations encoded in form suitable for fast appliance. Currently only x86-32 x86-64 is supported. v2 (by ldufour@): - Filter PIEGEN Signed-off-by:
-
Pavel Emelyanov authored
On restore we have several arrays of objects that get remapped into pie area and their number is also passed. Clean and shorten the remapping code a bit and bing their naming to common format. Signed-off-by: -
Andrew Vagin authored
systemd executes tests in subgroups: 9679 ? Ssl 0:41 /etc/alternatives/java -Dcom.sun.akuma.Daemon=daemonized -Djava.awt.headless=true -DJENKINS_HOME=/var/lib/jenkins -jar /usr/lib/jenkins/jenkins.war - [root@jenkins ~]# cat /proc/9679/cgroup 10:net_cls:/ 9:hugetlb:/ 8:cpuset:/ 7:blkio:/system.slice 6:freezer:/ 5:cpu,cpuacct:/system.slice 4:devices:/system.slice/jenkins.service 3:perf_event:/ 2:memory:/system.slice 1:name=systemd:/system.slice/jenkins.service Reported-by: Mr Jenkins Signed-off-by:
-
Andrew Vagin authored
Signed-off-by:
-
Adrian Reber authored
The criu(8) man-page is generated using asciidoc. The problem with asciidoc is that, due to its dependencies, it is not available on all distributions or it is undesired to install all asciidoc dependencies. The install target was unconditionally installing and thus building the man-page even if not explicitly specified with 'make docs'. With the new 'install-criu' target everything besides the man-page is installed and the target 'install-man' is only called by the target 'install'. Signed-off-by:
Adrian Reber <areber@redhat.com> Acked-by:
-
Pavel Emelyanov authored
On restore we will use the peer's name to connect() the socket back, so if there's no name dump should be aborted. This situation happens when we create a socketpair(), fork and dump only one task with one pair end. Reported-by:
Artem Kuzmitskiy <artem.kuzmitskiy@gmail.com> Signed-off-by:
-
