- 02 Apr, 2017 40 commits
-
-
Andrei Vagin authored
Signed-off-by:Andrei Vagin <avagin@virtuozzo.com>
-
Pavel Emelyanov authored
The contents of sysvshm areas is sometimes dumped twice. First time it goes via cr_dump_shmem(), since each attachment of the segment is noted by the add_shmem_area() and thus gets into the list scanned by the former call(). Second time it is dumped by dump_ipc_shm_pages(). Another thing is that in the 1st dump memory gets into the standard pagemap/pages image pair, while in the 2nd time it's all written as plain byte array into ipcns-shm.img. At restore contents is read from the ipcns-shm.img (2nd). While the first way is preferred, let's (temporarily) eliminate it. Historically these segments were dumped and restored the 2nd way. Signed-off-by:
Pavel Emelyanov <xemul@virtuozzo.com> Signed-off-by:
-
Kir Kolyshkin authored
This is just to prevent a compiler warning. Signed-off-by:
Kir Kolyshkin <kir@openvz.org> Signed-off-by:
-
Kir Kolyshkin authored
When this test is compiled on ARMv7, we get this: > maps03.c: In function 'main': > maps03.c:15:31: error: result of '10l << 30' requires 35 bits > to represent, but 'long int' only has 32 bits [-Werror=shift-overflow=] > mem = (void *)mmap(NULL, (10L << 30), PROT_READ | PROT_WRITE, > ^~ Surely, the nature of the test looks like it needs 64-bit addressing. Signed-off-by:
-
Kir Kolyshkin authored
Make it possible to use a custom message telling why the test is skipped. Signed-off-by:
-
Kir Kolyshkin authored
Makefile knows that a binary depends on .o file, and the .o file depends on .c file, and adds the per-target variables even in the indirect case. Therefore, .o suffix is not needed here. Note that it not hurts to leave it as is, so this commit is done only for the sake of keeping things simple. Signed-off-by:
-
Kir Kolyshkin authored
This test name was added by commit 1ce1cca2, but there is no test. We do not use these lists since zdtm.py was introduced, therefore the error went unnoticed. I found it while running "make dep". Signed-off-by:
-
Kir Kolyshkin authored
This commit expands the success with lib/Makefile to the rest of ztdm/Makefiles. In particular, it moves the common part to Makefile.inc and let {lib,static,transition}/Makefile use it. This results in: - dependencies being handled properly; - reduction of makefiles length due to unification; - possibly more proper dependencies and builds. While at it, let's also: - make rules for criu-rtc.so silent; - set default CC=gcc; - remove duplicate -m32 from CFLAGS. Signed-off-by: -
Kir Kolyshkin authored
In GNU make, 'override' means change a variable even if it was set from the make command line (as in "make VAR=value"). As we do not intend to modify CFLAGS/LDFLAGS from the command line, 'override' can be dropped. Signed-off-by:
-
Kir Kolyshkin authored
Brevity is the soul of wit. Signed-off-by:
-
Kir Kolyshkin authored
When optimizing our Travis runs, I noticed that test/zdtm/lib is built 3 times in a row -- first time the binaries, 2nd and 3rd time the deps, in parallel (sic!). I took a look at Makefile and found this, and some more problems, like: - dependencies are generated but not included - cleandep does not depend on clean - groups.c dependencies are not generated (or used) - "make all" does not make groups binary - overcomplicated and unused groups.{pid,out,cleanout} targets - extra -f flag to rm - etc. This patch (hopefully) fixes all of the above. In addition, it provides a few enhancements: - enables "silent make" rules, so now lib/ is in line with the rest of zdtm - adds "dep" target (not used, just for convenience) Signed-off-by: -
Kir Kolyshkin authored
Commit 299e4b4 ("travis: enable ccache for docker/qemu builds") combined multiple RUN statements in Dockerfiles into a single one, which is good for performance (as there is an intermediate cache created after each statement). Unfortunately, it partially did so by combining statements with use of semicolons, meaning we are ignoring non-zero exit codes. As a result, failure from make are ignored and Travis builds are all green. To fix, replace all occurences of semicolon with &&. Signed-off-by: -
Kir Kolyshkin authored
Since commit 93db22b ("compel: no -r for ARM ldflags") and commit ce043d3 ("compel: make plugins .a archives") the generated parasite/restorer blobs for 32-bit ARM are no longer relocatable, and so their ELF header has e_type of ET_EXEC (rather than ET_REL). As a result, we have "Unsupported header detected" message and a failure from compel hgen. This commit allows e_type == ET_EXEC in case NO_RELOCS is defined. I am not sure if it's OK to still allow ET_REL -- maybe Dmitry can chime in on that. Cc: Dmitry Safonov <dsafonov@virtuozzo.com> Signed-off-by: -
Kirill Tkhai authored
revert_unix_sk_cwd() may rewrite errno, so call pr_perror() firstly. Signed-off-by:
Kirill Tkhai <ktkhai@virtuozzo.com> Signed-off-by:
-
Andrei Vagin authored
A static test has to be stopped, because zdtm.py compares file descriptors before and after c/r. Signed-off-by: -
Kir Kolyshkin authored
Commit d9486bd7 ("arm/pie/build: do not produce relocatable parasite object") removed -r from LDFLAGS used to compile criu pie. This functionality somehow never made it to criu-dev, and was also lost in master then compel was ported to it. Make it work with compel. Unfortunately it was not as simple as I initially thought, as -r flag to ld was built into nmk. This patch removes it, and adds it to all places that need intermediate linking. Cc: Dmitry Safonov <dsafonov@virtuozzo.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Acked-by:
Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by:
-
Kir Kolyshkin authored
The objective is to only do parasite code linking once -- when we link parasite objects with compel plugin(s). So, let's use ar (rather than ld) here. This way we'll have a single ld invocation with the proper flags (from compel ldflags) etc. There are two tricks in doing it: 1. The order of objects while linking is important. Therefore, compel plugins should be the last to add to ld command line. 2. Somehow ld doesn't want to include parasite-head.o in the output (probably because no one else references it), so we have to force it in with the modification to our linker scripts. NB: compel makefiles are still a big mess, but I'll get there. Acked-by:
-
Kir Kolyshkin authored
When performing static linking, the order of objects is important (when the linker finds an undefined reference, it looks forward, never back). Therefore, sorting objects breaks things. Required for the next patch. Cc: Cyrill Gorcunov <gorcunov@openvz.org> Acked-by:
-
Kir Kolyshkin authored
1. Remove .FORCE, it's not used. 2. Consolidate CFLAGS stripping into a single line. Acked-by:
-
Kir Kolyshkin authored
The way criu/pie/Makefile is currently written, ld is run twice: 1. link $(NAME-obj-y) objects to NAME.built-in.o 2. link NAME.built-in.o, pie.lib.a, and compel plugins to NAME.built-in.bin.o (with compel ldflags and linker script) There is absolutely no need for such two-stage linking, but it was OK. It is not OK now, as "compel ldflags" for ARM doesn't need -r, and we can't run the first stage with -r and the second stage without it. So, let's simplify linking using a single ld invocation. This is my third attempt in doing it, I think I nailed it this time -- it is now clean and (relatively) simple. While at it: - fix compel linker script dependency (it was not working); - rearrange the Makefile so variables goes first, then rules; - remove a comment about mount implementation in restorer. NOTE that compel is called with ./ prefix so the file paths it prints are also prefixed with ./, which is needed for objectify macro to ignore those. Acked-by: -
Kir Kolyshkin authored
1. Remove useless LDFLAGS -- dynamic linker is not used here. 2. Remove $(iquotes) -- there are none defined here. 3. Consolidate filter-out statements into a single one. Acked-by:
-
Kir Kolyshkin authored
This is an addition to commit 3a7e027b ("criu: pie -- Switch to use compel shipped lds scripts"). Acked-by:
-
Kir Kolyshkin authored
There are too many directories in the include path. Let's remove this one, and refer to the blobs with pie/ prefix. Acked-by:
-
Kir Kolyshkin authored
This is a missing hunk from commit 85b04c8b "Makefiles: nuke $(SRC_DIR)" which has somehow escaped from my attention. Acked-by:
-
Stanislav Kinsburskiy authored
It can be passed as link to real object. Signed-off-by:
Stanislav Kinsburskiy <skinsbursky@virtuozzo.com> Signed-off-by:
-
Kir Kolyshkin authored
Usually we print the PID we wait for, let's do the same here. Probably just a typo. Cc: Kirill Tkhai <ktkhai@virtuozzo.com> Reviewed-by:
-
Kir Kolyshkin authored
Signed-off-by:
-
Kir Kolyshkin authored
I guess this is a followup to commit 59e80d4f. 1 s/determinate/figure out/ (there is no "determinate" verb in English) 2 moved 'with' to the end Cc: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by:
-
Stanislav Kinsburskiy authored
These helpers are valuable and can be used outside. Signed-off-by:
-
Pavel Tikhomirov authored
Before dump: 438 437 0:51 /mtest /zdtm/static/mnt_ext_manual.test rw,relatime - tmpfs zdtm_auto_ext_mnt rw 439 437 0:51 /mtest /zdtm/static/mnt_ext_manual_private_shared_bind.test rw,relatime shared:144 - tmpfs zdtm_auto_ext_mnt rw 440 437 0:51 /mtest /zdtm/static/mnt_ext_manual_bind.test rw,relatime shared:144 - tmpfs zdtm_auto_ext_mnt rw 441 437 0:51 /mtest /zdtm/static/mnt_ext_manual_slave_shared_bind.test rw,relatime shared:145 master:144 - tmpfs zdtm_auto_ext_mnt rw 442 437 0:51 /mtest /zdtm/static/mnt_ext_manual_slave_bind.test rw,relatime master:145 - tmpfs zdtm_auto_ext_mnt rw >From log: mount mnt_ext_manual_slave_bind is restored as slave of mnt_ext_manual_slave_shared_bind, which is restored as slave of mnt_ext_manual_private_shared_bind, which is restored as shared of mnt_ext_manual_bind, which is restored as private bind of external, that is exactly the same way mnt_is_external() looks for external. v2: fix test to cover all cases in mnt_is_external() Signed-off-by:
Pavel Tikhomirov <ptikhomirov@virtuozzo.com> Signed-off-by:
-
Pavel Tikhomirov authored
If container has external bindmount given to criu through --ext-mount-map option by admin, container user can bindmount subdirs of these external bindmount to somewhere else inside container creating secondary external bindmounts. Criu we will fail to restore them as having unreachable sharing. But we can restore secondary external bindmounts bindmounting them from primary external bindmount. v2: s/external_bind/mnt_is_external/, make mnt_is_external bool, do mnt_is_external without recursion v3: add debug message on propagate_mount when bind is set v5: remove hunk of v4 which goes to previous patch, so same as v3 Signed-off-by:
-
Pavel Tikhomirov authored
need it to check if we can bindmount from external mount note: when migrating from criu with patch to criu without, external mount mapping won't work, we do not support it. v2: s/real_root/ext_real_root/ v4: add comment v5: use ext_key field for mapping, put NO_ROOT_MOUNT in root for old externals for which we do not have it. Signed-off-by:
-
Kirill Tkhai authored
"return" and "exit" are mixed in this function, and this is wrong. Must be "return" only, because its callers don't want exit. Signed-off-by:
-
Pavel Tikhomirov authored
mi->external is always false in these places Signed-off-by:
-
Pavel Tikhomirov authored
s/find_widest_shared/find_wider_shared/ Signed-off-by:
-
Kir Kolyshkin authored
"make uninstall" is supposed to remove all the files that "make install" (with the same arguments) have created. This is a test to check that. PS ideally, "make uninstall" should also remove any empty directories, but let's not care about it for now. Signed-off-by:
-
Kir Kolyshkin authored
As we compile-test non-x86_64 architectures under qemu emulation, it works pretty slow. Dmitry Safonov suggested, and Andrey Vagin initially implemented supporting ccache for such builds. This patch is based heavily on Andrey's work -- all the bugs added are purely mine though. Performance results: in an ideal environment (two builds of the same code, one with cold (empty) ccache, another with the hot one) I saw compile time improvements of 4x to 5x, and total test run time improvement up to 2x to 2.5x. In layman terms, the complete test run that was taking more than 50 minutes now takes about 25! Notes on handling .ccache directory: 1. Before running docker build, .ccache directory (saved in between runs by Travis) is moved to criu source code root, from where it is copied by docker together with criu sources. 2. In Dockerfile, .ccache gets moved to /tmp, and CCACHE_DIR is set accordingly. 3. After running docker build, .ccache is copied out from docker container back to the host (from where it is saved by Travis). Ccache envorinment notes: 1. CCACHE_NOCOMPRESS is needed because otherwise tons of time is spent on compression/decompression (big performance hit under qemu). 2. CCACHE_CPP2 is required with clang, see detailed explanation at http://petereisentraut.blogspot.com/2011/09/ccache-and-clang-part-2.html The logic of setting CCACHE_CPP2 in Dockerfile is somewhat fancy; unfortunately I was not able to come up with a simpler approach. Misc: 1. Travis runs "ccache -s" after the build is completed. A call to "ccache -s" is called to Dockerfile before make, so one can compare before/after numbers. 2. make invocations are surrounded by "date" calls so one can get the compilation times. Changes in v2: - consolidate Dockerfile statements (ENV, RUN) - single object make test is no longer commented out - simplify "make mrproper" test Signed-off-by:
Dmitry Safonov <dsafonov@virtuozzo.com> Signed-off-by:
-
Kir Kolyshkin authored
For compatibility with ccache (so it can cache the compilation results), let's compile and link separately. For this, we have to - disable the implicit make rules - write the explicit ones While at it, do use the "silent make" stuff that is already here FIXME: figure out if it helps to speed up ccache build Signed-off-by:
-
Kir Kolyshkin authored
For now, it's done for local builds only (i.e. no per-arch Docker builds yet). The reason is, it's easier to play with ccache when the compiles are (relatively) fast. Performance: there is 2x to 3x improvement in build speeds for ideal cases (same code, hot cache), but the absolute savings are negligible (example: 7 seconds instead of 23). Note that ccache is not compatible with gcov, so we only enable it for non-gcov build (which happens to be the one with clang). Signed-off-by:
-
Kir Kolyshkin authored
Unify alpine and non-alpine builds. The only difference is foreign arch builds need some preparation -- separate that to a dependency. Unfortunately we can't use wildcard targets ("%: ") as non-wildcard ones are prevaling. Therefore, a somewhat ugly hack to generate $arch: Dockerfile.$arch dependency is added. While at it: - rename DB_ARGS to DB_CC - mark clean as phony - make "all" really run all the builds we can Nice side effect: autocompletion ("make -C scripts/build <TAB>") now works! Signed-off-by:
-
