An example, that steals stderr descriptor from victim task.
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This is just export by reasonable name of the existing
code, that sends and receives FDs via compel RPC socket.
v2:
Rebase on recent criu-dev
Fix parallel build
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
CID 73371 (#1 of 1): Big parameter passed by value (PASS_BY_VALUE)
pass_by_value: Passing parameter regs of type user_regs_struct_t
(size 224 bytes) by value.
Suggesting to do this until compel is released and API is cut in stone.
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Found by Coverity error:
> CID 172193 (#1 of 1): Bad bit shift operation (BAD_SHIFT)
> 1. large_shift: In expression 1 << sig % 64, left shifting
> by more than 31 bits has undefined behavior. The shift amount,
> sig % 64, is as much as 63.
That is:
1. yes, UB
2. while adding a signal to mask, this has flushed all other
signals from mask.
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The purpose is to unblock previously blocked SIGCHLD, not all the
signals. IIRC, this block-unblock dancing is about a race where SIGCHLD
may come during switching to restorer blob, where CRIU handler function
is already unmapped - which will result in segmentation violation if
SIGCHLD wasn't blocked.
Fixes:3a61c38a5832 ("restore: block sigchld to remap task_entries")
Cc: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The statement like
#include <compel/plugins/plugin-std.h>
looks a bit tautological. I think the single "plugins" word is enough:
#include <compel/plugins/std.h>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
First, for building compel plugins, we already have
"-I compel/include/uapi" in ccflags and asflags, so there is
no need to add "-iquote include/uapi".
Second, let's refer to compel plugin uapi includes in a uniform way,
choosing the same way the external code does, i.e. #include <compel/...>.
Third, in a few cases simplify #include statements by including
compel/plugins/plugin-std.h instead of a number of plugins/std/*.h files.
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This header should have everything we have in std/
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
* install
'cp' can copy recursively, create directories, and even
dereference symlinks. Everything we have in uapi/ is to be installed.
NOTE we can't use -r for compel includes, as there is some extra stuff
in there we don't want to take with us (in particular, plugins/
and 'compel -> .' symlinks).
* uninstall
rm -rf everything under compel includedir
While at it, fix some minor things here and there.
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It is not used anywhere, so unless someone has any plans, let's kill it.
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
A slash after a directory was missing in a number of places, that
resulted in "make uninstall" leaving a number of files behind.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
GNU make tries to rebuild any makefiles it uses. While in general it's
a good idea (and it is used e.g. in autoconf-based builds), in our case
it is not necessary, as all the makefiles are static.
More to say, as we have a few "match anything" rules for subdirectories,
Makefiles in these subdirs are also matching these rules, which leads to
excessive (re)building while a particular makefile is needed.
Protect such Makefiles with explicit (or pattern) rules, so make knows
it should do nothing to rebuild those.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Apparently @true was used as a rule for any target which should have
a non-empty rule doing nothing. For this, there's a "<target>: ;"
syntax, let's use this and eliminate an unnecessary fork/exec :)
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Mostly this is done in order to not repeat the recursive clean
lines twice.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
As all builds are done from top source dir, there is no need
to have SRC_DIR.
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Our whole system of Makefiles are integrated from top to bottom,
meaning:
1. The paths in sub-makefiles are relative to the top source dir.
2. Sub-makefiles are executed via make $(build)=<SUBDIR> <TARGET>
For some reason, makefiles under lib/ are the exclusion. Let's fix it.
Side effect: you can now build any individual target under lib/,
for example, "make lib/c/libcriu.so" works.
[v2: use the .FORCE, thanks to dsafonov@]
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
As compel is an external tool now, let's treat it as such -- remove the
hardcoded paths to compel files, they are now provided via compel CLI.
Details:
1. The compel std plugin dependency is dropped from the intermediate
criu/Makefile. It was there in order to make sure plugin is built before
pie, but the top level Makefile and Makefile.compel take care about it.
2. The compel std plugin logic is simplified in criu/pie/Makefile.
3. The compel linker script dependency is kept as-is for now (except
the linker script file name no longer appears in the list of objects
to link).
4. Linking of compel std plugin is moved to a second linking stage
in order to simplify the Makefile.
Side note: we can actually avoid running the linker twice!
5. The compel binary dependency is dropped from criu/pie/Makefile
as this is already there in top-level Makefile (look for criu-deps).
[v2: reworked after dsafonov@ comments; keep most compel deps]
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
To check that jump trampolines to rt-vdso works.
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Reduce __export_restore_task(), which is fat and too long,
drop needless parameter from vdso_proxify() and drop also
that find-tuple logic, which isn't needed really.
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It's always "dumpee" and is used in two pr_err() messages,
put it right there.
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
We should log whom we're remapping, but ~5 lines upper or so
we've unmapped dumpee vdso and vvar. Here we're mremapping runtime
host-provided vdso/vvar blobs to the dumpee position.
Correct the messages to reflect that we've remapped rt-vdso/vvar,
having the same naming as in vdso_do_park().
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
For 32-bit at this moment there are follwing entries:
o __vdso_clock_gettime
o __vdso_gettimeofday
o __vdso_time
o __kernel_vsyscall
o __kernel_sigreturn
o __kernel_rt_sigreturn
So, there isn't __vdso_getcpu(), which is present in 64-bit vDSO,
and 64-bit vDSO doesn't have those __kernel_*.
This is fine as two vdso blobs with the same not present symbols
are considered to be the same in comparison.
I didn't introduce ARCH_VDSO_SYMBOLS_32, as it would have different
size and that will result in quite painful conversion of struct
vdso_symtable (it's fixed size and e.g., inside parasite's parameters)
Which is not needed by the described behavior of vdso blobs comparison.
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Forward CRIU's log fd into helper, so it can actually print errors.
This will help to debug troubles with helper, like #273.
Log fd can be safely written to, as it's opened in CRIU with
O_APPEND or it's line-buffered (stdout) or not buffered (stderr)
stream. Anyway, it's double safe, as there is simple synchronization
between helper and CRIU, so only one of them will log at a time.
Also expanded the comment about a helper.
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This macro is not in use since mid-2012 (commit 4806e13
"protobuf: Convert vma_entry to PB format v3"), so I guess
it is about time to retire it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Let's rename the printing functions so their names look more like
the standard ones.
1. putc/puts with a file descriptor.
__std_putc -> std_dputc
__std_puts -> std_dputs
There are no standard putc/puts that accept fd as an argument,
but the libc convention is to use d prefix for such. Therefore:
NOTE we keep the order of the arguments intact, to be in line
with the rest of the functions.
2. *printf
__std_printk -> std_vdprintf
__std_printf -> std_dprintf
The reason is, these are the names of libc functions with similar
functionality/arguments.
Cc: Dmitry Safonov <dsafonov@virtuozzo.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This function works like printf, and it helps the compiler
to know that, so it can check whether arguments fit the
format string.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This function works like printf, and it helps the compiler
to know that, so it can check whether arguments fit the
format string.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
These are part of compel UAPI so should be prefixed with COMPEL_
in order to not pollute the namespace. While at it, move from
set of defines to an enum, which looks a bit cleaner.
Also, kill LOG_UNDEF as it's not used anywhere.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Those macros look twice as long as they should be on my 80-columns
terminal. As there is nothing here to justify such width, go ahead
and remove the extra tabs, keeping the code within 80 cols.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
We support compat mode now (at least partially) so
the test should be allowed.
Acked-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
To support ia32 compat mode on x86-64 we need to things
- extend fpu_state_t type to carry ia32 specifics
- fill up additional members in fpu_state_ia32_t type
before calling sigreturn (this also requires the
uc_mcontext::fpstate won't be aligned on 32 bytes)
Because we touches base types in compel the criu
has been updated accordingly.
Acked-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
In compat mode the fpu frame won't be aligned on 32 bit
due to kernel specifics.
Acked-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It was never functional neither we plan to support
native ia32 mode, so drop these incomplete code
pieces out.
- Presumably we will need TASK_SIZE for compat
mode so I provide TASK_SIZE_IA32 for this sake
- 32 bit syscalls are remaining for a while
Acked-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It's now obsoleted by compel library.
Maybe-TODO: Add compel tool exec action?
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This is the difference between two commits
criu-dev/b0f6f293/Unify own memcpy/memset/memcmp
master/0367a1fe/Drop prefix from own memcpy/memset/memcmp
that makes criu-dev after rebase on master with latter commit
be the same as it was with former commit before rebase.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
First, TASK_* defines provided by compel should be prefixed
with COMPEL_. The complication is, same constants are also used
by CRIU, some are even writted into images (meaning we should
not change their values).
One way to solve this would be to untie compel values from CRIU ones,
using some mapping between the two sets when needed (i.e. in calls to
compel_wait_task() and compel_resume_task()).
Fortunately, we can avoid implementing this mapping by separating
the ranges used by compel and criu. With this patch, compel is using
values in range 0x01..0x7f, and criu is reusing those, plus adding
more values in range 0x80..0xff for its own purposes.
Note tha the values that are used inside images are not changed
(as, luckily, they were all used by compel).
travis-ci: success for compel uapi cleanups (rev2)
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
infect.h includes compel.h, and compel.h includes infect.h.
Surely, due to include guards it will be sorted out, but
we'd rather just include what we need.
travis-ci: success for compel uapi cleanups
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
We have ptrace defines and functions that are part of UAPI,
and we have some internal stuff not to be exposed. Split
ptrace.h into two files accordingly.
While at it, do some cleanups:
- add ptrace_ prefix to some functions and macros
- remove (duplicated) PTRACE_* defines from .c files
- rename ptrace_seccomp(), remove its duplicate
- remove unused ptrace defines
- remove unneeded (ptrace-related) includes
travis-ci: success for compel uapi cleanups
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Reviewed-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
To run CRIU with ASan we have to use some non-default options:
- detect_leaks=0 - We have to many leaks for know. Let's disable until fixed.
- disable_coredump=0 - without this ASan library changes RLIMIT_CORE which
break rlmimits00 tests.
- log_path=asan.log - For some reason default output to stderr sometimes
doesn't work in CRIU. So error log will be stored
in asan.log.<pid> file instead.
travis-ci: success for series starting with [1/6] compel/infect: fix out-of-bounds parasite memcpy()
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
ASan doesn't play nicely with clone if we use current stack for
child task. ASan puts local variables on the fake stack
to catch use-after-return bug:
https://github.com/google/sanitizers/wiki/AddressSanitizerUseAfterReturn#algorithm
So it's become easy to overflow this fake stack frame in cloned child.
We need a real stack for clone().
To workaround this we add clone_noasan() not-instrumented wrapper for
clone(). Unfortunately we can't use __attrbute__((no_sanitize_addresss))
for this because of bug in GCC > 6:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69863
So the only way is to put this wrapper in separate non-instrumented file.
travis-ci: success for series starting with [1/6] compel/infect: fix out-of-bounds parasite memcpy()
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Dmitry Safonov <dsafonov@virtuozzo.com>
Signed-off-by: Pavel Emelyanov <xemul@virtuozzo.com>
