Currently, install_service_fd() always creates a new file descriptor,
but it is going to be reworked to reuse a file descriptor when it is
possible.
Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
For a long time we've been demanding from cpus to be compatible
on fpu frame level, but growing list of cpu features triggers
inability for restored programs to proceed after restore due
to specific intructions execution (such as avx2). Note the
fpu frame may carry same size but not on instruction level
where SIGILL may happen after the restore.
Thus lets require instruction mode to be set and verified by default.
Still one can drop this option via command line or rpc request.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmail.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Since we're using bintmaps for options don't exit early
if --cpu-cap=ins specified because there might be a
combination of options.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmail.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Since now validation is bit based there is no longer need to
exit early if fpu matches, we rather need over all possible
options where several capability modes can be set.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmaill.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
To preserve backward compatibility because
strict mode now lays under own block.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmaill.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Current scheme of when to write and verify cpuinfo
images is unclear: we define default mode to 'fpu'
and this implies that no cpuinfo image get written
on dump and read on restore. In turn if 'cpu' or
'ins' mode specified in command line we write and
verify this image.
This is all wrong -- we need to produce image if
any mode ecxept 'none' is specified. The 'none'
mode is designed exactly for skipping cpuinfo
image production or verification.
Because any non-default mode requires explicit
--cpu-cap option to be passed we can use this
and setup new CPU_CAP_IMAGE telling criu to
write or read this image.
Note the default mode doesn't break backward
compatibility because by default we don't
write or read the image at all.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmaill.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It should be there from the beginning but happen to escape.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmaill.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
xsave sizes should sit in fpu capability block because
instruction and strick mode checkig is a different
thing.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmaill.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Usually people simply leave cpu checkin in default
mode (which is fpu level) but idea was to be able
to compose a mixture of settings, for this sake
CPU_CAP_ constants are bit shifts. Thus use bitwise
operator for this.
Same time define CPU_CAP_ as bit shifts explicitly
and use explicit CPU_CAP_NONE compare where needed.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmaill.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Keep xsave sizes in image to be sure that on restore the application
won't override memory out of xsave frame size if been calling xsave
instruction directly.
Here are some details: while been testing vz7 containers migration
we've noticed that sometime applications do crash after restore,
what is worse such execution abort may happen not immediately
after the restore but after passing some time. After spending
a lot of time we discovered that it is due to the fact that
the migration is directed from an old cpu to a modern one
which has extensions such as mpx. In result libc has cached
small xsave size and then after restore any direct call to
xsave instruction overwrite memory which is allocated with
size less than needed.
Thus we save xsave frame size in image and require it to
match to prevent such situation.
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Dmitry Safonov <0x7f454c46@gmaill.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Before fix issubpath("./zdtm/static/mntns_shared_vs_private.test",
"./") returned false, though should return true.
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
colon operators are there obviously by mistake instead of semicolons
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
As mnt_roots always has a slash at the beginning we don't need another:
(00.156035) 1: mnt: Move the root to //tmp/.criu.mntns.ZKVJHl/15-0000000000
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
for i in $(find . -name "*.[ch]" -type f); do
sed -i 's/\(pr_warn(".*[^n]\)\("[),]\)/\1\\n\2/' $i
done
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
In RPC mode the log_level was always reset to the default log level of 2, if no
configuration file was used.
This saves the log level set via RPC in opts.log_level and a configuration file
can overwrite it later, but if it is only set via RPC this value is not ignored
now.
Signed-off-by: Adrian Reber <areber@redhat.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
The -c and -f options were introduced on December 2011 with
commit 485c4b50d1
show: Rework show procedure
Added
- ability to parse every single file crtools
understands via magic number
- dumped memory pages can be rendered in two
modes -- as a simple set of pages' virtual addresses
or in hexdummp fashion (via -c command line key).
For example
./crtools -s -c -f pages-2557.img
CR_FD_PAGES: pages-2557.img
----------------------------------------
7fffb44ace20: 00 00 00 00 00 00 00 00 00 00 00 74 65 73 74 2f |. . . . . . . . . . . t e s t / |
7fffb44ace30: 74 65 73 74 2d 73 69 67 61 63 74 69 6f 6e 00 48 |t e s t - s i g a c t i o n . H |
7fffb44ace40: 4f 53 54 4e 41 4d 45 3d 6c 6f 63 61 6c 68 6f 73 |O S T N A M E = l o c a l h o s |
7fffb44ace50: 74 2e 6c 6f 63 61 6c 64 6f 6d 61 69 6e 00 54 45 |t . l o c a l d o m a i n . T E |
The -p/--pid option was added on April 2013 with
commit 73d7e283a7
show: Filter -D output with --pid option
Show only image info relevant to given pid. Also reuse
the introduced --pid argument for exec action.
And the -F option was introduced on December 2013 with
commit 7ab8a3261b
show: Implement simple images filtering
The -F|--fields option specifies which fields (by name, comma
separated) should be printed.
For nested fields all names in path should be specified.
All four options were used in combination with the deprecated --show
command which has been decommissioned on February 2016 with
commit a501b4804b
show: Remove deprecated code
The 'show' action has been deprecated since 1.6, let's finally drop it.
The print_data() routine is kept for yet another (to be deprecated too)
feature called 'criu exec'.
Signed-off-by: Radostin Stoyanov <rstoyanov1@gmail.com>
Acked-by: Adrian Reber <areber@redhat.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
>>> CID 164715: (BUFFER_SIZE_WARNING)
>>> Calling strncpy with a maximum size argument of 16 bytes on
destination array "thread_args[i].comm" of size 16 bytes might
leave the destination string unterminated.
3473 strncpy(thread_args[i].comm, core->tc->comm, TASK_COMM_LEN);
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
Acked-by: Cyrill Gorcunov <gorcunov@gmail.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
When zdtm.py is executed with `list` sub-command the 'criu_bin'
option is not defined and criu.available() fails.
$ python test/zdtm.py list
Traceback (most recent call last):
File "test/zdtm.py", line 2243, in <module>
criu.available()
File "test/zdtm.py", line 1185, in available
if not os.access(opts['criu_bin'], os.X_OK):
KeyError: u'criu_bin'
However, we don't need to check the existence of criu_bin
unless we use the `run` action.
Signed-off-by: Radostin Stoyanov <rstoyanov1@gmail.com>
Acked-by: Adrian Reber <areber@redhat.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
It is especially important when accessing a hash --
there must be no negative indices ever.
https://jira.sw.ru/browse/PSBM-82945
Signed-off-by: Cyrill Gorcunov <gorcunov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
In commit 23b769498c ("dump: optimize dead_pid_conflict by searching
in rbtree") we'd replaced for_each_pstree_item and thread walk search
with pstree_pid_by_virt (rb-tree search). The latter already found us
the thread with conflicting pid if it exists, the only thing left is to
skip if the thread is also a main thread of the thread group (as it was
before patch). But some leftover left which checks something wrong: we
index ->threads array with "i", but "i" is not a number of the thread
it is a number of current dead pid.
Not sure if it helps with the initial bug and duplicates, but it might:
https://jira.sw.ru/browse/PSBM-55217
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
This can happen when running in user namespace with auto-dedup enabled.
Right now this means auto-dedup gets disabled.
Signed-off-by: Pawel Stradomski <pstradomski@google.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
call variables of type RemapFilePathEntry - "rpe" everywhere, similar as
we already name them in oher places
while on it remove unused second argument of open_remap_linked
Signed-off-by: Pavel Tikhomirov <ptikhomirov@virtuozzo.com>
Signed-off-by: Andrei Vagin <avagin@virtuozzo.com>
