First of all, this should be done strictly after we've stopped accessing
files by their paths, even absolute. This place is right before going
into restorer.
And the second thing is that we want to re-use the open_fd_by_id engine,
since it handles various tricky cases of open-file-by-path. And since
there's no such thing as fchroot(int fd), we emulate it using the
/proc/self/fd/ links.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
For the root task the clone syscall returns the pid in criu's pidns,
but for other processes the clone syscall returns PID in the restored
namespace.
The /proc/self link contains the PID value of the current process, so if
we want to determing the PID in a criu's pidns, we should use criu's
/proc.
v2: readlink() does not append a null byte to buf, so we must do that
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Currently pos has type unsigned long, so its size depends on
architecture. pos is saved as 64-bit value in the image file and it
isn't restored, if it is equal to -1. Due to convertation on 32-bit
platforms -1 is converted into UINT_MAX and we get error on restore.
$ zdtm.sh ns/static/tun
...
(00.398513) 5: Error (files-reg.c:534): Can't restore file pos: Illegal seek
(00.398888) 5: Error (files-reg.c:489): Can't open file /dev/net/tun: Illegal seek
...
id: 0x15 flags: 0x2 pos: 0x000000ffffffff fown: { uid: 0 euid: 0 signum: 0 pid_type: 0 pid: 0 } name: "/dev/net/tun"
crtools is compiled with _FILE_OFFSET_BITS=64, so off_t is always 64-bit.
Signed-off-by: Andrew Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The recv-fd stage is only required if we created any transport fd.
The post-open one is only required if there's at least one fdtype
with post-open callback in ops.
Both things can be evaluated in advance and unneeded stages can
be skipped.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The major issue with dump is -- some info id get via netlink,
some via sysfs and some (!) via opened and attached tun file.
But the latter cannot be created, if there's another one attached
(or the mq device is full with threads).
Thus we have to dump this info via existing tun file and keep one
in memory till the link dump code takes place.
Opposite situation is also possible -- we can have a persistent
unattached device. In this case we have to attach to it, dump
things and detach back.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
CID 1042292 (#1 of 1): Unsigned compared against 0 (NO_EFFECT)
unsigned_compare: This less-than-zero comparison of an unsigned value is
never true. "len < 0UL"
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Since *all* of them just call do_dump_gen_file with proper ops,
just call one directly. Compacts the code.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Based on work done by Cyrill Corcunov (many thanks for that).
In this commit we implement c/r for files which have opened
/proc/$pid/ns/$ids entries.
The idea is rather simple one
Checkpoint
==========
- Check if the file name is the one of known to be ns ref
- If match then write protobuf entry
Restore
=======
- Read all ns entries from the image
- When criu tries to open one we lookup over process
tree to figure out which PID should be used in path
and then just open it
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
To be able to dump procfs namespace entries we will need to
analyze the link name first and then figure out if the file
referred indeed a procfs entry or it's plain regular file.
This means we read link early, and to escape double reading
of same link, once we read it we remember it in fd_parms
structure which we pass to dump_one_reg_file.
Still the dump_one_reg_file is not solely used from inside
of dump_one_file, but from a number of other places (fifo,
special files) where fd_parms is filled only partially
and fill_fd_params is not even called. Thus dump_one_reg_file
must be ready for such scenario and read the link name by own
if the caller has not provided one.
To achieve all this we do
- extend fd_parms structure to have a reference on a new
fd_link structure, thus if caller already read the link
name it might assign the reference and call for dump_one_reg_file
- tune up dump_one_reg_file to fill own copy of fd_link
structure if the caller has not provied one
[ xemul: Added const to fill_fdlink arg ]
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We usually print resources IDs in hex format,
so fix the places where we don't (it confuses
otherwise).
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Currenly crtools supports a case when a child shared a fd table
with parent.
Here is only two interesting things.
* Service descriptors should be cloned for each process
who shared one fd table.
* One task should restore files and other tasks should sleep in this
* time.
v2: * allocate fdt_lock from shared memory
* don't wait a child, if it doesn't share fdtable
v3: * don't move ids on the pstree image
v4: * save ids in a separate image
* save fdinfo per id instead of pid
v5: fix alignment of service_fd_id
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
A few processes can share one fd table. Each process has own set of
service file descriptors and a process knows nothing about servic fds
of another processes. So if two process share one fd table,
close_old_fds will close servic descriptors of another process.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Many image files opened by open_image_ro weren't closed before return, fix
them all in this patch.
Signed-off-by: Huang Qiang <h.huangqiang@huawei.com>
Acked-by: Andrew Vagin <avagin@parallels.com>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We will need it for slave ttys migration. They serve for one purpose --
to clone self stdio descriptor and use it with tty layer, which will
be addressed in further patches.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It reads /proc/PID/fd and close all descriptors except service fds.
v2: s/is_one_of_service_fds/is_any_service_fd
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Don't do explicit switch by file type in files.c and don't export
intimate knowledge of pty being master/slave. Use a file desc op
for that.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The idea behind is pretty simple -- once we find
that there is a controlling terminal present we
do call ioctl on appropriate /dev/pts/N.
This is done in a bit unusuall manner. When we
find that there is a controling terminal present
we do create an additional FdinfoEntry for it
with object id taken from existing master peer.
The file engine stack this new FdinfoEntry on
fd_info_head head list. Thus we will have at
least two entries on this list. One for real
Fdinfo associated with master peer and one for
our new generated Fdfinfo entry, it depends on
pid which one become a file master.
Finally we do use post_open_fd hook in our
tty code which allows us to open controlling
terminal and yield proper ioctl on it.
v2:
- restore control terminals via service fd,
still need to speedup service fd retrieval.
v3:
- use prepare_ctl_tty() helper to generate
control terminal fdinfo entry
v4:
- use post_open_fd
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Usually the PTYs represent a pair of links -- master peer and slave
peer. Master peer must be opened before slave. Internally, when kernel
creates master peer it also generates a slave interface in a form of
/dev/pts/N, where N is that named pty "index". Master/slave connection
unambiguously identified by this index.
Still, one master can carry multiple slaves -- for example a user opens
one master via /dev/ptmx and appropriate /dev/pts/N in sequence.
The result will be the following
master
`- slave 1
`- slave 2
both slave will have same master index but different file descriptors.
Still inside the kernel pty parameters are same for both slaves. Thus
only one slave parameters should be restored, there is no need to carry
all parameters for every slave peer we've found.
Not yet addressed problems:
- At moment of restore the master peer might be already closed for
any reason so to resolve such problem we need to open a fake master
peer with proper index and hook a slave on it, then we close
master peer.
- Need to figure out how to deal with ttys which have some
data in buffers not yet flushed, at moment this data will
be simply lost during c/r
- Need to restore control terminals
- Need to fetch tty flags such as exclusive/packet-mode,
this can't be done without kernel patching
[ avagin@:
- ideas on contol terminals restore
- overall code redesign and simplification
]
v4:
- drop redundant pid from dump_chrdev
- make sure optional fown is passed on regular ptys
- add a comments about zeroifying termios
- get rid of redundant empty line in files.c
v5 (by avagin@):
- complete rework of tty image format, now we have
two files -- tty.img and tty-info.img. The idea
behind to reduce data being stored.
v6 (by xemul@):
- packet mode should be set to true in image,
until properly fetched from the kernel
- verify image data on retrieval
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
CC: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
When we will be restoring ttys we need that restore
procedure for master peers will be yielded earlier
than for slave peers due to ttys specifics. With this
commit we introduce @tty_slaves list which will allow
us to order tty file restore procesure.
Because we need to fetch which list to be used depending
on tty type this patch extend select_ps_list with fdinfo_list_entry
parameter.
v2 (by xemul@):
- make sure the epoll list is still last
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
It will be used for restoring epollfd.
Currently a transport fd may be added to epollfd.
epollfd should be populated, when all descriptors were already received.
Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
