The test for fpu transition will require to do
a runtime check for the cpu features it's running
on. For this sake we need to use cpuid. Thus make
it more widely available by providing in the general
header.
The code is adopted from the linux kernel code.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Since at moment we stick with sigreturn restore
we need to form a proper FPU frame and set a pointer
to it inside sigreturn frame.
For this sake we read the FPU image and here are two
cases are possible
- no fpu data at all -- nothing to restore, simpliest
case
- xsave frame is present but the host cpu supports only
fxsave instruction: we refuse to continue, since it means
there are no ymm registers on the machine where we're trying
to restore
- fxsave frame is present but the host cpu has xsave feature:
at moment we refuse to continue, requiring complete match
between "checkpoint and restore hosts", but in real we could
extend logic and form complete xsave frame from fxsave and
continue processing
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
For code simplicity we reserve the maximum size which
might be needed to form an FPU frame (ie for both
xsave and fxsave operations).
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The dumping of FPU state is done with help of ptrace
facility. There are two cases which we need to handle
depending on which features are available on host machine
1) The dump via ptrace(PTRACE_GETFPREGS ...)
In this case the kernel will use fxsave approach
inside the kenrel and provides us back the data
encoded in i387_fxsave_struct format.
2) The dump via ptrace(PTRACE_GETREGSET ...)
In this case the kernel will use xsave approach
inside the kernel and provides us back the data
encoded in xsave_struct format.
In any case we decode data and save it in protobuf format.
This is why core.proto file has been extended to keep new
entries.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
And don't forget to undef them once they are not needed.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Actually it was never used, just drop it.
Because of backward compatibility problem we
can't just zap it in protofile.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The cpu we're running on must at least support fxsave feature.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This patch add ability to test /proc/cpuinfo data
we're interested in at the moment.
The code provides the following functionality
- cpu_init, to parse cpuinfo and check if the
host cpu we're running on is suitable enough
for FPU checkpoint/restore. If FPU present then
there must be at least fxsave capability present
- cpu_set_feature/cpu_has_feature helpers which
provides to test certain bits and set them where
needed (we need to set bits when parse cpuinfo)
Note, we reserve space for all cpuinfo bits known
by the kernel at moment, while use only three FPU
related bits for a while. This is done because we might
need to use or find out other features in future.
After all it's just 40 bytes of memory needed to keep
all possible bits.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We will need these structures for restore FPU
state via sigframe, as as we decode data provided
by ptrace.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Wrong size was used in mapping and unmapping. But since
they are more-less close to each other, nobody noticed it.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Syntax is
crtools exec -t <pid> <syscall_name> [<arguments>]
Two types of arguments are supported -- plain, treated
as number, passed as is to the syscall. Buffer, started
with '&' -- the rest of the string is pushed to the tgt
task's memory and pointer to one is passed as syscall
argument.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The fist part prepares ctl to controll the seized task.
The 2nd one mmaps shareb buffer for data exchange.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This is mostly bugfix and improvements release.
Nonetheless, some new features exists, the most interesting are:
* proper COW mappings handling
* full packet sockets support, thus supporting the tcpdump tool
* the --shell-job option, which makes it possible to dump apps
launched from one shell and restore them in another
Some features are available with the custome kernel, but the good
news is that now _all_ of the patches we need are in one of the
-next trees or in the -mm one, and thus have good chances to get
merged in 3.8 (or soon after it).
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We need to test every even index (which represents a slave
peer) thus use modulo operation here.
Reported-by: Alexander Kartashov <alekskartashov@parallels.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
We have 3 calls for that, two called directly from cr-restore,
one -- from one the former. Clean this mess and write short
comments on each stage.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
This thing on pstre_item was created to carry task-specific
information across the "restore" code-flow.
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The watch target might be a deleted file. In this case we can't
re-create it on restore procedure.
A typical scenario
------------------
fd = inotify_init1()
wd = open(path)
inotify_add_watch(path)
unlink(path)
... checkpoint ...
here we have a @path which is unlinked but still present in inotify
watch list because inode is not yet freed. And if the program is killed
after checkpoint the last reference to a path get eliminated and inode
get freed from the kernel memory.
Thus any furher attempts to open the path via file handle (note that file
handle can't be used to create new file, the kernel doesn't permit that)
will simply fail.
So if plain opening via file handle failed we assume that at least
the file might be in our ghost files list (we're optimists, right?)
and if we successed we simply use a path generated by ghost file
engine for us.
v2 [by xemul@]:
- gather reference to ghost file remap early
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Because we need to lookup for ghost files from
inotify system where we only have device/inode
as a key, we save dev/ino in ghost image entry.
Note we use in-kernel format for device to be
consistent with inotify and mount related
code base.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
The remap_put will be needed to defer unlinking of ghost files
if they are referred from inotify system.
The lookup_remap is needed to figure out if the watch
target the inotify has present in ghost files list.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
Since fdinfo patches were merged to -mm
tree the output format has been slightly
changed. So update our tool accordingly.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Pavel Emelyanov <xemul@parallels.com>
