[RFC v7][PATCH 0/9] Kernel based checkpoint/restart

[Oren Laadan]

I'm re-posting this with a couple of fixes: one fixes the FPU
save/restore and another fixes argument to kunmap_atomic().

There is now a git tree at:
	git://gorgona.ncl.cs.columbia.edu/pub/git/linux-cr.git

And another tree to track development and older versions:
	git://gorgona.ncl.cs.columbia.edu/pub/git/linux-cr-dev.git

I could not have told it better than Dave Hansen:

--

I'd like to see these merged into -mm and on the way to mainline.  The
entire freakin' world is cc'd.  So sue me.  :) 

Why do we want it?  It allows containers to be moved between physical
machines' kernels in the same way that VMWare can move VMs between
physical machines' hypervisors.  There are currently at least two
out-of-tree implementations of this in the commercial world (IBM's
Metacluster and Parallels' OpenVZ/Virtuozzo) and several in the academic
world like Zap.

Why do we need it in mainline now?  Because we already have plenty of
out-of-tree ones, and  want to know what an in-tree one will be like.  :) 
What *I* want right now is the extra review and scrutiny that comes with
a mainline submission to make sure we're not going in a direction
contrary to the community.

This only supports pretty simple apps.  But, I trust Ingo when he says:
> > Generally, if something works for simple apps already (in a robust, 
> > compatible and supportable way) and users find it "very cool", then 
> > support for more complex apps is not far in the future.  but if you
> > want to support more complex apps straight away, it takes forever and
> > gets ugly.

We're *certainly* going to be changing the ABI (which is the format of
the checkpoint).  I'd like to follow the model that we used for
ext4-dev, which is to make it very clear that this is a development-only
feature for now.  Perhaps we do that by making the interface only
available through debugfs or something similar for now.  Or, reserving
the syscall numbers but require some runtime switch to be thrown before
they can be used.  I'm open to suggestions here.

These patches are Oren Laadan's baby.  Virtually all this code is his,
but he's a bit busy at the moment finishing up his PhD.  

There's a plethora of old history and some userspace tools below if you
want some more detail, but please ignore them and look at the kernel
code.  :) 

--

These patches implement basic checkpoint-restart [CR]. This version
(v7) supports basic tasks with simple private memory, and open files
(regular files and directories only). See original announcements below.

Oren.

--
Todo:
- Add support for x86-64 and improve ABI
- Refine or change syscall interface
- Extend to handle (multiple) tasks in a container
- Handle multiple namespaces in a container (e.g. save the filesystem
  namespaces state with the file descriptors)
- Security (without CAPS_SYS_ADMIN files restore may fail)

Changelog:

[2008-Oct-17] v7:
  - Fix save/restore state of FPU
  - Fix argument given to kunmap_atomic() in memory dump/restore

[2008-Oct-07] v6:
  - Balance all calls to cr_hbuf_get() with matching cr_hbuf_put()
    (even though it's not really needed)
  - Add assumptions and what's-missing to documentation
  - Misc fixes and cleanups

[2008-Sep-11] v5:
  - Config is now 'def_bool n' by default
  - Improve memory dump/restore code (following Dave Hansen's comments)
  - Change dump format (and code) to allow chunks of <vaddrs, pages>
    instead of one long list of each
  - Fix use of follow_page() to avoid faulting in non-present pages
  - Memory restore now maps user pages explicitly to copy data into them,
    instead of reading directly to user space; got rid of mprotect_fixup()
  - Remove preempt_disable() when restoring debug registers
  - Rename headers files s/ckpt/checkpoint/
  - Fix misc bugs in files dump/restore
  - Fixes and cleanups on some error paths
  - Fix misc coding style

[2008-Sep-09] v4:
  - Various fixes and clean-ups
  - Fix calculation of hash table size
  - Fix header structure alignment
  - Use stand list_... for cr_pgarr

[2008-Aug-29] v3:
  - Various fixes and clean-ups
  - Use standard hlist_... for hash table
  - Better use of standard kmalloc/kfree

[2008-Aug-20] v2:
  - Added Dump and restore of open files (regular and directories)
  - Added basic handling of shared objects, and improve handling of
    'parent tag' concept
  - Added documentation
  - Improved ABI, 64bit padding for image data
  - Improved locking when saving/restoring memory
  - Added UTS information to header (release, version, machine)
  - Cleanup extraction of filename from a file pointer
  - Refactor to allow easier reviewing
  - Remove requirement for CAPS_SYS_ADMIN until we come up with a
    security policy (this means that file restore may fail)
  - Other cleanup and response to comments for v1

[2008-Jul-29] v1:
  - Initial version: support a single task with address space of only
    private anonymous or file-mapped VMAs; syscalls ignore pid/crid
    argument and act on current process.

--
(Dave Hansen's announcement)

At the containers mini-conference before OLS, the consensus among
all the stakeholders was that doing checkpoint/restart in the kernel
as much as possible was the best approach.  With this approach, the
kernel will export a relatively opaque 'blob' of data to userspace
which can then be handed to the new kernel at restore time.

This is different than what had been proposed before, which was
that a userspace application would be responsible for collecting
all of this data.  We were also planning on adding lots of new,
little kernel interfaces for all of the things that needed
checkpointing.  This unites those into a single, grand interface.

The 'blob' will contain copies of select portions of kernel
structures such as vmas and mm_structs.  It will also contain
copies of the actual memory that the process uses.  Any changes
in this blob's format between kernel revisions can be handled by
an in-userspace conversion program.

This is a similar approach to virtually all of the commercial
checkpoint/restart products out there, as well as the research
project Zap.

These patches basically serialize internel kernel state and write
it out to a file descriptor.  The checkpoint and restore are done
with two new system calls: sys_checkpoint and sys_restart.

In this incarnation, they can only work checkpoint and restore a
single task. The task's address space may consist of only private,
simple vma's - anonymous or file-mapped. The open files may consist
of only simple files and directories.

--
(Original announcement)

In the recent mini-summit at OLS 2008 and the following days it was
agreed to tackle the checkpoint/restart (CR) by beginning with a very
simple case: save and restore a single task, with simple memory
layout, disregarding other task state such as files, signals etc.

Following these discussions I coded a prototype that can do exactly
that, as a starter. This code adds two system calls - sys_checkpoint
and sys_restart - that a task can call to save and restore its state
respectively. It also demonstrates how the checkpoint image file can
be formatted, as well as show its nested nature (e.g. cr_write_mm()
-> cr_write_vma() nesting).

The state that is saved/restored is the following:
* some of the task_struct
* some of the thread_struct and thread_info
* the cpu state (including FPU)
* the memory address space

In the current code, sys_checkpoint will checkpoint the current task,
although the logic exists to checkpoint other tasks (not in the
checkpointee's execution context). A simple loop will extend this to
handle multiple processes. sys_restart restarts the current tasks, and
with multiple tasks each task will call the syscall independently.
(Actually, to checkpoint outside the context of a task, it is also
necessary to also handle restart-block logic when saving/restoring the
thread data).

It takes longer to describe what isn't implemented or supported by
this prototype ... basically everything that isn't as simple as the
above.

As for containers - since we still don't have a representation for a
container, this patch has no notion of a container. The tests for
consistent namespaces (and isolation) are also omitted.