Re: [PATCH 18/38] C/R: core stuff

[Alexey Dobriyan <adobriyan@gmail.com>]

On Thu, May 28, 2009 at 06:20:25PM -0400, Oren Laadan wrote:
> Alexey Dobriyan wrote:
> > Price in loose case is strictly bigger than price in strict case
> > because set of loose states is strict superset of set of strict cases.
> 
> The code to enforce full-container approach is a *superset* of the
> code without this requirement.
> 
> If you can checkpoint a full container, surely you can checkpoint a
> sub-hierarchy of processes.

No!

To checkpoint uts_ns reliably, in strict case, one doesn't even need
uts_sem.

In loose case, one needs uts_sem.

	static int dump_uts_ns(struct kstate_context *ctx, struct kstate_object *obj)
	{
		struct uts_namespace *uts_ns = obj->o_obj;
		struct kstate_image_uts_ns *i;
		int rv;
	
		i = kstate_prepare_image(KSTATE_OBJ_UTS_NS, sizeof(*i));
		if (!i)
			return -ENOMEM;
	
		if (ctx->dump_live)
			down_read(&uts_sem);
		strncpy(i->sysname, (const char *)uts_ns->name.sysname, 64);
		strncpy(i->nodename, (const char *)uts_ns->name.nodename, 64);
		strncpy(i->release, (const char *)uts_ns->name.release, 64);
		strncpy(i->version, (const char *)uts_ns->name.version, 64);
		strncpy(i->machine, (const char *)uts_ns->name.machine, 64);
		strncpy(i->domainname, (const char *)uts_ns->name.domainname, 64);
		if (ctx->dump_live)
			up_read(&uts_sem);

		rv = kstate_write_image(ctx, i, sizeof(*i), obj);
		kfree(i);
		return rv;
	}