[CRIU] [PATCH 4/4] arch: x86 -- Add proxification of vDSO calls

[Cyrill Gorcunov]

On Wed, Apr 17, 2013 at 05:32:30PM +0400, Pavel Emelyanov wrote:
> > +	if (vdso_proxy.proxify) {
> > +		if (vdso_proxy.sym_from) {
> > +			struct vma_area *last;
> > +			struct vma_area *vma;
> > +
> > +			ret = -1;
> > +			vma = alloc_vma_area();
> > +			if (!vma)
> > +				goto out;
> > +
> > +			last = list_entry(rst_vmas.h.prev, struct vma_area, list);
> > +			rst_vmas.nr++;
> > +			list_add_tail(&vma->list, &rst_vmas.h);
> 
> > +                       vma->vma.start  = last->vma.end;
> > +                       vma->vma.end    = vma->vma.start + symtable_vma_size(&vdso_proxy.sym_rt);
> > +                       vma->vma.prot   = PROT_READ | PROT_EXEC;
> > +                       vma->vma.flags  = MAP_PRIVATE | MAP_ANONYMOUS;
> > +                       vma->vma.status = VMA_AREA_REGULAR | VMA_AREA_VDSO | VMA_ANON_PRIVATE;
> 
> Who will then make sure that real VDSO would be at that place?

Wait, I'm confused now. At moment the new VDSO with kernel supplied content
is placed after the last VMA found in image (which is a bit incorrect as
said in comment and need to be fixed, but work for now).

So we do

 - read all vmas in image
 - append new vdso
 - patch old vdso to redirect calls to new vdso

so I somehow don't understand what you mean by "who will make sure".