Re: [PATCH 2/6] x86/vdso: introduce do_map_vdso() and vdso_type enum

[Dmitry Safonov]

On 06/03/2016 12:50 PM, Cyrill Gorcunov wrote:
> On Wed, Jun 01, 2016 at 04:11:33PM +0300, Dmitry Safonov wrote:
> > Make in-kernel API to map vDSO blobs on x86.
> >
> > Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> > Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> > Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > Cc: "H. Peter Anvin" <hpa@xxxxxxxxx>
> > Signed-off-by: Dmitry Safonov <dsafonov@xxxxxxxxxxxxx>
>
> Hi! Sorry for delay in reply. Dima, here is a moment I somehow missing:
> when do_map_vdso is called it is implied that old vdso is unmapped already,
> or both can be present in the system?

Both can be present. There are some limitations: if first vDSO was
32-bit, syscalls through it wouldn't work, only on the last mapped.
It's because of mm->context.vdso pointer, which for 32-bit vDSO
should point on it. (and this code make it point to a new mapped vDSO)

On RFC I did patch that unmaps previous vDSO, but I have though
it's better to simplify this code and drop unmapping.