On Thu, 15 Aug 2019, Dmitry Safonov wrote:
> From: Andrei Vagin <avagin@xxxxxxxxx>
>
> As it has been discussed on timens RFC, adding a new conditional branch
> `if (inside_time_ns)` on VDSO for all processes is undesirable.
>
> Addressing those problems, there are two versions of VDSO's .so:
> for host tasks (without any penalty) and for processes inside of time
> namespace with clk_to_ns() that subtracts offsets from host's time.
>
> Introduce vdso_static_branch_unlikely(), which is similar to
> static_branch_unlikely(); alias it with timens_static_branch_unlikely()
> under CONFIG_TIME_NS.
>
> The timens code in vdso will look like this:
>
> if (timens_static_branch_unlikely()) {
> clk_to_ns(clk, ts);
Please name that clk_to_namespace(). _ns() is widely used for nanoseconds.
> }
>
> The version of vdso which is compiled from sources will never execute
> clk_to_ns(). And then we can patch the 'no-op' in the straight-line
> codepath with a 'jump' instruction to the out-of-line true branch and
> get the timens version of the vdso library.
Colour me confused. Why do we need that static branch at all?
Why don't we compile VDSO_NO_NAMESPACE and VDSO_NAMESPACE right away? One
has the clk_to_namespace() one does not. The you can spare the whole static
key patching and the NO_NAMESPACE variant does not have extra 5 NOPS.
The VDSO is one page IIRC, so having the extra namespace variant around
does really not matter at all.
Thanks,
tglx
