Re: [PATCH bpf-next] bpf, doc: Update bpf_jit_enable limitation for CONFIG_BPF_JIT_ALWAYS_ON

[Leo Yan <leo.yan@linaro.org>]

On Fri, Apr 27, 2018 at 11:44:44AM +0200, Daniel Borkmann wrote:
> On 04/26/2018 04:26 AM, Leo Yan wrote:
> > When CONFIG_BPF_JIT_ALWAYS_ON is enabled, kernel has limitation for
> > bpf_jit_enable, so it has fixed value 1 and we cannot set it to 2
> > for JIT opcode dumping; this patch is to update the doc for it.
> > 
> > Signed-off-by: Leo Yan <leo.yan@linaro.org>
> > ---
> >  Documentation/networking/filter.txt | 6 ++++++
> >  1 file changed, 6 insertions(+)
> > 
> > diff --git a/Documentation/networking/filter.txt b/Documentation/networking/filter.txt
> > index fd55c7d..feddab9 100644
> > --- a/Documentation/networking/filter.txt
> > +++ b/Documentation/networking/filter.txt
> > @@ -483,6 +483,12 @@ Example output from dmesg:
> >  [ 3389.935851] JIT code: 00000030: 00 e8 28 94 ff e0 83 f8 01 75 07 b8 ff ff 00 00
> >  [ 3389.935852] JIT code: 00000040: eb 02 31 c0 c9 c3
> >  
> > +When CONFIG_BPF_JIT_ALWAYS_ON is enabled, bpf_jit_enable is set to 1 by default
> > +and it returns failure if change to any other value from proc node; this is
> > +for security consideration to avoid leaking info to unprivileged users. In this
> > +case, we can't directly dump JIT opcode image from kernel log, alternatively we
> > +need to use bpf tool for the dumping.
> > +
> 
> Could you change this doc text a bit, I think it's slightly misleading. From the first
> sentence one could also interpret that value 0 would leaking info to unprivileged users
> whereas here we're only talking about the case of value 2. Maybe something roughly like
> this to make it more clear:
> 
>   When CONFIG_BPF_JIT_ALWAYS_ON is enabled, bpf_jit_enable is permanently set to 1 and
>   setting any other value than that will return in failure. This is even the case for
>   setting bpf_jit_enable to 2, since dumping the final JIT image into the kernel log
>   is discouraged and introspection through bpftool (under tools/bpf/bpftool/) is the
>   generally recommended approach instead.

Yeah, your rephrasing is more clear and better.  Will do this and send
new patch soon.  Thanks for your helping.

> Thanks,
> Daniel