Re: bpf and local lock

[David Miller]

From: Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx>
Date: Mon, 2 Dec 2019 13:52:38 -0800

> On Mon, Nov 25, 2019 at 11:14:33AM -0800, David Miller wrote:
>> 
>> Thomas,
>> 
>> I am working on eliminating the explicit softirq disables around BPF
>> program invocation and replacing it with local lock usage instead.
>> 
>> We would really need to at least have the non-RT stubs upstream to
>> propagate this cleanly, do you think this is possible?
> 
> Hi Thomas,
> 
> seconding the same question: any chance local lock api can be sent upstream
> soon? If api skeleton can get in during this merge window we will have the next
> bpf-next/net-next cycle to sort out details. If not the bpf+rt would need to
> wait one more release. Not a big deal. Just trying to figure out a time line
> when can we start working on concrete bpf+rt patches.

FWIW, I have some simple patches I'm working on that start to annotate
the bpf function invocation call sites.

And as part of that I add the non-RT stubs plus some new interfaces I
think might be necessary.

I've been told Thomas is going to be offline for another week so I'll
just keep working on this and post when I have something concrete.