Re: [RFC PATCH 0/5] Make eBPF programs output data to perf event

From	Peter Zijlstra <peterz@infradead.org>
Newsgroups	linux.kernel
Subject	Re: [RFC PATCH 0/5] Make eBPF programs output data to perf event
Date	2015-07-02 11:40 +0200
Message-ID	<pHJ8m-U3-13@gated-at.bofh.it> (permalink)
References	<pHgpI-50p-5@gated-at.bofh.it> <pHj4e-yA-5@gated-at.bofh.it> <pHjGV-1xX-13@gated-at.bofh.it> <pHoQi-4Hd-21@gated-at.bofh.it> <pHCJz-5iM-3@gated-at.bofh.it>
Organization	linux.* mail to news gateway

Show all headers | View raw

On Wed, Jul 01, 2015 at 07:48:25PM -0700, Alexei Starovoitov wrote:
> On 7/1/15 4:58 AM, Peter Zijlstra wrote:
> >
> >But why create a separate trace buffer, it should go into the regular
> >perf buffer.
> 
> +1
> 
> I think
> +static char __percpu *perf_extra_trace_buf[PERF_NR_CONTEXTS];
> is redundant.
> It adds quite a bit of unnecessary complexity to the whole patch set.

Ah, that is what he does. I through he was creating a trace buffer, as
in the stuff ftrace uses, to output stuff concurrently to the regular
perf buffer, which would be insane ;-)

So you need these temporary buffers if you cannot tell a-priory how much
data is going to get written. The perf buffer can only deal with fixed
size events, you request a buffer of @size, you write into it, you
commit.

So if there's variable sized data, you first need to generate it in
order to tell how long it is, so that you can reserve your record and
copy it in. Sucks, but that's the way it is.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Back to linux.kernel | Previous | Next | Find similar | Unroll thread

Thread

Re: [RFC PATCH 0/5] Make eBPF programs output data to perf event Peter Zijlstra <peterz@infradead.org> - 2015-07-02 11:40 +0200

csiph-web