linux-kernel - Re: Test throughput on PCIe interface

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20160802132248.GJ6862@twins.programming.kicks-ass.net>
Date:	Tue, 2 Aug 2016 15:22:48 +0200
From:	Peter Zijlstra <peterz@...radead.org>
To:	Bjorn Helgaas <helgaas@...nel.org>
Cc:	Muni Sekhar <munisekharrms@...il.com>, linux-pci@...r.kernel.org,
	Ingo Molnar <mingo@...hat.com>,
	Arnaldo Carvalho de Melo <acme@...nel.org>,
	Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
	linux-kernel@...r.kernel.org, Stephane Eranian <eranian@...gle.com>
Subject: Re: Test throughput on PCIe interface

On Fri, Jul 22, 2016 at 11:14:06AM -0500, Bjorn Helgaas wrote:
> [+cc perf folks]
> 
> On Fri, Jul 22, 2016 at 05:59:51PM +0530, Muni Sekhar wrote:
> > Hi All,
> > 
> > 
> > We have xilinx PCIe endpoint with DMA reference block.
> > 
> > 
> > The DMA-REF block provides a mechanism to DMA(also supports
> > Scatter-Gather DMA) transfer data at the maximum rate between host
> > (CPU) memory and a FIFO in the DMA-REF block.
> > 
> > 
> > DMA-REF block provides the loopback. The intention is to use this
> > block in production and self-test, to check that the PCIe bus is
> > operating at the expected maximum transfer speed.
> > 
> > 
> > Does kernel has any standard utilities to test throughput on PCIe interface?
> 
> I know some PCIe controllers devices do have performance monitors, and
> perf might support some of them, but I don't know any details.

At least some of the Intel Uncore drivers have boxes specific to PCIe
(for instance my IVB-EP has an uncore_r2pcie PMU).

I'm not aware of any endpoint drivers, but it should not be too hard to
create uncore drivers for them.