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] [thread-next>] [day] [month] [year] [list]
Date:   Mon, 17 Dec 2018 09:08:54 +1100
From:   NeilBrown <neil@...wn.name>
To:     Bjørn Mork <bjorn@...k.no>, gerg@...nel.org
Cc:     sean.wang@...iatek.com, andrew@...n.ch,
        vivien.didelot@...oirfairelinux.com, f.fainelli@...il.com,
        netdev@...r.kernel.org, blogic@...nwrt.org,
        René van Dorst <opensource@...rst.com>
Subject: Re: [PATCH 0/3]: net: dsa: mt7530: support MT7530 in the MT7621 SoC

On Tue, Dec 11 2018, NeilBrown wrote:

>
> I got your patch working on 4.20-rc5 and did a performance comparison.
> With the staging driver (using iperf3) I get
>   220 MBit/sec in
>   680 MBit/sec out
>
> with the patched mainline driver I get
>   190 MBit/sec in
>    93 MBit/sec out
>
> (numbers are a bit rubbery, but within 10%)
>
> I haven't looked into why this might be, but thought I would mention it.
>
> Strangely when I test with scp, I get about 10MB/sec in both directions
> with both drivers.  Maybe the CPU limits encryption speed.
>
> I have a 4.4-based kernel where I get 940MBit/sec both ways - using a
> precursor of the current staging driver.

Just FYI, I've been looking further into this, and I don't think the
problem is (entirely) related to the driver.

In my 4.4 kernel, the build_skb() call in (the equivalent of)
mtk_poll_rx() takes about 1.2usec and the call to napi_gro_receive()
takes about 3usec.

In my 4.20 kernel, these calls take about 30 and 24 usec respectively.
This easily explains the slowdown.
I don't yet know why, and won't have time to look for a few days.
I haven't looked into how this affects the
drivers/net/ethernet/mediatek driver in 4.20.

If anyone has ideas about why these might be so slow, I'd love to hear
them.

Thanks,
NeilBrown

Download attachment "signature.asc" of type "application/pgp-signature" (833 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ