netdev - Re: Is 3.18 patch "The tcp: TSO packets automatic sizing" will cause to Troughput drop when enable NO_HZ_IDEL or HIGH_RES

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:	Wed, 07 May 2014 08:45:52 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	V JobNickname <workofv@...il.com>
Cc:	netdev@...r.kernel.org
Subject: Re: Is 3.18 patch "The tcp: TSO packets automatic sizing" will
 cause to Troughput drop when enable NO_HZ_IDEL or HIGH_RES_TIMERS ?

On Wed, 2014-05-07 at 22:27 +0800, V JobNickname wrote:
> 2014-05-07 0:41 GMT+08:00 Eric Dumazet <eric.dumazet@...il.com>:
> > On Tue, 2014-05-06 at 09:30 -0700, Eric Dumazet wrote:
> >
> >> I cant see any effect for this patch explaining a 50% reduction in
> >> speed, unless a special qdisc setup maybe.
> >
> > Oh wait, a buggy driver could explain the problem.
> >
> > Some drivers really expected TX completion to happen as a side effect of
> > subsequent transmits.
> >
> > You'll have to give more details on the configuration you are using.
> >
> >
> >
> Does the "buggy driver" here mean the NIC driver or some higher level
> network stack driver?
> The xmit function of the used NIC driver is to stuff DMA channel, is
> there have "expected Tx completion" behavior in the driver's flow ?
> 
> The more detail:
> 
> 1. The platform's NIC is compilable with dwmac, use driver
> ethernet/stmicro/stmmac
>      some feature controls in plat_stmmacenet_data are set as
>         .phy_bus_name = "stmmac",
>         .bus_id = 0,
>         .phy_addr = -1, /*auto phy by pooling by mdio bus */
>         .has_gmac = 1,
>         .enh_desc = 0,
>         .tx_coe = 1,
>         .rx_coe = STMMAC_RX_COE_TYPE2,

Right, this driver seems to implement a 'coalescing tx strategy' using a
40ms timer.

Try to lower STMMAC_COAL_TX_TIMER ?



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html