On Tue, Sep 10, 2019 at 10:35:34PM +0200, Matthias Schmidt wrote:
> Hi Stefan,
>
> * Stefan Sperling wrote:
> >
> > I think I see why. I forgot to convert some existing ieee80211_input()
> > calls to ieee80211_inputm(), in ieee80211_input.c.
> > These calls are related to buffered aggregated fram
Hi Stefan,
* Stefan Sperling wrote:
>
> I think I see why. I forgot to convert some existing ieee80211_input()
> calls to ieee80211_inputm(), in ieee80211_input.c.
> These calls are related to buffered aggregated frames, so aggregated
> frames triggered multiple if_input() calls per interrupt aga
On Tue, Sep 10, 2019 at 07:08:14PM +0200, Matthias Schmidt wrote:
> Hi Stefan,
>
> * Stefan Sperling wrote:
> >
> > New diff with above changes:
>
> I tested your new diff with two different systems:
>
> * Thinkpad T450s with iwm (8265, same as yesterday)
> * Thinkpad X220 with iwn (6205)
>
>
Hi Stefan,
* Stefan Sperling wrote:
>
> New diff with above changes:
I tested your new diff with two different systems:
* Thinkpad T450s with iwm (8265, same as yesterday)
* Thinkpad X220 with iwn (6205)
and on both systems I see a drastic regression compared to yesterday's
patch. The downloa
On Mon, Sep 09, 2019 at 06:17:34PM -0300, Martin Pieuchot wrote:
> On 09/09/19(Mon) 16:37, Stefan Sperling wrote:
> > On Mon, Sep 09, 2019 at 03:10:04PM +0200, Stefan Sperling wrote:
> > > The wifi stack currently calls if_input once per packet instead of once
> > > per interrupt. To make the wifi
On 09/09/19(Mon) 16:37, Stefan Sperling wrote:
> On Mon, Sep 09, 2019 at 03:10:04PM +0200, Stefan Sperling wrote:
> > The wifi stack currently calls if_input once per packet instead of once
> > per interrupt. To make the wifi layer play nicely with the network stack
> > we can split ieee80211_input
Hi Stefan,
* Stefan Sperling wrote:
> On Mon, Sep 09, 2019 at 03:10:04PM +0200, Stefan Sperling wrote:
> > The wifi stack currently calls if_input once per packet instead of once
> > per interrupt. To make the wifi layer play nicely with the network stack
> > we can split ieee80211_input() into tw
On Mon 09/09/2019 16:37, Stefan Sperling wrote:
> On Mon, Sep 09, 2019 at 03:10:04PM +0200, Stefan Sperling wrote:
> > The wifi stack currently calls if_input once per packet instead of once
> > per interrupt. To make the wifi layer play nicely with the network stack
> > we can split ieee80211_inpu
this seems to work fine on
iwm0 at pci2 dev 0 function 0 "Intel Dual Band Wireless AC 7260" rev 0x83, msi
download is considerably less wobbly, upload also benefits a bit(?):
before:
[florian@x1:~]$ tcpbench -t 10 192.168.178.73
elapsed_ms bytes mbps bwidth
1001
On Mon, Sep 09, 2019 at 03:10:04PM +0200, Stefan Sperling wrote:
> The wifi stack currently calls if_input once per packet instead of once
> per interrupt. To make the wifi layer play nicely with the network stack
> we can split ieee80211_input() into two parts:
Updated diff which avoids purging t
robert@ noticed that iwm(4) Rx throughput increased after running:
sysctl net.link.ifrxq.pressure_drop=32
The default value for this sysctl is 8.
dlg@ explained the problem to me:
[[[
the pressure stuff counts the number of times ifiq_input (which is
called from if_input) gets called befor
11 matches
Mail list logo
