em JumboFrame improovement and PCIe addon-card regression [Was: Re: em regression, UDP LOR followed by ssh stall]

[Harald Schmalzbauer]

Brandon Gooch schrieb am 16.04.2010 17:32 (localtime):
...

Thanks Jack! Your work is very appreciated.

It's extremely appreciated!


    vendor     = 'Intel Corporation'
    class      = network
    subclass   = ethernet

em0: Using MSIX interrupts with 5 vectors
em0: [ITHREAD]
em0: [ITHREAD]
em0: [ITHREAD]
em0: [ITHREAD]
em0: [ITHREAD]
em0: Ethernet address: 00:1b:21:3e:90:52

Maybe it's something MSIX related?





ping -f -l 8972



Antwort von 192.168.147.11: Bytes=8972 Zeit<1ms TTL=64
Antwort von 192.168.147.11: Bytes=8972 Zeit<1ms TTL=64

GREAT! Thank you very much for that improovement!

-Harry

I tried another semi-productive system and since that hadn't exposed any peculiarity I also upgraded one not too important productive machine. All have onboard 82566DM and 82541EI chips and all are working fine so far. It seems only my addon-card is affected, which doesn't exhibit a part number: e...@pci0:1:0:0: class=0x020000 card=0xa01f8086 chip=0x10d38086 rev=0x00 hdr=0x00 em0: <Intel(R) PRO/1000 Network Connection 7.0.0> port 0x2000-0x201f mem 0xe1a80000-0xe1a9ffff,0xe1a00000-0xe1a7ffff,0xe1aa0000-0xe1aa3fff irq 16 at device 0.0 on pci1 And I took the chance to reactivate jumbo-frame testings on one of the machines at so far it looks promising. I can get icmp relpys with 8972 bytes payload requests. That hasn't been working with "PRO/1000 Network Connection 6.9.14". Now it reports "PRO/1000 Network Connection 7.0.0" and with known working windows-driven intel mtu-9014 setted cards I get response (which always worked between the windows machines but never with my FreeBSD RELENG_8): Ping wird ausgeführt für banana.r2k.rzhp.spsnetz.de [192.168.147.11] mit 8972 Bytes Daten:

signature.asc
Description: OpenPGP digital signature