Piotr Jasiukajtis wrote: > On Fri, Jul 18, 2008 at 1:20 PM, James Carlson <[EMAIL PROTECTED]> wrote: >> Piotr Jasiukajtis writes: >>> On 7/17/08, Rao Shoaib <[EMAIL PROTECTED]> wrote: >>>> e1000g is a generic driver for lots of intel chipsets. As we are learning >>>> in the performance work not all of them behave the same, in case of e1000g >>>> you really have to look at the chipset to make any comparison. >>> This is my NIC (Lenovo T61): >>> >>> Jul 19 20:57:35 nx0 pcplusmp: [ID 803547 kern.info] pcplusmp: >>> pci8086,1049 (e1000g) instance 0 vector 0x18 ioapic 0xff intin 0xff is >>> bound to cpu 1 >>> Jul 19 20:57:35 nx0 mac: [ID 469746 kern.info] NOTICE: e1000g0 registered >>> Jul 19 20:57:35 nx0 e1000g: [ID 766679 kern.info] Intel(R) PRO/1000 >>> Network Connection, Driver Ver. 5.2.9 >>> Jul 19 20:57:37 nx0 mac: [ID 435574 kern.info] NOTICE: e1000g0 link >>> up, 1000 Mbps, full duplex >> My home system (which seems to work well and is running snv_92) >> reports this driver as: >> >> Jun 27 08:51:16 carlson pcplusmp: [ID 803547 kern.info] pcplusmp: >> pci8086,1079 (e1000g) instance 1 vector 0x1b ioapic 0x2 intin 0x3 is bound >> to cpu 0 >> Jun 27 08:51:16 carlson e1000g: [ID 766679 kern.info] Intel(R) PRO/1000 >> Network Connection, Driver Ver. 5.2.8 >> Jun 27 08:51:18 carlson mac: [ID 435574 kern.info] NOTICE: e1000g1 link up, >> 100 Mbps, half duplex >> >> The differences here include the PCI IDs (I have 10bd and 1079, you >> have 1049), the driver version (!), and the fact that I've got some >> really, really old hubs. ;-} >> >> On the driver version number, something strange is afoot. Looking at >> the source versioning in the gate, 5.2.9 corresponds with CR 6713685, >> which integrated in snv_93. Are you sure you're running snv_91? >> >> In any event, both the different chipset and the type of interface >> could make a difference here. > Ok I have just used belenix 0.7.1 test2 which is snv91 based and the > problem exists. > > Jul 19 16:11:35 belenix pcplusmp: [ID 803547 kern.info] pcplusmp: > pci8086,1049 (e1000g) instance 0 vector 0x18 ioapic 0xff intin 0xff is > bound to cpu 0 > Jul 19 16:11:35 belenix mac: [ID 469746 kern.info] NOTICE: e1000g0 registered > Jul 19 16:11:35 belenix e1000g: [ID 766679 kern.info] Intel(R) > PRO/1000 Network Connection, Driver Ver. 5.2.7 > Jul 19 16:12:35 belenix mac: [ID 435574 kern.info] NOTICE: e1000g0 > link up, 1000 Mbps, full duplex
A datapoint from a Thinkpad R51 with a working e1000g0: Jul 18 09:25:12 shrek mac: [ID 469746 kern.info] NOTICE: e1000g0 registered Jul 18 09:25:12 shrek unix: [ID 954099 kern.info] NOTICE: IRQ10 is being shared by drivers with different interrupt levels. Jul 18 09:25:12 shrek This may result in reduced system performance. Jul 18 09:25:12 shrek e1000g: [ID 766679 kern.info] Intel(R) PRO/1000 Network Connection, Driver Ver. 5.2.9 [...] Jul 18 09:25:23 shrek mac: [ID 435574 kern.info] NOTICE: e1000g0 link up, 1000 Mbps, full duplex I don't get the initial line of logging (pcplusmp pci8086...), maybe that's because this is just a single core Pentium M. scanpci reports: pci bus 0x0002 cardnum 0x01 function 0x00: vendor 0x8086 device 0x101e Intel Corporation 82540EP Gigabit Ethernet Controller (Mobile) Cheers, Chris _______________________________________________ networking-discuss mailing list networking-discuss@opensolaris.org
