Re: [Etherboot-users] Re: Problem booting e1000 card : No IP address - Workaround
Brought to you by:
marty_connor,
stefanhajnoczi
From: Timothy L. <tl...@ro...> - 2005-07-28 01:52:28
|
Georg Baum wrote: > Am Dienstag, 26. Juli 2005 19:44 schrieb Timothy Legge: > > >>From a quick review, I would say that the hw->mac_type might be more >>useful to you than the pci device id. It would probably be faster to >>track what is missing. > > > You are right. However, for the cards in question this is a 1:1 mapping: > > Mathias Ulrich Ulisses Georg > dev id 0x1008 0x1076 0x100E 0x1019 > 82544EI_COPPER 82541GI 82540EM 82547EI > mac_type e1000_82544 e1000_82541_rev_2 e1000_82540 e1000_82547 > phy_id M88E1000_I IGP01E1000_I M88E1011_I IGP01E1000_I > phy_type e1000_phy_m88 e1000_phy_igp e1000_phy_m88 e1000_phy_igp > > It seems that the problem of Mathias, Ulrich and Ulisses is indeed the > same: All three mac types are listed in the reset switch case that calls > E1000_WRITE_REG_IO instead of E1000_WRITE_REG, and the reports show that > this is the failing place in the code. My (working) card uses > E1000_WRITE_REG. > > The switch in the reset routine was added last year because of a report > from James Pearson that it was needed for the 82546EB (mac type > e1000_82546), so we can't simply switch everything back to > E1000_WRITE_REG again. > > I checked again with the current Linux driver, and could not find any > obvious reason for this difference, so I propose the attached fix for > now. There are a few places in the EB driver where things are ifdefed out. Are any of them important? Can you identify any this alike (are all the problem cards PCI-X?)? There may be work arounds for those specific things. Good luck. Tim |