Ronald G Minnich wrote: >On Thu, 14 Mar 2002, Jeff Garzik wrote: > >>A few notes, blindly enabling IO and MEM for all devices is bad... >>recall that video devices for example try to decode the same VGA region >>at the same time. Disabling the IO decode bit in PCI config space for >>VGA devices is the only generic way to avoid this. >> > >Does "blind" mean "enabling if a device has IO address space". Basically, >I only enable IO and MEM on devices that have addressable IO and MEM. >Except for VGA, is this a problem? >
It might be :/ The kernel already does this for you via pci_enable_device() called by the Linux driver. So, I would advise not turning on IO/MEM decoding unless absolutely necessary. I can certainly see some devices reacting to the enabling of decoding... by increased power consumption if nothing else, right? Even the IDE driver calls pci_enable_device in the kernel... A tangent, which I just thought of: you guys interested in PCI should pick up this book, _PCI System Architecture_: http://www.amazon.com/exec/obidos/ASIN/0201309742/qid=1016152267/sr=8-1/ref=sr_8_7_1/103-4649482-0295858 This is an excellent reference, and distills the PCI specifications into something understandable. I believe that one could possibly avoid purchasing the PCI specifications, with this book... It looks like there is also a _PCI-X System Architecture_ book, but I don't know anything about that one. >>WRT to bus mastering, it is very very evil to blindly enable >>busmastering. >> > >Understand. I take this to mean my original decision to NOT enable bus >master in linuxbios was a good decision. I wonder what we should do for >9load then. I'm seriously thinking of dumping 9load and having linux load >the plan9 kernel. That I know will work. 9load just can't seem to function >without a normal BIOS there. > >It's another reason I like using an OS as my boot loader ... > I wonder if you will change your tune when elfBIOS exists and is useable.... ;-) Jeff
