Konrad Rzeszutek wrote:
I am suggesting to have a global flag called iSCSI while installing(like other flags "askmethod", "updates"), so when installing the user should tell the Anaconda explicitly to get the values from iBFT.Would it be easier if there was a /sysfs entry that told anacondaabout the presence of iBFT? And if there was any pertinent information then anaconda would extract the information from /sysfs instead of searching in/dev/mem?
There's already something sortof like this. The current ACPI spec reserves the table name "IBFT", and Intel has a patch (I _think_ they've posted it already, but I'm not sure) that gives you access to all the ACPI tables that exist on the system as binary blobs in /sys/firmware/acpi/table/ . So if the system firmware is implemented correctly, then you shouldn't need to ever touch /dev/mem (which as we've mentioned is dangerous and unacceptable.)
What Prasanna is telling us, though, is that the system firmware (on some system he's concerned with; I'm not sure which that is) doesn't actually implement this.
This could also have the extra benefit of initrd shell script being able to parse the /sysfs and constructing the pertient iSCSI target information dynamically (and running iscsiadm with those gathereddate) instead of having the values hard-coded as they are now.
That's the high-level plan, yes. -- Peter