stephan, oille,
Please check the memhole patch.
when the hole_startk == one node basek, 1. it will modify base reg instead of hole reg 2. or if K8_HW_MEM_HOLE_SIZE_AUTO =1, then hole_startk will be moved to pri node mem middle....
this could solve the problem such as 1. two nodes, and first cpu have 2G installed, and memhole size is set to 2G, or mmio_basek is 2G, ( 2 IB installed). 2. four nodes. and every node has 1 G installed, and mmio will be 1G, and memhole is 1G.....
New Mem hole behavoir will be 1. if K8_HW_MEM_HOLE_SIZEK == 0, then don't set mem_hole 2. if it is set, and it make hole_startk <= mmio_basek, will not reset mem hole in northbridge.c 3. if it is set, and it make hole_startk > mmio_basek, memhole will be reset in northbeidge.c 4. if hole_startk met with basek, it will use base reg instead of increase the hole_sizek to make hole_startk to the middle of pri node.
YH
has this been added via the issue tracker? Please, no one, no commits unless from the issue tracker. This is really important to get right.
Ying Hai, I apologize for not being on top of your patches. Things got crazy the last month with SC 2005 and I am now digging out.
If you could, summarize for me in a seperate post what I should do to sync our trees up. I can use the tarball you sent, but I need a little direction.
I am trying to wrap up as many of the issues as I can, as fast as I can, though I expect the LNXI patches will be tough.
thanks
ron
I have added this to the issue tracker.
ron