[Bug 12483] Re: laptop-mode/IDE-APM hang on various laptops

[Bart Samwel]

OK, the IDE controllers are virtually the same:

T41 and T42:
82801DBM (ICH4-M) IDE Controller (Rev 01)
82801DBM (ICH4-M) IDE Controller (rev 01)

Asus M6Ne:
82801DBM (ICH4-M) IDE Controller (rev 03)

R51:
82801DBM (ICH4) Ultra ATA Storage Controller (rev 01)

According to ThinkWiki this chip may be found in Thinkpads G40, G41, R40,     R50, R50e, R50p, R51, T40, T40p, T41, T41p, T42, T42p, X31, X32, X40. Reference:

http://www.thinkwiki.org/wiki/Intel_82801DBM

The thing is, AgenT reported that the problem did _not_ occur on his Thinkpad, which contains the exact same chip. And I know a lot more laptops containing this chip that _don't_ have the problem. Combo with some other chip? Or with a specific HD series?

Anyway, I googled "system hang 82801DBM" and found the 82801DBM specs, which contained on erratum that may or may not be our problem, I'm not very well-versed in PCI protocol speak. If anybody else cares to interpret this and see if it may be the problem, please be my guest. :-) The doc is here:

http://download.intel.com/design/chipsets/specupdt/25266307.pdf

Quote:

"5. PCI Non-Linear Addressing

Problem: If a PCI Memory Read Multiple or Memory Read Line transaction falls on the last DWORD of a 32 byte cache line boundary and non-linear addressing (cache-line wrap mode) is used, the ICH4-M will pre-fetch data past the cache line boundary. All subsequent PCI bus master reads will get incorrect data. Subsequent processor cycles to PCI/LPC will get blocked behind the surplus data resulting in a system hang.

Implication: System hang only seen in synthetic test environment. No known commercial PCI devices support cache-line wrap mode using Memory Read Multiple or Memory Read Line.

Workaround: None

Status: There are no plans to fix this erratum."

-- 
laptop-mode/IDE-APM hang on various laptops
https://launchpad.net/bugs/12483