OpenSolaris

Printable Version Enter a New Search
Bug ID 6672051
Synopsis FMA appears to implement rule 4B incorrectly
State 10-Fix Delivered (Fix available in build)
Category:Subcategory fma:mem
Keywords spsw-s10u8-goal
Responsible Engineer Tom Pothier
Reported Against solaris_10u4
Duplicate Of
Introduced In solaris_nevada
Commit to Fix snv_103
Fixed In snv_103
Release Fixed solaris_nevada(snv_103) , solaris_10u8(s10u8_01) (Bug ID:2169185)
Related Bugs 6346507 , 6714311 , 6747961 , 6749403
Submit Date 6-March-2008
Last Update Date 9-June-2009
Description
The cpumem diagnosis engine apparently implements rule 4B incorrectly.
It declared a DIMM faulty when it saw 6 CEs from the same DIMM, where
three were on were on bit 41 checkword 3 (all different AFARs as required),
two were on bit 41 checkword 2 (again different AFARs), and one was on bit
40 checkword 1.  It needed to see one more CE on bit 40 checkword 1 before
declaring a rule 4B violation.

Now, this DIMM did indeed have a Rule 4B violation, it's just that
the DE didn't seem to associate the ereports causing the 4B violation
with the fault record.

I'm attaching the errlog, fltlog, and some fmdump outputs.
I have a second example, and am uploading the fltlog and some fmdump outputs
for that one.  However, the errlog file, even when compressed, is too large
to upload.
Work Around
N/A
Comments
N/A