EDAC, sb_edac: Fix reporting wrong DIMM when patrol scrubber finds error
Borislav Petkov
maintainerPATCH 3/3 EDAC, sb_edac: Fix reporting wrong DIMM when patrol scrubber finds error
Quoted 1 line · Qiuxu
The
sb_edacEDAC driver this patches fix for.
I know which driver! Your commit message formulation needs fixing.
Quoted 10 lines
Sorry for the ambiguity. "ha" is the acronym for "Home Agent" that is a hardware component interfaces with cache and directly with its owner iMC (Integrated Memory Controller). We use the property that the index of "ha" equals to iMC's.
Quoted from below datasheet section 5 (here "Bbox" is home agent, and "Mbox" is iMC):
"The Bboxes receive Home channel request and snoop responses from caching agents, and provides data response and completion to the system via Bbox to Router connection. Read and write commands to memory go out of the Bbox to the Memory Controller (MBox)."
https://www.intel.com/content/dam/www/public/us/en/documents/datasheets/xeon-e7-8800-4800-2800-families-vol-2-datasheet.pdfAnd there is also block diagram Figure 5-1 there, hope it's useful for understanding.
Now take the gist of all that knowledge and synthesize in a proper comment above it, *shortly* explaining what this function does.
Quoted 3 lines
Quoted 1 line
You need to check the
get_hapointer before calling it becauseKNIGHTS_LANDINGassigns NULL to it.There is already a check whether it's KNIGHT_LANDING case before calling "get_memory_error_data_from_mce() -> get_ha()".
Regardless, pls add that check anyway in case someone adds new CPU support and forgets to add that function.
Thx.