This reverts commit c86da67436.
Alas, I have to disagree with this in every single line. The comment
added to the top of the file only applies to a single function therein
which sits over a hundred lines below. That's not much helpful. More-
over, the link in the comment is already down ofc.
The comment is also irritating as it doesn't state in which way (enco-
ding!) it applies to the code, which presumably led to the wrong in-
terpretation of the IDs.
At last, if anything should have changed it is the strings, the IDs
are resolved to. `smbios_fill_dimm_manufacturer_from_id()` has to
resolve the IDs it gets actually fed and not a random selection from
any spec.
Since I digged into it, here's why the numbers are correct: The func-
tion started with the SPD encoding of DDR3 in mind. There, the lower
byte is the number of a "bank" of IDs with an odd-parity in the upper
most bit. The upper byte is the ID within the bank. The "correction"
was to clear the parity bit for naught. The function was later exten-
ded with IDs in the DDR2-SPD encoding (which is actually 64-bit not
16). There, a byte, starting from the lowest, is either an ID below
127 plus odd-parity, or 127 which means look in the next byte/bank.
Unused bytes seem to be filled with 0xff, I guess from the 0xff2c.
Change-Id: Icdb48e4f2c102f619fbdca856e938e85135cfb18
Reviewed-on: https://review.coreboot.org/17873
Tested-by: build bot (Jenkins)
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-by: HAOUAS Elyes <ehaouas@noos.fr>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Martin Roth <martinroth@google.com>
a01695bf9a