diff options
author | Nico Huber <nico.h@gmx.de> | 2016-12-14 23:48:09 +0100 |
---|---|---|
committer | Nico Huber <nico.h@gmx.de> | 2016-12-17 17:53:42 +0100 |
commit | a01695bf9ac5ad401950af96682b494127ad8a8d (patch) | |
tree | 953dfa5fc09ba55995f122a56f06907166b6e842 | |
parent | 9e70ce0c3ec823a41abc1f44b642d1bdfb14b4cb (diff) | |
download | coreboot-a01695bf9ac5ad401950af96682b494127ad8a8d.tar.xz |
Revert "arch/x86/smbios: Correct manufacturer ID"
This reverts commit c86da67436827c25919a2f5966049485a58fc984.
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>
-rw-r--r-- | src/arch/x86/smbios.c | 21 |
1 files changed, 9 insertions, 12 deletions
diff --git a/src/arch/x86/smbios.c b/src/arch/x86/smbios.c index 0c7f3acd95..864020001e 100644 --- a/src/arch/x86/smbios.c +++ b/src/arch/x86/smbios.c @@ -15,13 +15,6 @@ * GNU General Public License for more details. */ -/* - * Standard Manufacturer's Identification Code - * JEP106AS (Revision of JEP106AR, October 2015) - * MAY 2016 - * http://www.jedec.org/standards-documents/results/JEP106AS - */ - #include <stdlib.h> #include <string.h> #include <smbios.h> @@ -134,7 +127,7 @@ static int smbios_processor_name(char *start) void smbios_fill_dimm_manufacturer_from_id(uint16_t mod_id, struct smbios_type17 *t) { switch (mod_id) { - case 0x9b05: + case 0x2c80: t->manufacturer = smbios_add_string(t->eos, "Crucial"); break; @@ -150,7 +143,7 @@ void smbios_fill_dimm_manufacturer_from_id(uint16_t mod_id, struct smbios_type17 t->manufacturer = smbios_add_string(t->eos, "Kingston"); break; - case 0xad00: + case 0x987f: t->manufacturer = smbios_add_string(t->eos, "Hynix"); break; @@ -162,7 +155,11 @@ void smbios_fill_dimm_manufacturer_from_id(uint16_t mod_id, struct smbios_type17 t->manufacturer = smbios_add_string(t->eos, "OCZ"); break; - case 0x3406: + case 0xad80: + t->manufacturer = smbios_add_string(t->eos, + "Hynix/Hyundai"); + break; + case 0xb502: t->manufacturer = smbios_add_string(t->eos, "SuperTalent"); break; @@ -170,7 +167,7 @@ void smbios_fill_dimm_manufacturer_from_id(uint16_t mod_id, struct smbios_type17 t->manufacturer = smbios_add_string(t->eos, "GSkill"); break; - case 0xce00: + case 0xce80: t->manufacturer = smbios_add_string(t->eos, "Samsung"); break; @@ -178,7 +175,7 @@ void smbios_fill_dimm_manufacturer_from_id(uint16_t mod_id, struct smbios_type17 t->manufacturer = smbios_add_string(t->eos, "Elpida"); break; - case 0x2c00: + case 0xff2c: t->manufacturer = smbios_add_string(t->eos, "Micron"); break; |