Date: Sun, 06 Apr 2008 14:22:17 +0300 From: Andriy Gapon <avg@icyb.net.ua> To: Nate Lawson <nate@root.org> Cc: freebsd-acpi@freebsd.org Subject: Re: mismatch between FACP and chipset spec Message-ID: <47F8B269.6030606@icyb.net.ua> In-Reply-To: <47C70B87.5060301@icyb.net.ua> References: <47C67001.20101@icyb.net.ua> <47C6FACC.9090502@root.org> <47C70B87.5060301@icyb.net.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
This is a multi-part message in MIME format.
--------------000400010100020701020907
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
on 28/02/2008 21:29 Andriy Gapon said the following:
> on 28/02/2008 20:17 Nate Lawson said the following:
>> Andriy Gapon wrote:
>>> But addresses given for PM1a_CNT_BLK are not documented at all! On the
>>> other hand ACPI description of that register perfectly matches
>>> description of PIIX4 PMCNTRL register that is located at 0x4004 with a
>>> given base address. So, this is 0x4040 vs. 0x4004, looks like a possible
>>> typo/mistake by an author of ACPI tables for this motherboard.
>> 440BX is pretty old, and typos like that were common back then. ACPI
>> was less used and drivers would just hardcode 0x4004 or whatever.
>>
>>> Question: is there any way I can way override the address of
>>> PM1a_CNT_BLK? My guess is that there is zero chance that there would be
>>> any BIOS updates for this old and exotic motherboard (MP2-BX-X).
>> No generic way. You'd have to add a quirk to acpi.c/acpi_quirks and use
>> that to override FADT.
>>
>>> I think that this register is mostly useful for BM_RLD bit which is used
>>> in C3 support. I don't use C3 (there is an errata for C3 with this
>>> chipset and there is no PM2_CNT register defined anyway), but I am
>>> curios anyway.
>> Seems like a lot of effort for no gain. Since you are getting good at
>> debugging, can we get you newer hardware to play with? :)
>
> Nate,
>
> thank you for pointing me to acpi_quirks and for confirming my doubts
> over worthwhileness of this endeavor.
I knew that I wouldn't be able to keep my hands off it :-)
Actually that PM1a_CNT_BLK typo caused one very minor annoyance for me
(but still an annoyance!) - 'shutdown -p' would cause the machine to
hang right after 'Powering system off using ACPI' line.
It seems that ACPI power off requires a write to PM1a_CNT_BLK area and
the hardware doesn't tolerate a write to a bogus IO register.
Attached is small patch that fixes the issue for my particular system.
It is a bit ugly but I couldn't think of another way for a quirk like this.
--
Andriy Gapon
--------------000400010100020701020907
Content-Type: text/x-patch;
name="mp2bxx.piix4.PM1a_CNT_BLK-poweroff.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
filename="mp2bxx.piix4.PM1a_CNT_BLK-poweroff.patch"
--- sys/dev/acpica/acpi_quirk.c.orig 2008-04-05 01:01:08.000000000 +0300
+++ sys/dev/acpica/acpi_quirk.c 2008-04-05 01:55:54.000000000 +0300
@@ -71,6 +71,7 @@
static int aq_strcmp(char *actual, char *possible);
static int aq_match_header(ACPI_TABLE_HEADER *hdr,
const struct acpi_q_rule *match);
+static void aq_special(void);
static int
aq_revcmp(int revision, enum ops_t op, int value)
@@ -154,6 +155,7 @@
bzero(&fadt, sizeof(xsdt));
/* Then, override the quirks with any matched from table signatures. */
+ done = FALSE;
for (entry = acpi_quirks_table; entry->match; entry++) {
done = TRUE;
for (match = entry->match; match->sig[0] != '\0'; match++) {
@@ -180,5 +182,48 @@
}
}
+ /* Special quirks that can not be expressed in a generic form */
+ if (!done)
+ aq_special();
+
return (0);
}
+
+void aq_special(void)
+{
+ static const struct acpi_q_rule MP2_BX_X[] = {
+ { "FADT", OEM, {"AWARD "}, {"AWRDACPI"} },
+ { "FADT", OEM_REV, {.op = OP_EQL}, {.rev = 0x42302e31} },
+ { "FADT", CREATOR, {"AWRD"} },
+ { "FADT", CREATOR_REV, {.op = OP_EQL}, {.rev = 0x0} },
+ { "" }
+ };
+
+ const struct acpi_q_rule *match;
+ ACPI_TABLE_HEADER fadt;
+ int done;
+
+ if (ACPI_FAILURE(AcpiGetTableHeader(ACPI_SIG_FADT, 0, &fadt)))
+ bzero(&fadt, sizeof(fadt));
+
+ /* Code specific to MP2-BX-X motherboard. */
+ done = TRUE;
+ for (match = &MP2_BX_X[0]; match->sig[0] != '\0'; match++) {
+ if (aq_match_header(&fadt, match) == FALSE) {
+ done = FALSE;
+ break;
+ }
+ }
+ if (done) {
+ /* This MP2-BX-X FADT specifies PM1a_CNT_BLK=0x4040-0x4041,
+ * while in fact it should be 0x4004-0x4005 according to
+ * PIIX4E specification (PMCNTRL IO register).
+ */
+ printf("MP2-BX-X: changing PM1a_CNT_BLK from 0x%x to 0x4004\n",
+ AcpiGbl_FADT.Pm1aControlBlock);
+ AcpiGbl_FADT.Pm1aControlBlock = 0x4004;
+ AcpiGbl_FADT.Pm1ControlLength = 2;
+ return;
+ }
+}
+
--------------000400010100020701020907--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?47F8B269.6030606>