[coreboot-gerrit] Patch set updated for coreboot: 64a23d6 baytrail: HDA function disable workaround

[Aaron Durbin (adurbin@google.com)]

Aaron Durbin (adurbin at google.com) just uploaded a new patch set to gerrit, which you can find at http://review.coreboot.org/4926

-gerrit

commit 64a23d670690b2e30d2fa458fbf882d6e94a9a04
Author: Aaron Durbin <adurbin at chromium.org>
Date:   Wed Oct 30 17:07:46 2013 -0500

    baytrail: HDA function disable workaround
    
    Parts of the audio path are common between the HDA and LPE.
    However, those parts are power-controlled by the D-state of
    the HDA device. Therefore, one cannot put the HDA into D3Hot
    because those audio paths will be shutdown.
    
    BUG=chrome-os-partner:22871
    BRANCH=None
    TEST=Built and booted through depthcharge. Disabling HDA still
         causes a shutdown when performing warm reset, however I
         was able to verify the magic sequence was being performed.
    
    Change-Id: I3b01356d85a4b7b902bd896b8eb9e7bc509fcc42
    Signed-off-by: Aaron Durbin <adurbin at chromium.org>
    Reviewed-on: https://chromium-review.googlesource.com/175491
    Reviewed-by: Duncan Laurie <dlaurie at chromium.org>
---
 src/soc/intel/baytrail/baytrail/iomap.h |  1 +
 src/soc/intel/baytrail/southcluster.c   | 27 +++++++++++++++++++++++++++
 2 files changed, 28 insertions(+)

diff --git a/src/soc/intel/baytrail/baytrail/iomap.h b/src/soc/intel/baytrail/baytrail/iomap.h
index fa34105..d46a0fe 100644
--- a/src/soc/intel/baytrail/baytrail/iomap.h
+++ b/src/soc/intel/baytrail/baytrail/iomap.h
@@ -31,6 +31,7 @@
 #define PUNIT_BASE_ADDRESS		0xfed05000
 #define RCBA_BASE_ADDRESS		0xfed1c000
 #define HPET_BASE_ADDRESS		0xfed00000
+#define TEMP_BASE_ADDRESS		0xfd000000
 
 /* IO Port base */
 #define ACPI_BASE_ADDRESS		0x0400
diff --git a/src/soc/intel/baytrail/southcluster.c b/src/soc/intel/baytrail/southcluster.c
index 42349fa..c71ab7e 100644
--- a/src/soc/intel/baytrail/southcluster.c
+++ b/src/soc/intel/baytrail/southcluster.c
@@ -238,10 +238,37 @@ static inline void set_d3hot_bits(device_t dev, int offset)
 	pci_write_config8(dev, offset + 4, reg8);
 }
 
+/* Parts of the audio subsystem are powered by the HDA device. Therefore, one
+ * cannot put HDA into D3Hot. Instead perform this workaround to make some of
+ * the audio paths work for LPE audio. */
+static void hda_work_around(device_t dev)
+{
+	unsigned long gctl = TEMP_BASE_ADDRESS + 0x8;
+
+	/* Need to set magic register 0x43 to 0xd7 in config space. */
+	pci_write_config8(dev, 0x43, 0xd7);
+
+	/* Need to set bit 0 of GCTL to take the device out of reset. However,
+	 * that requires setting up the 64-bit BAR. */
+	pci_write_config32(dev, PCI_BASE_ADDRESS_0, TEMP_BASE_ADDRESS);
+	pci_write_config32(dev, PCI_BASE_ADDRESS_1, 0);
+	pci_write_config8(dev, PCI_COMMAND, PCI_COMMAND_MEMORY);
+	write32(gctl, read32(gctl) | 0x1);
+	pci_write_config8(dev, PCI_COMMAND, 0);
+	pci_write_config32(dev, PCI_BASE_ADDRESS_0, 0);
+}
+
 static int place_device_in_d3hot(device_t dev)
 {
 	unsigned offset;
 
+	/* Parts of the HDA block are used for LPE audio as well.
+	 * Therefore assume the HDA will never be put into D3Hot. */
+	if (dev->path.pci.devfn == PCI_DEVFN(HDA_DEV, HDA_FUNC)) {
+		hda_work_around(dev);
+		return 0;
+	}
+
 	offset = pci_find_capability(dev, PCI_CAP_ID_PM);
 
 	if (offset != 0) {