Skip to contentSkip to conversation

x86, quirks: Add workaround for AMD F16h Erratum792

PATCH New Aravind Gopalakrishnan 2 participants 2 messages Jan 23, 2014 – Jan 25, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 x86, quirks: Add workaround for AMD F16h Erratum792

The workaround for this Erratum is included in AGESA. But BIOSes spun only after Jan2014 will have the fix (atleast server versions of the chip). The erratum affects both embedded and server platforms and since we cannot say with certainity that ALL BIOSes on systems out in the field will have the fix, we should probably insulate ourselves in case BIOS does not do the right thing or someone is using old BIOSes.

Refer Revision Guide for AMD F16h models 00h-0fh, document 51810 Rev. 3.04, November2013 for details on the Erratum.

Tested the patch on Fam16h server platform and works fine.

Changes from V1:
- modify commit msg to say it affects embedded and server systems..
- per Ingo's suggestion: reword comment, function name
- rebase off latest tip/master

---
 arch/x86/kernel/quirks.c |   36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)
arch/x86/kernel/quirks.c +36 -0
index 04ee1e2..50ce52f 100644
@@ -571,3 +571,39 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_15H_NB_F5,
 			quirk_amd_nb_node);
 
 #endif
+
+#if defined(CONFIG_PCI)
+/*
+ * Processor does not ensure DRAM scrub read/write sequence
+ * is atomic wrt accesses to CC6 save state area. Therefore
+ * if a concurrent scrub read/write access is to same address
+ * the entry may appear as if it is not written. This quirk
+ * applies to Fam16h models 00h-0Fh
+ *
+ * see Revision Guide for AMD F16h models 00h-0fh,
+ * document 51810 rev. 3.04, Nov 2013
+ */
+static void amd_disable_seq_and_redirect_scrub(struct pci_dev *dev)
+{
+	u32 val;
+
+	/* Suggested workaround:
+	 * set D18F3x58[4:0] = 00h and set D18F3x5C[0] = 0b
+	 */
+	pci_read_config_dword(dev, 0x58, &val);
+	if (val & 0x1F) {
+		val &= ~(0x1F);
+		pci_write_config_dword(dev, 0x58, val);
+	}
+
+	pci_read_config_dword(dev, 0x5C, &val);
+	if (val & BIT(0)) {
+		val &= ~BIT(0);
+		pci_write_config_dword(dev, 0x5c, val);
+	}
+}
+
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_16H_NB_F3,
+			amd_disable_seq_and_redirect_scrub);
+
+#endif
Signature · 1 line
1.7.9.5
#1 lore raw

[tip:x86/urgent] x86/quirks: Add workaround for AMD F16h Erratum792

x86/quirks: Add workaround for AMD F16h Erratum792
Commit-ID:  fb53a1ab88d14848dc292842e35c3bda3a665997
Gitweb:     http://git.kernel.org/tip/fb53a1ab88d14848dc292842e35c3bda3a665997
Author:     Aravind Gopalakrishnan <[email protected]>
AuthorDate: Thu, 23 Jan 2014 16:13:32 -0600
Committer:  Ingo Molnar <[email protected]>
CommitDate: Sat, 25 Jan 2014 08:44:19 +0100

x86/quirks: Add workaround for AMD F16h Erratum792

The workaround for this Erratum is included in AGESA. But BIOSes spun only after Jan2014 will have the fix (atleast server versions of the chip). The erratum affects both embedded and server platforms and since we cannot say with certainity that ALL BIOSes on systems out in the field will have the fix, we should probably insulate ourselves in case BIOS does not do the right thing or someone is using old BIOSes.

Refer to Revision Guide for AMD F16h models 00h-0fh, document 51810 Rev. 3.04, November2013 for details on the Erratum.

Tested the patch on Fam16h server platform and it works fine.

---
 arch/x86/kernel/quirks.c | 37 +++++++++++++++++++++++++++++++++++++
 1 file changed, 37 insertions(+)
arch/x86/kernel/quirks.c +37 -0
index 04ee1e2..7c6acd4 100644
@@ -571,3 +571,40 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_15H_NB_F5,
 			quirk_amd_nb_node);
 
 #endif
+
+#ifdef CONFIG_PCI
+/*
+ * Processor does not ensure DRAM scrub read/write sequence
+ * is atomic wrt accesses to CC6 save state area. Therefore
+ * if a concurrent scrub read/write access is to same address
+ * the entry may appear as if it is not written. This quirk
+ * applies to Fam16h models 00h-0Fh
+ *
+ * See "Revision Guide" for AMD F16h models 00h-0fh,
+ * document 51810 rev. 3.04, Nov 2013
+ */
+static void amd_disable_seq_and_redirect_scrub(struct pci_dev *dev)
+{
+	u32 val;
+
+	/*
+	 * Suggested workaround:
+	 * set D18F3x58[4:0] = 00h and set D18F3x5C[0] = 0b
+	 */
+	pci_read_config_dword(dev, 0x58, &val);
+	if (val & 0x1F) {
+		val &= ~(0x1F);
+		pci_write_config_dword(dev, 0x58, val);
+	}
+
+	pci_read_config_dword(dev, 0x5C, &val);
+	if (val & BIT(0)) {
+		val &= ~BIT(0);
+		pci_write_config_dword(dev, 0x5c, val);
+	}
+}
+
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_16H_NB_F3,
+			amd_disable_seq_and_redirect_scrub);
+
+#endif