Fix slowdowns and crashes for AMD GPUs in pre-PCIe-v3 slots

2 files changed, 86 insertions, 0 deletions

diff --git a/PCI-Fix-incorrect-value-returned-from-pcie_get_speed.patch b/PCI-Fix-incorrect-value-returned-from-pcie_get_speed.patch
new file mode 100644
index 000000000..6735da020
--- /dev/null
+++ b/PCI-Fix-incorrect-value-returned-from-pcie_get_speed.patch
@@ -0,0 +1,81 @@
+From b8f56153d6d66e01e52f9f703ab0c6f8c8637354 Mon Sep 17 00:00:00 2001
+From: Mikulas Patocka <mpatocka@redhat.com>
+Date: Thu, 29 Nov 2018 17:24:17 -0600
+Subject: [PATCH] PCI: Fix incorrect value returned from pcie_get_speed_cap()
+
+The macros PCI_EXP_LNKCAP_SLS_*GB are values, not bit masks.  We must mask
+the register and compare it against them.
+
+This fixes errors like this:
+
+  amdgpu: [powerplay] failed to send message 261 ret is 0
+
+when a PCIe-v3 card is plugged into a PCIe-v1 slot, because the slot is
+being incorrectly reported as PCIe-v3 capable.
+
+6cf57be0f78e, which appeared in v4.17, added pcie_get_speed_cap() with the
+incorrect test of PCI_EXP_LNKCAP_SLS as a bitmask.  5d9a63304032, which
+appeared in v4.19, changed amdgpu to use pcie_get_speed_cap(), so the
+amdgpu bug reports below are regressions in v4.19.
+
+Fixes: 6cf57be0f78e ("PCI: Add pcie_get_speed_cap() to find max supported link speed")
+Fixes: 5d9a63304032 ("drm/amdgpu: use pcie functions for link width and speed")
+Link: https://bugs.freedesktop.org/show_bug.cgi?id=108704
+Link: https://bugs.freedesktop.org/show_bug.cgi?id=108778
+Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
+[bhelgaas: update comment, remove use of PCI_EXP_LNKCAP_SLS_8_0GB and
+PCI_EXP_LNKCAP_SLS_16_0GB since those should be covered by PCI_EXP_LNKCAP2,
+remove test of PCI_EXP_LNKCAP for zero, since that register is required]
+Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
+Acked-by: Alex Deucher <alexander.deucher@amd.com>
+Cc: stable@vger.kernel.org	# v4.17+
+Signed-off-by: Jeremy Cline <jcline@redhat.com>
+---
+ drivers/pci/pci.c | 24 +++++++++++-------------
+ 1 file changed, 11 insertions(+), 13 deletions(-)
+
+diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
+index d068f11d08a7..c9d8e3c837de 100644
+--- a/drivers/pci/pci.c
++++ b/drivers/pci/pci.c
+@@ -5556,9 +5556,13 @@ enum pci_bus_speed pcie_get_speed_cap(struct pci_dev *dev)
+ 	u32 lnkcap2, lnkcap;
+ 
+ 	/*
+-	 * PCIe r4.0 sec 7.5.3.18 recommends using the Supported Link
+-	 * Speeds Vector in Link Capabilities 2 when supported, falling
+-	 * back to Max Link Speed in Link Capabilities otherwise.
++	 * Link Capabilities 2 was added in PCIe r3.0, sec 7.8.18.  The
++	 * implementation note there recommends using the Supported Link
++	 * Speeds Vector in Link Capabilities 2 when supported.
++	 *
++	 * Without Link Capabilities 2, i.e., prior to PCIe r3.0, software
++	 * should use the Supported Link Speeds field in Link Capabilities,
++	 * where only 2.5 GT/s and 5.0 GT/s speeds were defined.
+ 	 */
+ 	pcie_capability_read_dword(dev, PCI_EXP_LNKCAP2, &lnkcap2);
+ 	if (lnkcap2) { /* PCIe r3.0-compliant */
+@@ -5574,16 +5578,10 @@ enum pci_bus_speed pcie_get_speed_cap(struct pci_dev *dev)
+ 	}
+ 
+ 	pcie_capability_read_dword(dev, PCI_EXP_LNKCAP, &lnkcap);
+-	if (lnkcap) {
+-		if (lnkcap & PCI_EXP_LNKCAP_SLS_16_0GB)
+-			return PCIE_SPEED_16_0GT;
+-		else if (lnkcap & PCI_EXP_LNKCAP_SLS_8_0GB)
+-			return PCIE_SPEED_8_0GT;
+-		else if (lnkcap & PCI_EXP_LNKCAP_SLS_5_0GB)
+-			return PCIE_SPEED_5_0GT;
+-		else if (lnkcap & PCI_EXP_LNKCAP_SLS_2_5GB)
+-			return PCIE_SPEED_2_5GT;
+-	}
++	if ((lnkcap & PCI_EXP_LNKCAP_SLS) == PCI_EXP_LNKCAP_SLS_5_0GB)
++		return PCIE_SPEED_5_0GT;
++	else if ((lnkcap & PCI_EXP_LNKCAP_SLS) == PCI_EXP_LNKCAP_SLS_2_5GB)
++		return PCIE_SPEED_2_5GT;
+ 
+ 	return PCI_SPEED_UNKNOWN;
+ }
+-- 
+2.19.2
+
diff --git a/kernel.spec b/kernel.spec
index 661641eda..3c58c9178 100644
--- a/kernel.spec
+++ b/kernel.spec
@@ -641,6 +641,10 @@ Patch513: CVE-2018-19407.patch
 # rhbz 1650984, in linux-next and Cc'd for stable
 Patch514: net-phy-add-workaround-for-issue-where-PHY-driver-do.patch
 
+# In the PCI tree and Cc'd for stable, fixes an issue with amdgpu
+# https://patchwork.freedesktop.org/patch/259364/
+Patch515: PCI-Fix-incorrect-value-returned-from-pcie_get_speed.patch
+
 # END OF PATCH DEFINITIONS
 
 %endif
@@ -1904,6 +1908,7 @@ fi
 %changelog
 * Thu Nov 29 2018 Jeremy Cline <jeremy@jcline.org>
 - Fix a problem with some rtl8168 chips (rhbz 1650984)
+- Fix slowdowns and crashes for AMD GPUs in pre-PCIe-v3 slots
 
 * Tue Nov 27 2018 Jeremy Cline <jcline@redhat.com> - 4.19.5-300
 - Linux v4.19.5