If this happens the device crashes and all the config space reports 0xFF
forever. That can make a nasty mess if anything then tries to process the
resulting PCI headers.
If the staging driver is loaded the device is managed via the PUNIT
interface to avoid this erratum.
Signed-off-by: Alan Cox <[email protected]>
---
arch/x86/pci/fixup.c | 13 +++++++++++++
1 file changed, 13 insertions(+)
diff --git a/arch/x86/pci/fixup.c b/arch/x86/pci/fixup.c
index 8f5a4f3d08dc..513f9eb565fe 100644
--- a/arch/x86/pci/fixup.c
+++ b/arch/x86/pci/fixup.c
@@ -637,6 +637,19 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x2031, quirk_no_aersid);
DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x2032, quirk_no_aersid);
DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_INTEL, 0x2033, quirk_no_aersid);
+/*
+ * The BYT and CHT ISP advertises PCI runtime PM but it doesn't work and crashes
+ * the device. Disable so the kernel framework doesn't hang the device trying.
+ * The device itself has to be managed by the PUNIT directly.
+ */
+static void isp_pm_cap_fixup(struct pci_dev *dev)
+{
+ dev_info(&dev->dev, "Disabling PCI power management on camera ISP\n");
+ dev->pm_cap = 0;
+}
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x0f38, isp_pm_cap_fixup);
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x22b8, isp_pm_cap_fixup);
+
#ifdef CONFIG_PHYS_ADDR_T_64BIT
#define AMD_141b_MMIO_BASE(x) (0x80 + (x) * 0x8)
From 1583362661066902184@xxx Mon Nov 06 23:51:43 +0000 2017
X-GM-THRID: 1583362661066902184
X-Gmail-Labels: Inbox,Category Forums,HistoricalUnread