We have a couple of workarounds available for the ECAM ghosting issue
that affects the Synopsys Designware PCIe RCs. First of all, we can be
optimistic and hope that the silicon gets fixed at some point. Then,
there is a SCP firmware hack that hides these ghosts by remapping the
ECAM region using the SMMU sitting between the CPU and the PCIe RC
slave interface. Finally, we have a workaround involving stage 2
translation tables that may be enabled at will using a DIP switch on
the board.

Instead of adding elaborate logic to infer which of these situations we
may find ourselves in, let's just test for the symptom directly in the
_STA method implementation of the PNP0A08 devices, and deactivate the
device if the ECAM space does not appear sane.

Contributed-under: TianoCore Contribution Agreement 1.1
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
 Silicon/Socionext/SynQuacer/AcpiTables/AcpiSsdtRootPci.asl | 40 ++++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/Silicon/Socionext/SynQuacer/AcpiTables/AcpiSsdtRootPci.asl b/Silicon/Socionext/SynQuacer/AcpiTables/AcpiSsdtRootPci.asl
index 1735264f09a3..5ffed663e17d 100644
--- a/Silicon/Socionext/SynQuacer/AcpiTables/AcpiSsdtRootPci.asl
+++ b/Silicon/Socionext/SynQuacer/AcpiTables/AcpiSsdtRootPci.asl
@@ -31,6 +31,26 @@ DefinitionBlock ("SsdtPci.aml", "SSDT", 1, "SNI", "SYNQUACR",
         Name (_BBN, Zero)                // PCI Base Bus Number
         Name (_CCA, 1)                   // Cache Coherency Attribute
 
+        OperationRegion (BDF1, SystemMemory, SYNQUACER_PCI_SEG0_CONFIG_BASE, 0x10000)
+        Field (BDF1, DWordAcc, NoLock, Preserve) {
+            Offset (0x8000),
+            VPID, 16,
+        }
+
+        Method (_STA, 0x0, Serialized) {
+            //
+            // Check whether the VID/PID of device #1 on bus #0 equals 0xffff.
+            // If this is not the case, we are dealing with a ghost device,
+            // which means we are running with outdated SCP firmware and we
+            // should keep this PCIe RC disabled.
+            //
+            Store (VPID, local1)
+            If (!LEqual (local1, 0xffff)) {
+                Return (0x0)
+            }
+            Return (0xf)
+        }
+
         // PCI Routing Table
         Name (_PRT, Package () {
             Package () { 0xFFFF, 0, Zero, 222 },   // INTA
@@ -149,6 +169,26 @@ DefinitionBlock ("SsdtPci.aml", "SSDT", 1, "SNI", "SYNQUACR",
         Name (_BBN, Zero)                // PCI Base Bus Number
         Name (_CCA, 1)                   // Cache Coherency Attribute
 
+        OperationRegion (BDF1, SystemMemory, SYNQUACER_PCI_SEG1_CONFIG_BASE, 0x10000)
+        Field (BDF1, DWordAcc, NoLock, Preserve) {
+            Offset (0x8000),
+            VPID, 16,
+        }
+
+        Method (_STA, 0x0, Serialized) {
+            //
+            // Check whether the VID/PID of device #1 on bus #0 equals 0xffff.
+            // If this is not the case, we are dealing with a ghost device,
+            // which means we are running with outdated SCP firmware and we
+            // should keep this PCIe RC disabled.
+            //
+            Store (VPID, local1)
+            If (!LEqual (local1, 0xffff)) {
+                Return (0x0)
+            }
+            Return (0xf)
+        }
+
         // PCI Routing Table
         Name (_PRT, Package () {
             Package () { 0xFFFF, 0, Zero, 214 },   // INTA
-- 
2.11.0

_______________________________________________
edk2-devel mailing list
edk2-devel@lists.01.org
https://lists.01.org/mailman/listinfo/edk2-devel