svn commit: r350494 - head/sys/powerpc/powernv

[Justin Hibbits]

Author: jhibbits
Date: Thu Aug  1 03:59:25 2019
New Revision: 350494
URL: https://svnweb.freebsd.org/changeset/base/350494

Log:
  powerpc/powernv: Only clear EEH freeze for some errors
  
  Only clear an EEH freeze if an error occurs.  However, if an OPAL_HARDWARE
  error is returned, this indicates a hardware failure which cannot be
  unfrozen, and instead needs a hardware reset.  Attempting to unfreeze a
  broken PCH will result in console spam for each attempt.  To avoid the spam,
  just don't do it.

Modified:
  head/sys/powerpc/powernv/opal.h
  head/sys/powerpc/powernv/opal_pci.c

Modified: head/sys/powerpc/powernv/opal.h
==============================================================================
--- head/sys/powerpc/powernv/opal.h	Thu Aug  1 03:55:58 2019	(r350493)
+++ head/sys/powerpc/powernv/opal.h	Thu Aug  1 03:59:25 2019	(r350494)
@@ -53,6 +53,7 @@ int opal_call(uint64_t token, ...);
 #define	OPAL_PCI_CONFIG_WRITE_BYTE	16
 #define	OPAL_PCI_CONFIG_WRITE_HALF_WORD	17
 #define	OPAL_PCI_CONFIG_WRITE_WORD	18
+#define	OPAL_PCI_EEH_FREEZE_STATUS	23
 #define	OPAL_PCI_EEH_FREEZE_CLEAR	26
 #define	OPAL_PCI_PHB_MMIO_ENABLE	27
 #define	OPAL_PCI_SET_PHB_MEM_WINDOW	28

Modified: head/sys/powerpc/powernv/opal_pci.c
==============================================================================
--- head/sys/powerpc/powernv/opal_pci.c	Thu Aug  1 03:55:58 2019	(r350493)
+++ head/sys/powerpc/powernv/opal_pci.c	Thu Aug  1 03:59:25 2019	(r350494)
@@ -117,6 +117,8 @@ static bus_dma_tag_t opalpci_get_dma_tag(device_t dev,
 #define	OPAL_EEH_ACTION_CLEAR_FREEZE_DMA	2
 #define	OPAL_EEH_ACTION_CLEAR_FREEZE_ALL	3
 
+#define	OPAL_EEH_STOPPED_NOT_FROZEN		0
+
 /*
  * Constants
  */
@@ -501,10 +503,11 @@ opalpci_read_config(device_t dev, u_int bus, u_int slo
 {
 	struct opalpci_softc *sc;
 	uint64_t config_addr;
-	uint8_t byte;
+	uint8_t byte, eeh_state;
 	uint16_t half;
 	uint32_t word;
 	int error;
+	uint16_t err_type;
 
 	sc = device_get_softc(dev);
 
@@ -536,11 +539,19 @@ opalpci_read_config(device_t dev, u_int bus, u_int slo
 	 *
 	 * XXX: Make this conditional on the existence of a freeze
 	 */
-	opal_call(OPAL_PCI_EEH_FREEZE_CLEAR, sc->phb_id, OPAL_PCI_DEFAULT_PE,
-	    OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
 	
-	if (error != OPAL_SUCCESS)
+	if (error != OPAL_SUCCESS) {
+		if (error != OPAL_HARDWARE) {
+			opal_call(OPAL_PCI_EEH_FREEZE_STATUS, sc->phb_id,
+			    OPAL_PCI_DEFAULT_PE, vtophys(&eeh_state),
+			    vtophys(&err_type), NULL);
+			if (eeh_state != OPAL_EEH_STOPPED_NOT_FROZEN)
+				opal_call(OPAL_PCI_EEH_FREEZE_CLEAR,
+				    sc->phb_id, OPAL_PCI_DEFAULT_PE,
+				    OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
+		}
 		word = 0xffffffff;
+	}
 
 	return (word);
 }
@@ -552,6 +563,8 @@ opalpci_write_config(device_t dev, u_int bus, u_int sl
 	struct opalpci_softc *sc;
 	uint64_t config_addr;
 	int error = OPAL_SUCCESS;
+	uint16_t err_type;
+	uint8_t eeh_state;
 
 	sc = device_get_softc(dev);
 
@@ -577,8 +590,15 @@ opalpci_write_config(device_t dev, u_int bus, u_int sl
 		 * Poking config state for non-existant devices can make
 		 * the host bridge hang up. Clear any errors.
 		 */
-		opal_call(OPAL_PCI_EEH_FREEZE_CLEAR, sc->phb_id,
-		    OPAL_PCI_DEFAULT_PE, OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
+		if (error != OPAL_HARDWARE) {
+			opal_call(OPAL_PCI_EEH_FREEZE_STATUS, sc->phb_id,
+			    OPAL_PCI_DEFAULT_PE, vtophys(&eeh_state),
+			    vtophys(&err_type), NULL);
+			if (eeh_state != OPAL_EEH_STOPPED_NOT_FROZEN)
+				opal_call(OPAL_PCI_EEH_FREEZE_CLEAR,
+				    sc->phb_id, OPAL_PCI_DEFAULT_PE,
+				    OPAL_EEH_ACTION_CLEAR_FREEZE_ALL);
+		}
 	}
 }