Module Name:    src
Committed By:   martin
Date:           Tue Oct 15 19:33:24 UTC 2019

Modified Files:
        src/sys/dev/pci [netbsd-9]: pci.c

Log Message:
Pull up following revision(s) (requested by jmcneill in ticket #331):

        sys/dev/pci/pci.c: revision 1.155

The PCI Local Bus Specification says that we should probe devices by
reading PCI_ID_REG. pci_enumerate_bus was incorrectly reading PCI_BHLC_REG
first, which surprisingly has worked for the past 16 years. Unfortunately,
this undefined behavior does the wrong thing on Amazon's Annapurna Labs
PCIe host controller.


To generate a diff of this commit:
cvs rdiff -u -r1.154 -r1.154.4.1 src/sys/dev/pci/pci.c

Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.

Modified files:

Index: src/sys/dev/pci/pci.c
diff -u src/sys/dev/pci/pci.c:1.154 src/sys/dev/pci/pci.c:1.154.4.1
--- src/sys/dev/pci/pci.c:1.154	Sat Dec 15 05:38:45 2018
+++ src/sys/dev/pci/pci.c	Tue Oct 15 19:33:23 2019
@@ -1,4 +1,4 @@
-/*	$NetBSD: pci.c,v 1.154 2018/12/15 05:38:45 msaitoh Exp $	*/
+/*	$NetBSD: pci.c,v 1.154.4.1 2019/10/15 19:33:23 martin Exp $	*/
 
 /*
  * Copyright (c) 1995, 1996, 1997, 1998
@@ -36,7 +36,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: pci.c,v 1.154 2018/12/15 05:38:45 msaitoh Exp $");
+__KERNEL_RCSID(0, "$NetBSD: pci.c,v 1.154.4.1 2019/10/15 19:33:23 martin Exp $");
 
 #ifdef _KERNEL_OPT
 #include "opt_pci.h"
@@ -723,10 +723,6 @@ pci_enumerate_bus(struct pci_softc *sc, 
 
 		tag = pci_make_tag(pc, sc->sc_bus, device, 0);
 
-		bhlcr = pci_conf_read(pc, tag, PCI_BHLC_REG);
-		if (PCI_HDRTYPE_TYPE(bhlcr) > 2)
-			continue;
-
 		id = pci_conf_read(pc, tag, PCI_ID_REG);
 
 		/* Invalid vendor ID value? */
@@ -736,6 +732,10 @@ pci_enumerate_bus(struct pci_softc *sc, 
 		if (PCI_VENDOR(id) == 0)
 			continue;
 
+		bhlcr = pci_conf_read(pc, tag, PCI_BHLC_REG);
+		if (PCI_HDRTYPE_TYPE(bhlcr) > 2)
+			continue;
+
 		qd = pci_lookup_quirkdata(PCI_VENDOR(id), PCI_PRODUCT(id));
 
 		if (qd != NULL &&

Reply via email to