6878294 FMA is too aggressive in marking the disk as faulty for zfs relocated the fatal read media error

6895028 libtopo does not work on Netra-X4200-M2 platform
6892864 setserdsuffix() should work with integer type

7 files changed, 168 insertions, 6 deletions

diff --git a/usr/src/cmd/fm/eversholt/files/common/disk.esc b/usr/src/cmd/fm/eversholt/files/common/disk.esc
index 823c605690..ce2a1d8854 100644
--- a/usr/src/cmd/fm/eversholt/files/common/disk.esc
+++ b/usr/src/cmd/fm/eversholt/files/common/disk.esc
@@ -19,27 +19,43 @@
  * CDDL HEADER END
  */
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
 #pragma dictionary "DISK"
 
-#define	P	disk
+#define	P			disk
 
 fru P;
 asru P;
 
 /*
- * disk-as-detector: The disk-as-detector DE provides the mapping between
+ * Over all comments for this file:
+ * <disk-as-detector> The disk-as-detector DE provides the mapping between
  * ereports generated by a kernel disk driver sd(7D) and resulting faults.
  */
 
 /*
+ * SERD engine for media error fault propagation:
+ *
+ * This strategy is designed to give a file system, like ZFS, the
+ * ability to attempt data recovery/relocation without faulting a disk.
+ * This implementation depends on a file system retry to the same lba
+ * to trigger a fault when recovery/relocation is not possible.
+ *
+ * We let the engine propagate one error only once every 1 minute and then if we
+ * still get 2 or more * errors within 24 hours for the same LBA, there is a fault.
+ */
+engine serd.io.scsi.cmd.disk.dev.rqs.merr@P, N=1, T=24h;
+
+/*
  * disk-as-detector: fault events.
  */
 event fault.io.scsi.cmd.disk.dev.rqs.derr@P;
-event fault.io.scsi.cmd.disk.dev.rqs.merr@P;
+event fault.io.scsi.cmd.disk.dev.rqs.merr@P,
+    engine=serd.io.scsi.cmd.disk.dev.rqs.merr@P;
+
 /*
  * The uderr fault will be defined at some future time.
  * event fault.io.scsi.cmd.disk.dev.uderr@P;
@@ -96,14 +112,30 @@ prop fault.io.scsi.cmd.disk.dev.rqs.derr@P->
     setpayloadprop("asc", payloadprop("asc")) &&
     setpayloadprop("ascq", payloadprop("ascq"))};
 
+/*
+ * Utilize setserdsuffix with specific LBA, 
+ * the serd engine would only trigger if the fault recurred on the same LBA
+ */
 prop fault.io.scsi.cmd.disk.dev.rqs.merr@P->
     ereport.io.scsi.cmd.disk.dev.rqs.merr@P{ DRIVER_ASSESSMENT_FATAL &&
+    setserdsuffix(payloadprop("lba")) &&
     setpayloadprop("key", payloadprop("key")) &&
     setpayloadprop("asc", payloadprop("asc")) &&
     setpayloadprop("ascq", payloadprop("ascq")) &&
     setpayloadprop("lba", payloadprop("lba"))};
 
 /*
+ * NOTE: this propagation uses the "may" propagation of eversholt.
+ * The ereport need never exist. It's just a way of making
+ * the diagnosis wait for the within time on that ereport
+ * to complete. Once it has completed the diagnosis continues
+ * even though the dummy ereport didn't occur.
+ */
+event ereport.io.scsi.cmd.disk.dev.rqs.merr.dummy@P {within(60s)};
+prop fault.io.scsi.cmd.disk.dev.rqs.merr@P (0) ->
+	ereport.io.scsi.cmd.disk.dev.rqs.merr.dummy@P;
+
+/*
  * The uderr fault will be propagated at some future time.
  * prop fault.io.scsi.cmd.disk.dev.uderr@P->
  *     ereport.io.scsi.cmd.disk.dev.uderr@P{ DRIVER_ASSESSMENT_FATAL };
diff --git a/usr/src/cmd/fm/modules/common/eversholt/eval.c b/usr/src/cmd/fm/modules/common/eversholt/eval.c
index f7bd53227c..54b24f8513 100644
--- a/usr/src/cmd/fm/modules/common/eversholt/eval.c
+++ b/usr/src/cmd/fm/modules/common/eversholt/eval.c
@@ -649,6 +649,18 @@ eval_func(struct node *funcnp, struct lut *ex, struct node *events[],
 				serdvalp->v = (uintptr_t)stable(buf);
 				FREE(buf);
 			}
+			if (funcname == L_setserdsuffix &&
+			    serdvalp->t == UINT64) {
+				int len = snprintf(NULL, 0, "%lld",
+				    serdvalp->v);
+				char *buf = MALLOC(len + 1);
+
+				(void) snprintf(buf, len + 1, "%lld",
+				    serdvalp->v);
+				serdvalp->t = STRING;
+				serdvalp->v = (uintptr_t)stable(buf);
+				FREE(buf);
+			}
 
 			if (serdvalp->t == UINT64)
 				out(O_ALTFP|O_VERB2, " (%llu)", serdvalp->v);
diff --git a/usr/src/lib/fm/topo/maps/Makefile b/usr/src/lib/fm/topo/maps/Makefile
index c22647c404..de6d9f5e0f 100644
--- a/usr/src/lib/fm/topo/maps/Makefile
+++ b/usr/src/lib/fm/topo/maps/Makefile
@@ -45,6 +45,7 @@ i386_SUBDIRS = i86pc \
 		i386 \
 		SUNW,Sun-Fire-X4200-Server \
 		SUNW,Sun-Fire-X4200-M2 \
+		SUNW,Netra-X4200-M2 \
 		SUNW,Sun-Fire-X4500 \
 		SUNW,Sun-Fire-X4540 \
 		SUNW,Sun-Fire-X4600-M2
diff --git a/usr/src/lib/fm/topo/maps/SUNW,Netra-X4200-M2/Makefile b/usr/src/lib/fm/topo/maps/SUNW,Netra-X4200-M2/Makefile
new file mode 100644
index 0000000000..1a84db38f4
--- /dev/null
+++ b/usr/src/lib/fm/topo/maps/SUNW,Netra-X4200-M2/Makefile
@@ -0,0 +1,40 @@
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or http://www.opensolaris.org/os/licensing.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+# Use is subject to license terms.
+#
+
+# NOTE: The name of the xml file we are building is 'platform'
+# specific, but its build is structured as 'arch' specific since
+# 'uname -i' on all x86 platforms returns i86pc.
+
+ARCH = i86pc
+CLASS = arch
+DTDFILE = topology.dtd.1
+TOPOFILE = Netra-X4200-M2-disk-hc-topology.xml
+SRCDIR = ../SUNW,Netra-X4200-M2
+
+PLATFORM = Netra-X4200-M2
+TOPOBASE = ../i86pc/i86pc-hc-topology.xml
+
+include ../Makefile.map
diff --git a/usr/src/lib/fm/topo/maps/SUNW,Netra-X4200-M2/Netra-X4200-M2-disk-hc-topology.xmlgen b/usr/src/lib/fm/topo/maps/SUNW,Netra-X4200-M2/Netra-X4200-M2-disk-hc-topology.xmlgen
new file mode 100644
index 0000000000..90d37aac60
--- /dev/null
+++ b/usr/src/lib/fm/topo/maps/SUNW,Netra-X4200-M2/Netra-X4200-M2-disk-hc-topology.xmlgen
@@ -0,0 +1,69 @@
+#!/usr/local/bin/perl
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or http://www.opensolaris.org/os/licensing.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+#
+# Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
+# Use is subject to license terms.
+#
+
+$num_bays = 4;
+$bay_label = "HD";
+
+print <<EOF;
+<topology name='disk' scheme='hc'>
+  <range name='bay' min='0' max='3'>
+EOF
+
+$controller = 0;
+for ($bay = 0; $bay < $num_bays; $bay++) {
+	$hpath = "/pci\@79,0/pci1022,7458\@11/pci1000,3060\@2";
+	$tpath = sprintf("/sd\@%x,0", $bay);
+	$apoint = sprintf(":scsi::dsk/c%dt%dd0",
+	    $controller + 1, $bay);
+
+	print <<EOF;
+    <node instance='$bay'>
+      <propgroup name='protocol' version='1' name-stability='Private'
+        data-stability='Private'>
+	<propval name='label' type='string' value='$bay_label$bay' />
+      </propgroup>
+      <propgroup name='io' version='1' name-stability='Private'
+        data-stability='Private'>
+	<propval name='ap-path' type='string' value='/devices$hpath$apoint' />
+      </propgroup>
+      <propgroup name='binding' version='1' name-stability='Private'
+        data-stability='Private'>
+	<propval name='occupant-path' type='string' 
+	  value='$hpath$tpath' />
+      </propgroup>
+    </node>
+EOF
+}
+
+print <<EOF;
+    <dependents grouping='children'>
+      <range name='disk' min='0' max='0'>
+	<enum-method name='disk' version='1' />
+      </range>
+    </dependents>
+  </range>
+</topology>
+EOF
diff --git a/usr/src/lib/fm/topo/maps/i86pc/i86pc-legacy-hc-topology.xml b/usr/src/lib/fm/topo/maps/i86pc/i86pc-legacy-hc-topology.xml
index d6ef9b2e07..00b771f66d 100644
--- a/usr/src/lib/fm/topo/maps/i86pc/i86pc-legacy-hc-topology.xml
+++ b/usr/src/lib/fm/topo/maps/i86pc/i86pc-legacy-hc-topology.xml
@@ -34,7 +34,8 @@
 
   <range name='motherboard' min='0' max='0'>
     <node instance='0'>
-      <set type='product' setlist='Sun-Fire-X4100-Server|Sun-Fire-X4200-Server|Sun-Fire-X4100-M2|Sun-Fire-X4200-M2|Sun-Fire-X4140|Sun-Fire-X4240|Sun-Fire-X4440|Sun-Fire-X4500|Sun-Fire-X4540|Sun-Fire-X4600-M2'>
+      <set type='product'
+		  setlist='Sun-Fire-X4100-Server|Sun-Fire-X4200-Server|Sun-Fire-X4100-M2|Sun-Fire-X4200-M2|Netra-X4200-M2|Sun-Fire-X4140|Sun-Fire-X4240|Sun-Fire-X4440|Sun-Fire-X4500|Sun-Fire-X4540|Sun-Fire-X4600-M2'>
         <fac-enum provider='fac_prov_ipmi' />
         <propgroup name='protocol' version='1'
             name-stability='Private' data-stability='Private' >
@@ -86,7 +87,8 @@
     
     <dependents grouping='children'>
     
-    <set type='product' setlist='Sun-Fire-X4100-Server|Sun-Fire-X4200-Server|Sun-Fire-X4100-M2|Sun-Fire-X4200-M2|Sun-Fire-X4140|Sun-Fire-X4240|Sun-Fire-X4440'>
+    <set type='product'
+		setlist='Sun-Fire-X4100-Server|Sun-Fire-X4200-Server|Sun-Fire-X4100-M2|Sun-Fire-X4200-M2|Netra-X4200-M2|Sun-Fire-X4140|Sun-Fire-X4240|Sun-Fire-X4440'>
        <range name='psu' min='0' max='100'>
    	  <enum-method name='ipmi' version='1' />
    	  <propmap name='psu' />
@@ -104,6 +106,11 @@
              <propmap name='Sun-Fire-X4200-Server-disk' />
 	  </range>
        </set>
+       <set type='product' setlist='Netra-X4200-M2'>
+	  <range name='bay' min='0' max='3'>
+             <propmap name='Netra-X4200-M2-disk' />
+	  </range>
+       </set>
        <set type='product' setlist='default'>
 	  <range name='bay' min='0' max='1024'>
              <enum-method name='ses' version='1' />
diff --git a/usr/src/pkgdefs/SUNWfmd/prototype_i386 b/usr/src/pkgdefs/SUNWfmd/prototype_i386
index 764a92f156..6acf321c3f 100644
--- a/usr/src/pkgdefs/SUNWfmd/prototype_i386
+++ b/usr/src/pkgdefs/SUNWfmd/prototype_i386
@@ -92,6 +92,7 @@ f none usr/platform/i86pc/lib/fm/topo/maps/i86pc-legacy-hc-topology.xml 444 root
 f none usr/platform/i86pc/lib/fm/topo/maps/psu-hc-topology.xml 444 root bin
 f none usr/platform/i86pc/lib/fm/topo/maps/Sun-Fire-X4200-M2-disk-hc-topology.xml 444 root bin
 f none usr/platform/i86pc/lib/fm/topo/maps/Sun-Fire-X4200-Server-disk-hc-topology.xml 444 root bin
+f none usr/platform/i86pc/lib/fm/topo/maps/Netra-X4200-M2-disk-hc-topology.xml 444 root bin
 f none usr/platform/i86pc/lib/fm/topo/maps/Sun-Fire-X4500-disk-hc-topology.xml 444 root bin
 f none usr/platform/i86pc/lib/fm/topo/maps/Sun-Fire-X4540-disk-hc-topology.xml 444 root bin
 f none usr/platform/i86pc/lib/fm/topo/maps/Sun-Fire-X4600-M2-disk-hc-topology.xml 444 root bin