salvager: Error volumes on GetInodeSummary errors

When GetInodeSummary fails due to an internal failure (not from just
failing to find applicable inodes), currently it just returns an
error, and does not return the checked-out singleVolumeNumber back to
the fileserver.

When we fail to gather inodes, we should force the volume to an error
state, since we haven't salvaged the volume. But if we fail to find
any applicable inodes, we just want to VOL_DONE the volume, since the
header has possibly been destroyed, and the volume doesn't exist.

So, issue an FSYNC_VOL_FORCE_ERROR command when we encounter errors in
GetInodeSummary, except when we fail to find applicable inodes.

Reviewed-on: http://gerrit.openafs.org/4439
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 8084fc28a3069949bdc788f048fa8a96c012b96d)

Change-Id: I64e8e1ac8e3386d09a3bcd7696a2d0f70ded33b0
Reviewed-on: http://gerrit.openafs.org/9459
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Stephan Wiesand <stephan.wiesand@desy.de>

diff --git a/src/vol/vol-salvage.c b/src/vol/vol-salvage.c
index 4bcdc6d2844c69c3163387f413c4d300f5663041..0a712943e8f00bbd6038a47200929e1f69f8da9f 100644 (file)
--- a/src/vol/vol-salvage.c
+++ b/src/vol/vol-salvage.c
@@ -182,6 +182,7 @@ Vnodes with 0 inode pointers in RW volumes are now deleted.
 #include "volume.h"
 #include "partition.h"
 #include "daemon_com.h"
+#include "daemon_com_inline.h"
 #include "fssync.h"
 #include "fssync_inline.h"
 #include "volume_inline.h"
@@ -320,6 +321,7 @@ static int IsVnodeOrphaned(struct SalvInfo *salvinfo, VnodeId vnode);
 static int AskVolumeSummary(struct SalvInfo *salvinfo,
                             VolumeId singleVolumeNumber);
 static void MaybeAskOnline(struct SalvInfo *salvinfo, VolumeId volumeId);
+static void AskError(struct SalvInfo *salvinfo, VolumeId volumeId);
 
 #if defined(AFS_DEMAND_ATTACH_FS) || defined(AFS_DEMAND_ATTACH_UTIL)
 static int LockVolume(struct SalvInfo *salvinfo, VolumeId volumeId);
@@ -1186,6 +1188,8 @@ GetInodeSummary(struct SalvInfo *salvinfo, FILE *inodeFile, VolumeId singleVolum
     char *part = salvinfo->fileSysPath;
     char *tdir;
     int i;
+    int retcode = 0;
+    int deleted = 0;
 
     /* This file used to come from vfsck; cobble it up ourselves now... */
     if ((err =
@@ -1194,7 +1198,8 @@ GetInodeSummary(struct SalvInfo *salvinfo, FILE *inodeFile, VolumeId singleVolum
                        singleVolumeNumber, &forceSal, forceR, wpath, NULL)) < 0) {
        if (err == -2) {
            Log("*** I/O error %d when writing a tmp inode file; Not salvaged %s ***\nIncrease space on partition or use '-tmpdir'\n", errno, dev);
-           return -1;
+           retcode = -1;
+           goto error;
        }
        Abort("Unable to get inodes for \"%s\"; not salvaged\n", dev);
     }
@@ -1248,17 +1253,38 @@ GetInodeSummary(struct SalvInfo *salvinfo, FILE *inodeFile, VolumeId singleVolum
            else {
                struct VolumeSummary *vsp;
                int i;
+               int foundSVN = 0;
 
                GetVolumeSummary(salvinfo, singleVolumeNumber);
 
                for (i = 0, vsp = salvinfo->volumeSummaryp; i < salvinfo->nVolumes; i++) {
-                   if (vsp->fileName)
+                   if (vsp->fileName) {
+                       if (vsp->header.id == singleVolumeNumber) {
+                           foundSVN = 1;
+                       }
                        DeleteExtraVolumeHeaderFile(salvinfo, vsp);
+                   }
+               }
+
+               if (!foundSVN) {
+                   if (Testing) {
+                       MaybeAskOnline(salvinfo, singleVolumeNumber);
+                   } else {
+                       /* make sure we get rid of stray .vol headers, even if
+                        * they're not in our volume summary (might happen if
+                        * e.g. something else created them and they're not in the
+                        * fileserver VGC) */
+                       VDestroyVolumeDiskHeader(salvinfo->fileSysPartition,
+                                                singleVolumeNumber, 0 /*parent*/);
+                       AskDelete(salvinfo, singleVolumeNumber);
+                   }
                }
            }
            Log("%s vice inodes on %s; not salvaged\n",
                singleVolumeNumber ? "No applicable" : "No", dev);
-           return -1;
+           retcode = -1;
+           deleted = 1;
+           goto error;
        }
        ip = (struct ViceInodeInfo *)malloc(nInodes*sizeof(struct ViceInodeInfo));
        if (ip == NULL) {
@@ -1284,7 +1310,8 @@ GetInodeSummary(struct SalvInfo *salvinfo, FILE *inodeFile, VolumeId singleVolum
            if (fwrite(&summary, sizeof(summary), 1, summaryFile) != 1) {
                Log("Difficulty writing summary file (errno = %d); %s not salvaged\n", errno, dev);
                fclose(summaryFile);
-               return -1;
+               retcode = -1;
+               goto error;
            }
            summary.index += (summary.nInodes);
            nInodes -= summary.nInodes;
@@ -1296,7 +1323,8 @@ GetInodeSummary(struct SalvInfo *salvinfo, FILE *inodeFile, VolumeId singleVolum
        if (fflush(summaryFile) == EOF || fsync(fileno(summaryFile)) == -1) {
            Log("Unable to write summary file (errno = %d); %s not salvaged\n", errno, dev);
            fclose(summaryFile);
-           return -1;
+           retcode = -1;
+           goto error;
        }
        if (canfork && !debug) {
            ShowLog = 0;
@@ -1325,7 +1353,13 @@ GetInodeSummary(struct SalvInfo *salvinfo, FILE *inodeFile, VolumeId singleVolum
     }
     Log("%d nVolumesInInodeFile %lu \n",salvinfo->nVolumesInInodeFile,(unsigned long)(status.st_size));
     fclose(summaryFile);
-    return 0;
+
+ error:
+    if (retcode && singleVolumeNumber && !deleted) {
+       AskError(salvinfo, singleVolumeNumber);
+    }
+
+    return retcode;
 }
 
 /* Comparison routine for volume sort.
@@ -4380,6 +4414,21 @@ LockVolume(struct SalvInfo *salvinfo, VolumeId volumeId)
 }
 #endif /* AFS_DEMAND_ATTACH_FS || AFS_DEMAND_ATTACH_UTIL */
 
+static void
+AskError(struct SalvInfo *salvinfo, VolumeId volumeId)
+{
+#if defined(AFS_DEMAND_ATTACH_FS) || defined(AFS_DEMAND_ATTACH_UTIL)
+    afs_int32 code;
+    code = FSYNC_VolOp(volumeId, salvinfo->fileSysPartition->name,
+                       FSYNC_VOL_FORCE_ERROR, FSYNC_WHATEVER, NULL);
+    if (code != SYNC_OK) {
+       Log("AskError: failed to force volume %lu into error state; "
+           "SYNC error code %ld (%s)\n", (long unsigned)volumeId,
+           (long)code, SYNC_res2string(code));
+    }
+#endif /* AFS_DEMAND_ATTACH_FS || AFS_DEMAND_ATTACH_UTIL */
+}
+
 void
 AskOffline(struct SalvInfo *salvinfo, VolumeId volumeId)
 {