Blame SOURCES/0039-devices-use-dev-cache-aliases-handling-from-label-sc.patch

4d51e5
From 932b9720bb074e49ac920642b3fe9c3d84019787 Mon Sep 17 00:00:00 2001
4d51e5
From: David Teigland <teigland@redhat.com>
4d51e5
Date: Mon, 28 Feb 2022 17:37:12 -0600
4d51e5
Subject: [PATCH 39/54] devices: use dev-cache aliases handling from label scan
4d51e5
 functions
4d51e5
4d51e5
The label scan functions where doing some device alias validation
4d51e5
which is now better handled by the dev-cache layer, so just use
4d51e5
that.
4d51e5
---
4d51e5
 lib/device/dev-cache.c |   4 +-
4d51e5
 lib/device/dev-cache.h |   1 +
4d51e5
 lib/label/label.c      | 143 ++++++++++-------------------------------
4d51e5
 3 files changed, 36 insertions(+), 112 deletions(-)
4d51e5
4d51e5
diff --git a/lib/device/dev-cache.c b/lib/device/dev-cache.c
4d51e5
index 58e67e130..b0759b06c 100644
4d51e5
--- a/lib/device/dev-cache.c
4d51e5
+++ b/lib/device/dev-cache.c
4d51e5
@@ -1410,7 +1410,7 @@ static void _remove_alias(struct device *dev, const char *name)
4d51e5
  * deactivated LV.  Those old paths are all invalid and are dropped here.
4d51e5
  */
4d51e5
 
4d51e5
-static void _verify_aliases(struct device *dev)
4d51e5
+void dev_cache_verify_aliases(struct device *dev)
4d51e5
 {
4d51e5
 	struct dm_str_list *strl, *strl2;
4d51e5
 	struct stat st;
4d51e5
@@ -1459,7 +1459,7 @@ static struct device *_dev_cache_get(struct cmd_context *cmd, const char *name,
4d51e5
 			_remove_alias(dev, name);
4d51e5
 
4d51e5
 			/* Remove any other names in dev->aliases that are incorrect. */
4d51e5
-			_verify_aliases(dev);
4d51e5
+			dev_cache_verify_aliases(dev);
4d51e5
 		}
4d51e5
 		return NULL;
4d51e5
 	}
4d51e5
diff --git a/lib/device/dev-cache.h b/lib/device/dev-cache.h
4d51e5
index 51c3fc6c3..321a56d7b 100644
4d51e5
--- a/lib/device/dev-cache.h
4d51e5
+++ b/lib/device/dev-cache.h
4d51e5
@@ -55,6 +55,7 @@ int dev_cache_add_dir(const char *path);
4d51e5
 struct device *dev_cache_get(struct cmd_context *cmd, const char *name, struct dev_filter *f);
4d51e5
 struct device *dev_cache_get_existing(struct cmd_context *cmd, const char *name, struct dev_filter *f);
4d51e5
 struct device *dev_cache_get_by_devt(struct cmd_context *cmd, dev_t devt);
4d51e5
+void dev_cache_verify_aliases(struct device *dev);
4d51e5
 
4d51e5
 struct device *dev_hash_get(const char *name);
4d51e5
 
4d51e5
diff --git a/lib/label/label.c b/lib/label/label.c
4d51e5
index cf707f7a3..06958b502 100644
4d51e5
--- a/lib/label/label.c
4d51e5
+++ b/lib/label/label.c
4d51e5
@@ -458,7 +458,6 @@ static int _scan_dev_open(struct device *dev)
4d51e5
 	const char *name;
4d51e5
 	const char *modestr;
4d51e5
 	struct stat sbuf;
4d51e5
-	int retried = 0;
4d51e5
 	int flags = 0;
4d51e5
 	int fd, di;
4d51e5
 
4d51e5
@@ -478,14 +477,23 @@ static int _scan_dev_open(struct device *dev)
4d51e5
 		return 0;
4d51e5
 	}
4d51e5
 
4d51e5
+ next_name:
4d51e5
 	/*
4d51e5
 	 * All the names for this device (major:minor) are kept on
4d51e5
 	 * dev->aliases, the first one is the primary/preferred name.
4d51e5
+	 *
4d51e5
+	 * The default name preferences in dev-cache mean that the first
4d51e5
+	 * name in dev->aliases is not a symlink for scsi devices, but is
4d51e5
+	 * the /dev/mapper/ symlink for mpath devices.
4d51e5
+	 *
4d51e5
+	 * If preferred names are set to symlinks, should this
4d51e5
+	 * first attempt to open using a non-symlink?
4d51e5
+	 *
4d51e5
+	 * dm_list_first() returns NULL if the list is empty.
4d51e5
 	 */
4d51e5
 	if (!(name_list = dm_list_first(&dev->aliases))) {
4d51e5
-		/* Shouldn't happen */
4d51e5
-		log_error("Device open %s %d:%d has no path names.",
4d51e5
-			  dev_name(dev), (int)MAJOR(dev->dev), (int)MINOR(dev->dev));
4d51e5
+		log_error("Device open %d:%d has no path names.",
4d51e5
+			  (int)MAJOR(dev->dev), (int)MINOR(dev->dev));
4d51e5
 		return 0;
4d51e5
 	}
4d51e5
 	name_sl = dm_list_item(name_list, struct dm_str_list);
4d51e5
@@ -513,50 +521,34 @@ static int _scan_dev_open(struct device *dev)
4d51e5
 		modestr = "ro";
4d51e5
 	}
4d51e5
 
4d51e5
-retry_open:
4d51e5
-
4d51e5
 	fd = open(name, flags, 0777);
4d51e5
-
4d51e5
 	if (fd < 0) {
4d51e5
 		if ((errno == EBUSY) && (flags & O_EXCL)) {
4d51e5
 			log_error("Can't open %s exclusively.  Mounted filesystem?",
4d51e5
 				  dev_name(dev));
4d51e5
+			return 0;
4d51e5
 		} else {
4d51e5
-			int major, minor;
4d51e5
-
4d51e5
 			/*
4d51e5
-			 * Shouldn't happen, if it does, print stat info to help figure
4d51e5
-			 * out what's wrong.
4d51e5
+			 * drop name from dev->aliases and use verify_aliases to
4d51e5
+			 * drop any other invalid aliases before retrying open with
4d51e5
+			 * any remaining valid paths.
4d51e5
 			 */
4d51e5
-
4d51e5
-			major = (int)MAJOR(dev->dev);
4d51e5
-			minor = (int)MINOR(dev->dev);
4d51e5
-
4d51e5
-			log_error("Device open %s %d:%d failed errno %d", name, major, minor, errno);
4d51e5
-
4d51e5
-			if (stat(name, &sbuf)) {
4d51e5
-				log_debug_devs("Device open %s %d:%d stat failed errno %d",
4d51e5
-					       name, major, minor, errno);
4d51e5
-			} else if (sbuf.st_rdev != dev->dev) {
4d51e5
-				log_debug_devs("Device open %s %d:%d stat %d:%d does not match.",
4d51e5
-					       name, major, minor,
4d51e5
-					       (int)MAJOR(sbuf.st_rdev), (int)MINOR(sbuf.st_rdev));
4d51e5
-			}
4d51e5
-
4d51e5
-			if (!retried) {
4d51e5
-				/*
4d51e5
-				 * FIXME: remove this, the theory for this retry is that
4d51e5
-				 * there may be a udev race that we can sometimes mask by
4d51e5
-				 * retrying.  This is here until we can figure out if it's
4d51e5
-				 * needed and if so fix the real problem.
4d51e5
-				 */
4d51e5
-				usleep(5000);
4d51e5
-				log_debug_devs("Device open %s retry", dev_name(dev));
4d51e5
-				retried = 1;
4d51e5
-				goto retry_open;
4d51e5
-			}
4d51e5
+			log_debug("Drop alias for %d:%d failed open %s (%d)",
4d51e5
+				  (int)MAJOR(dev->dev), (int)MINOR(dev->dev), name, errno);
4d51e5
+			dev_cache_failed_path(dev, name);
4d51e5
+			dev_cache_verify_aliases(dev);
4d51e5
+			goto next_name;
4d51e5
 		}
4d51e5
-		return 0;
4d51e5
+	}
4d51e5
+
4d51e5
+	/* Verify that major:minor from the path still match dev. */
4d51e5
+	if ((fstat(fd, &sbuf) < 0) || (sbuf.st_rdev != dev->dev)) {
4d51e5
+		log_warn("Invalid path %s for device %d:%d, trying different path.",
4d51e5
+			 name, (int)MAJOR(dev->dev), (int)MINOR(dev->dev));
4d51e5
+		(void)close(fd);
4d51e5
+		dev_cache_failed_path(dev, name);
4d51e5
+		dev_cache_verify_aliases(dev);
4d51e5
+		goto next_name;
4d51e5
 	}
4d51e5
 
4d51e5
 	dev->flags |= DEV_IN_BCACHE;
4d51e5
@@ -604,37 +596,6 @@ static int _scan_dev_close(struct device *dev)
4d51e5
 	return 1;
4d51e5
 }
4d51e5
 
4d51e5
-static void _drop_bad_aliases(struct device *dev)
4d51e5
-{
4d51e5
-	struct dm_str_list *strl, *strl2;
4d51e5
-	const char *name;
4d51e5
-	struct stat sbuf;
4d51e5
-	int major = (int)MAJOR(dev->dev);
4d51e5
-	int minor = (int)MINOR(dev->dev);
4d51e5
-	int bad;
4d51e5
-
4d51e5
-	dm_list_iterate_items_safe(strl, strl2, &dev->aliases) {
4d51e5
-		name = strl->str;
4d51e5
-		bad = 0;
4d51e5
-
4d51e5
-		if (stat(name, &sbuf)) {
4d51e5
-			bad = 1;
4d51e5
-			log_debug_devs("Device path check %d:%d %s stat failed errno %d",
4d51e5
-					major, minor, name, errno);
4d51e5
-		} else if (sbuf.st_rdev != dev->dev) {
4d51e5
-			bad = 1;
4d51e5
-			log_debug_devs("Device path check %d:%d %s stat %d:%d does not match.",
4d51e5
-				       major, minor, name,
4d51e5
-				       (int)MAJOR(sbuf.st_rdev), (int)MINOR(sbuf.st_rdev));
4d51e5
-		}
4d51e5
-
4d51e5
-		if (bad) {
4d51e5
-			log_debug_devs("Device path check %d:%d dropping path %s.", major, minor, name);
4d51e5
-			dev_cache_failed_path(dev, name);
4d51e5
-		}
4d51e5
-	}
4d51e5
-}
4d51e5
-
4d51e5
 // Like bcache_invalidate, only it throws any dirty data away if the
4d51e5
 // write fails.
4d51e5
 static void _invalidate_di(struct bcache *cache, int di)
4d51e5
@@ -662,10 +623,8 @@ static int _scan_list(struct cmd_context *cmd, struct dev_filter *f,
4d51e5
 	char headers_buf[HEADERS_BUF_SIZE];
4d51e5
 	struct dm_list wait_devs;
4d51e5
 	struct dm_list done_devs;
4d51e5
-	struct dm_list reopen_devs;
4d51e5
 	struct device_list *devl, *devl2;
4d51e5
 	struct block *bb;
4d51e5
-	int retried_open = 0;
4d51e5
 	int scan_read_errors = 0;
4d51e5
 	int scan_process_errors = 0;
4d51e5
 	int scan_failed_count = 0;
4d51e5
@@ -676,7 +635,6 @@ static int _scan_list(struct cmd_context *cmd, struct dev_filter *f,
4d51e5
 
4d51e5
 	dm_list_init(&wait_devs);
4d51e5
 	dm_list_init(&done_devs);
4d51e5
-	dm_list_init(&reopen_devs);
4d51e5
 
4d51e5
 	log_debug_devs("Scanning %d devices for VG info", dm_list_size(devs));
4d51e5
 
4d51e5
@@ -700,9 +658,9 @@ static int _scan_list(struct cmd_context *cmd, struct dev_filter *f,
4d51e5
 
4d51e5
 		if (!_in_bcache(devl->dev)) {
4d51e5
 			if (!_scan_dev_open(devl->dev)) {
4d51e5
-				log_debug_devs("Scan failed to open %s.", dev_name(devl->dev));
4d51e5
+				log_debug_devs("Scan failed to open %d:%d %s.",
4d51e5
+					       (int)MAJOR(devl->dev->dev), (int)MINOR(devl->dev->dev), dev_name(devl->dev));
4d51e5
 				dm_list_del(&devl->list);
4d51e5
-				dm_list_add(&reopen_devs, &devl->list);
4d51e5
 				devl->dev->flags |= DEV_SCAN_NOT_READ;
4d51e5
 				continue;
4d51e5
 			}
4d51e5
@@ -786,41 +744,6 @@ static int _scan_list(struct cmd_context *cmd, struct dev_filter *f,
4d51e5
 	if (!dm_list_empty(devs))
4d51e5
 		goto scan_more;
4d51e5
 
4d51e5
-	/*
4d51e5
-	 * We're done scanning all the devs.  If we failed to open any of them
4d51e5
-	 * the first time through, refresh device paths and retry.  We failed
4d51e5
-	 * to open the devs on the reopen_devs list.
4d51e5
-	 *
4d51e5
-	 * FIXME: it's not clear if or why this helps.
4d51e5
-	 */
4d51e5
-	if (!dm_list_empty(&reopen_devs)) {
4d51e5
-		if (retried_open) {
4d51e5
-			/* Don't try again. */
4d51e5
-			scan_failed_count += dm_list_size(&reopen_devs);
4d51e5
-			dm_list_splice(&done_devs, &reopen_devs);
4d51e5
-			goto out;
4d51e5
-		}
4d51e5
-		retried_open = 1;
4d51e5
-
4d51e5
-		dm_list_iterate_items_safe(devl, devl2, &reopen_devs) {
4d51e5
-			_drop_bad_aliases(devl->dev);
4d51e5
-
4d51e5
-			if (dm_list_empty(&devl->dev->aliases)) {
4d51e5
-				log_warn("WARNING: Scan ignoring device %d:%d with no paths.",
4d51e5
-					 (int)MAJOR(devl->dev->dev),
4d51e5
-					 (int)MINOR(devl->dev->dev));
4d51e5
-					 
4d51e5
-				dm_list_del(&devl->list);
4d51e5
-				lvmcache_del_dev(devl->dev);
4d51e5
-				scan_failed_count++;
4d51e5
-			}
4d51e5
-		}
4d51e5
-
4d51e5
-		/* Put devs that failed to open back on the original list to retry. */
4d51e5
-		dm_list_splice(devs, &reopen_devs);
4d51e5
-		goto scan_more;
4d51e5
-	}
4d51e5
-out:
4d51e5
 	log_debug_devs("Scanned devices: read errors %d process errors %d failed %d",
4d51e5
 			scan_read_errors, scan_process_errors, scan_failed_count);
4d51e5
 
4d51e5
-- 
4d51e5
2.34.3
4d51e5