Skip to contentSkip to conversation

tracing: fix test for blockdev inode in wb error handling tracepoints

PATCH Superseded Jeff Layton 1 participant 2 messages Jan 30, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tracing: fix test for blockdev inode in wb error handling tracepoints

Al pointed out that the test for whether a mapping refers to a blockdev inode in these tracepoints is garbage, as the superblock is always non-NULL.

Add a new mapping_to_dev helper function that determines this the correct way, and change the tracepoints to use it. Also, fix up the indentation in this file.

---
 include/linux/fs.h             | 20 +++++++
 include/trace/events/filemap.h | 99 +++++++++++++++-------------------
 2 files changed, 64 insertions(+), 55 deletions(-)
include/linux/fs.h +20 -0
index 98e0349adb52..a0e50fcc3cab 100644
@@ -2590,6 +2590,21 @@ static inline bool sb_is_blkdev_sb(struct super_block *sb)
 {
 	return sb == blockdev_superblock;
 }
+
+/*
+ * Return the device associated with this mapping.
+ *
+ * For normal inodes, return the sb of the backing filesystem, for blockdev
+ * inodes, return the block device number associated with the inode.
+ */
+static inline dev_t mapping_to_dev(struct address_space *mapping)
+{
+	struct inode *inode = mapping->host;
+
+	if (sb_is_blkdev_sb(inode->i_sb))
+		return inode->i_rdev;
+	return inode->i_sb->s_dev;
+}
 #else
 static inline void bd_forget(struct inode *inode) {}
 static inline int sync_blockdev(struct block_device *bdev) { return 0; }
@@ -2619,6 +2634,11 @@ static inline bool sb_is_blkdev_sb(struct super_block *sb)
 {
 	return false;
 }
+
+static inline dev_t mapping_to_dev(struct address_space *mapping)
+{
+	return mapping->host->i_sb->s_dev;
+}
 #endif
 extern int sync_filesystem(struct super_block *);
 extern const struct file_operations def_blk_fops;
include/trace/events/filemap.h +44 -55

114 lines not on this page · raw message

Signature · 1 line
2.24.1
#1 lore raw

PATCH v2 1/1 tracing: fix test for blockdev inode in wb error handling tracepoints

Al pointed out that the test for whether a mapping refers to a blockdev inode in these tracepoints is garbage, as the superblock is always non-NULL.

Add a new mapping_to_dev helper function that determines this the correct way, and change the tracepoints to use it. Also, fix up the indentation in this file.

---
 include/linux/fs.h             | 16 ++++++
 include/trace/events/filemap.h | 99 +++++++++++++++-------------------
 2 files changed, 60 insertions(+), 55 deletions(-)

Al, sorry.

I had based the original on an older tree. This one should merge more cleanly, and also just adds a single, non-conditionally compiled helper.

include/linux/fs.h +16 -0
index 40be2ccb87f3..264ac7d602a6 100644
@@ -2616,6 +2616,22 @@ static inline bool sb_is_blkdev_sb(struct super_block *sb)
 	return false;
 }
 #endif
+
+/*
+ * Return the device associated with this mapping.
+ *
+ * For normal inodes, return the sb of the backing filesystem, for blockdev
+ * inodes, return the block device number associated with the inode.
+ */
+static inline dev_t mapping_to_dev(struct address_space *mapping)
+{
+	struct inode *inode = mapping->host;
+
+	if (sb_is_blkdev_sb(inode->i_sb))
+		return inode->i_rdev;
+	return inode->i_sb->s_dev;
+}
+
 extern int sync_filesystem(struct super_block *);
 extern const struct file_operations def_blk_fops;
 extern const struct file_operations def_chr_fops;
include/trace/events/filemap.h +44 -55

114 lines not on this page · raw message

Signature · 1 line
2.24.1