ceph: fix inode number handling on arches with 32-bit ino_t

Tuan and Ulrich mentioned that they were hitting a problem on s390x,
which has a 32-bit ino_t value, even though it's a 64-bit arch (for
historical reasons).

I think the current handling of inode numbers in the ceph driver is
wrong. It tries to use 32-bit inode numbers on 32-bit arches, but that's
actually not a problem. 32-bit arches can deal with 64-bit inode numbers
just fine when userland code is compiled with LFS support (the common
case these days).

What we really want to do is just use 64-bit numbers everywhere, unless
someone has mounted with the ino32 mount option. In that case, we want
to ensure that we hash the inode number down to something that will fit
in 32 bits before presenting the value to userland.

Add new helper functions that do this, and only do the conversion before
presenting these values to userland in getattr and readdir.

The inode table hashvalue is changed to just cast the inode number to
unsigned long, as low-order bits are the most likely to vary anyway.

While it's not strictly required, we do want to put something in
inode->i_ino. Instead of basing it on BITS_PER_LONG, however, base it on
the size of the ino_t type.

NOTE: This is a user-visible change on 32-bit arches:

1/ inode numbers will be seen to have changed between kernel versions.
   32-bit arches will see large inode numbers now instead of the hashed
   ones they saw before.

2/ any really old software not built with LFS support may start failing
   stat() calls with -EOVERFLOW on inode numbers >2^32. Nothing much we
   can do about these, but hopefully the intersection of people running
   such code on ceph will be very small.

The workaround for both problems is to mount with "-o ino32".

[ idryomov: changelog tweak ]

URL: https://tracker.ceph.com/issues/46828
Reported-by: Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
Reported-and-Tested-by: Tuan Hoang1 <Tuan.Hoang1@ibm.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

1 files changed, 38 insertions, 35 deletions

diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index 4c3c964b1c54..a3995ebe0623 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -457,15 +457,7 @@ ceph_vino(const struct inode *inode)
 	return ceph_inode(inode)->i_vino;
 }
 
-/*
- * ino_t is <64 bits on many architectures, blech.
- *
- *               i_ino (kernel inode)   st_ino (userspace)
- * i386          32                     32
- * x86_64+ino32  64                     32
- * x86_64        64                     64
- */
-static inline u32 ceph_ino_to_ino32(__u64 vino)
+static inline u32 ceph_ino_to_ino32(u64 vino)
 {
 	u32 ino = vino & 0xffffffff;
 	ino ^= vino >> 32;
@@ -475,34 +467,17 @@ static inline u32 ceph_ino_to_ino32(__u64 vino)
 }
 
 /*
- * kernel i_ino value
+ * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
+ * some arches. We generally do not use this value inside the ceph driver, but
+ * we do want to set it to something, so that generic vfs code has an
+ * appropriate value for tracepoints and the like.
  */
-static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
+static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
 {
-#if BITS_PER_LONG == 32
-	return ceph_ino_to_ino32(vino.ino);
-#else
+	if (sizeof(ino_t) == sizeof(u32))
+		return ceph_ino_to_ino32(vino.ino);
 	return (ino_t)vino.ino;
-#endif
-}
-
-/*
- * user-visible ino (stat, filldir)
- */
-#if BITS_PER_LONG == 32
-static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
-{
-	return ino;
-}
-#else
-static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
-{
-	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
-		ino = ceph_ino_to_ino32(ino);
-	return ino;
 }
-#endif
-
 
 /* for printf-style formatting */
 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
@@ -511,11 +486,34 @@ static inline u64 ceph_ino(struct inode *inode)
 {
 	return ceph_inode(inode)->i_vino.ino;
 }
+
 static inline u64 ceph_snap(struct inode *inode)
 {
 	return ceph_inode(inode)->i_vino.snap;
 }
 
+/**
+ * ceph_present_ino - format an inode number for presentation to userland
+ * @sb: superblock where the inode lives
+ * @ino: inode number to (possibly) convert
+ *
+ * If the user mounted with the ino32 option, then the 64-bit value needs
+ * to be converted to something that can fit inside 32 bits. Note that
+ * internal kernel code never uses this value, so this is entirely for
+ * userland consumption.
+ */
+static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
+{
+	if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
+		return ceph_ino_to_ino32(ino);
+	return ino;
+}
+
+static inline u64 ceph_present_inode(struct inode *inode)
+{
+	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
+}
+
 static inline int ceph_ino_compare(struct inode *inode, void *data)
 {
 	struct ceph_vino *pvino = (struct ceph_vino *)data;
@@ -524,11 +522,16 @@ static inline int ceph_ino_compare(struct inode *inode, void *data)
 		ci->i_vino.snap == pvino->snap;
 }
 
+
 static inline struct inode *ceph_find_inode(struct super_block *sb,
 					    struct ceph_vino vino)
 {
-	ino_t t = ceph_vino_to_ino(vino);
-	return ilookup5(sb, t, ceph_ino_compare, &vino);
+	/*
+	 * NB: The hashval will be run through the fs/inode.c hash function
+	 * anyway, so there is no need to squash the inode number down to
+	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
+	 */
+	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
 }