[PATCH] 32-bit dev_t: internal use

Starting the conversion:
	* internal dev_t made 32bit.
	* new helpers - new_encode_dev(), new_decode_dev(), huge_encode_dev(),
huge_decode_dev(), new_valid_dev().  They do encoding/decoding of 32bit and
64bit values; for now huge_... are aliases for new_... and new_valid_dev()
is always true.  We do 12:20 for 32bit; representation is compatible with
16bit one - we have major in bits 19--8 and minor in 31--20,7--0.  That's
what the userland sees; internally we have (major << 20)|minor, of course.
	* MKDEV(), MAJOR() and MINOR() updated.
	* several places used to handle Missed'em'V dev_t (14:18 split)
manually; that stuff had been taken into common helpers.
	Now we can start replacing old_... with new_... and huge_..., depending
on the width available.  MKDEV() callers should (for now) make sure that major
and minor are within 12:20.  That's what the next chunk will do.

arch/mips/kernel/sysirix.c

@@ -1187,13 +1187,7 @@ asmlinkage int irix_uname(struct iuname *buf)
 
 #undef DEBUG_XSTAT
 
-static inline u32
-linux_to_irix_dev_t(dev_t t)
-{
-	return MAJOR (t) << 18 | MINOR (t);
-}
-
-static inline int irix_xstat32_xlate(struct kstat *stat, void *ubuf)
+static int irix_xstat32_xlate(struct kstat *stat, void *ubuf)
 {
 	struct xstat32 {
 		u32 st_dev, st_pad1[3], st_ino, st_mode, st_nlink, st_uid, st_gid;
@@ -1206,13 +1200,15 @@ static inline int irix_xstat32_xlate(struct kstat *stat, void *ubuf)
 		u32 st_pad4[8];
 	} ub;
 
-	ub.st_dev     = linux_to_irix_dev_t(stat->dev);
+	if (!sysv_valid_dev(stat->dev) || !sysv_valid_dev(stat->rdev))
+		return -EOVERFLOW;
+	ub.st_dev     = sysv_encode_dev(stat->dev);
 	ub.st_ino     = stat->ino;
 	ub.st_mode    = stat->mode;
 	ub.st_nlink   = stat->nlink;
 	SET_STAT_UID(ub, stat->uid);
 	SET_STAT_GID(ub, stat->gid);
-	ub.st_rdev    = linux_to_irix_dev_t(stat->rdev);
+	ub.st_rdev    = sysv_encode_dev(stat->rdev);
 #if BITS_PER_LONG == 32
 	if (stat->size > MAX_NON_LFS)
 		return -EOVERFLOW;
@@ -1231,7 +1227,7 @@ static inline int irix_xstat32_xlate(struct kstat *stat, void *ubuf)
 	return copy_to_user(ubuf, &ub, sizeof(ub)) ? -EFAULT : 0;
 }
 
-static inline void irix_xstat64_xlate(struct kstat *stat, void *ubuf)
+static int irix_xstat64_xlate(struct kstat *stat, void *ubuf)
 {
 	struct xstat64 {
 		u32 st_dev; s32 st_pad1[3];
@@ -1248,14 +1244,17 @@ static inline void irix_xstat64_xlate(struct kstat *stat, void *ubuf)
 		s32 st_pad4[8];
 	} ks;
 
-	ks.st_dev = linux_to_irix_dev_t(stat->dev);
+	if (!sysv_valid_dev(stat->dev) || !sysv_valid_dev(stat->rdev))
+		return -EOVERFLOW;
+
+	ks.st_dev = sysv_encode_dev(stat->dev);
 	ks.st_pad1[0] = ks.st_pad1[1] = ks.st_pad1[2] = 0;
 	ks.st_ino = (unsigned long long) stat->ino;
 	ks.st_mode = (u32) stat->mode;
 	ks.st_nlink = (u32) stat->nlink;
 	ks.st_uid = (s32) stat->uid;
 	ks.st_gid = (s32) stat->gid;
-	ks.st_rdev = linux_to_irix_dev_t (stat->rdev);
+	ks.st_rdev = sysv_encode_dev (stat->rdev);
 	ks.st_pad2[0] = ks.st_pad2[1] = 0;
 	ks.st_size = (long long) stat->size;
 	ks.st_pad3 = 0;
@@ -1275,7 +1274,7 @@ static inline void irix_xstat64_xlate(struct kstat *stat, void *ubuf)
 	ks.st_pad4[4] = ks.st_pad4[5] = ks.st_pad4[6] = ks.st_pad4[7] = 0;
 
 	/* Now write it all back. */
-	copy_to_user(ubuf, &ks, sizeof(struct xstat64));
+	return copy_to_user(ubuf, &ub, sizeof(ub)) ? -EFAULT : 0;
 }
 
 asmlinkage int irix_xstat(int version, char *filename, struct stat *statbuf)
@@ -1295,8 +1294,7 @@ asmlinkage int irix_xstat(int version, char *filename, struct stat *statbuf)
 				retval = irix_xstat32_xlate(&stat, statbuf);
 				break;
 			case 3:
-				irix_xstat64_xlate(&stat, statbuf);
-				retval = 0; /* Really? */
+				retval = irix_xstat64_xlate(&stat, statbuf);
 				break;
 			default:
 				retval = -EINVAL;
@@ -1323,8 +1321,7 @@ asmlinkage int irix_lxstat(int version, char *filename, struct stat *statbuf)
 				error = irix_xstat32_xlate(&stat, statbuf);
 				break;
 			case 3:
-				irix_xstat64_xlate(&stat, statbuf);
-				error = 0;
+				error = irix_xstat64_xlate(&stat, statbuf);
 				break;
 			default:
 				error = -EINVAL;
@@ -1350,8 +1347,7 @@ asmlinkage int irix_fxstat(int version, int fd, struct stat *statbuf)
 				error = irix_xstat32_xlate(&stat, statbuf);
 				break;
 			case 3:
-				irix_xstat64_xlate(&stat, statbuf);
-				error = 0;
+				error = irix_xstat64_xlate(&stat, statbuf);
 				break;
 			default:
 				error = -EINVAL;

arch/sparc64/solaris/fs.c

@@ -28,21 +28,6 @@
 
 #include "conv.h"
 
-static inline u32 R4_DEV(dev_t DEV)
-{
-	return MINOR(DEV) | (MAJOR(DEV) << 18);
-}
-
-static inline unsigned R4_MAJOR(u32 DEV)
-{
-	return (DEV >> 18) & 0x3fff;
-}
-
-static inline unsigned R4_MINOR(u32 DEV)
-{
-	return DEV & 0x3ffff;
-}
-
 #define R3_VERSION	1
 #define R4_VERSION	2
 
@@ -96,15 +81,17 @@ struct sol_stat64 {
 
 static inline int putstat(struct sol_stat *ubuf, struct kstat *kbuf)
 {
-	if (kbuf->size > MAX_NON_LFS)
+	if (kbuf->size > MAX_NON_LFS ||
+	    !sysv_valid_dev(kbuf->dev) ||
+	    !sysv_valid_dev(kbuf->rdev))
 		return -EOVERFLOW;
-	if (put_user (R4_DEV(kbuf->dev), &ubuf->st_dev)	||
+	if (put_user (sysv_encode_dev(kbuf->dev), &ubuf->st_dev)	||
 	    __put_user (kbuf->ino, &ubuf->st_ino)		||
 	    __put_user (kbuf->mode, &ubuf->st_mode)		||
 	    __put_user (kbuf->nlink, &ubuf->st_nlink)	||
 	    __put_user (kbuf->uid, &ubuf->st_uid)		||
 	    __put_user (kbuf->gid, &ubuf->st_gid)		||
-	    __put_user (R4_DEV(kbuf->rdev), &ubuf->st_rdev)	||
+	    __put_user (sysv_encode_dev(kbuf->rdev), &ubuf->st_rdev)	||
 	    __put_user (kbuf->size, &ubuf->st_size)		||
 	    __put_user (kbuf->atime.tv_sec, &ubuf->st_atime.tv_sec)	||
 	    __put_user (kbuf->atime.tv_nsec, &ubuf->st_atime.tv_nsec)	||
@@ -121,13 +108,15 @@ static inline int putstat(struct sol_stat *ubuf, struct kstat *kbuf)
 
 static inline int putstat64(struct sol_stat64 *ubuf, struct kstat *kbuf)
 {
-	if (put_user (R4_DEV(kbuf->dev), &ubuf->st_dev)	||
+	if (!sysv_valid_dev(kbuf->dev) || !sysv_valid_dev(kbuf->rdev))
+		return -EOVERFLOW;
+	if (put_user (sysv_encode_dev(kbuf->dev), &ubuf->st_dev)	||
 	    __put_user (kbuf->ino, &ubuf->st_ino)		||
 	    __put_user (kbuf->mode, &ubuf->st_mode)		||
 	    __put_user (kbuf->nlink, &ubuf->st_nlink)	||
 	    __put_user (kbuf->uid, &ubuf->st_uid)		||
 	    __put_user (kbuf->gid, &ubuf->st_gid)		||
-	    __put_user (R4_DEV(kbuf->rdev), &ubuf->st_rdev)	||
+	    __put_user (sysv_encode_dev(kbuf->rdev), &ubuf->st_rdev)	||
 	    __put_user (kbuf->size, &ubuf->st_size)		||
 	    __put_user (kbuf->atime.tv_sec, &ubuf->st_atime.tv_sec)	||
 	    __put_user (kbuf->atime.tv_nsec, &ubuf->st_atime.tv_nsec)	||
@@ -261,8 +250,8 @@ asmlinkage int solaris_mknod(u32 path, u32 mode, s32 dev)
 		(int (*)(const char *,int,unsigned))SYS(mknod);
 	int major, minor;
 
-	if ((major = R4_MAJOR(dev)) > 255 || 
-	    (minor = R4_MINOR(dev)) > 255) return -EINVAL;
+	if ((major = sysv_major(dev)) > 255 || 
+	    (minor = sysv_minor(dev)) > 255) return -EINVAL;
 	return sys_mknod((const char *)A(path), mode, old_encode_dev(MKDEV(major,minor)));
 }
 
@@ -415,6 +404,8 @@ static int report_statvfs(struct vfsmount *mnt, struct inode *inode, u32 buf)
 		if (j > 15) j = 15;
 		if (IS_RDONLY(inode)) i = 1;
 		if (mnt->mnt_flags & MNT_NOSUID) i |= 2;
+		if (!sysv_valid_dev(inode->i_sb->s_dev))
+			return -EOVERFLOW;
 		if (put_user (s.f_bsize, &ss->f_bsize)		||
 		    __put_user (0, &ss->f_frsize)		||
 		    __put_user (s.f_blocks, &ss->f_blocks)	||
@@ -423,7 +414,7 @@ static int report_statvfs(struct vfsmount *mnt, struct inode *inode, u32 buf)
 		    __put_user (s.f_files, &ss->f_files)	||
 		    __put_user (s.f_ffree, &ss->f_ffree)	||
 		    __put_user (s.f_ffree, &ss->f_favail)	||
-		    __put_user (R4_DEV(inode->i_sb->s_dev), &ss->f_fsid) ||
+		    __put_user (sysv_encode_dev(inode->i_sb->s_dev), &ss->f_fsid) ||
 		    __copy_to_user (ss->f_basetype,p,j)		||
 		    __put_user (0, (char *)&ss->f_basetype[j])	||
 		    __put_user (s.f_namelen, &ss->f_namemax)	||
@@ -449,6 +440,8 @@ static int report_statvfs64(struct vfsmount *mnt, struct inode *inode, u32 buf)
 		if (j > 15) j = 15;
 		if (IS_RDONLY(inode)) i = 1;
 		if (mnt->mnt_flags & MNT_NOSUID) i |= 2;
+		if (!sysv_valid_dev(inode->i_sb->s_dev))
+			return -EOVERFLOW;
 		if (put_user (s.f_bsize, &ss->f_bsize)		||
 		    __put_user (0, &ss->f_frsize)		||
 		    __put_user (s.f_blocks, &ss->f_blocks)	||
@@ -457,7 +450,7 @@ static int report_statvfs64(struct vfsmount *mnt, struct inode *inode, u32 buf)
 		    __put_user (s.f_files, &ss->f_files)	||
 		    __put_user (s.f_ffree, &ss->f_ffree)	||
 		    __put_user (s.f_ffree, &ss->f_favail)	||
-		    __put_user (R4_DEV(inode->i_sb->s_dev), &ss->f_fsid) ||
+		    __put_user (sysv_encode_dev(inode->i_sb->s_dev), &ss->f_fsid) ||
 		    __copy_to_user (ss->f_basetype,p,j)		||
 		    __put_user (0, (char *)&ss->f_basetype[j])	||
 		    __put_user (s.f_namelen, &ss->f_namemax)	||

drivers/s390/block/dasd_int.h

@@ -14,7 +14,8 @@
 
 #ifdef __KERNEL__
 
-#define DASD_PER_MAJOR ( 1U<<(MINORBITS-DASD_PARTN_BITS))
+/* we keep old device allocation scheme; IOW, minors are still in 0..255 */
+#define DASD_PER_MAJOR ( 1U<<(8-DASD_PARTN_BITS))
 #define DASD_PARTN_MASK ((1 << DASD_PARTN_BITS) - 1)
 
 /*

fs/libfs.c

@@ -429,22 +429,3 @@ void simple_release_fs(struct vfsmount **mount, int *count)
 	spin_unlock(&pin_fs_lock);
 	mntput(mnt);
 }
-
-/* acceptable for old filesystems */
-int old_valid_dev(dev_t dev)
-{
-	return MAJOR(dev) < 256 && MINOR(dev) < 256;
-}
-EXPORT_SYMBOL(old_valid_dev);
-
-u16 old_encode_dev(dev_t dev)
-{
-	return (MAJOR(dev) << 8) | MINOR(dev);
-}
-EXPORT_SYMBOL(old_encode_dev);
-
-dev_t old_decode_dev(u16 val)
-{
-	return MKDEV((val >> 8) & 255, val & 255);
-}
-EXPORT_SYMBOL(old_decode_dev);

fs/xfs/linux/xfs_iops.c

@@ -113,7 +113,11 @@ linvfs_mknod(
 	xattr_exists_t	test_default_acl = _ACL_DEFAULT_EXISTS;
 	int		error;
 
-	if (!old_valid_dev(rdev))
+	/*
+	 * Irix uses Missed'em'V split, but doesn't want to see
+	 * the upper 5 bits of (14bit) major.
+	 */
+	if (!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff)
 		return -EINVAL;
 
 	if (test_default_acl && test_default_acl(dvp)) {
@@ -135,7 +139,7 @@ linvfs_mknod(
 
 	switch (mode & S_IFMT) {
 	case S_IFCHR: case S_IFBLK: case S_IFIFO: case S_IFSOCK:
-		va.va_rdev = XFS_MKDEV(MAJOR(rdev), MINOR(rdev));
+		va.va_rdev = sysv_encode_dev(rdev);
 		va.va_mask |= XFS_AT_RDEV;
 		/*FALLTHROUGH*/
 	case S_IFREG:

fs/xfs/linux/xfs_super.c

@@ -174,7 +174,7 @@ xfs_revalidate_inode(
 		inode->i_rdev = 0;
 	} else {
 		xfs_dev_t dev = ip->i_df.if_u2.if_rdev;
-		inode->i_rdev = XFS_DEV_TO_DEVT(dev);
+		inode->i_rdev = MKDEV(sysv_major(dev) & 0x1ff, sysv_minor(dev));
 	}
 	inode->i_blksize = PAGE_CACHE_SIZE;
 	inode->i_generation = ip->i_d.di_gen;

fs/xfs/xfs_types.h

@@ -197,20 +197,4 @@ typedef enum {
 	XFS_BTNUM_MAX
 } xfs_btnum_t;
 
-
-/*
- * Juggle IRIX device numbers - still used in ondisk structures
- */
-#define XFS_DEV_BITSMAJOR	14
-#define XFS_DEV_BITSMINOR	18
-#define XFS_DEV_MAXMAJ		0x1ff
-#define XFS_DEV_MAXMIN		0x3ffff
-#define XFS_DEV_MAJOR(dev)	((int)(((unsigned)(dev)>>XFS_DEV_BITSMINOR) \
-				    & XFS_DEV_MAXMAJ))
-#define XFS_DEV_MINOR(dev)	((int)((dev)&XFS_DEV_MAXMIN))
-#define XFS_MKDEV(major,minor) ((xfs_dev_t)(((major)<<XFS_DEV_BITSMINOR) \
-				    | (minor&XFS_DEV_MAXMIN)))
-
-#define XFS_DEV_TO_DEVT(dev)	MKDEV(XFS_DEV_MAJOR(dev),XFS_DEV_MINOR(dev))
-
 #endif	/* !__XFS_TYPES_H */

include/linux/fs.h

@@ -1392,9 +1392,6 @@ struct tree_descr { char *name; struct file_operations *ops; int mode; };
 extern int simple_fill_super(struct super_block *, int, struct tree_descr *);
 extern int simple_pin_fs(char *name, struct vfsmount **mount, int *count);
 extern void simple_release_fs(struct vfsmount **mount, int *count);
-extern int old_valid_dev(dev_t);
-extern u16 old_encode_dev(dev_t);
-extern dev_t old_decode_dev(u16);
 
 extern int inode_change_ok(struct inode *, struct iattr *);
 extern int inode_setattr(struct inode *, struct iattr *);

include/linux/kdev_t.h

 #ifndef _LINUX_KDEV_T_H
 #define _LINUX_KDEV_T_H
 #ifdef __KERNEL__
-/* These are for user-level "dev_t" */
-#define MINORBITS	8
+#define MINORBITS	20
 #define MINORMASK	((1U << MINORBITS) - 1)
 
 #define MAJOR(dev)	((unsigned int) ((dev) >> MINORBITS))
@@ -18,6 +17,77 @@
 		buffer;							\
 	})
 
+/* acceptable for old filesystems */
+static inline int old_valid_dev(dev_t dev)
+{
+	return MAJOR(dev) < 256 && MINOR(dev) < 256;
+}
+
+static inline u16 old_encode_dev(dev_t dev)
+{
+	return (MAJOR(dev) << 8) | MINOR(dev);
+}
+
+static inline dev_t old_decode_dev(u16 val)
+{
+	return MKDEV((val >> 8) & 255, val & 255);
+}
+
+static inline int new_valid_dev(dev_t dev)
+{
+	return 1;
+}
+
+static inline u32 new_encode_dev(dev_t dev)
+{
+	unsigned major = MAJOR(dev);
+	unsigned minor = MINOR(dev);
+	return (minor & 0xff) | (major << 8) | ((minor & ~0xff) << 12);
+}
+
+static inline dev_t new_decode_dev(u32 dev)
+{
+	unsigned major = (dev & 0xfff00) >> 8;
+	unsigned minor = (dev & 0xff) | ((dev >> 12) & 0xfff00);
+	return MKDEV(major, minor);
+}
+
+static inline int huge_valid_dev(dev_t dev)
+{
+	return 1;
+}
+
+static inline u64 huge_encode_dev(dev_t dev)
+{
+	return new_encode_dev(dev);
+}
+
+static inline dev_t huge_decode_dev(u64 dev)
+{
+	return new_decode_dev(dev);
+}
+
+static inline int sysv_valid_dev(dev_t dev)
+{
+	return MAJOR(dev) < (1<<14) && MINOR(dev) < (1<<18);
+}
+
+static inline u32 sysv_encode_dev(dev_t dev)
+{
+	return MINOR(dev) | (MAJOR(dev) << 18);
+}
+
+static inline unsigned sysv_major(u32 dev)
+{
+	return (dev >> 18) & 0x3fff;
+}
+
+static inline unsigned sysv_minor(u32 dev)
+{
+	return dev & 0x3ffff;
+}
+
+
 #else /* __KERNEL__ */
 
 /*

include/linux/raid/md_k.h

@@ -64,11 +64,7 @@ static inline int level_to_pers (int level)
 typedef struct mddev_s mddev_t;
 typedef struct mdk_rdev_s mdk_rdev_t;
 
-#if (MINORBITS != 8)
-#error MD does not handle bigger kdev yet
-#endif
-
-#define MAX_MD_DEVS  (1<<MINORBITS)	/* Max number of md dev */
+#define MAX_MD_DEVS  256	/* Max number of md dev */
 
 /*
  * options passed in raidrun:

include/linux/types.h

@@ -17,7 +17,7 @@
 
 #ifndef __KERNEL_STRICT_NAMES
 
-typedef __u16 __kernel_dev_t;
+typedef __u32 __kernel_dev_t;
 
 typedef __kernel_fd_set		fd_set;
 typedef __kernel_dev_t		dev_t;