Remove msdosfs -o large support.

Its purpose was to translate the values for msdosfs inode numbers,
which is calculated from the msdosfs structures describing the file,
into the range representable by 32bit ino_t.  The translation acted
for filesystems larger than 128Gb, it reserved the range 0xf0000000
(FILENO_FIRST_DYN) to UINT32_MAX and remembered some arbitrary
translation of ino >= FILENO_FIRST_DYN into this range.  It consumed
memory that could be only freed by unmount, and the translation was
not stable across remounts.

With ino_t type extended to 64 bit, there is no such issue and values
can be returned without compaction to 32bit.  That is, for the native
environments, the translation layer is not necessary and adds
significant undeserved code complexity.  For compat ABIs which use
32bit ino_t, the vfs.ino64_trunc_error sysctl provides some measures
to soften the failure mode when inode numbers truncation is not safe.

Discussed with:	bde
Sponsored by:	The FreeBSD Foundation

sbin/mount_msdosfs/mount_msdosfs.8

@@ -30,7 +30,7 @@
 .\"
 .\" $FreeBSD$
 .\"
-.Dd April 3, 2017
+.Dd May 28, 2017
 .Dt MOUNT_MSDOSFS 8
 .Os
 .Sh NAME
@@ -73,11 +73,6 @@ as described in
 .Xr mount 8 .
 The following MSDOS file system-specific options are available:
 .Bl -tag -width indent
-.It Cm large
-Support file systems larger than 128 gigabytes at the expense
-of 32 bytes of kernel memory for each file on disk.
-This memory will not be reclaimed until the file system has
-been unmounted.
 .It Cm longnames
 Force Windows 95 long filenames to be visible.
 .It Cm shortnames

sys/conf/files

@@ -3323,7 +3323,6 @@ fs/fuse/fuse_vnops.c		optional fuse
 fs/msdosfs/msdosfs_conv.c	optional msdosfs
 fs/msdosfs/msdosfs_denode.c	optional msdosfs
 fs/msdosfs/msdosfs_fat.c	optional msdosfs
-fs/msdosfs/msdosfs_fileno.c	optional msdosfs
 fs/msdosfs/msdosfs_iconv.c	optional msdosfs_iconv
 fs/msdosfs/msdosfs_lookup.c	optional msdosfs
 fs/msdosfs/msdosfs_vfsops.c	optional msdosfs

sys/fs/msdosfs/msdosfs_fileno.c

@@ -1,152 +0,0 @@
-/*-
- * Copyright (c) 2003-2004 Tim J. Robbins.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- * 1. Redistributions of source code must retain the above copyright
- *    notice, this list of conditions and the following disclaimer.
- * 2. Redistributions in binary form must reproduce the above copyright
- *    notice, this list of conditions and the following disclaimer in the
- *    documentation and/or other materials provided with the distribution.
- *
- * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
- * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
- * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
- * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
- * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
- * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
- * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
- * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
- * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
- * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
- * SUCH DAMAGE.
- */
-/*
- * Compress 64-bit file numbers into temporary, unique 32-bit file numbers.
- * This is needed because the algorithm we use to calculate these numbers
- * generates 64-bit quantities, but struct dirent's d_fileno member and
- * struct vnodeattr's va_fileid member only have space for 32 bits.
- *
- * 32-bit file numbers are generated sequentially, and stored in a
- * red-black tree, indexed on 64-bit file number. The mappings do not
- * persist across reboots (or unmounts); anything that relies on this
- * (e.g. NFS) will not work correctly. This scheme consumes 32 bytes
- * of kernel memory per file (on i386), and it may be possible for a user
- * to cause a panic by creating millions of tiny files.
- *
- * As an optimization, we split the file number space between statically
- * allocated and dynamically allocated. File numbers less than
- * FILENO_FIRST_DYN are left unchanged and do not have any tree nodes
- * allocated to them.
- */
-
-#include <sys/cdefs.h>
-__FBSDID("$FreeBSD$");
-
-#include <sys/param.h>
-#include <sys/systm.h>
-#include <sys/kernel.h>
-#include <sys/lock.h>
-#include <sys/malloc.h>
-#include <sys/mount.h>
-
-#include <fs/msdosfs/bpb.h>
-#include <fs/msdosfs/direntry.h>
-#include <fs/msdosfs/msdosfsmount.h>
-
-static MALLOC_DEFINE(M_MSDOSFSFILENO, "msdosfs_fileno", "MSDOSFS fileno mapping node");
-
-RB_PROTOTYPE(msdosfs_filenotree, msdosfs_fileno, mf_tree,
-    msdosfs_fileno_compare)
-
-static int msdosfs_fileno_compare(struct msdosfs_fileno *,
-    struct msdosfs_fileno *);
-
-#define	FILENO_FIRST_DYN	0xf0000000
-
-/* Initialize file number mapping structures. */
-void
-msdosfs_fileno_init(struct mount *mp)
-{
-	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
-
-	RB_INIT(&pmp->pm_filenos);
-	pmp->pm_nfileno = FILENO_FIRST_DYN;
-	if (pmp->pm_HugeSectors > 0xffffffff /
-	    (pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1)
-		pmp->pm_flags |= MSDOSFS_LARGEFS;
-}
-
-/* Free 32-bit file number generation structures. */
-void
-msdosfs_fileno_free(struct mount *mp)
-{
-	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
-	struct msdosfs_fileno *mf, *next;
-
-	for (mf = RB_MIN(msdosfs_filenotree, &pmp->pm_filenos); mf != NULL;
-	    mf = next) {
-		next = RB_NEXT(msdosfs_filenotree, &pmp->pm_filenos, mf);
-		RB_REMOVE(msdosfs_filenotree, &pmp->pm_filenos, mf);
-		free(mf, M_MSDOSFSFILENO);
-	}
-}
-
-/* Map a 64-bit file number into a 32-bit one. */
-uint32_t
-msdosfs_fileno_map(struct mount *mp, uint64_t fileno)
-{
-	struct msdosfsmount *pmp = VFSTOMSDOSFS(mp);
-	struct msdosfs_fileno key, *mf, *tmf;
-	uint32_t mapped;
-
-	if ((pmp->pm_flags & MSDOSFS_LARGEFS) == 0) {
-		KASSERT((uint32_t)fileno == fileno,
-		    ("fileno >32 bits but not a large fs?"));
-		return ((uint32_t)fileno);
-	}
-	if (fileno < FILENO_FIRST_DYN)
-		return ((uint32_t)fileno);
-	MSDOSFS_LOCK_MP(pmp);
-	key.mf_fileno64 = fileno;
-	mf = RB_FIND(msdosfs_filenotree, &pmp->pm_filenos, &key);
-	if (mf != NULL) {
-		mapped = mf->mf_fileno32;
-		MSDOSFS_UNLOCK_MP(pmp);
-		return (mapped);
-	}
-	if (pmp->pm_nfileno < FILENO_FIRST_DYN)
-		panic("msdosfs_fileno_map: wraparound");
-	MSDOSFS_UNLOCK_MP(pmp);
-	mf = malloc(sizeof(*mf), M_MSDOSFSFILENO, M_WAITOK);
-	MSDOSFS_LOCK_MP(pmp);
-	tmf = RB_FIND(msdosfs_filenotree, &pmp->pm_filenos, &key);
-	if (tmf != NULL) {
-		mapped = tmf->mf_fileno32;
-		MSDOSFS_UNLOCK_MP(pmp);
-		free(mf, M_MSDOSFSFILENO);
-		return (mapped);
-	}
-	mf->mf_fileno64 = fileno;
-	mapped = mf->mf_fileno32 = pmp->pm_nfileno++;
-	RB_INSERT(msdosfs_filenotree, &pmp->pm_filenos, mf);
-	MSDOSFS_UNLOCK_MP(pmp);
-	return (mapped);
-}
-
-/* Compare by 64-bit file number. */
-static int
-msdosfs_fileno_compare(struct msdosfs_fileno *fa, struct msdosfs_fileno *fb)
-{
-
-	if (fa->mf_fileno64 > fb->mf_fileno64)
-		return (1);
-	else if (fa->mf_fileno64 < fb->mf_fileno64)
-		return (-1);
-	return (0);
-}
-
-RB_GENERATE(msdosfs_filenotree, msdosfs_fileno, mf_tree,
-    msdosfs_fileno_compare)

sys/fs/msdosfs/msdosfs_vfsops.c

@@ -82,7 +82,7 @@ static const char *msdosfs_opts[] = {
 	"async", "noatime", "noclusterr", "noclusterw",
 	"export", "force", "from", "sync",
 	"cs_dos", "cs_local", "cs_win", "dirmask",
-	"gid", "kiconv", "large", "longname",
+	"gid", "kiconv", "longname",
 	"longnames", "mask", "shortname", "shortnames",
 	"uid", "win95", "nowin95",
 	NULL
@@ -242,17 +242,6 @@ msdosfs_mount(struct mount *mp)
 	 */
 	if (mp->mnt_flag & MNT_UPDATE) {
 		pmp = VFSTOMSDOSFS(mp);
-		if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) {
-			/*
-			 * Forbid export requests if filesystem has
-			 * MSDOSFS_LARGEFS flag set.
-			 */
-			if ((pmp->pm_flags & MSDOSFS_LARGEFS) != 0) {
-				vfs_mount_error(mp,
-				    "MSDOSFS_LARGEFS flag set, cannot export");
-				return (EOPNOTSUPP);
-			}
-		}
 		if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
 		    vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0)) {
 			error = VFS_SYNC(mp, MNT_WAIT);
@@ -466,20 +455,6 @@ mountmsdosfs(struct vnode *devvp, struct mount *mp)
 	pmp->pm_mask = pmp->pm_dirmask = S_IXUSR | S_IXGRP | S_IXOTH |
 	    S_IRUSR | S_IRGRP | S_IROTH | S_IWUSR;
 
-	/*
-	 * Experimental support for large MS-DOS filesystems.
-	 * WARNING: This uses at least 32 bytes of kernel memory (which is not
-	 * reclaimed until the FS is unmounted) for each file on disk to map
-	 * between the 32-bit inode numbers used by VFS and the 64-bit
-	 * pseudo-inode numbers used internally by msdosfs. This is only
-	 * safe to use in certain controlled situations (e.g. read-only FS
-	 * with less than 1 million files).
-	 * Since the mappings do not persist across unmounts (or reboots), these
-	 * filesystems are not suitable for exporting through NFS, or any other
-	 * application that requires fixed inode numbers.
-	 */
-	vfs_flagopt(mp->mnt_optnew, "large", &pmp->pm_flags, MSDOSFS_LARGEFS);
-
 	/*
 	 * Compute several useful quantities from the bpb in the
 	 * bootsector.  Copy in the dos 5 variant of the bpb then fix up
@@ -521,20 +496,6 @@ mountmsdosfs(struct vnode *devvp, struct mount *mp)
 		pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
 		pmp->pm_HugeSectors = pmp->pm_Sectors;
 	}
-	if (!(pmp->pm_flags & MSDOSFS_LARGEFS)) {
-		if (pmp->pm_HugeSectors > 0xffffffff /
-		    (pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1) {
-			/*
-			 * We cannot deal currently with this size of disk
-			 * due to fileid limitations (see msdosfs_getattr and
-			 * msdosfs_readdir)
-			 */
-			error = EINVAL;
-			vfs_mount_error(mp,
-			    "Disk too big, try '-o large' mount option");
-			goto error_exit;
-		}
-	}
 
 	if (pmp->pm_RootDirEnts == 0) {
 		if (pmp->pm_FATsecs
@@ -745,10 +706,7 @@ mountmsdosfs(struct vnode *devvp, struct mount *mp)
 	mp->mnt_kern_flag |= MNTK_USES_BCACHE | MNTK_NO_IOPF;
 	MNT_IUNLOCK(mp);
 
-	if (pmp->pm_flags & MSDOSFS_LARGEFS)
+	return (0);
-		msdosfs_fileno_init(mp);
-
-	return 0;
 
 error_exit:
 	if (bp)
@@ -835,8 +793,6 @@ msdosfs_unmount(struct mount *mp, int mntflags)
 	vrele(pmp->pm_devvp);
 	dev_rel(pmp->pm_dev);
 	free(pmp->pm_inusemap, M_MSDOSFSFAT);
-	if (pmp->pm_flags & MSDOSFS_LARGEFS)
-		msdosfs_fileno_free(mp);
 	lockdestroy(&pmp->pm_fatlock);
 	free(pmp, M_MSDOSFSMNT);
 	mp->mnt_data = NULL;

sys/fs/msdosfs/msdosfs_vnops.c

@@ -284,11 +284,7 @@ msdosfs_getattr(struct vop_getattr_args *ap)
 			fileid = (uint64_t)roottobn(pmp, 0) * dirsperblk;
 		fileid += (uoff_t)dep->de_diroffset / sizeof(struct direntry);
 	}
-
+	vap->va_fileid = fileid;
-	if (pmp->pm_flags & MSDOSFS_LARGEFS)
-		vap->va_fileid = msdosfs_fileno_map(pmp->pm_mountp, fileid);
-	else
-		vap->va_fileid = (long)fileid;
 
 	mode = S_IRWXU|S_IRWXG|S_IRWXO;
 	vap->va_mode = mode & 
@@ -1472,7 +1468,6 @@ msdosfs_readdir(struct vop_readdir_args *ap)
 	int blsize;
 	long on;
 	u_long cn;
-	uint64_t fileno;
 	u_long dirsperblk;
 	long bias = 0;
 	daddr_t bn, lbn;
@@ -1543,20 +1538,9 @@ msdosfs_readdir(struct vop_readdir_args *ap)
 		if (offset < bias) {
 			for (n = (int)offset / sizeof(struct direntry);
 			     n < 2; n++) {
-				if (FAT32(pmp))
+				dirbuf.d_fileno = FAT32(pmp) ?
-					fileno = (uint64_t)cntobn(pmp,
+				    (uint64_t)cntobn(pmp, pmp->pm_rootdirblk) *
-								 pmp->pm_rootdirblk)
+				    dirsperblk : 1;
-							  * dirsperblk;
-				else
-					fileno = 1;
-				if (pmp->pm_flags & MSDOSFS_LARGEFS) {
-					dirbuf.d_fileno =
-					    msdosfs_fileno_map(pmp->pm_mountp,
-					    fileno);
-				} else {
-
-					dirbuf.d_fileno = (uint32_t)fileno;
-				}
 				dirbuf.d_type = DT_DIR;
 				switch (n) {
 				case 0:
@@ -1661,31 +1645,24 @@ msdosfs_readdir(struct vop_readdir_args *ap)
 			 * msdosfs_getattr.
 			 */
 			if (dentp->deAttributes & ATTR_DIRECTORY) {
-				fileno = getushort(dentp->deStartCluster);
+				cn = getushort(dentp->deStartCluster);
-				if (FAT32(pmp))
+				if (FAT32(pmp)) {
-					fileno |= getushort(dentp->deHighClust) << 16;
+					cn |= getushort(dentp->deHighClust) <<
-				/* if this is the root directory */
+					    16;
-				if (fileno == MSDOSFSROOT)
+					if (cn == MSDOSFSROOT)
-					if (FAT32(pmp))
+						cn = pmp->pm_rootdirblk;
-						fileno = (uint64_t)cntobn(pmp,
+				}
-								pmp->pm_rootdirblk)
+				if (cn == MSDOSFSROOT && !FAT32(pmp))
-							 * dirsperblk;
+					dirbuf.d_fileno = 1;
-					else
-						fileno = 1;
 				else
-					fileno = (uint64_t)cntobn(pmp, fileno) *
+					dirbuf.d_fileno = cntobn(pmp, cn) *
 					    dirsperblk;
 				dirbuf.d_type = DT_DIR;
 			} else {
-				fileno = (uoff_t)offset /
+				dirbuf.d_fileno = (uoff_t)offset /
 				    sizeof(struct direntry);
 				dirbuf.d_type = DT_REG;
 			}
-			if (pmp->pm_flags & MSDOSFS_LARGEFS) {
-				dirbuf.d_fileno =
-				    msdosfs_fileno_map(pmp->pm_mountp, fileno);
-			} else
-				dirbuf.d_fileno = (uint32_t)fileno;
 
 			if (chksum != winChksum(dentp->deName)) {
 				dirbuf.d_namlen = dos2unixfn(dentp->deName,

sys/fs/msdosfs/msdosfsmount.h

@@ -108,10 +108,7 @@ struct msdosfsmount {
 	void *pm_w2u;	/* Unicode->Local iconv handle */
 	void *pm_u2d;	/* Unicode->DOS iconv handle */
 	void *pm_d2u;	/* DOS->Local iconv handle */
-	uint32_t pm_nfileno;	/* next 32-bit fileno */
+	struct lock pm_fatlock;	/* lockmgr protecting allocations */
-	RB_HEAD(msdosfs_filenotree, msdosfs_fileno)
-	    pm_filenos; /* 64<->32-bit fileno mapping */
-	struct lock pm_fatlock;	/* lockmgr protecting allocations and rb tree */
 };
 
 /*
@@ -216,10 +213,6 @@ struct msdosfs_fileno {
 	 ? roottobn((pmp), (dirofs)) \
 	 : cntobn((pmp), (dirclu)))
 
-void msdosfs_fileno_init(struct mount *);
-void msdosfs_fileno_free(struct mount *);
-uint32_t msdosfs_fileno_map(struct mount *, uint64_t);
-
 #define	MSDOSFS_LOCK_MP(pmp) \
 	lockmgr(&(pmp)->pm_fatlock, LK_EXCLUSIVE, NULL)
 #define	MSDOSFS_UNLOCK_MP(pmp) \
@@ -261,7 +254,6 @@ struct msdosfs_args {
 #define	MSDOSFSMNT_RONLY	0x80000000	/* mounted read-only	*/
 #define	MSDOSFSMNT_WAITONFAT	0x40000000	/* mounted synchronous	*/
 #define	MSDOSFS_FATMIRROR	0x20000000	/* FAT is mirrored */
-#define	MSDOSFS_LARGEFS		0x10000000	/* perform fileno mapping */
 #define	MSDOSFS_FSIMOD		0x01000000
 
 #endif /* !_MSDOSFS_MSDOSFSMOUNT_H_ */

sys/modules/msdosfs/Makefile

@@ -4,7 +4,7 @@
 
 KMOD=	msdosfs
 SRCS=	vnode_if.h \
-	msdosfs_conv.c msdosfs_denode.c msdosfs_fat.c msdosfs_fileno.c \
+	msdosfs_conv.c msdosfs_denode.c msdosfs_fat.c \
 	msdosfs_lookup.c msdosfs_vfsops.c msdosfs_vnops.c
 EXPORT_SYMS=	msdosfs_iconv