summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authortjr <tjr@FreeBSD.org>2004-07-03 13:22:38 +0000
committertjr <tjr@FreeBSD.org>2004-07-03 13:22:38 +0000
commitab16560f3305829961b71dbe71233abcd50f0aba (patch)
tree651a802dbc4eb055206f7c5ccba7348f998b53bd
parent158b00dfc0a2916d3a3e23aa2bd11bd84d6c0841 (diff)
downloadFreeBSD-src-ab16560f3305829961b71dbe71233abcd50f0aba.zip
FreeBSD-src-ab16560f3305829961b71dbe71233abcd50f0aba.tar.gz
By popular request, add a workaround that allows large (>128GB or so)
FAT32 filesystems to be mounted, subject to some fairly serious limitations. This works by extending the internal pseudo-inode-numbers generated from the file's starting cluster number to 64-bits, then creating a table mapping these into arbitrary 32-bit inode numbers, which can fit in struct dirent's d_fileno and struct vattr's va_fileid fields. The mappings do not persist across unmounts or reboots, so it's not possible to export these filesystems through NFS. The mapping table may grow to be rather large, and may grow large enough to exhaust kernel memory on filesystems with millions of files. Don't enable this option unless you understand the consequences.
-rw-r--r--sys/conf/NOTES12
-rw-r--r--sys/conf/files1
-rw-r--r--sys/conf/options3
-rw-r--r--sys/fs/msdosfs/msdosfs_fileno.c163
-rw-r--r--sys/fs/msdosfs/msdosfs_vfsops.c14
-rw-r--r--sys/fs/msdosfs/msdosfs_vnops.c46
-rw-r--r--sys/fs/msdosfs/msdosfsmount.h24
-rw-r--r--sys/modules/msdosfs/Makefile5
8 files changed, 253 insertions, 15 deletions
diff --git a/sys/conf/NOTES b/sys/conf/NOTES
index f944152..fd1b42a 100644
--- a/sys/conf/NOTES
+++ b/sys/conf/NOTES
@@ -843,6 +843,18 @@ options MSDOSFS_ICONV
options NTFS_ICONV
options UDF_ICONV
+# 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.
+options MSDOSFS_LARGE
+
#####################################################################
# POSIX P1003.1B
diff --git a/sys/conf/files b/sys/conf/files
index a25b963..7237d64 100644
--- a/sys/conf/files
+++ b/sys/conf/files
@@ -880,6 +880,7 @@ fs/msdosfs/msdosfs_fat.c optional msdosfs
fs/msdosfs/msdosfs_lookup.c optional msdosfs
fs/msdosfs/msdosfs_vfsops.c optional msdosfs
fs/msdosfs/msdosfs_vnops.c optional msdosfs
+fs/msdosfs/msdosfs_fileno.c optional msdosfs_large
fs/msdosfs/msdosfs_iconv.c optional msdosfs_iconv
fs/ntfs/ntfs_compr.c optional ntfs
fs/ntfs/ntfs_ihash.c optional ntfs
diff --git a/sys/conf/options b/sys/conf/options
index 2fb6582..b593318 100644
--- a/sys/conf/options
+++ b/sys/conf/options
@@ -219,6 +219,9 @@ NFS_ROOT opt_nfsroot.h
NETSMB opt_netsmb.h
NETSMBCRYPTO opt_netsmb.h
+# Experimental support for large MS-DOS filesystems; SEE WARNING IN "NOTES"!
+MSDOSFS_LARGE opt_msdosfs.h
+
# Options used only in subr_param.c.
HZ opt_param.h
MAXFILES opt_param.h
diff --git a/sys/fs/msdosfs/msdosfs_fileno.c b/sys/fs/msdosfs/msdosfs_fileno.c
new file mode 100644
index 0000000..ad6d4a7
--- /dev/null
+++ b/sys/fs/msdosfs/msdosfs_fileno.c
@@ -0,0 +1,163 @@
+/*-
+ * 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/conf.h>
+#include <sys/kernel.h>
+#include <sys/mount.h>
+#include <sys/malloc.h>
+#include <sys/lock.h>
+#include <sys/mutex.h>
+
+#include <fs/msdosfs/bpb.h>
+#include <fs/msdosfs/bootsect.h>
+#include <fs/msdosfs/msdosfsmount.h>
+#include <fs/msdosfs/direntry.h>
+
+static MALLOC_DEFINE(M_MSDOSFSFILENO, "MSDOSFS fileno", "MSDOSFS fileno mapping node");
+
+static struct mtx fileno_mtx;
+MTX_SYSINIT(fileno, &fileno_mtx, "MSDOSFS fileno", MTX_DEF);
+
+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(mp)
+ 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(mp)
+ 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(mp, fileno)
+ 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);
+ mtx_lock(&fileno_mtx);
+ key.mf_fileno64 = fileno;
+ mf = RB_FIND(msdosfs_filenotree, &pmp->pm_filenos, &key);
+ if (mf != NULL) {
+ mapped = mf->mf_fileno32;
+ mtx_unlock(&fileno_mtx);
+ return (mapped);
+ }
+ if (pmp->pm_nfileno < FILENO_FIRST_DYN)
+ panic("msdosfs_fileno_map: wraparound");
+ mtx_unlock(&fileno_mtx);
+ mf = malloc(sizeof(*mf), M_MSDOSFSFILENO, M_WAITOK);
+ mtx_lock(&fileno_mtx);
+ tmf = RB_FIND(msdosfs_filenotree, &pmp->pm_filenos, &key);
+ if (tmf != NULL) {
+ mapped = tmf->mf_fileno32;
+ mtx_unlock(&fileno_mtx);
+ 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);
+ mtx_unlock(&fileno_mtx);
+ return (mapped);
+}
+
+/* Compare by 64-bit file number. */
+static int
+msdosfs_fileno_compare(fa, fb)
+ struct msdosfs_fileno *fa, *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)
diff --git a/sys/fs/msdosfs/msdosfs_vfsops.c b/sys/fs/msdosfs/msdosfs_vfsops.c
index 3ac0fb6..5d0b3c5 100644
--- a/sys/fs/msdosfs/msdosfs_vfsops.c
+++ b/sys/fs/msdosfs/msdosfs_vfsops.c
@@ -71,6 +71,8 @@
#include <fs/msdosfs/denode.h>
#include <fs/msdosfs/fat.h>
+#include "opt_msdosfs.h"
+
#define MSDOSFS_DFLTBSIZE 4096
#if 1 /*def PC98*/
@@ -227,6 +229,9 @@ msdosfs_mount(mp, path, data, ndp, td)
/*
* Process export requests.
*/
+ if ((args.export.ex_flags & MNT_EXPORTED) != 0 &&
+ (pmp->pm_flags & MSDOSFS_LARGEFS) != 0)
+ return (EOPNOTSUPP);
return (vfs_export(mp, &args.export));
}
}
@@ -415,6 +420,7 @@ mountmsdosfs(devvp, mp, td, argp)
pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
pmp->pm_HugeSectors = pmp->pm_Sectors;
}
+#ifndef MSDOSFS_LARGE
if (pmp->pm_HugeSectors > 0xffffffff /
(pmp->pm_BytesPerSec / sizeof(struct direntry)) + 1) {
/*
@@ -426,6 +432,7 @@ mountmsdosfs(devvp, mp, td, argp)
printf("mountmsdosfs(): disk too big, sorry\n");
goto error_exit;
}
+#endif /* !MSDOSFS_LARGE */
if (pmp->pm_RootDirEnts == 0) {
if (bsp->bs710.bsBootSectSig2 != BOOTSIG2
@@ -628,6 +635,10 @@ mountmsdosfs(devvp, mp, td, argp)
mp->mnt_flag |= MNT_LOCAL;
devvp->v_rdev->si_mountpoint = mp;
+#ifdef MSDOSFS_LARGE
+ msdosfs_fileno_init(mp);
+#endif
+
return 0;
error_exit:
@@ -720,6 +731,9 @@ msdosfs_unmount(mp, mntflags, td)
#endif
vrele(pmp->pm_devvp);
free(pmp->pm_inusemap, M_MSDOSFSFAT);
+#ifdef MSDOSFS_LARGE
+ msdosfs_fileno_free(mp);
+#endif
free(pmp, M_MSDOSFSMNT);
mp->mnt_data = (qaddr_t)0;
mp->mnt_flag &= ~MNT_LOCAL;
diff --git a/sys/fs/msdosfs/msdosfs_vnops.c b/sys/fs/msdosfs/msdosfs_vnops.c
index 728fd10..05a296e 100644
--- a/sys/fs/msdosfs/msdosfs_vnops.c
+++ b/sys/fs/msdosfs/msdosfs_vnops.c
@@ -77,6 +77,8 @@
#include <fs/msdosfs/denode.h>
#include <fs/msdosfs/fat.h>
+#include "opt_msdosfs.h"
+
#define DOS_FILESIZE_MAX 0xffffffff
/*
@@ -284,7 +286,7 @@ msdosfs_getattr(ap)
mode_t mode;
struct timespec ts;
u_long dirsperblk = pmp->pm_BytesPerSec / sizeof(struct direntry);
- u_long fileid;
+ uint64_t fileid;
getnanotime(&ts);
DETIMES(dep, &ts, &ts, &ts);
@@ -295,16 +297,22 @@ msdosfs_getattr(ap)
* doesn't work.
*/
if (dep->de_Attributes & ATTR_DIRECTORY) {
- fileid = cntobn(pmp, dep->de_StartCluster) * dirsperblk;
+ fileid = (uint64_t)cntobn(pmp, dep->de_StartCluster) *
+ dirsperblk;
if (dep->de_StartCluster == MSDOSFSROOT)
fileid = 1;
} else {
- fileid = cntobn(pmp, dep->de_dirclust) * dirsperblk;
+ fileid = (uint64_t)cntobn(pmp, dep->de_dirclust) *
+ dirsperblk;
if (dep->de_dirclust == MSDOSFSROOT)
- fileid = roottobn(pmp, 0) * dirsperblk;
- fileid += dep->de_diroffset / sizeof(struct direntry);
+ fileid = (uint64_t)roottobn(pmp, 0) * dirsperblk;
+ fileid += (uint64_t)dep->de_diroffset / sizeof(struct direntry);
}
- vap->va_fileid = fileid;
+#ifdef MSDOSFS_LARGE
+ vap->va_fileid = msdosfs_fileno_map(pmp->pm_mountp, fileid);
+#else
+ vap->va_fileid = (long)fileid;
+#endif
if ((dep->de_Attributes & ATTR_READONLY) == 0)
mode = S_IRWXU|S_IRWXG|S_IRWXO;
else
@@ -1453,7 +1461,7 @@ msdosfs_readdir(ap)
int blsize;
long on;
u_long cn;
- u_long fileno;
+ uint64_t fileno;
u_long dirsperblk;
long bias = 0;
daddr_t bn, lbn;
@@ -1525,11 +1533,17 @@ msdosfs_readdir(ap)
for (n = (int)offset / sizeof(struct direntry);
n < 2; n++) {
if (FAT32(pmp))
- dirbuf.d_fileno = cntobn(pmp,
+ fileno = (uint64_t)cntobn(pmp,
pmp->pm_rootdirblk)
* dirsperblk;
else
- dirbuf.d_fileno = 1;
+ fileno = 1;
+#ifdef MSDOSFS_LARGE
+ dirbuf.d_fileno = msdosfs_fileno_map(
+ pmp->pm_mountp, fileno);
+#else
+ dirbuf.d_fileno = (uint32_t)fileno;
+#endif
dirbuf.d_type = DT_DIR;
switch (n) {
case 0:
@@ -1640,19 +1654,25 @@ msdosfs_readdir(ap)
/* if this is the root directory */
if (fileno == MSDOSFSROOT)
if (FAT32(pmp))
- fileno = cntobn(pmp,
+ fileno = (uint64_t)cntobn(pmp,
pmp->pm_rootdirblk)
* dirsperblk;
else
fileno = 1;
else
- fileno = cntobn(pmp, fileno) * dirsperblk;
- dirbuf.d_fileno = fileno;
+ fileno = (uint64_t)cntobn(pmp, fileno) *
+ dirsperblk;
dirbuf.d_type = DT_DIR;
} else {
- dirbuf.d_fileno = offset / sizeof(struct direntry);
+ fileno = (uint64_t)offset / sizeof(struct direntry);
dirbuf.d_type = DT_REG;
}
+#ifdef MSDOSFS_LARGE
+ dirbuf.d_fileno = msdosfs_fileno_map(pmp->pm_mountp,
+ fileno);
+#else
+ dirbuf.d_fileno = (uint32_t)fileno;
+#endif
if (chksum != winChksum(dentp->deName)) {
dirbuf.d_namlen = dos2unixfn(dentp->deName,
(u_char *)dirbuf.d_name,
diff --git a/sys/fs/msdosfs/msdosfsmount.h b/sys/fs/msdosfs/msdosfsmount.h
index 0ff712f..46f9cc5 100644
--- a/sys/fs/msdosfs/msdosfsmount.h
+++ b/sys/fs/msdosfs/msdosfsmount.h
@@ -53,10 +53,14 @@
#ifdef _KERNEL
+#include <sys/tree.h>
+
#ifdef MALLOC_DECLARE
MALLOC_DECLARE(M_MSDOSFSMNT);
#endif
+struct msdosfs_fileno;
+
/*
* Layout of the mount control block for a msdos filesystem.
*/
@@ -99,7 +103,20 @@ struct msdosfsmount {
void *pm_w2u; /* Unicode->Local iconv handle */
void *pm_u2d; /* Unicode->DOS iconv handle */
void *pm_d2u; /* DOS->Local iconv handle */
+ u_int32_t pm_nfileno; /* next 32-bit fileno */
+ RB_HEAD(msdosfs_filenotree, msdosfs_fileno) pm_filenos; /* 64<->32-bit fileno mapping */
+};
+
+/*
+ * A 64-bit file number and the 32-bit file number to which it is mapped,
+ * in a red-black tree node.
+ */
+struct msdosfs_fileno {
+ RB_ENTRY(msdosfs_fileno) mf_tree;
+ uint32_t mf_fileno32;
+ uint64_t mf_fileno64;
};
+
/* Byte offset in FAT on filesystem pmp, cluster cn */
#define FATOFS(pmp, cn) ((cn) * (pmp)->pm_fatmult / (pmp)->pm_fatdiv)
@@ -202,6 +219,10 @@ int msdosfs_init(struct vfsconf *vfsp);
int msdosfs_uninit(struct vfsconf *vfsp);
int msdosfs_mountroot(void);
+void msdosfs_fileno_init(struct mount *);
+void msdosfs_fileno_free(struct mount *);
+uint32_t msdosfs_fileno_map(struct mount *, uint64_t);
+
#endif /* _KERNEL */
/*
@@ -219,7 +240,7 @@ struct msdosfs_args {
char *cs_win; /* Windows(Unicode) Charset */
char *cs_dos; /* DOS Charset */
char *cs_local; /* Local Charset */
- mode_t dirmask; /* dir mask to be applied for msdosfs perms */
+ mode_t dirmask; /* dir mask to be applied for msdosfs perms */
};
/*
@@ -236,6 +257,7 @@ 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_ARGSMAGIC 0xe4eff300
diff --git a/sys/modules/msdosfs/Makefile b/sys/modules/msdosfs/Makefile
index ef8989f..9ee8451 100644
--- a/sys/modules/msdosfs/Makefile
+++ b/sys/modules/msdosfs/Makefile
@@ -3,9 +3,12 @@
.PATH: ${.CURDIR}/../../fs/msdosfs
KMOD= msdosfs
-SRCS= vnode_if.h \
+SRCS= opt_msdosfs.h vnode_if.h \
msdosfs_conv.c msdosfs_denode.c msdosfs_fat.c msdosfs_lookup.c \
msdosfs_vfsops.c msdosfs_vnops.c
EXPORT_SYMS= msdosfs_iconv
+opt_msdosfs.h:
+ touch ${.TARGET}
+
.include <bsd.kmod.mk>
OpenPOWER on IntegriCloud