From: Raphael S. Carvalho <raphael.scarv at gmail.com>
Change since v1:
* Fix bug on dentry structure (thank you specification; btw, sarcasm), and
consequently a bug on ufs_readdir.
* Add readlink support (applied tests for symlinks whose destionation path
were stored in blk pointers and the file itself).
* Several improvements.
Wrote the documentation below. I think it would be good to push the doc to
the wiki as soon as the UFS support gets merged.
Unix Fast File System (UFS/FFS) 1/2 on Syslinux - (usage/install)
-----
There is a confusion about the name of this file system, then I decided to
contact the author, Kirk McKusick, who replied:
"The name has always been confusing (my bad). The code is broken into two
parts, the part that handles naming (UFS where the U stands for Unix), and the
part that handles disk layout (FFS where the F stands for Fast).
When the two parts are put together they are called UFS/FFS or more commonly
just UFS."
Dependencies:
- Environment:
Require a *GNU/Linux* system to follow these steps.
- Package:
ufsutils (UFS filesystems utilities)
- Device driver:
UFS support on Linux: you need a kernel with the option CONFIG_UFS_FS_WRITE
enabled or at least configured as a module.
For the latter case, load the module ufs.ko with modprobe (probably located
at: /lib/modules/).
If this option wasn't even configured as a module, then you need to
recompile
your kernel with that option enabled.
Further information can be found in the link below:
http://oz123.github.io/writings/2012-11-26-enable-ufs-write-support-debian/
Creating an UFS image:
sudo dd if=/dev/zero of=<image> bs= count=;
- UFS1:
sudo mkfs.ufs -O 1 <image>;
- UFS2:
sudo mkfs.ufs <image>;
Mounting an UFS image:
- UFS1:
sudo mount -t ufs -o rw,loop,ufstype=44bsd <image> <mountpoint>;
- UFS2:
sudo mount -t ufs -o rw,loop,ufstype=ufs2 <image> <mountpoint>;
Installing syslinux on an UFS image:
- Install extlinux into your mount point:
sudo extlinux -i <mountpoint>;
- Finally, set up the syslinux config file.
Thanks for pcacjr who introduced me to Syslinux :-)
Raphael S. Carvalho (2):
installer: Add UFS1/2 support for Extlinux installer
core/fs: Add support to Unix File system 1/2.
core/fs/ufs/bmap.c | 202 ++++++++++++++++++++
core/fs/ufs/ufs.c | 486 ++++++++++++++++++++++++++++++++++++++++++++++++
core/fs/ufs/ufs.h | 254 +++++++++++++++++++++++++
core/ldlinux.asm | 2 +
extlinux/main.c | 51 ++++-
extlinux/ufs.h | 26 +++
extlinux/ufs_fs.h | 307 ++++++++++++++++++++++++++++++
libinstaller/syslxfs.h | 4 +-
8 files changed, 1322 insertions(+), 10 deletions(-)
create mode 100644 core/fs/ufs/bmap.c
create mode 100644 core/fs/ufs/ufs.c
create mode 100644 core/fs/ufs/ufs.h
create mode 100644 extlinux/ufs.h
create mode 100644 extlinux/ufs_fs.h
--
1.7.2.5
Raphael S.Carvalho
2014-May-29 23:35 UTC
[syslinux] [PATCH v2 1/2] installer: Add UFS1/2 support for Extlinux installer
From: Raphael S. Carvalho <raphael.scarv at gmail.com>
It's needed to enumerate both UFS1 and UFS2 as they have different magic
numbers
and super block offsets.
Neither UFS1 nor UFS2 have COW feature, therefore, ldlinux.sys can be installed
as a regular file.
Signed-off-by: Raphael S. Carvalho <raphael.scarv at gmail.com>
---
extlinux/main.c | 51 +++++++--
extlinux/ufs.h | 26 ++++
extlinux/ufs_fs.h | 307 ++++++++++++++++++++++++++++++++++++++++++++++++
libinstaller/syslxfs.h | 4 +-
4 files changed, 378 insertions(+), 10 deletions(-)
create mode 100644 extlinux/ufs.h
create mode 100644 extlinux/ufs_fs.h
diff --git a/extlinux/main.c b/extlinux/main.c
index 47c1e68..6b9a04b 100644
--- a/extlinux/main.c
+++ b/extlinux/main.c
@@ -14,8 +14,8 @@
/*
* extlinux.c
*
- * Install the syslinux boot block on an fat, ntfs, ext2/3/4, btrfs and xfs
- * filesystem.
+ * Install the syslinux boot block on an fat, ntfs, ext2/3/4, btrfs, xfs,
+ * and ufs1/2 filesystem.
*/
#define _GNU_SOURCE /* Enable everything */
@@ -50,6 +50,8 @@
#include "xfs.h"
#include "xfs_types.h"
#include "xfs_sb.h"
+#include "ufs.h"
+#include "ufs_fs.h"
#include "misc.h"
#include "version.h"
#include "syslxint.h"
@@ -308,7 +310,7 @@ static int patch_file_and_bootblock(int fd, const char *dir,
int devfd)
nsect += 2; /* Two sectors for the ADV */
sectp = alloca(sizeof(sector_t) * nsect);
if (fs_type == EXT2 || fs_type == VFAT || fs_type == NTFS ||
- fs_type == XFS) {
+ fs_type == XFS || fs_type == UFS1 || fs_type == UFS2) {
if (sectmap(fd, sectp, nsect)) {
perror("bmap");
exit(1);
@@ -342,6 +344,7 @@ int install_bootblock(int fd, const char *device)
struct fat_boot_sector sb3;
struct ntfs_boot_sector sb4;
xfs_sb_t sb5;
+ struct ufs_super_block sb6;
bool ok = false;
if (fs_type == EXT2) {
@@ -395,11 +398,25 @@ int install_bootblock(int fd, const char *device)
ok = true;
}
+ } else if (fs_type == UFS1 || fs_type == UFS2) {
+ uint32_t sblock_off = (fs_type == UFS1) ?
+ SBLOCK_UFS1 : SBLOCK_UFS2;
+ uint32_t ufs_smagic = (fs_type == UFS1) ?
+ UFS1_SUPER_MAGIC : UFS2_SUPER_MAGIC;
+
+ if (xpread(fd, &sb6, sizeof sb6, sblock_off) != sizeof sb6) {
+ perror("reading superblock");
+ return 1;
+ }
+
+ if (sb6.fs_magic == ufs_smagic)
+ ok = true;
}
if (!ok) {
fprintf(stderr,
- "no fat, ntfs, ext2/3/4, btrfs or xfs superblock found on %s\n",
+ "no fat, ntfs, ext2/3/4, btrfs, xfs "
+ "or ufs1/2 superblock found on %s\n",
device);
return 1;
}
@@ -959,7 +976,8 @@ static char * get_default_subvol(char * rootdir, char *
subvol)
static int install_file(const char *path, int devfd, struct stat *rst)
{
- if (fs_type == EXT2 || fs_type == VFAT || fs_type == NTFS)
+ if (fs_type == EXT2 || fs_type == VFAT || fs_type == NTFS
+ || fs_type == UFS1 || fs_type == UFS2)
return ext2_fat_install_file(path, devfd, rst);
else if (fs_type == BTRFS)
return btrfs_install_file(path, devfd, rst);
@@ -987,7 +1005,7 @@ static int validate_device(const char *path, int devfd)
struct statfs sfs;
int pfd;
int rv = -1;
-
+
pfd = open(path, O_RDONLY|O_DIRECTORY);
if (pfd < 0)
goto err;
@@ -1067,6 +1085,16 @@ static const char *find_device(const char *mtab_file,
dev_t dev)
done = true;
break;
}
+
+ break;
+ case UFS1:
+ case UFS2:
+ if (!strcmp(mnt->mnt_type, "ufs") &&
!stat(mnt->mnt_fsname, &dst) &&
+ dst.st_rdev == dev) {
+ done = true;
+ }
+
+ break;
case NONE:
break;
}
@@ -1191,7 +1219,7 @@ static int validate_device_btrfs(int pfd, int dfd)
return -1;
return 0; /* It's good! */
-}
+}
static const char *find_device_btrfs(const char *path)
{
@@ -1280,7 +1308,7 @@ static const char *get_devname(const char *path)
fprintf(stderr, "%s: cannot create device %s\n", program,
devname);
return devname;
}
-
+
atexit(device_cleanup); /* unlink the device node on exit */
devname = devname_buf;
}
@@ -1334,10 +1362,15 @@ static int open_device(const char *path, struct stat
*st, const char **_devname)
fs_type = NTFS;
else if (sfs.f_type == XFS_SUPER_MAGIC)
fs_type = XFS;
+ else if (sfs.f_type == UFS1_SUPER_MAGIC)
+ fs_type = UFS1;
+ else if (sfs.f_type == UFS2_SUPER_MAGIC)
+ fs_type = UFS2;
if (!fs_type) {
fprintf(stderr,
- "%s: not a fat, ntfs, ext2/3/4, btrfs or xfs filesystem: %s\n",
+ "%s: not a fat, ntfs, ext2/3/4, btrfs, xfs or"
+ "ufs1/2 filesystem: %s\n",
program, path);
return -1;
}
diff --git a/extlinux/ufs.h b/extlinux/ufs.h
new file mode 100644
index 0000000..d324699
--- /dev/null
+++ b/extlinux/ufs.h
@@ -0,0 +1,26 @@
+/*
+ * Copyright (c) 2013 Raphael S. Carvalho <raphael.scarv at gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the
+ * Free Software Foundation, Inc.,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+*/
+
+#ifndef UFS_H_
+#define UFS_H_
+
+#define UFS1_SUPER_MAGIC 0x00011954
+#define UFS2_SUPER_MAGIC 0x19540119
+
+#endif /* UFS_H_ */
diff --git a/extlinux/ufs_fs.h b/extlinux/ufs_fs.h
new file mode 100644
index 0000000..ff8a4e7
--- /dev/null
+++ b/extlinux/ufs_fs.h
@@ -0,0 +1,307 @@
+/*
+ * Taken from Linux kernel tree (linux/fs/ufs)
+ * linux/include/linux/ufs_fs.h
+ *
+ * Copyright (C) 1996
+ * Adrian Rodriguez (adrian at franklins-tower.rutgers.edu)
+ * Laboratory for Computer Science Research Computing Facility
+ * Rutgers, The State University of New Jersey
+ *
+ * Copyright (c) 2013 Raphael S. Carvalho <raphael.scarv at gmail.com>
+ *
+ * Clean swab support by Fare <fare at tunes.org>
+ * just hope no one is using NNUUXXI on __?64 structure elements
+ * 64-bit clean thanks to Maciej W. Rozycki <macro at ds2.pg.gda.pl>
+ *
+ * 4.4BSD (FreeBSD) support added on February 1st 1998 by
+ * Niels Kristian Bech Jensen <nkbj at image.dk> partially based
+ * on code by Martin von Loewis <martin at mira.isdn.cs.tu-berlin.de>.
+ *
+ * NeXTstep support added on February 5th 1998 by
+ * Niels Kristian Bech Jensen <nkbj at image.dk>.
+ *
+ * Write support by Daniel Pirkl <daniel.pirkl at email.cz>
+ *
+ * HP/UX hfs filesystem support added by
+ * Martin K. Petersen <mkp at mkp.net>, August 1999
+ *
+ * UFS2 (of FreeBSD 5.x) support added by
+ * Niraj Kumar <niraj17 at iitbombay.org> , Jan 2004
+ *
+ */
+
+#ifndef __LINUX_UFS_FS_H
+#define __LINUX_UFS_FS_H
+
+#include <inttypes.h>
+
+typedef uint64_t __fs64;
+typedef uint32_t __fs32;
+typedef uint16_t __fs16;
+
+#define UFS_BBLOCK 0
+#define UFS_BBSIZE 8192
+#define UFS_SBLOCK 8192
+#define UFS_SBSIZE 8192
+
+#define UFS_SECTOR_SIZE 512
+#define UFS_SECTOR_BITS 9
+#define UFS_MAGIC 0x00011954
+#define UFS_MAGIC_BW 0x0f242697
+#define UFS2_MAGIC 0x19540119
+#define UFS_CIGAM 0x54190100 /* byteswapped MAGIC */
+
+/* Copied from FreeBSD */
+/*
+ * Each disk drive contains some number of filesystems.
+ * A filesystem consists of a number of cylinder groups.
+ * Each cylinder group has inodes and data.
+ *
+ * A filesystem is described by its super-block, which in turn
+ * describes the cylinder groups. The super-block is critical
+ * data and is replicated in each cylinder group to protect against
+ * catastrophic loss. This is done at `newfs' time and the critical
+ * super-block data does not change, so the copies need not be
+ * referenced further unless disaster strikes.
+ *
+ * For filesystem fs, the offsets of the various blocks of interest
+ * are given in the super block as:
+ * [fs->fs_sblkno] Super-block
+ * [fs->fs_cblkno] Cylinder group block
+ * [fs->fs_iblkno] Inode blocks
+ * [fs->fs_dblkno] Data blocks
+ * The beginning of cylinder group cg in fs, is given by
+ * the ``cgbase(fs, cg)'' macro.
+ *
+ * Depending on the architecture and the media, the superblock may
+ * reside in any one of four places. For tiny media where every block
+ * counts, it is placed at the very front of the partition. Historically,
+ * UFS1 placed it 8K from the front to leave room for the disk label and
+ * a small bootstrap. For UFS2 it got moved to 64K from the front to leave
+ * room for the disk label and a bigger bootstrap, and for really piggy
+ * systems we check at 256K from the front if the first three fail. In
+ * all cases the size of the superblock will be SBLOCKSIZE. All values are
+ * given in byte-offset form, so they do not imply a sector size. The
+ * SBLOCKSEARCH specifies the order in which the locations should be searched.
+ */
+#define SBLOCK_FLOPPY 0
+#define SBLOCK_UFS1 8192
+#define SBLOCK_UFS2 65536
+#define SBLOCK_PIGGY 262144
+#define SBLOCKSIZE 8192
+#define SBLOCKSEARCH \
+ { SBLOCK_UFS2, SBLOCK_UFS1, SBLOCK_FLOPPY, SBLOCK_PIGGY, -1 }
+
+#define UFS_MAXNAMLEN 255
+#define UFS_MAXMNTLEN 512
+#define UFS2_MAXMNTLEN 468
+#define UFS2_MAXVOLLEN 32
+#define UFS_MAXCSBUFS 31
+#define UFS_LINK_MAX 32000
+/*
+#define UFS2_NOCSPTRS ((128 / sizeof(void *)) - 4)
+*/
+#define UFS2_NOCSPTRS 28
+
+/*
+ * UFS_DIR_PAD defines the directory entries boundaries
+ * (must be a multiple of 4)
+ */
+#define UFS_DIR_PAD 4
+#define UFS_DIR_ROUND (UFS_DIR_PAD - 1)
+#define UFS_DIR_REC_LEN(name_len) (((name_len) + 1 + 8 + UFS_DIR_ROUND) &
~UFS_DIR_ROUND)
+
+struct ufs_timeval {
+ __fs32 tv_sec;
+ __fs32 tv_usec;
+};
+
+struct ufs_dir_entry {
+ __fs32 d_ino; /* inode number of this entry */
+ __fs16 d_reclen; /* length of this entry */
+ union {
+ __fs16 d_namlen; /* actual length of d_name */
+ struct {
+ __u8 d_type; /* file type */
+ __u8 d_namlen; /* length of string in d_name */
+ } d_44;
+ } d_u;
+ __u8 d_name[UFS_MAXNAMLEN + 1]; /* file name */
+};
+
+struct ufs_csum {
+ __fs32 cs_ndir; /* number of directories */
+ __fs32 cs_nbfree; /* number of free blocks */
+ __fs32 cs_nifree; /* number of free inodes */
+ __fs32 cs_nffree; /* number of free frags */
+};
+struct ufs2_csum_total {
+ __fs64 cs_ndir; /* number of directories */
+ __fs64 cs_nbfree; /* number of free blocks */
+ __fs64 cs_nifree; /* number of free inodes */
+ __fs64 cs_nffree; /* number of free frags */
+ __fs64 cs_numclusters; /* number of free clusters */
+ __fs64 cs_spare[3]; /* future expansion */
+};
+
+struct ufs_csum_core {
+ __u64 cs_ndir; /* number of directories */
+ __u64 cs_nbfree; /* number of free blocks */
+ __u64 cs_nifree; /* number of free inodes */
+ __u64 cs_nffree; /* number of free frags */
+ __u64 cs_numclusters; /* number of free clusters */
+};
+
+struct ufs_super_block {
+ union {
+ struct {
+ __fs32 fs_link; /* UNUSED */
+ } fs_42;
+ struct {
+ __fs32 fs_state; /* file system state flag */
+ } fs_sun;
+ } fs_u0;
+ __fs32 fs_rlink; /* UNUSED */
+ __fs32 fs_sblkno; /* addr of super-block in filesys */
+ __fs32 fs_cblkno; /* offset of cyl-block in filesys */
+ __fs32 fs_iblkno; /* offset of inode-blocks in filesys */
+ __fs32 fs_dblkno; /* offset of first data after cg */
+ __fs32 fs_cgoffset; /* cylinder group offset in cylinder */
+ __fs32 fs_cgmask; /* used to calc mod fs_ntrak */
+ __fs32 fs_time; /* last time written -- time_t */
+ __fs32 fs_size; /* number of blocks in fs */
+ __fs32 fs_dsize; /* number of data blocks in fs */
+ __fs32 fs_ncg; /* number of cylinder groups */
+ __fs32 fs_bsize; /* size of basic blocks in fs */
+ __fs32 fs_fsize; /* size of frag blocks in fs */
+ __fs32 fs_frag; /* number of frags in a block in fs */
+/* these are configuration parameters */
+ __fs32 fs_minfree; /* minimum percentage of free blocks */
+ __fs32 fs_rotdelay; /* num of ms for optimal next block */
+ __fs32 fs_rps; /* disk revolutions per second */
+/* these fields can be computed from the others */
+ __fs32 fs_bmask; /* ``blkoff'' calc of blk offsets */
+ __fs32 fs_fmask; /* ``fragoff'' calc of frag offsets */
+ __fs32 fs_bshift; /* ``lblkno'' calc of logical blkno */
+ __fs32 fs_fshift; /* ``numfrags'' calc number of frags */
+/* these are configuration parameters */
+ __fs32 fs_maxcontig; /* max number of contiguous blks */
+ __fs32 fs_maxbpg; /* max number of blks per cyl group */
+/* these fields can be computed from the others */
+ __fs32 fs_fragshift; /* block to frag shift */
+ __fs32 fs_fsbtodb; /* fsbtodb and dbtofsb shift constant */
+ __fs32 fs_sbsize; /* actual size of super block */
+ __fs32 fs_csmask; /* csum block offset */
+ __fs32 fs_csshift; /* csum block number */
+ __fs32 fs_nindir; /* value of NINDIR */
+ __fs32 fs_inopb; /* value of INOPB */
+ __fs32 fs_nspf; /* value of NSPF */
+/* yet another configuration parameter */
+ __fs32 fs_optim; /* optimization preference, see below */
+/* these fields are derived from the hardware */
+ union {
+ struct {
+ __fs32 fs_npsect; /* # sectors/track including spares */
+ } fs_sun;
+ struct {
+ __fs32 fs_state; /* file system state time stamp */
+ } fs_sunx86;
+ } fs_u1;
+ __fs32 fs_interleave; /* hardware sector interleave */
+ __fs32 fs_trackskew; /* sector 0 skew, per track */
+/* a unique id for this filesystem (currently unused and unmaintained) */
+/* In 4.3 Tahoe this space is used by fs_headswitch and fs_trkseek */
+/* Neither of those fields is used in the Tahoe code right now but */
+/* there could be problems if they are. */
+ __fs32 fs_id[2]; /* file system id */
+/* sizes determined by number of cylinder groups and their sizes */
+ __fs32 fs_csaddr; /* blk addr of cyl grp summary area */
+ __fs32 fs_cssize; /* size of cyl grp summary area */
+ __fs32 fs_cgsize; /* cylinder group size */
+/* these fields are derived from the hardware */
+ __fs32 fs_ntrak; /* tracks per cylinder */
+ __fs32 fs_nsect; /* sectors per track */
+ __fs32 fs_spc; /* sectors per cylinder */
+/* this comes from the disk driver partitioning */
+ __fs32 fs_ncyl; /* cylinders in file system */
+/* these fields can be computed from the others */
+ __fs32 fs_cpg; /* cylinders per group */
+ __fs32 fs_ipg; /* inodes per cylinder group */
+ __fs32 fs_fpg; /* blocks per group * fs_frag */
+/* this data must be re-computed after crashes */
+ struct ufs_csum fs_cstotal; /* cylinder summary information */
+/* these fields are cleared at mount time */
+ __s8 fs_fmod; /* super block modified flag */
+ __s8 fs_clean; /* file system is clean flag */
+ __s8 fs_ronly; /* mounted read-only flag */
+ __s8 fs_flags;
+ union {
+ struct {
+ __s8 fs_fsmnt[UFS_MAXMNTLEN];/* name mounted on */
+ __fs32 fs_cgrotor; /* last cg searched */
+ __fs32 fs_csp[UFS_MAXCSBUFS];/*list of fs_cs info buffers */
+ __fs32 fs_maxcluster;
+ __fs32 fs_cpc; /* cyl per cycle in postbl */
+ __fs16 fs_opostbl[16][8]; /* old rotation block list head */
+ } fs_u1;
+ struct {
+ __s8 fs_fsmnt[UFS2_MAXMNTLEN]; /* name mounted on */
+ __u8 fs_volname[UFS2_MAXVOLLEN]; /* volume name */
+ __fs64 fs_swuid; /* system-wide uid */
+ __fs32 fs_pad; /* due to alignment of fs_swuid */
+ __fs32 fs_cgrotor; /* last cg searched */
+ __fs32 fs_ocsp[UFS2_NOCSPTRS]; /*list of fs_cs info buffers */
+ __fs32 fs_contigdirs;/*# of contiguously allocated dirs */
+ __fs32 fs_csp; /* cg summary info buffer for fs_cs */
+ __fs32 fs_maxcluster;
+ __fs32 fs_active;/* used by snapshots to track fs */
+ __fs32 fs_old_cpc; /* cyl per cycle in postbl */
+ __fs32 fs_maxbsize;/*maximum blocking factor permitted */
+ __fs64 fs_sparecon64[17];/*old rotation block list head */
+ __fs64 fs_sblockloc; /* byte offset of standard superblock */
+ struct ufs2_csum_total fs_cstotal;/*cylinder summary information*/
+ struct ufs_timeval fs_time; /* last time written */
+ __fs64 fs_size; /* number of blocks in fs */
+ __fs64 fs_dsize; /* number of data blocks in fs */
+ __fs64 fs_csaddr; /* blk addr of cyl grp summary area */
+ __fs64 fs_pendingblocks;/* blocks in process of being freed */
+ __fs32 fs_pendinginodes;/*inodes in process of being freed */
+ } fs_u2;
+ } fs_u11;
+ union {
+ struct {
+ __fs32 fs_sparecon[53];/* reserved for future constants */
+ __fs32 fs_reclaim;
+ __fs32 fs_sparecon2[1];
+ __fs32 fs_state; /* file system state time stamp */
+ __fs32 fs_qbmask[2]; /* ~usb_bmask */
+ __fs32 fs_qfmask[2]; /* ~usb_fmask */
+ } fs_sun;
+ struct {
+ __fs32 fs_sparecon[53];/* reserved for future constants */
+ __fs32 fs_reclaim;
+ __fs32 fs_sparecon2[1];
+ __fs32 fs_npsect; /* # sectors/track including spares */
+ __fs32 fs_qbmask[2]; /* ~usb_bmask */
+ __fs32 fs_qfmask[2]; /* ~usb_fmask */
+ } fs_sunx86;
+ struct {
+ __fs32 fs_sparecon[50];/* reserved for future constants */
+ __fs32 fs_contigsumsize;/* size of cluster summary array */
+ __fs32 fs_maxsymlinklen;/* max length of an internal symlink */
+ __fs32 fs_inodefmt; /* format of on-disk inodes */
+ __fs32 fs_maxfilesize[2]; /* max representable file size */
+ __fs32 fs_qbmask[2]; /* ~usb_bmask */
+ __fs32 fs_qfmask[2]; /* ~usb_fmask */
+ __fs32 fs_state; /* file system state time stamp */
+ } fs_44;
+ } fs_u2;
+ __fs32 fs_postblformat; /* format of positional layout tables */
+ __fs32 fs_nrpos; /* number of rotational positions */
+ __fs32 fs_postbloff; /* (__s16) rotation block list head */
+ __fs32 fs_rotbloff; /* (__u8) blocks for each rotation */
+ __fs32 fs_magic; /* magic number */
+ __u8 fs_space[1]; /* list of blocks for each rotation */
+}; /*struct ufs_super_block*/
+
+#endif
diff --git a/libinstaller/syslxfs.h b/libinstaller/syslxfs.h
index 4d8f3b2..7e46e23 100644
--- a/libinstaller/syslxfs.h
+++ b/libinstaller/syslxfs.h
@@ -12,7 +12,7 @@
#ifndef _SYSLXFS_H_
#define _SYSLXFS_H_
-/* Global fs_type for handling fat, ntfs, ext2/3/4, btrfs and xfs */
+/* Global fs_type for handling fat, ntfs, ext2/3/4, btrfs, xfs and ufs1/2 */
enum filesystem {
NONE,
EXT2,
@@ -20,6 +20,8 @@ enum filesystem {
VFAT,
NTFS,
XFS,
+ UFS1,
+ UFS2,
};
extern int fs_type;
--
1.7.2.5
Raphael S.Carvalho
2014-May-29 23:35 UTC
[syslinux] [PATCH v2 2/2] core/fs: Add support to Unix File system 1/2.
From: Raphael S. Carvalho <raphael.scarv at gmail.com>
It's already loading modules successfully, booting Linux, and both
UFS version 1 and 2 seem to be working correctly.
Signed-off-by: Raphael S. Carvalho <raphael.scarv at gmail.com>
---
core/fs/ufs/bmap.c | 202 ++++++++++++++++++++++
core/fs/ufs/ufs.c | 486 ++++++++++++++++++++++++++++++++++++++++++++++++++++
core/fs/ufs/ufs.h | 254 +++++++++++++++++++++++++++
core/ldlinux.asm | 2 +
4 files changed, 944 insertions(+), 0 deletions(-)
create mode 100644 core/fs/ufs/bmap.c
create mode 100644 core/fs/ufs/ufs.c
create mode 100644 core/fs/ufs/ufs.h
diff --git a/core/fs/ufs/bmap.c b/core/fs/ufs/bmap.c
new file mode 100644
index 0000000..7d8490f
--- /dev/null
+++ b/core/fs/ufs/bmap.c
@@ -0,0 +1,202 @@
+/*
+ * Copyright (C) 2013 Raphael S. Carvalho <raphael.scarv at gmail.com>
+ *
+ * Partially taken from fs/ext2/bmap.c
+ * This file was modified according UFS1/2 needs.
+ *
+ * Copyright (C) 2009 Liu Aleaxander -- All rights reserved. This file
+ * may be redistributed under the terms of the GNU Public License.
+ */
+
+#include <stdio.h>
+#include <dprintf.h>
+#include <fs.h>
+#include <disk.h>
+#include <cache.h>
+#include "ufs.h"
+
+/*
+ * Copy blk address into buffer, this was needed since UFS1/2 addr size
+ * in blk maps differs from each other (32/64 bits respectivelly).
+ */
+static inline uint64_t
+get_blkaddr (const uint8_t *blk, uint32_t index, uint32_t shift)
+{
+ uint64_t addr = 0;
+
+ memcpy((uint8_t *) &addr,
+ (uint8_t *) blk + (index << shift),
+ 1 << shift);
+
+ return addr;
+}
+
+/*
+ * Scan forward in a range of blocks to see if they are contiguous,
+ * then return the initial value.
+ */
+static uint64_t
+scan_set_nblocks(const uint8_t *map, uint32_t index, uint32_t addr_shift,
+ unsigned int count, size_t *nblocks)
+{
+ uint64_t addr;
+ uint64_t blk = get_blkaddr(map, index, addr_shift);
+
+ /*
+ * Block spans 8 fragments, then address is interleaved by 8.
+ * This code works for either 32/64 sized addresses.
+ */
+ if (nblocks) {
+ uint32_t skip = blk ? FRAGMENTS_PER_BLK : 0;
+ uint32_t next = blk + skip;
+ size_t cnt = 1;
+
+ /* Get address of starting blk pointer */
+ map += (index << addr_shift);
+
+ ufs_debug("[scan] start blk: %u\n", blk);
+ ufs_debug("[scan] count (nr of blks): %u\n", count);
+ /* Go up to the end of blk map */
+ while (--count) {
+ map += 1 << addr_shift;
+ addr = get_blkaddr(map, 0, addr_shift);
+#if 0
+ /* Extra debugging info (Too much prints) */
+ ufs_debug("[scan] addr: %u next: %u\n", addr, next);
+#endif
+ if (addr == next) {
+ cnt++;
+ next += skip;
+ } else {
+ break;
+ }
+ }
+ *nblocks = cnt;
+ ufs_debug("[scan] nblocks: %u\n", cnt);
+ ufs_debug("[scan] end blk: %u\n", next - FRAGMENTS_PER_BLK);
+ }
+
+ return blk;
+}
+
+/*
+ * The actual indirect block map handling - the block passed in should
+ * be relative to the beginning of the particular block hierarchy.
+ *
+ * @shft_per_blk: shift to get nr. of addresses in a block.
+ * @mask_per_blk: mask to limit the max nr. of addresses in a block.
+ * @addr_count: nr. of addresses in a block.
+ */
+static uint64_t
+bmap_indirect(struct fs_info *fs, uint64_t start, uint32_t block,
+ int levels, size_t *nblocks)
+{
+ uint32_t shft_per_blk = fs->block_shift - UFS_SB(fs)->addr_shift;
+ uint32_t addr_count = (1 << shft_per_blk);
+ uint32_t mask_per_blk = addr_count - 1;
+ const uint8_t *blk = NULL;
+ uint32_t index = 0;
+
+ while (levels--) {
+ if (!start) {
+ if (nblocks)
+ *nblocks = addr_count << (levels * shft_per_blk);
+ return 0;
+ }
+
+ blk = get_cache(fs->fs_dev, frag_to_blk(fs, start));
+ index = (block >> (levels * shft_per_blk)) & mask_per_blk;
+ start = get_blkaddr(blk, index, UFS_SB(fs)->addr_shift);
+ }
+
+ return scan_set_nblocks(blk, index, UFS_SB(fs)->addr_shift,
+ addr_count - index, nblocks);
+}
+
+/*
+ * Handle the traditional block map, like indirect, double indirect
+ * and triple indirect
+ */
+uint64_t ufs_bmap (struct inode *inode, block_t block, size_t *nblocks)
+{
+ uint32_t shft_per_blk, ptrs_per_blk;
+ static uint32_t indir_blks, double_blks, triple_blks;
+ struct fs_info *fs = inode->fs;
+
+ /* Initialize static values */
+ if (!indir_blks) {
+ shft_per_blk = fs->block_shift - UFS_SB(fs)->addr_shift;
+ ptrs_per_blk = fs->block_size >> UFS_SB(fs)->addr_shift;
+
+ indir_blks = ptrs_per_blk;
+ double_blks = ptrs_per_blk << shft_per_blk;
+ triple_blks = double_blks << shft_per_blk;
+ }
+
+ /*
+ * direct blocks
+ * (UFS2_ADDR_SHIFT) is also used for UFS1 because its direct ptr array
+ * was extended to 64 bits.
+ */
+ if (block < UFS_DIRECT_BLOCKS)
+ return scan_set_nblocks((uint8_t *) PVT(inode)->direct_blk_ptr,
+ block, UFS2_ADDR_SHIFT,
+ UFS_DIRECT_BLOCKS - block, nblocks);
+
+ /* indirect blocks */
+ block -= UFS_DIRECT_BLOCKS;
+ if (block < indir_blks)
+ return bmap_indirect(fs, PVT(inode)->indirect_blk_ptr,
+ block, 1, nblocks);
+
+ /* double indirect blocks */
+ block -= indir_blks;
+ if (block < double_blks)
+ return bmap_indirect(fs, PVT(inode)->double_indirect_blk_ptr,
+ block, 2, nblocks);
+
+ /* triple indirect blocks */
+ block -= double_blks;
+ if (block < triple_blks)
+ return bmap_indirect(fs, PVT(inode)->triple_indirect_blk_ptr,
+ block, 3, nblocks);
+
+ /* This can't happen... */
+ return 0;
+}
+
+/*
+ * Next extent for getfssec
+ * "Remaining sectors" means (lstart & blkmask).
+ */
+int ufs_next_extent(struct inode *inode, uint32_t lstart)
+{
+ struct fs_info *fs = inode->fs;
+ int blktosec = BLOCK_SHIFT(fs) - SECTOR_SHIFT(fs);
+ int frag_shift = BLOCK_SHIFT(fs) - UFS_SB(fs)->c_blk_frag_shift;
+ int blkmask = (1 << blktosec) - 1;
+ block_t block;
+ size_t nblocks = 0;
+
+ ufs_debug("ufs_next_extent:\n");
+ block = ufs_bmap(inode, lstart >> blktosec, &nblocks);
+ ufs_debug("blk: %u\n", block);
+
+ if (!block) // Sparse block
+ inode->next_extent.pstart = EXTENT_ZERO;
+ else
+ /*
+ * Convert blk into sect addr and add the remaining
+ * sectors into pstart (sector start address).
+ */
+ inode->next_extent.pstart + ((sector_t) (block << (frag_shift -
SECTOR_SHIFT(fs)))) |
+ (lstart & blkmask);
+
+ /*
+ * Subtract the remaining sectors from len since these sectors
+ * were added to pstart (sector start address).
+ */
+ inode->next_extent.len = (nblocks << blktosec) - (lstart &
blkmask);
+ return 0;
+}
\ No newline at end of file
diff --git a/core/fs/ufs/ufs.c b/core/fs/ufs/ufs.c
new file mode 100644
index 0000000..a7ef28f
--- /dev/null
+++ b/core/fs/ufs/ufs.c
@@ -0,0 +1,486 @@
+/*
+ * Copyright (C) 2013 Raphael S. Carvalho <raphael.scarv at gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the
+ * Free Software Foundation, Inc.,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+*/
+
+#include <dprintf.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/dirent.h>
+#include <cache.h>
+#include <disk.h>
+#include <fs.h>
+#include <minmax.h>
+#include "core.h"
+#include "ufs.h"
+
+/*
+ * Read the super block and check magic fields based on
+ * passed paramaters.
+ */
+static bool
+do_checksb(struct ufs_super_block *sb, struct disk *disk,
+ const uint32_t sblock_off, const uint32_t ufs_smagic)
+{
+ uint32_t lba;
+ static uint32_t count;
+
+ /* How many sectors are needed to fill sb struct */
+ if (!count)
+ count = sizeof *sb >> disk->sector_shift;
+ /* Get lba address based on sector size of disk */
+ lba = sblock_off >> (disk->sector_shift);
+ /* Read super block */
+ disk->rdwr_sectors(disk, sb, lba, count, 0);
+
+ if (sb->magic == ufs_smagic)
+ return true;
+
+ return false;
+}
+
+/*
+ * Go through all possible ufs superblock offsets.
+ * TODO: Add UFS support to removable media (sb offset: 0).
+ */
+static int
+ufs_checksb(struct ufs_super_block *sb, struct disk *disk)
+{
+ /* Check for UFS1 sb */
+ if (do_checksb(sb, disk, UFS1_SBLOCK_OFFSET, UFS1_SUPER_MAGIC))
+ return UFS1;
+ /* Check for UFS2 sb */
+ if (do_checksb(sb, disk, UFS2_SBLOCK_OFFSET, UFS2_SUPER_MAGIC))
+ return UFS2;
+ /* UFS2 may also exist in 256k-, but this isn't the default */
+ if (do_checksb(sb, disk, UFS2_SBLOCK2_OFFSET, UFS2_SUPER_MAGIC))
+ return UFS2_PIGGY;
+
+ return NONE;
+}
+
+/*
+ * lblock stands for linear block address,
+ * whereas pblock is the actual blk ptr to get data from.
+ *
+ * UFS1/2 use frag addrs rather than blk ones, then
+ * the offset into the block must be calculated.
+ */
+static const void *
+ufs_get_cache(struct inode *inode, block_t lblock)
+{
+ const void *data;
+ struct fs_info *fs = inode->fs;
+ struct ufs_sb_info *sb = UFS_SB(inode->fs);
+ uint64_t frag_addr, frag_offset;
+ uint32_t frag_shift;
+ block_t pblock;
+
+ frag_addr = ufs_bmap(inode, lblock, NULL);
+ if (!frag_addr)
+ return NULL;
+
+ frag_shift = fs->block_shift - sb->c_blk_frag_shift;
+ /* Get fragment byte address */
+ frag_offset = frag_addr << frag_shift;
+ /* Convert frag addr to blk addr */
+ pblock = frag_to_blk(fs, frag_addr);
+ /* Read the blk */
+ data = get_cache(fs->fs_dev, pblock);
+
+ /* Return offset into block */
+ return data + (frag_offset & (fs->block_size - 1));
+}
+
+/*
+ * Based on fs/ext2/ext2.c
+ * find a dir entry, return it if found, or return NULL.
+ */
+static const struct ufs_dir_entry *
+ufs_find_entry(struct fs_info *fs, struct inode *inode, const char *dname)
+{
+ const struct ufs_dir_entry *dir;
+ const char *data;
+ int32_t i, offset, maxoffset;
+ block_t index = 0;
+
+ ufs_debug("ufs_find_entry: dname: %s ", dname);
+ for (i = 0; i < inode->size; i += fs->block_size) {
+ data = ufs_get_cache(inode, index++);
+ offset = 0;
+ maxoffset = min(inode->size-i, fs->block_size);
+
+ /* The smallest possible size is 9 bytes */
+ while (offset < maxoffset-8) {
+ dir = (const struct ufs_dir_entry *)(data + offset);
+ if (dir->dir_entry_len > maxoffset - offset)
+ break;
+
+ /*
+ * Name fields are variable-length and null terminated,
+ * then it's possible to use strcmp directly.
+ */
+ if (dir->inode_value && !strcmp(dname, (const char
*)dir->name)) {
+ ufs_debug("(found)\n");
+ return dir;
+ }
+ offset += dir->dir_entry_len;
+ }
+ }
+ ufs_debug("(not found)\n");
+ return NULL;
+}
+
+/*
+ * Get either UFS1/2 inode structures.
+ */
+static const void *
+ufs_get_inode(struct fs_info *fs, int inr)
+{
+ const char *data;
+ uint32_t group, inode_offset, inode_table;
+ uint32_t block_num, block_off;
+
+ /* Get cylinder group nr. */
+ group = inr / UFS_SB(fs)->inodes_per_cg;
+ /*
+ * Ensuring group will not exceed the range 0:groups_count-1.
+ * By the way, this should *never* happen.
+ * Unless the (on-disk) fs structure is corrupted!
+ */
+ if (group >= UFS_SB(fs)->groups_count) {
+ printf("ufs_get_inode: "
+ "group(%d) exceeded the avail. range (0:%d)\n",
+ group, UFS_SB(fs)->groups_count - 1);
+ return NULL;
+ }
+
+ /* Offset into inode table of the cylinder group */
+ inode_offset = inr % UFS_SB(fs)->inodes_per_cg;
+ /* Get inode table blk addr respective to cylinder group */
+ inode_table = (group * UFS_SB(fs)->blocks_per_cg) +
+ UFS_SB(fs)->off_inode_tbl;
+ /* Calculating staggering offset (UFS1 only!) */
+ if (UFS_SB(fs)->fs_type == UFS1)
+ inode_table += UFS_SB(fs)->ufs1.delta_value *
+ (group & UFS_SB(fs)->ufs1.cycle_mask);
+
+ /* Get blk nr and offset into the blk */
+ block_num = inode_table + inode_offset / UFS_SB(fs)->inodes_per_block;
+ block_off = inode_offset % UFS_SB(fs)->inodes_per_block;
+
+ /*
+ * Read the blk from the blk addr previously computed;
+ * Calc the inode struct offset into the read block.
+ */
+ data = get_cache(fs->fs_dev, block_num);
+ return data + block_off * UFS_SB(fs)->inode_size;
+}
+
+static struct inode *
+ufs1_iget_by_inr(struct fs_info *fs, uint32_t inr)
+{
+ const struct ufs1_inode *ufs_inode;
+ struct inode *inode;
+ uint64_t *dest;
+ uint32_t *source;
+ int i;
+
+ ufs_inode = (struct ufs1_inode *) ufs_get_inode(fs, inr);
+ if (!ufs_inode)
+ return NULL;
+
+ if (!(inode = alloc_inode(fs, inr, sizeof(struct ufs_inode_pvt))))
+ return NULL;
+
+ /* UFS1 doesn't support neither creation nor deletion times */
+ inode->refcnt = ufs_inode->link_count;
+ inode->mode = IFTODT(ufs_inode->file_mode);
+ inode->size = ufs_inode->size;
+ inode->atime = ufs_inode->a_time;
+ inode->mtime = ufs_inode->m_time;
+ inode->blocks = ufs_inode->blocks_held;
+ inode->flags = ufs_inode->flags;
+
+ /*
+ * Copy and extend blk pointers to 64 bits, so avoid
+ * having two structures for inode private.
+ */
+ dest = (uint64_t *) inode->pvt;
+ source = (uint32_t *) ufs_inode->direct_blk_ptr;
+ for (i = 0; i < UFS_NBLOCKS; i++)
+ dest[i] = ((uint64_t) source[i]) & 0xFFFFFFFF;
+
+ return inode;
+}
+
+static struct inode *
+ufs2_iget_by_inr(struct fs_info *fs, uint32_t inr)
+{
+ const struct ufs2_inode *ufs_inode;
+ struct inode *inode;
+
+ ufs_inode = (struct ufs2_inode *) ufs_get_inode(fs, inr);
+ if (!ufs_inode)
+ return NULL;
+
+ if (!(inode = alloc_inode(fs, inr, sizeof(struct ufs_inode_pvt))))
+ return NULL;
+
+ /* UFS2 doesn't support deletion time */
+ inode->refcnt = ufs_inode->link_count;
+ inode->mode = IFTODT(ufs_inode->file_mode);
+ inode->size = ufs_inode->size;
+ inode->atime = ufs_inode->a_time;
+ inode->ctime = ufs_inode->creat_time;
+ inode->mtime = ufs_inode->m_time;
+ inode->blocks = ufs_inode->bytes_held >> fs->block_shift;
+ inode->flags = ufs_inode->flags;
+ memcpy(inode->pvt, ufs_inode->direct_blk_ptr,
+ sizeof(uint64_t) * UFS_NBLOCKS);
+
+ return inode;
+}
+
+/*
+ * Both ufs_iget_root and ufs_iget callback based on ufs type.
+ */
+static struct inode *
+ufs_iget_root(struct fs_info *fs)
+{
+ return UFS_SB(fs)->ufs_iget_by_inr(fs, UFS_ROOT_INODE);
+}
+
+static struct inode *
+ufs_iget(const char *dname, struct inode *parent)
+{
+ const struct ufs_dir_entry *dir;
+ struct fs_info *fs = parent->fs;
+
+ dir = ufs_find_entry(fs, parent, dname);
+ if (!dir)
+ return NULL;
+
+ return UFS_SB(fs)->ufs_iget_by_inr(fs, dir->inode_value);
+}
+
+static void ufs1_read_blkaddrs(struct inode *inode, char *buf)
+{
+ uint32_t dest[UFS_NBLOCKS];
+ const uint64_t *source = (uint64_t *) (inode->pvt);
+ int i;
+
+ /* Convert ufs_inode_pvt uint64_t fields into uint32_t
+ * Upper-half part of ufs1 private blk addrs are always supposed to be
+ * zero (it's previosuly extended by us), thus data isn't being
lost. */
+ for (i = 0; i < UFS_NBLOCKS; i++) {
+ if ((source[i] >> 32) != 0) {
+ /* This should never happen, but will not prevent anything
+ * from working. */
+ ufs_debug("ufs1: inode->pvt[%d]: warning!\n", i);
+ }
+
+ dest[i] = (uint32_t)(source[i] & 0xFFFFFFFF);
+ }
+ memcpy(buf, (const char *) dest, inode->size);
+}
+
+static void ufs2_read_blkaddrs(struct inode *inode, char *buf)
+{
+ memcpy(buf, (const char *) (inode->pvt), inode->size);
+}
+
+/*
+ * Taken from ext2/ext2.c.
+ * Read the entire contents of an inode into a memory buffer
+ */
+static int cache_get_file(struct inode *inode, void *buf, size_t bytes)
+{
+ struct fs_info *fs = inode->fs;
+ size_t block_size = BLOCK_SIZE(fs);
+ uint32_t index = 0; /* Logical block number */
+ size_t chunk;
+ const char *data;
+ char *p = buf;
+
+ if (inode->size > bytes)
+ bytes = inode->size;
+
+ while (bytes) {
+ chunk = min(bytes, block_size);
+ data = ufs_get_cache(inode, index++);
+ memcpy(p, data, chunk);
+
+ bytes -= chunk;
+ p += chunk;
+ }
+
+ return 0;
+}
+
+static int ufs_readlink(struct inode *inode, char *buf)
+{
+ struct fs_info *fs = inode->fs;
+ uint32_t i_symlink_limit;
+
+ if (inode->size > BLOCK_SIZE(fs))
+ return -1; /* Error! */
+
+ // TODO: use UFS_SB(fs)->maxlen_isymlink instead.
+ i_symlink_limit = ((UFS_SB(fs)->fs_type == UFS1) ?
+ sizeof(uint32_t) : sizeof(uint64_t)) * UFS_NBLOCKS;
+ ufs_debug("UFS_SB(fs)->maxlen_isymlink=%d",
UFS_SB(fs)->maxlen_isymlink);
+
+ if (inode->size <= i_symlink_limit)
+ UFS_SB(fs)->ufs_read_blkaddrs(inode, buf);
+ else
+ cache_get_file(inode, buf, inode->size);
+
+ return inode->size;
+}
+
+static inline enum dir_type_flags get_inode_mode(uint8_t type)
+{
+ switch(type) {
+ case UFS_DTYPE_FIFO: return DT_FIFO;
+ case UFS_DTYPE_CHARDEV: return DT_CHR;
+ case UFS_DTYPE_DIR: return DT_DIR;
+ case UFS_DTYPE_BLOCK: return DT_BLK;
+ case UFS_DTYPE_RFILE: return DT_REG;
+ case UFS_DTYPE_SYMLINK: return DT_LNK;
+ case UFS_DTYPE_SOCKET: return DT_SOCK;
+ case UFS_DTYPE_WHITEOUT: return DT_WHT;
+ default: return DT_UNKNOWN;
+ }
+}
+
+/*
+ * Read one directory entry at a time
+ */
+static int ufs_readdir(struct file *file, struct dirent *dirent)
+{
+ struct fs_info *fs = file->fs;
+ struct inode *inode = file->inode;
+ const struct ufs_dir_entry *dir;
+ const char *data;
+ block_t index = file->offset >> fs->block_shift;
+
+ if (file->offset >= inode->size)
+ return -1; /* End of file */
+
+ data = ufs_get_cache(inode, index);
+ dir = (const struct ufs_dir_entry *)
+ (data + (file->offset & (BLOCK_SIZE(fs) - 1)));
+
+ dirent->d_ino = dir->inode_value;
+ dirent->d_off = file->offset;
+ dirent->d_reclen = offsetof(struct dirent, d_name) + dir->name_length
+ 1;
+ dirent->d_type = get_inode_mode(dir->file_type & 0x0F);
+ memcpy(dirent->d_name, dir->name, dir->name_length);
+ dirent->d_name[dir->name_length] = '\0';
+
+ file->offset += dir->dir_entry_len; /* Update for next reading */
+
+ return 0;
+}
+
+static inline struct ufs_sb_info *
+set_ufs_info(struct ufs_super_block *sb, int ufs_type)
+{
+ struct ufs_sb_info *sbi;
+
+ sbi = malloc(sizeof *sbi);
+ if (!sbi)
+ malloc_error("ufs_sb_info structure");
+
+ /* Setting up UFS-dependent info */
+ if (ufs_type == UFS1) {
+ sbi->inode_size = sizeof (struct ufs1_inode);
+ sbi->groups_count = sb->ufs1.nr_frags / sb->frags_per_cg;
+ sbi->ufs1.delta_value = sb->ufs1.delta_value;
+ sbi->ufs1.cycle_mask = sb->ufs1.cycle_mask;
+ sbi->ufs_iget_by_inr = ufs1_iget_by_inr;
+ sbi->ufs_read_blkaddrs = ufs1_read_blkaddrs;
+ sbi->addr_shift = UFS1_ADDR_SHIFT;
+ } else { // UFS2 or UFS2_PIGGY
+ sbi->inode_size = sizeof (struct ufs2_inode);
+ sbi->groups_count = sb->ufs2.nr_frags / sb->frags_per_cg;
+ sbi->ufs_iget_by_inr = ufs2_iget_by_inr;
+ sbi->ufs_read_blkaddrs = ufs2_read_blkaddrs;
+ sbi->addr_shift = UFS2_ADDR_SHIFT;
+ }
+ sbi->inodes_per_block = sb->block_size / sbi->inode_size;
+ sbi->inodes_per_cg = sb->inodes_per_cg;
+ sbi->blocks_per_cg = sb->frags_per_cg >>
sb->c_blk_frag_shift;
+ sbi->off_inode_tbl = sb->off_inode_tbl >>
sb->c_blk_frag_shift;
+ sbi->c_blk_frag_shift = sb->c_blk_frag_shift;
+ sbi->maxlen_isymlink = sb->maxlen_isymlink;
+ sbi->fs_type = ufs_type;
+
+ return sbi;
+}
+
+/*
+ * Init the fs metadata and return block size
+ */
+static int ufs_fs_init(struct fs_info *fs)
+{
+ struct disk *disk = fs->fs_dev->disk;
+ struct ufs_super_block sb;
+ struct cache *cs;
+
+ int ufs_type = ufs_checksb(&sb, disk);
+ if (ufs_type == NONE)
+ return -1;
+
+ ufs_debug("%s SB FOUND!\n", ufs_type == UFS1 ? "UFS1" :
"UFS2");
+ ufs_debug("Block size: %u\n", sb.block_size);
+
+ fs->fs_info = (struct ufs_sb_info *) set_ufs_info(&sb, ufs_type);
+ fs->sector_shift = disk->sector_shift;
+ fs->sector_size = disk->sector_size;
+ fs->block_shift = sb.block_shift;
+ fs->block_size = sb.block_size;
+
+ /* Initialize the cache, and force a clean on block zero */
+ cache_init(fs->fs_dev, sb.block_shift);
+ cs = _get_cache_block(fs->fs_dev, 0);
+ memset(cs->data, 0, fs->block_size);
+ cache_lock_block(cs);
+
+ /* For debug purposes */
+ //ufs_checking(fs);
+
+ //return -1;
+ return fs->block_shift;
+}
+
+const struct fs_ops ufs_fs_ops = {
+ .fs_name = "ufs",
+ .fs_flags = FS_USEMEM | FS_THISIND,
+ .fs_init = ufs_fs_init,
+ .searchdir = NULL,
+ .getfssec = generic_getfssec,
+ .close_file = generic_close_file,
+ .mangle_name = generic_mangle_name,
+ .open_config = generic_open_config,
+ .readlink = ufs_readlink,
+ .readdir = ufs_readdir,
+ .iget_root = ufs_iget_root,
+ .iget = ufs_iget,
+ .next_extent = ufs_next_extent,
+};
diff --git a/core/fs/ufs/ufs.h b/core/fs/ufs/ufs.h
new file mode 100644
index 0000000..04e9f84
--- /dev/null
+++ b/core/fs/ufs/ufs.h
@@ -0,0 +1,254 @@
+/*
+ * Copyright (C) 2013 Raphael S. Carvalho <raphael.scarv at gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the
+ * Free Software Foundation, Inc.,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+*/
+
+#ifndef _UFS_H_
+#define _UFS_H_
+
+#include <stdint.h>
+
+/* Sector addresses */
+#define UFS1_SBLOCK_OFFSET 8192
+#define UFS2_SBLOCK_OFFSET 65536
+#define UFS2_SBLOCK2_OFFSET 262144
+
+#define UFS1_ADDR_SHIFT 2
+#define UFS2_ADDR_SHIFT 3
+
+/* Super magic numbers */
+#define UFS1_SUPER_MAGIC (0x011954)
+#define UFS2_SUPER_MAGIC (0x19540119)
+
+#define UFS_ROOT_INODE 2
+
+#define UFS_DIRECT_BLOCKS 12
+#define UFS_INDIRECT_BLOCK 1
+#define UFS_DOUBLE_INDIRECT_BLOCK 1
+#define UFS_TRIPLE_INDIRECT_BLOCK 1
+/* Total number of block addr hold by inodes */
+#define UFS_NBLOCKS 15
+
+/* Blocks span 8 fragments */
+#define FRAGMENTS_PER_BLK 8
+
+/* UFS types */
+typedef enum {
+ NONE,
+ UFS1,
+ UFS2,
+ UFS2_PIGGY,
+} ufs_t;
+
+/*
+ * UFS1/UFS2 SUPERBLOCK structure
+ * CG stands for Cylinder Group.
+ *
+ * Variables prepended with off store offsets relative to
+ * base address of a Cylinder Group (CG).
+ */
+struct ufs_super_block { // supporting either ufs1 or ufs2
+ uint8_t unused[8];
+ /* Offset values */
+ uint32_t off_backup_sb; // Backup super block
+ uint32_t off_group_desc; // Group Descriptor
+ uint32_t off_inode_tbl; // Inode table
+ uint32_t off_data_block; // First data block
+ union {
+ struct { /* Used for UFS1 */
+ uint32_t delta_value; // For calc staggering offset
+ uint32_t cycle_mask; // Mask for staggering offset
+ uint32_t last_written; // Last written time
+ uint32_t nr_frags; // Number of frags in FS
+ uint32_t storable_frags_nr; // Nr of frags that can store data
+ } ufs1;
+ uint8_t unused1[20];
+ };
+ uint32_t nr_cyl_groups; // Number of cylinder groups.
+ uint32_t block_size; // Block size in bytes.
+ uint32_t fragment_size; // Fragment size in bytes.
+ uint8_t unused2[16];
+ uint32_t block_addr_mask; // to calculate the address
+ uint32_t frag_addr_mask;
+ uint32_t block_shift; // to calculate byte address
+ uint32_t frag_shift;
+ uint32_t nr_contiguous_blk; // max number of continuous blks to alloc
+ uint32_t nr_blks_per_cg; // max number of blks per cylinder group
+ uint32_t c_blk_frag_shift; // Bits to convert blk and frag address.
+ uint32_t c_frag_sect_shift; // Bits to convert frag and sect address.
+ uint32_t superblk_size; // Superblock size.
+ uint8_t unused3[76];
+ uint32_t inodes_per_cg; // Inodes per cylinder group
+ uint32_t frags_per_cg; // Fragments per cylinder group
+ union {
+ struct { /* Used for UFS2 */
+ uint8_t unused[888];
+ uint64_t nr_frags; // Number of fragments in FS
+ uint8_t unused1[232];
+ } ufs2;
+ uint8_t unused4[1128];
+ };
+ uint32_t maxlen_isymlink; // Max length of internal symlink
+ uint32_t inodes_format; // Format of inodes
+ uint8_t unused5[44];
+ uint32_t magic; // Magic value
+ uint8_t pad[160]; // padding up to sector (512 bytes) boundary
+} __attribute__((__packed__));
+
+/*
+ * Info about UFS1/2 super block.
+ */
+struct ufs_sb_info {
+ uint32_t blocks_per_cg; // Blocks per cylinder group
+ uint32_t inodes_per_cg; // Inodes per cylinder group
+ uint32_t inode_size;
+ uint32_t inodes_per_block; // Inodes per block
+ struct { /* UFS1 only! */
+ /* Values for calculating staggering offset */
+ uint32_t delta_value;
+ uint32_t cycle_mask;
+ } ufs1;
+ uint32_t off_inode_tbl; // Inode table offset.
+ uint32_t groups_count; // Number of groups in the fs
+ uint32_t addr_shift; // 2 ^ addr_shift = size in bytes of default addr.
+ uint32_t c_blk_frag_shift; // Convert blk/frag addr (vice-versa)
+ uint32_t maxlen_isymlink; // Max length of internal symlink
+ struct inode *(*ufs_iget_by_inr)(struct fs_info *, uint32_t);
+ void (*ufs_read_blkaddrs)(struct inode *, char *);
+ ufs_t fs_type; // { UFS1, UFS2, UFS2_PIGGY }
+};
+
+/*
+ * Get super block info struct
+ */
+static inline struct ufs_sb_info *UFS_SB(struct fs_info *fs)
+{
+ return fs->fs_info;
+}
+
+/*
+ * Convert frag addr to blk addr
+ */
+static inline block_t frag_to_blk(struct fs_info *fs, uint64_t frag)
+{
+ return frag >> UFS_SB(fs)->c_blk_frag_shift;
+}
+
+/*
+ * UFS1 inode structures
+ */
+struct ufs1_inode {
+ uint16_t file_mode;
+ uint16_t link_count;
+ uint8_t unused[4];
+ uint64_t size;
+ uint32_t a_time; // Access time
+ uint32_t a_time_nanosec;
+ uint32_t m_time; // Modified time
+ uint32_t m_time_nanosec;
+ uint32_t ch_time; // Change time
+ uint32_t ch_time_nanosec;
+ uint32_t direct_blk_ptr[12];
+ uint32_t indirect_blk_ptr;
+ uint32_t double_indirect_blk_ptr;
+ uint32_t triple_indirect_blk_ptr;
+ uint32_t flags; // Status flags
+ uint32_t blocks_held; // Blocks held
+ uint32_t generation_nrb; // (NFS)
+ uint32_t used_id;
+ uint32_t group_id;
+ uint8_t unused1[8];
+} __attribute__((__packed__));
+
+/*
+ * UFS2 inode structures
+ */
+struct ufs2_inode {
+ uint16_t file_mode;
+ uint16_t link_count;
+ uint32_t user_id;
+ uint32_t group_id;
+ uint32_t inode_blocksize;
+ uint64_t size;
+ uint64_t bytes_held;
+ uint64_t a_time; // Access time
+ uint64_t m_time; // Modified time
+ uint64_t ch_time; // Change time
+ uint64_t creat_time; // Creation time
+ uint32_t a_time_nanosec;
+ uint32_t m_time_nanosec;
+ uint32_t ch_time_nanosec;
+ uint32_t creat_time_nanosec;
+ uint32_t generation_nrb; // (NFS)
+ uint32_t kernel_flags;
+ uint32_t flags;
+ uint32_t ext_attr_size; // Extended attrib size.
+ uint64_t ext_direct_blk_ptrs[2]; // Ext. attrib blk pointers.
+ uint64_t direct_blk_ptr[12];
+ uint64_t indirect_blk_ptr;
+ uint64_t double_indirect_blk_ptr;
+ uint64_t triple_indirect_blk_ptr;
+ uint8_t unused[24];
+} __attribute__((__packed__));
+
+#define PVT(p) ((struct ufs_inode_pvt *) p->pvt)
+
+struct ufs_inode_pvt {
+ uint64_t direct_blk_ptr[12];
+ uint64_t indirect_blk_ptr;
+ uint64_t double_indirect_blk_ptr;
+ uint64_t triple_indirect_blk_ptr;
+};
+
+struct ufs_dir_entry {
+ uint32_t inode_value;
+ uint16_t dir_entry_len;
+ uint8_t file_type;
+ uint8_t name_length;
+ uint8_t name[1]; // Dir names are null terminated!!!
+} __attribute__((__packed__));
+
+enum inode_type_flags {
+ UFS_INO_FIFO = 0x1000,
+ UFS_INO_CHARDEV = 0x2000,
+ UFS_INO_DIRECTORY = 0x4000,
+ UFS_INO_BLOCKDEV = 0x6000,
+ UFS_INO_RFILE = 0x8000,
+ UFS_INO_SYMLINK = 0xA000,
+ UFS_INO_UNIXSOCKET = 0xC000,
+};
+
+enum dir_type_flags {
+ UFS_DTYPE_UNKNOWN = 0,
+ UFS_DTYPE_FIFO = 1,
+ UFS_DTYPE_CHARDEV = 2,
+ UFS_DTYPE_DIR = 4,
+ UFS_DTYPE_BLOCK = 6,
+ UFS_DTYPE_RFILE = 8,
+ UFS_DTYPE_SYMLINK = 10,
+ UFS_DTYPE_SOCKET = 12,
+ UFS_DTYPE_WHITEOUT = 14,
+};
+
+/* Functions from bmap.c */
+extern uint64_t ufs_bmap (struct inode *, block_t, size_t *);
+extern int ufs_next_extent(struct inode *, uint32_t);
+
+#define ufs_debug dprintf
+//extern void ufs_checking (struct fs_info *);
+
+#endif /* _UFS_H_ */
diff --git a/core/ldlinux.asm b/core/ldlinux.asm
index a1f96b7..050f503 100644
--- a/core/ldlinux.asm
+++ b/core/ldlinux.asm
@@ -43,6 +43,8 @@ ROOT_FS_OPS:
dd xfs_fs_ops
extern btrfs_fs_ops
dd btrfs_fs_ops
+ extern ufs_fs_ops
+ dd ufs_fs_ops
dd 0
%include "diskfs.inc"
--
1.7.2.5
Op 2014-05-29 om 20:35 schreef Raphael S.Carvalho:> From: Raphael S. Carvalho <raphael.scarv at gmail.com> > > Change since v1: > * Fix bug on dentry structure (thank you specification; btw, sarcasm), and > consequently a bug on ufs_readdir. > * Add readlink support (applied tests for symlinks whose destionation path > were stored in blk pointers and the file itself). > * Several improvements. > > Wrote the documentation below. I think it would be good to push the doc to > the wiki as soon as the UFS support gets merged. > > Unix Fast File System (UFS/FFS) 1/2 on Syslinux - (usage/install) > ----- > There is a confusion about the name of this file system, then I decided to > contact the author, Kirk McKusick, who replied: > "The name has always been confusing (my bad). The code is broken into two > parts, the part that handles naming (UFS where the U stands for Unix), and the > part that handles disk layout (FFS where the F stands for Fast). > When the two parts are put together they are called UFS/FFS or more commonly > just UFS." > > Dependencies: > - Environment: > Require a *GNU/Linux* system to follow these steps. > - Package: > ufsutils (UFS filesystems utilities) > - Device driver: > UFS support on Linux: you need a kernel with the option CONFIG_UFS_FS_WRITE > enabled or at least configured as a module. > For the latter case, load the module ufs.ko with modprobe (probably located > at: /lib/modules/). > If this option wasn't even configured as a module, then you need to recompile > your kernel with that option enabled. > Further information can be found in the link below: > http://oz123.github.io/writings/2012-11-26-enable-ufs-write-support-debian/ > > Creating an UFS image: > sudo dd if=/dev/zero of=<image> bs= count=; > - UFS1: > sudo mkfs.ufs -O 1 <image>; > - UFS2: > sudo mkfs.ufs <image>; > > Mounting an UFS image: > - UFS1: > sudo mount -t ufs -o rw,loop,ufstype=44bsd <image> <mountpoint>; > - UFS2: > sudo mount -t ufs -o rw,loop,ufstype=ufs2 <image> <mountpoint>; > > Installing syslinux on an UFS image: > - Install extlinux into your mount point: > sudo extlinux -i <mountpoint>; > > - Finally, set up the syslinux config file. > > > Thanks for pcacjr who introduced me to Syslinux :-) > > Raphael S. Carvalho (2): > installer: Add UFS1/2 support for Extlinux installer > core/fs: Add support to Unix File system 1/2. > > core/fs/ufs/bmap.c | 202 ++++++++++++++++++++ > core/fs/ufs/ufs.c | 486 ++++++++++++++++++++++++++++++++++++++++++++++++ > core/fs/ufs/ufs.h | 254 +++++++++++++++++++++++++ > core/ldlinux.asm | 2 + > extlinux/main.c | 51 ++++- > extlinux/ufs.h | 26 +++ > extlinux/ufs_fs.h | 307 ++++++++++++++++++++++++++++++ > libinstaller/syslxfs.h | 4 +- > 8 files changed, 1322 insertions(+), 10 deletions(-) > create mode 100644 core/fs/ufs/bmap.c > create mode 100644 core/fs/ufs/ufs.c > create mode 100644 core/fs/ufs/ufs.h > create mode 100644 extlinux/ufs.h > create mode 100644 extlinux/ufs_fs.h >FYI: Both patches are commited. Thanks for the code contritubion to Syslinux
Possibly Parallel Threads
- [PATCH 0/2] UFS1/2 support series
- [PATCH 2/2] core/fs: Add support for Unix File system 1/2.
- [PATCH 001/001] core/fs: Add support to Unix File system 1/2.
- [PATCH 001/001] Add UFS1/2 support to Extlinux installer.
- [PATCH 1/1 v2] Add UFS1/2 support to Extlinux installer.