#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: lfs_vfsops.c,v 1.399 2026/01/05 05:02:47 perseant Exp $");
#if defined(_KERNEL_OPT)
#include "opt_lfs.h"
#include "opt_quota.h"
#include "opt_uvmhist.h"
#endif
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/namei.h>
#include <sys/proc.h>
#include <sys/kernel.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <sys/kthread.h>
#include <sys/buf.h>
#include <sys/device.h>
#include <sys/file.h>
#include <sys/fstypes.h>
#include <sys/disklabel.h>
#include <sys/ioctl.h>
#include <sys/errno.h>
#include <sys/malloc.h>
#include <sys/pool.h>
#include <sys/socket.h>
#include <sys/syslog.h>
#include <sys/sysctl.h>
#include <sys/conf.h>
#include <sys/kauth.h>
#include <sys/module.h>
#include <sys/syscallvar.h>
#include <sys/syscall.h>
#include <sys/syscallargs.h>
#include <miscfs/specfs/specdev.h>
#include <ufs/lfs/ulfs_quotacommon.h>
#include <ufs/lfs/ulfs_inode.h>
#include <ufs/lfs/ulfsmount.h>
#include <ufs/lfs/ulfs_bswap.h>
#include <ufs/lfs/ulfs_extern.h>
#ifdef UVMHIST
#include <uvm/uvm.h>
#endif
#include <uvm/uvm_extern.h>
#include <uvm/uvm_object.h>
#include <uvm/uvm_page.h>
#include <uvm/uvm_stat.h>
#include <ufs/lfs/lfs.h>
#include <ufs/lfs/lfs_accessors.h>
#include <ufs/lfs/lfs_kernel.h>
#include <ufs/lfs/lfs_extern.h>
#include <miscfs/genfs/genfs.h>
#include <miscfs/genfs/genfs_node.h>
MODULE(MODULE_CLASS_VFS, lfs, NULL);
static int lfs_gop_write(struct vnode *, struct vm_page **, int, int);
static int lfs_mountfs(struct vnode *, struct mount *, struct lwp *);
static int lfs_flushfiles(struct mount *, int);
static int compare_nodes_sd(void *, const void *, const void *);
static int compare_key_sd(void *, const void *, const void *);
extern const struct vnodeopv_desc lfs_vnodeop_opv_desc;
extern const struct vnodeopv_desc lfs_specop_opv_desc;
extern const struct vnodeopv_desc lfs_fifoop_opv_desc;
extern int locked_queue_rcount;
extern long locked_queue_rbytes;
static const rb_tree_ops_t lfs_rbtree_ops = {
.rbto_compare_nodes = compare_nodes_sd,
.rbto_compare_key = compare_key_sd,
.rbto_node_offset = offsetof(struct segdelta, rb_entry),
.rbto_context = NULL
};
struct lwp * lfs_writer_daemon = NULL;
kcondvar_t lfs_writerd_cv;
struct workqueue *lfs_cluster_wq = NULL;
struct workqueue *lfs_super_wq = NULL;
int lfs_do_flush = 0;
int lfs_do_rfw = 1;
const struct vnodeopv_desc * const lfs_vnodeopv_descs[] = {
&lfs_vnodeop_opv_desc,
&lfs_specop_opv_desc,
&lfs_fifoop_opv_desc,
NULL,
};
struct vfsops lfs_vfsops = {
.vfs_name = MOUNT_LFS,
.vfs_min_mount_data = sizeof (struct ulfs_args),
.vfs_mount = lfs_mount,
.vfs_start = ulfs_start,
.vfs_unmount = lfs_unmount,
.vfs_root = ulfs_root,
.vfs_quotactl = ulfs_quotactl,
.vfs_statvfs = lfs_statvfs,
.vfs_sync = lfs_sync,
.vfs_vget = lfs_vget,
.vfs_loadvnode = lfs_loadvnode,
.vfs_newvnode = lfs_newvnode,
.vfs_fhtovp = lfs_fhtovp,
.vfs_vptofh = lfs_vptofh,
.vfs_init = lfs_init,
.vfs_reinit = lfs_reinit,
.vfs_done = lfs_done,
.vfs_mountroot = lfs_mountroot,
.vfs_snapshot = (void *)eopnotsupp,
.vfs_extattrctl = lfs_extattrctl,
.vfs_suspendctl = genfs_suspendctl,
.vfs_renamelock_enter = genfs_renamelock_enter,
.vfs_renamelock_exit = genfs_renamelock_exit,
.vfs_fsync = (void *)eopnotsupp,
.vfs_opv_descs = lfs_vnodeopv_descs
};
const struct genfs_ops lfs_genfsops = {
.gop_size = lfs_gop_size,
.gop_alloc = ulfs_gop_alloc,
.gop_write = lfs_gop_write,
.gop_markupdate = ulfs_gop_markupdate,
.gop_putrange = genfs_gop_putrange,
};
struct shortlong {
const char *sname;
const char *lname;
};
static int
sysctl_lfs_dostats(SYSCTLFN_ARGS)
{
extern struct lfs_stats lfs_stats;
extern int lfs_dostats;
int error;
error = sysctl_lookup(SYSCTLFN_CALL(rnode));
if (error || newp == NULL)
return (error);
if (lfs_dostats == 0)
memset(&lfs_stats, 0, sizeof(lfs_stats));
return (0);
}
SYSCTL_SETUP(lfs_sysctl_setup, "lfs sysctl")
{
int i;
extern int lfs_writeindir, lfs_dostats, lfs_clean_vnhead,
lfs_fs_pagetrip, lfs_ignore_lazy_sync;
#ifdef DEBUG
extern int lfs_debug_log_subsys[DLOG_MAX];
struct shortlong dlog_names[DLOG_MAX] = {
{ "rollforward", "Debug roll-forward code" },
{ "alloc", "Debug inode allocation and free list" },
{ "avail", "Debug space-available-now accounting" },
{ "flush", "Debug flush triggers" },
{ "lockedlist", "Debug locked list accounting" },
{ "vnode_verbose", "Verbose per-vnode-written debugging" },
{ "vnode", "Debug vnode use during segment write" },
{ "segment", "Debug segment writing" },
{ "seguse", "Debug segment used-bytes accounting" },
{ "cleaner", "Debug cleaning routines" },
{ "mount", "Debug mount/unmount routines" },
{ "pagecache", "Debug UBC interactions" },
{ "dirop", "Debug directory-operation accounting" },
{ "malloc", "Debug private malloc accounting" },
};
#endif
struct shortlong stat_names[] = {
{ "segsused", "Number of new segments allocated" },
{ "psegwrites", "Number of partial-segment writes" },
{ "psyncwrites", "Number of synchronous partial-segment"
" writes" },
{ "pcleanwrites", "Number of partial-segment writes by the"
" cleaner" },
{ "blocktot", "Number of blocks written" },
{ "cleanblocks", "Number of blocks written by the cleaner" },
{ "ncheckpoints", "Number of checkpoints made" },
{ "nwrites", "Number of whole writes" },
{ "nsync_writes", "Number of synchronous writes" },
{ "wait_exceeded", "Number of times writer waited for"
" cleaner" },
{ "write_exceeded", "Number of times writer invoked flush" },
{ "flush_invoked", "Number of times flush was invoked" },
{ "vflush_invoked", "Number of time vflush was called" },
{ "clean_inlocked", "Number of vnodes skipped for being dead" },
{ "clean_vnlocked", "Number of vnodes skipped for vget failure" },
{ "segs_reclaimed", "Number of segments reclaimed" },
};
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT,
CTLTYPE_NODE, "lfs",
SYSCTL_DESCR("Log-structured file system"),
NULL, 0, NULL, 0,
CTL_VFS, 5, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "flushindir", NULL,
NULL, 0, &lfs_writeindir, 0,
CTL_VFS, 5, LFS_WRITEINDIR, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "clean_vnhead", NULL,
NULL, 0, &lfs_clean_vnhead, 0,
CTL_VFS, 5, LFS_CLEAN_VNHEAD, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "dostats",
SYSCTL_DESCR("Maintain statistics on LFS operations"),
sysctl_lfs_dostats, 0, &lfs_dostats, 0,
CTL_VFS, 5, LFS_DOSTATS, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "pagetrip",
SYSCTL_DESCR("How many dirty pages in fs triggers"
" a flush"),
NULL, 0, &lfs_fs_pagetrip, 0,
CTL_VFS, 5, LFS_FS_PAGETRIP, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "ignore_lazy_sync",
SYSCTL_DESCR("Lazy Sync is ignored entirely"),
NULL, 0, &lfs_ignore_lazy_sync, 0,
CTL_VFS, 5, LFS_IGNORE_LAZY_SYNC, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "rfw",
SYSCTL_DESCR("Use in-kernel roll-forward on mount"),
NULL, 0, &lfs_do_rfw, 0,
CTL_VFS, 5, LFS_DO_RFW, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "rfw_limit",
SYSCTL_DESCR("Maximum number of partial segments"
" to roll forward"),
NULL, 0, &lfs_rfw_max_psegs, 0,
CTL_VFS, 5, LFS_RFW_LIMIT, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT,
CTLTYPE_NODE, "stats",
SYSCTL_DESCR("Statistics"),
NULL, 0, NULL, 0,
CTL_VFS, 5, LFS_STATS, CTL_EOL);
for (i = 0; i < sizeof(struct lfs_stats) / sizeof(u_int); i++) {
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READONLY,
CTLTYPE_INT, stat_names[i].sname,
SYSCTL_DESCR(stat_names[i].lname),
NULL, 0, &(((u_int *)&lfs_stats.segsused)[i]),
0, CTL_VFS, 5, LFS_STATS, i, CTL_EOL);
}
#ifdef DEBUG
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT,
CTLTYPE_NODE, "debug",
SYSCTL_DESCR("Debug options"),
NULL, 0, NULL, 0,
CTL_VFS, 5, LFS_DEBUG, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "freelist",
SYSCTL_DESCR("Track consistency of inode freelist"),
NULL, 0, &lfs_do_check_freelist, 0,
CTL_VFS, 5, LFS_DEBUG, LFS_DEBUG_FREELIST, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT,
CTLTYPE_NODE, "log",
SYSCTL_DESCR("Verbose logging"),
NULL, 0, NULL, 0,
CTL_VFS, 5, LFS_DEBUG, LFS_DEBUG_LOG, CTL_EOL);
for (i = 0; i < DLOG_MAX; i++) {
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, dlog_names[i].sname,
SYSCTL_DESCR(dlog_names[i].lname),
NULL, 0, &(lfs_debug_log_subsys[i]), 0,
CTL_VFS, 5, LFS_DEBUG, LFS_DEBUG_LOG, i, CTL_EOL);
}
#endif
}
static const struct syscall_package lfs_syscalls[] = {
{ SYS_lfs_bmapv, 0, (sy_call_t *)sys_lfs_bmapv },
{ SYS_lfs_markv, 0, (sy_call_t *)sys_lfs_markv },
{ SYS___lfs_segwait50, 0, (sy_call_t *)sys___lfs_segwait50 },
{ SYS_lfs_segclean, 0, (sy_call_t *)sys_lfs_segclean },
{ 0, 0, NULL },
};
static int
lfs_modcmd(modcmd_t cmd, void *arg)
{
int error;
switch (cmd) {
case MODULE_CMD_INIT:
error = syscall_establish(NULL, lfs_syscalls);
if (error)
return error;
error = vfs_attach(&lfs_vfsops);
if (error != 0) {
syscall_disestablish(NULL, lfs_syscalls);
break;
}
cv_init(&lfs_allclean_wakeup, "segment");
break;
case MODULE_CMD_FINI:
error = vfs_detach(&lfs_vfsops);
if (error != 0)
break;
syscall_disestablish(NULL, lfs_syscalls);
cv_destroy(&lfs_allclean_wakeup);
break;
default:
error = ENOTTY;
break;
}
return (error);
}
struct pool lfs_inode_pool;
struct pool lfs_dinode_pool;
struct pool lfs_inoext_pool;
struct pool lfs_lbnentry_pool;
static void
lfs_writerd(void *arg)
{
mount_iterator_t *iter;
struct mount *mp;
struct lfs *fs;
struct vfsops *vfs = NULL;
int fsflags;
int lfsc;
int wrote_something = 0;
mutex_enter(&lfs_lock);
KASSERTMSG(lfs_writer_daemon == NULL, "more than one LFS writer daemon");
lfs_writer_daemon = curlwp;
mutex_exit(&lfs_lock);
vfs = vfs_getopsbyname(MOUNT_LFS);
mutex_enter(&lfs_lock);
for (;;) {
KASSERT(mutex_owned(&lfs_lock));
if (wrote_something == 0)
cv_timedwait(&lfs_writerd_cv, &lfs_lock, hz/10 + 1);
KASSERT(mutex_owned(&lfs_lock));
wrote_something = 0;
if (lfs_do_flush || locked_queue_count + locked_queue_rcount > LFS_MAX_BUFS ||
locked_queue_bytes + locked_queue_rbytes > LFS_MAX_BYTES ||
lfs_subsys_pages > LFS_MAX_PAGES) {
if (lfs_do_flush) {
DLOG((DLOG_FLUSH, "lfs_writerd: lfs_do_flush\n"));
}
if (locked_queue_count + locked_queue_rcount > LFS_MAX_BUFS) {
DLOG((DLOG_FLUSH, "lfs_writerd: lqc+lqrc = %d, max %d\n",
locked_queue_count, LFS_MAX_BUFS));
}
if (locked_queue_bytes + locked_queue_rbytes > LFS_MAX_BYTES) {
DLOG((DLOG_FLUSH, "lfs_writerd: lqb + lqrb = %ld, max %ld\n",
locked_queue_bytes, LFS_MAX_BYTES));
}
if (lfs_subsys_pages > LFS_MAX_PAGES) {
DLOG((DLOG_FLUSH, "lfs_writerd: lssp = %d, max %d\n",
lfs_subsys_pages, LFS_MAX_PAGES));
}
lfs_flush(NULL, SEGM_WRITERD, 0);
lfs_do_flush = 0;
KASSERT(mutex_owned(&lfs_lock));
continue;
}
KASSERT(mutex_owned(&lfs_lock));
mutex_exit(&lfs_lock);
mountlist_iterator_init(&iter);
lfsc = 0;
while ((mp = mountlist_iterator_next(iter)) != NULL) {
KASSERT(!mutex_owned(&lfs_lock));
if (strncmp(mp->mnt_stat.f_fstypename, MOUNT_LFS,
sizeof(mp->mnt_stat.f_fstypename)) == 0) {
++lfsc;
fs = VFSTOULFS(mp)->um_lfs;
daddr_t ooffset = 0;
fsflags = SEGM_SINGLE;
mutex_enter(&lfs_lock);
ooffset = lfs_sb_getoffset(fs);
if (lfs_sb_getnextseg(fs) < lfs_sb_getcurseg(fs) && fs->lfs_nowrap) {
mutex_exit(&lfs_lock);
continue;
}
if (LFS_STARVED_FOR_SEGS(fs)) {
mutex_exit(&lfs_lock);
DLOG((DLOG_FLUSH, "lfs_writerd: need cleaning before writing possible\n"));
lfs_wakeup_cleaner(fs);
continue;
}
if ((fs->lfs_dirvcount > LFS_MAX_FSDIROP(fs) ||
lfs_dirvcount > LFS_MAX_DIROP) &&
fs->lfs_dirops == 0) {
fsflags &= ~SEGM_SINGLE;
fsflags |= SEGM_CKP;
DLOG((DLOG_FLUSH, "lfs_writerd: checkpoint\n"));
lfs_flush_fs(fs, fsflags);
} else if (fs->lfs_pdflush) {
DLOG((DLOG_FLUSH, "lfs_writerd: pdflush set\n"));
lfs_flush_fs(fs, fsflags);
} else if (!TAILQ_EMPTY(&fs->lfs_pchainhd)) {
DLOG((DLOG_FLUSH, "lfs_writerd: pchain non-empty\n"));
mutex_exit(&lfs_lock);
lfs_writer_enter(fs, "wrdirop");
lfs_flush_pchain(fs);
lfs_writer_leave(fs);
mutex_enter(&lfs_lock);
}
if (lfs_sb_getoffset(fs) != ooffset)
++wrote_something;
mutex_exit(&lfs_lock);
}
KASSERT(!mutex_owned(&lfs_lock));
}
if (lfsc == 0) {
mutex_enter(&lfs_lock);
lfs_writer_daemon = NULL;
mutex_exit(&lfs_lock);
mountlist_iterator_destroy(iter);
break;
}
mountlist_iterator_destroy(iter);
mutex_enter(&lfs_lock);
}
KASSERT(!mutex_owned(&lfs_lock));
mutex_enter(&vfs_list_lock);
if (vfs != NULL)
vfs->vfs_refcount--;
mutex_exit(&vfs_list_lock);
kthread_exit(0);
}
void
lfs_init(void)
{
malloc_type_attach(M_SEGMENT);
pool_init(&lfs_inode_pool, sizeof(struct inode), 0, 0, 0,
"lfsinopl", &pool_allocator_nointr, IPL_NONE);
pool_init(&lfs_dinode_pool, sizeof(union lfs_dinode), 0, 0, 0,
"lfsdinopl", &pool_allocator_nointr, IPL_NONE);
pool_init(&lfs_inoext_pool, sizeof(struct lfs_inode_ext), 8, 0, 0,
"lfsinoextpl", &pool_allocator_nointr, IPL_NONE);
pool_init(&lfs_lbnentry_pool, sizeof(struct lbnentry), 0, 0, 0,
"lfslbnpool", &pool_allocator_nointr, IPL_NONE);
ulfs_init();
#ifdef DEBUG
memset(lfs_log, 0, sizeof(lfs_log));
#endif
mutex_init(&lfs_lock, MUTEX_DEFAULT, IPL_NONE);
cv_init(&lfs_writerd_cv, "lfswrite");
cv_init(&locked_queue_cv, "lfsbuf");
cv_init(&lfs_writing_cv, "lfsflush");
workqueue_create(&lfs_cluster_wq, "lfscwq", lfs_cluster_work, NULL,
PRI_BIO, IPL_BIO, WQ_MPSAFE);
workqueue_create(&lfs_super_wq, "lfsswq", lfs_super_work, NULL,
PRI_BIO, IPL_BIO, WQ_MPSAFE);
}
void
lfs_reinit(void)
{
ulfs_reinit();
}
void
lfs_done(void)
{
ulfs_done();
mutex_destroy(&lfs_lock);
cv_destroy(&lfs_writerd_cv);
cv_destroy(&locked_queue_cv);
cv_destroy(&lfs_writing_cv);
workqueue_destroy(lfs_cluster_wq);
workqueue_destroy(lfs_super_wq);
pool_destroy(&lfs_inode_pool);
pool_destroy(&lfs_dinode_pool);
pool_destroy(&lfs_inoext_pool);
pool_destroy(&lfs_lbnentry_pool);
malloc_type_detach(M_SEGMENT);
}
int
lfs_mountroot(void)
{
extern struct vnode *rootvp;
struct lfs *fs = NULL;
struct mount *mp;
struct lwp *l = curlwp;
struct ulfsmount *ump;
int error;
if (device_class(root_device) != DV_DISK)
return (ENODEV);
if (rootdev == NODEV)
return (ENODEV);
if ((error = vfs_rootmountalloc(MOUNT_LFS, "root_device", &mp))) {
vrele(rootvp);
return (error);
}
if ((error = lfs_mountfs(rootvp, mp, l))) {
vfs_unbusy(mp);
vfs_rele(mp);
return (error);
}
mountlist_append(mp);
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
lfs_sb_setfsmnt(fs, mp->mnt_stat.f_mntonname);
(void)lfs_statvfs(mp, &mp->mnt_stat);
vfs_unbusy(mp);
setrootfstime((time_t)lfs_sb_gettstamp(VFSTOULFS(mp)->um_lfs));
return (0);
}
int
lfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct vnode *devvp;
struct ulfs_args *args = data;
struct ulfsmount *ump = NULL;
struct lfs *fs = NULL;
int error = 0, update;
mode_t accessmode;
if (args == NULL)
return EINVAL;
if (*data_len < sizeof *args)
return EINVAL;
if (mp->mnt_flag & MNT_GETARGS) {
ump = VFSTOULFS(mp);
if (ump == NULL)
return EIO;
args->fspec = NULL;
*data_len = sizeof *args;
return 0;
}
update = mp->mnt_flag & MNT_UPDATE;
if (args->fspec != NULL) {
error = namei_simple_user(args->fspec,
NSM_FOLLOW_NOEMULROOT, &devvp);
if (error != 0)
return (error);
if (!update) {
if (devvp->v_type != VBLK)
error = ENOTBLK;
else if (bdevsw_lookup(devvp->v_rdev) == NULL)
error = ENXIO;
} else {
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
if (devvp != fs->lfs_devvp) {
if (devvp->v_rdev != fs->lfs_devvp->v_rdev)
error = EINVAL;
else {
vrele(devvp);
devvp = fs->lfs_devvp;
vref(devvp);
}
}
}
} else {
if (!update) {
return (EINVAL);
} else {
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
devvp = fs->lfs_devvp;
vref(devvp);
}
}
if (error == 0) {
accessmode = VREAD;
if (update ?
(mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
(mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
KAUTH_ARG(accessmode));
VOP_UNLOCK(devvp);
}
if (error) {
vrele(devvp);
return (error);
}
if (!update) {
int flags;
if (mp->mnt_flag & MNT_RDONLY)
flags = FREAD;
else
flags = FREAD|FWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = VOP_OPEN(devvp, flags, FSCRED);
VOP_UNLOCK(devvp);
if (error)
goto fail;
error = lfs_mountfs(devvp, mp, l);
if (error) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
(void)VOP_CLOSE(devvp, flags, NOCRED);
VOP_UNLOCK(devvp);
goto fail;
}
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
} else {
vrele(devvp);
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
if (!fs->lfs_ronly && (mp->mnt_iflag & IMNT_WANTRDONLY)) {
int flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = lfs_flushfiles(mp, flags);
if (error)
return error;
fs->lfs_ronly = 1;
lfs_prelock(fs, 0);
mutex_enter(&lfs_lock);
while (fs->lfs_iocount > 1)
(void)mtsleep(&fs->lfs_iocount, PRIBIO + 1,
"lfs_roio", 0, &lfs_lock);
mutex_exit(&lfs_lock);
lfs_preunlock(fs);
} else if (fs->lfs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
if (fs->lfs_use_quota2) {
#ifdef LFS_QUOTA2
error = lfs_quota2_mount(mp);
#else
uprintf("%s: no kernel support for this "
"filesystem's quotas\n",
mp->mnt_stat.f_mntonname);
if (mp->mnt_flag & MNT_FORCE) {
uprintf("%s: mounting anyway; "
"fsck afterwards\n",
mp->mnt_stat.f_mntonname);
} else {
error = EINVAL;
}
#endif
if (error) {
return error;
}
}
fs->lfs_ronly = 0;
if (lfs_sb_getpflags(fs) & LFS_PF_CLEAN) {
lfs_sb_setpflags(fs, lfs_sb_getpflags(fs) & ~LFS_PF_CLEAN);
lfs_writesuper(fs, lfs_sb_getsboff(fs, 0));
lfs_writesuper(fs, lfs_sb_getsboff(fs, 1));
}
}
if (args->fspec == NULL)
return 0;
}
error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
if (error == 0)
lfs_sb_setfsmnt(fs, mp->mnt_stat.f_mntonname);
return error;
fail:
vrele(devvp);
return (error);
}
static int
lfs_checkmagic(struct lfs *fs)
{
switch (fs->lfs_dlfs_u.u_32.dlfs_magic) {
case LFS_MAGIC:
fs->lfs_is64 = false;
fs->lfs_dobyteswap = false;
break;
case LFS64_MAGIC:
fs->lfs_is64 = true;
fs->lfs_dobyteswap = false;
break;
#ifdef LFS_EI
case LFS_MAGIC_SWAPPED:
fs->lfs_is64 = false;
fs->lfs_dobyteswap = true;
break;
case LFS64_MAGIC_SWAPPED:
fs->lfs_is64 = true;
fs->lfs_dobyteswap = true;
break;
#endif
default:
return EINVAL;
}
return 0;
}
int
lfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
{
struct lfs *primarysb, *altsb, *thesb;
struct buf *primarybuf, *altbuf;
struct lfs *fs;
struct ulfsmount *ump;
struct vnode *vp;
dev_t dev;
int error, i, ronly, fsbsize;
kauth_cred_t cred;
CLEANERINFO *cip;
SEGUSE *sup;
daddr_t sb_addr;
ino_t *orphan;
size_t norphan;
cred = l ? l->l_cred : NOCRED;
__CTASSERT(sizeof(struct dlfs) == DEV_BSIZE);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
VOP_UNLOCK(devvp);
if (error)
return (error);
ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
primarybuf = NULL;
altbuf = NULL;
ump = NULL;
sb_addr = LFS_LABELPAD / DEV_BSIZE;
while (1) {
error = bread(devvp, sb_addr, LFS_SBPAD, 0, &primarybuf);
if (error)
goto out;
primarysb = (struct lfs *)primarybuf->b_data;
error = lfs_checkmagic(primarysb);
if (error) {
DLOG((DLOG_MOUNT, "lfs_mountfs: primary superblock wrong magic\n"));
goto out;
}
if (lfs_sb_getbsize(primarysb) > MAXBSIZE ||
lfs_sb_getversion(primarysb) > LFS_VERSION ||
lfs_sb_getbsize(primarysb) < sizeof(struct dlfs)) {
DLOG((DLOG_MOUNT, "lfs_mountfs: primary superblock sanity failed\n"));
error = EINVAL;
goto out;
}
if (lfs_sb_getinodefmt(primarysb) > LFS_MAXINODEFMT) {
DLOG((DLOG_MOUNT, "lfs_mountfs: unknown inode format %d\n",
lfs_sb_getinodefmt(primarysb)));
error = EINVAL;
goto out;
}
if (lfs_sb_getversion(primarysb) == 1)
fsbsize = DEV_BSIZE;
else {
fsbsize = 1 << lfs_sb_getffshift(primarysb);
if (sb_addr != (lfs_sb_getsboff(primarysb, 0) << (lfs_sb_getffshift(primarysb) - DEV_BSHIFT))) {
DLOG((DLOG_MOUNT, "lfs_mountfs: sb daddr"
" 0x%llx is not right, trying 0x%llx\n",
(long long)sb_addr,
(long long)(lfs_sb_getsboff(primarysb, 0) << (lfs_sb_getffshift(primarysb) - DEV_BSHIFT))));
sb_addr = lfs_sb_getsboff(primarysb, 0) << (lfs_sb_getffshift(primarysb) - DEV_BSHIFT);
brelse(primarybuf, BC_INVAL);
continue;
}
}
break;
}
if (lfs_sb_getsboff(primarysb, 1) &&
lfs_sb_getsboff(primarysb, 1) - LFS_LABELPAD / fsbsize > LFS_SBPAD / fsbsize)
{
error = bread(devvp, lfs_sb_getsboff(primarysb, 1) * (fsbsize / DEV_BSIZE),
LFS_SBPAD, 0, &altbuf);
if (error)
goto out;
altsb = (struct lfs *)altbuf->b_data;
error = lfs_checkmagic(altsb);
if (error)
goto out;
if (lfs_sb_getbsize(altsb) > MAXBSIZE ||
lfs_sb_getversion(altsb) > LFS_VERSION ||
lfs_sb_getbsize(altsb) < sizeof(struct dlfs)) {
DLOG((DLOG_MOUNT, "lfs_mountfs: alt superblock"
" sanity failed\n"));
error = EINVAL;
goto out;
}
if (lfs_sb_getversion(primarysb) == 1) {
if (lfs_sb_gettstamp(altsb) < lfs_sb_gettstamp(primarysb))
thesb = altsb;
else
thesb = primarysb;
} else {
if (lfs_sb_getserial(altsb) < lfs_sb_getserial(primarysb))
thesb = altsb;
else
thesb = primarysb;
}
} else {
DLOG((DLOG_MOUNT, "lfs_mountfs: invalid alt superblock location"
" daddr=0x%x\n", lfs_sb_getsboff(primarysb, 1)));
error = EINVAL;
goto out;
}
fs = kmem_zalloc(sizeof(struct lfs), KM_SLEEP);
memcpy(&fs->lfs_dlfs_u.u_32, &thesb->lfs_dlfs_u.u_32,
sizeof(struct dlfs));
fs->lfs_is64 = thesb->lfs_is64;
fs->lfs_dobyteswap = thesb->lfs_dobyteswap;
fs->lfs_hasolddirfmt = false;
if (lfs_sb_getversion(fs) < 2) {
lfs_sb_setsumsize(fs, LFS_V1_SUMMARY_SIZE);
lfs_sb_setibsize(fs, lfs_sb_getbsize(fs));
lfs_sb_sets0addr(fs, lfs_sb_getsboff(fs, 0));
lfs_sb_settstamp(fs, lfs_sb_getotstamp(fs));
lfs_sb_setfsbtodb(fs, 0);
}
if (lfs_sb_getresvseg(fs) == 0)
lfs_sb_setresvseg(fs, MIN(lfs_sb_getminfreeseg(fs) - 1, \
MAX(MIN_RESV_SEGS, lfs_sb_getminfreeseg(fs) / 2 + 1)));
if (lfs_fsbtob(fs, LFS_NRESERVE(fs)) > LFS_MAX_BYTES && !ronly) {
DLOG((DLOG_MOUNT, "lfs_mount: to mount this filesystem read/write,"
" we need BUFPAGES >= %lld\n",
(long long)((bufmem_hiwater / bufmem_lowater) *
LFS_INVERSE_MAX_BYTES(
lfs_fsbtob(fs, LFS_NRESERVE(fs))) >> PAGE_SHIFT)));
kmem_free(fs, sizeof(struct lfs));
error = EFBIG;
goto out;
}
if (l != NULL) {
fs->lfs_flags = LFS_NOTYET;
fs->lfs_rfpid = l->l_proc->p_pid;
}
ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
ump->um_lfs = fs;
ump->um_fstype = fs->lfs_is64 ? ULFS2 : ULFS1;
brelse(primarybuf, BC_INVAL);
brelse(altbuf, BC_INVAL);
primarybuf = NULL;
altbuf = NULL;
fs->lfs_devbsize = DEV_BSIZE;
fs->lfs_iocount = 0;
fs->lfs_diropwait = 0;
fs->lfs_activesb = 0;
lfs_sb_setuinodes(fs, 0);
fs->lfs_ravail = 0;
fs->lfs_favail = 0;
fs->lfs_sbactive = 0;
fs->lfs_doifile = 0;
fs->lfs_writer = 0;
fs->lfs_dirops = 0;
fs->lfs_nadirop = 0;
fs->lfs_prelock = 0;
fs->lfs_seglock = 0;
fs->lfs_cleanlock = 0;
fs->lfs_pdflush = 0;
fs->lfs_sleepers = 0;
fs->lfs_pages = 0;
fs->lfs_prelocklwp = NULL;
rw_init(&fs->lfs_fraglock);
rw_init(&fs->lfs_iflock);
cv_init(&fs->lfs_sleeperscv, "lfs_slp");
cv_init(&fs->lfs_diropscv, "lfs_dirop");
cv_init(&fs->lfs_stopcv, "lfsstop");
cv_init(&fs->lfs_nextsegsleep, "segment");
cv_init(&fs->lfs_cleanercv, "cleancv");
cv_init(&fs->lfs_prelockcv, "prelockcv");
cv_init(&fs->lfs_cleanquitcv, "cleanquit");
fs->lfs_ronly = ronly;
if (ronly == 0)
fs->lfs_fmod = 1;
dev = devvp->v_rdev;
fs->lfs_dev = dev;
fs->lfs_devvp = devvp;
fs->um_flags = 0;
fs->um_bptrtodb = lfs_sb_getffshift(fs) - DEV_BSHIFT;
fs->um_seqinc = lfs_sb_getfrag(fs);
fs->um_nindir = lfs_sb_getnindir(fs);
fs->um_lognindir = ffs(lfs_sb_getnindir(fs)) - 1;
fs->um_maxsymlinklen = lfs_sb_getmaxsymlinklen(fs);
fs->um_dirblksiz = LFS_DIRBLKSIZ;
fs->um_maxfilesize = lfs_sb_getmaxfilesize(fs);
fs->lfs_use_quota2 = 0;
fs->lfs_quota_magic = 0;
fs->lfs_quota_flags = 0;
fs->lfs_quotaino[0] = 0;
fs->lfs_quotaino[1] = 0;
mp->mnt_data = ump;
mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_LFS);
mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
mp->mnt_stat.f_namemax = LFS_MAXNAMLEN;
mp->mnt_stat.f_iosize = lfs_sb_getbsize(fs);
mp->mnt_flag |= MNT_LOCAL;
mp->mnt_iflag |= IMNT_SHRLOOKUP;
mp->mnt_fs_bshift = lfs_sb_getbshift(fs);
mp->mnt_iflag |= IMNT_CAN_RWTORO;
if (fs->um_maxsymlinklen > 0)
mp->mnt_iflag |= IMNT_DTYPE;
else
fs->lfs_hasolddirfmt = true;
ump->um_mountp = mp;
for (i = 0; i < ULFS_MAXQUOTAS; i++)
ump->um_quotas[i] = NULLVP;
spec_node_setmountedfs(devvp, mp);
lfs_setup_resblks(fs);
TAILQ_INIT(&fs->lfs_dchainhd);
TAILQ_INIT(&fs->lfs_cleanhd);
TAILQ_INIT(&fs->lfs_pchainhd);
rb_tree_init(&fs->lfs_segdhd, &lfs_rbtree_ops);
if ((error = VFS_VGET(mp, LFS_IFILE_INUM, LK_EXCLUSIVE, &vp)) != 0) {
DLOG((DLOG_MOUNT, "lfs_mountfs: ifile vget failed, error=%d\n", error));
goto out;
}
fs->lfs_ivnode = vp;
vref(vp);
lfs_order_freelist(fs, &orphan, &norphan);
fs->lfs_nactive = 0;
for (i = 0; i < lfs_sb_getnseg(fs); i++) {
int changed;
struct buf *bp;
LFS_SEGENTRY(sup, fs, i, bp);
changed = 0;
if (!ronly) {
uint32_t nflags = sup->su_flags;
nflags &= ~(SEGUSE_ACTIVE | SEGUSE_EMPTY
| SEGUSE_READY | SEGUSE_INVAL
| SEGUSE_ERROR);
if (sup->su_nbytes == 0) {
if (sup->su_flags & SEGUSE_DIRTY)
nflags |= SEGUSE_EMPTY | SEGUSE_READY;
} else {
nflags &= ~(SEGUSE_EMPTY | SEGUSE_READY);
}
if (nflags != sup->su_flags) {
sup->su_flags = nflags;
++changed;
}
}
if (changed)
LFS_WRITESEGENTRY(sup, fs, i, bp);
else
brelse(bp, 0);
}
fs->lfs_lastcleaned = (uint32_t)-1;
lfs_free_orphans(fs, orphan, norphan);
if (!ronly) {
lfs_roll_forward(fs, mp, l);
lfs_reset_avail(fs);
}
fs->lfs_rfpid = 0;
if (!ronly && fs->lfs_use_quota2) {
#ifdef LFS_QUOTA2
error = lfs_quota2_mount(mp);
#else
uprintf("%s: no kernel support for this filesystem's quotas\n",
mp->mnt_stat.f_mntonname);
if (mp->mnt_flag & MNT_FORCE) {
uprintf("%s: mounting anyway; fsck afterwards\n",
mp->mnt_stat.f_mntonname);
} else {
error = EINVAL;
}
#endif
if (error) {
printf("lfs_mountfs: sorry, leaking some memory\n");
goto out;
}
}
if (!fs->lfs_ronly) {
lfs_sb_setpflags(fs, lfs_sb_getpflags(fs) & ~LFS_PF_CLEAN);
lfs_writesuper(fs, lfs_sb_getsboff(fs, 0));
lfs_writesuper(fs, lfs_sb_getsboff(fs, 1));
}
fs->lfs_flags &= ~(LFS_NOTYET);
wakeup(&fs->lfs_flags);
{
struct buf *bp;
LFS_CLEANERINFO(cip, fs, bp);
lfs_ci_setclean(fs, cip, lfs_sb_getnclean(fs));
lfs_ci_setdirty(fs, cip, lfs_sb_getnseg(fs) - lfs_sb_getnclean(fs));
lfs_ci_setavail(fs, cip, lfs_sb_getavail(fs));
lfs_ci_setbfree(fs, cip, lfs_sb_getbfree(fs));
(void) LFS_BWRITE_LOG(bp);
}
{
struct buf *bp;
LFS_SEGENTRY(sup, fs, lfs_dtosn(fs, lfs_sb_getoffset(fs)), bp);
sup->su_flags |= SEGUSE_DIRTY | SEGUSE_ACTIVE;
fs->lfs_nactive++;
LFS_WRITESEGENTRY(sup, fs, lfs_dtosn(fs, lfs_sb_getoffset(fs)), bp);
}
vput(vp);
mutex_enter(&lfs_lock);
if (lfs_writer_daemon == NULL &&
kthread_create(PRI_BIO, 0, NULL,
lfs_writerd, NULL, NULL, "lfs_writer") != 0)
panic("fork lfs_writer");
mutex_exit(&lfs_lock);
printf("WARNING: the log-structured file system is experimental\n"
"WARNING: it may cause system crashes and/or corrupt data\n");
return (0);
out:
if (primarybuf)
brelse(primarybuf, BC_INVAL);
if (altbuf)
brelse(altbuf, BC_INVAL);
if (ump) {
kmem_free(ump->um_lfs, sizeof(struct lfs));
kmem_free(ump, sizeof(*ump));
mp->mnt_data = NULL;
}
return (error);
}
void
lfs_reset_avail(struct lfs *fs)
{
daddr_t avail;
int sn, nclean, nsb, labelcorrect, curr, mfs;
SEGUSE *sup;
struct buf *bp;
KASSERT(!fs->lfs_ronly);
avail = nclean = nsb = labelcorrect = 0;
for (sn = 0; sn < lfs_sb_getnseg(fs); ++sn) {
LFS_SEGENTRY(sup, fs, sn, bp);
if (!(sup->su_flags & SEGUSE_DIRTY)) {
++nclean;
avail += lfs_segtod(fs, 1);
if (sup->su_flags & SEGUSE_SUPERBLOCK)
++nsb;
if (sn == 0 && lfs_sb_getversion(fs) > 1
&& lfs_sb_gets0addr(fs) < lfs_btofsb(fs, LFS_LABELPAD))
labelcorrect = lfs_btofsb(fs, LFS_LABELPAD)
- lfs_sb_gets0addr(fs);
}
brelse(bp, 0);
}
sn = lfs_dtosn(fs, lfs_sb_getoffset(fs));
curr = lfs_sntod(fs, sn + 1) - lfs_sb_getoffset(fs);
mfs = lfs_segtod(fs, (lfs_sb_getminfreeseg(fs) -
(lfs_sb_getminfreeseg(fs) / 2)));
avail = nclean * lfs_segtod(fs, 1);
avail -= nsb * lfs_btofsb(fs, LFS_SBPAD);
avail -= labelcorrect;
avail += curr;
avail -= mfs;
DLOG((DLOG_AVAIL, "avail := %jd*%jd-%jd*%jd-%jd+%jd-%jd=%jd\n",
(intmax_t)nclean,
(intmax_t)lfs_segtod(fs, 1),
(intmax_t)nsb,
(intmax_t)lfs_btofsb(fs, LFS_SBPAD),
(intmax_t)labelcorrect,
(intmax_t)curr,
(intmax_t)mfs,
(intmax_t)avail));
lfs_sb_setavail(fs, avail);
lfs_sb_setnclean(fs, nclean);
}
int
lfs_unmount(struct mount *mp, int mntflags)
{
struct ulfsmount *ump;
struct lfs *fs;
int error, ronly;
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
error = lfs_flushfiles(mp, mntflags & MNT_FORCE ? FORCECLOSE : 0);
if (error)
return error;
DEBUG_CHECK_FREELIST(fs);
KASSERT(LIST_FIRST(&fs->lfs_ivnode->v_dirtyblkhd) == NULL);
vgone(fs->lfs_ivnode);
ronly = !fs->lfs_ronly;
if (fs->lfs_devvp->v_type != VBAD)
spec_node_setmountedfs(fs->lfs_devvp, NULL);
vn_lock(fs->lfs_devvp, LK_EXCLUSIVE | LK_RETRY);
error = VOP_CLOSE(fs->lfs_devvp,
ronly ? FREAD : FREAD|FWRITE, NOCRED);
vput(fs->lfs_devvp);
if (fs->lfs_pages > 0)
printf("lfs_unmount: still claim %d pages (%d in subsystem)\n",
fs->lfs_pages, lfs_subsys_pages);
mutex_enter(&lfs_lock);
while (fs->lfs_iocount)
mtsleep(&fs->lfs_iocount, PRIBIO + 1, "lfs_umount2", 0,
&lfs_lock);
mutex_exit(&lfs_lock);
free(fs->lfs_ino_bitmap, M_SEGMENT);
lfs_free_resblks(fs);
cv_destroy(&fs->lfs_sleeperscv);
cv_destroy(&fs->lfs_diropscv);
cv_destroy(&fs->lfs_stopcv);
cv_destroy(&fs->lfs_nextsegsleep);
rw_destroy(&fs->lfs_fraglock);
rw_destroy(&fs->lfs_iflock);
kmem_free(fs, sizeof(struct lfs));
kmem_free(ump, sizeof(*ump));
mp->mnt_data = NULL;
mp->mnt_flag &= ~MNT_LOCAL;
return (error);
}
static int
lfs_flushfiles(struct mount *mp, int flags)
{
struct lwp *l = curlwp;
struct ulfsmount *ump;
struct lfs *fs;
struct vnode *vp;
int error;
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
if (!fs->lfs_ronly) {
lfs_segwrite(mp, SEGM_CKP | SEGM_SYNC);
lfs_segwrite(mp, SEGM_CKP | SEGM_SYNC);
}
lfs_wakeup_cleaner(fs);
mutex_enter(&lfs_lock);
while (fs->lfs_sleepers)
cv_wait(&fs->lfs_sleeperscv, &lfs_lock);
mutex_exit(&lfs_lock);
#ifdef LFS_EXTATTR
if (ump->um_fstype == ULFS1) {
if (ump->um_extattr.uepm_flags & ULFS_EXTATTR_UEPM_STARTED) {
ulfs_extattr_stop(mp, curlwp);
}
if (ump->um_extattr.uepm_flags & ULFS_EXTATTR_UEPM_INITIALIZED) {
ulfs_extattr_uepm_destroy(&ump->um_extattr);
mp->mnt_flag &= ~MNT_EXTATTR;
}
}
#endif
#ifdef LFS_QUOTA
if ((error = lfsquota1_umount(mp, flags)) != 0)
return (error);
#endif
#ifdef LFS_QUOTA2
if ((error = lfsquota2_umount(mp, flags)) != 0)
return (error);
#endif
if ((error = vflush(mp, fs->lfs_ivnode, flags)) != 0)
return (error);
if ((error = VFS_SYNC(mp, MNT_WAIT, l->l_cred)) != 0)
return (error);
vp = fs->lfs_ivnode;
mutex_enter(vp->v_interlock);
if (LIST_FIRST(&vp->v_dirtyblkhd))
panic("lfs_flushfiles: still dirty blocks on ifile vnode");
mutex_exit(vp->v_interlock);
if (!fs->lfs_ronly) {
lfs_sb_setpflags(fs, lfs_sb_getpflags(fs) | LFS_PF_CLEAN);
lfs_writesuper(fs, lfs_sb_getsboff(fs, 0));
lfs_writesuper(fs, lfs_sb_getsboff(fs, 1));
}
mutex_enter(&lfs_lock);
while (fs->lfs_iocount)
mtsleep(&fs->lfs_iocount, PRIBIO + 1, "lfs_umount", 0,
&lfs_lock);
mutex_exit(&lfs_lock);
return 0;
}
int
lfs_statvfs(struct mount *mp, struct statvfs *sbp)
{
struct lfs *fs;
struct ulfsmount *ump;
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
sbp->f_bsize = lfs_sb_getbsize(fs);
sbp->f_frsize = lfs_sb_getfsize(fs);
sbp->f_iosize = lfs_sb_getbsize(fs);
sbp->f_blocks = LFS_EST_NONMETA(fs) - VTOI(fs->lfs_ivnode)->i_lfs_effnblks;
sbp->f_bfree = LFS_EST_BFREE(fs);
KASSERT(sbp->f_bfree <= lfs_sb_getdsize(fs));
#if 0
if (sbp->f_bfree < 0)
sbp->f_bfree = 0;
#endif
sbp->f_bresvd = LFS_EST_RSVD(fs);
if (sbp->f_bfree > sbp->f_bresvd)
sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
else
sbp->f_bavail = 0;
sbp->f_files = lfs_sb_getbfree(fs) / lfs_btofsb(fs, lfs_sb_getibsize(fs))
* LFS_INOPB(fs);
sbp->f_ffree = sbp->f_files - lfs_sb_getnfiles(fs);
sbp->f_favail = sbp->f_ffree;
sbp->f_fresvd = 0;
copy_statvfs_info(sbp, mp);
return (0);
}
int
lfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
{
int error, segflags;
struct lfs *fs;
fs = VFSTOULFS(mp)->um_lfs;
if (fs->lfs_ronly)
return 0;
mutex_enter(&lfs_lock);
if (fs->lfs_nowrap && lfs_sb_getnextseg(fs) < lfs_sb_getcurseg(fs)) {
mutex_exit(&lfs_lock);
return 0;
}
mutex_exit(&lfs_lock);
lfs_writer_enter(fs, "lfs_wsync");
DLOG((DLOG_FLUSH, "lfs_sync waitfor=%x at 0x%jx\n", waitfor,
(uintmax_t)lfs_sb_getoffset(fs)));
segflags = 0;
if (waitfor == MNT_LAZY)
segflags = 0;
else if (waitfor == MNT_NOWAIT)
segflags = SEGM_CKP;
else
segflags = SEGM_SYNC | SEGM_CKP;
error = lfs_segwrite(mp, segflags);
lfs_writer_leave(fs);
#ifdef LFS_QUOTA
lfs_qsync(mp);
#endif
return (error);
}
int
lfs_vget(struct mount *mp, ino_t ino, int lktype, struct vnode **vpp)
{
int error;
error = vcache_get(mp, &ino, sizeof(ino), vpp);
if (error)
return error;
error = vn_lock(*vpp, lktype);
if (error) {
vrele(*vpp);
*vpp = NULL;
return error;
}
return 0;
}
static int
compare_nodes_sd(void *context, const void *v1, const void *v2)
{
const struct segdelta *sd1, *sd2;
sd1 = (const struct segdelta *)v1;
sd2 = (const struct segdelta *)v2;
return sd1->segnum - sd2->segnum;
}
static int
compare_key_sd(void *context, const void *nv, const void *kv)
{
const struct segdelta *sd;
long key;
sd = (const struct segdelta *)nv;
key = *(const long *)kv;
return sd->segnum - key;
}
static void
lfs_init_vnode(struct ulfsmount *ump, ino_t ino, struct vnode *vp)
{
struct lfs *fs = ump->um_lfs;
struct inode *ip;
union lfs_dinode *dp;
ASSERT_NO_SEGLOCK(fs);
ip = pool_get(&lfs_inode_pool, PR_WAITOK);
memset(ip, 0, sizeof(*ip));
dp = pool_get(&lfs_dinode_pool, PR_WAITOK);
memset(dp, 0, sizeof(*dp));
ip->inode_ext.lfs = pool_get(&lfs_inoext_pool, PR_WAITOK);
memset(ip->inode_ext.lfs, 0, sizeof(*ip->inode_ext.lfs));
ip->i_din = dp;
ip->i_ump = ump;
ip->i_vnode = vp;
ip->i_dev = fs->lfs_dev;
lfs_dino_setinumber(fs, dp, ino);
ip->i_number = ino;
ip->i_lfs = fs;
ip->i_lfs_effnblks = 0;
SPLAY_INIT(&ip->i_lfs_lbtree);
ip->i_lfs_nbtree = 0;
rb_tree_init(&ip->i_lfs_segdhd, &lfs_rbtree_ops);
vp->v_tag = VT_LFS;
vp->v_op = lfs_vnodeop_p;
vp->v_data = ip;
}
static void
lfs_deinit_vnode(struct ulfsmount *ump, struct vnode *vp)
{
struct inode *ip = VTOI(vp);
pool_put(&lfs_inoext_pool, ip->inode_ext.lfs);
pool_put(&lfs_dinode_pool, ip->i_din);
pool_put(&lfs_inode_pool, ip);
vp->v_data = NULL;
}
int
lfs_loadvnode(struct mount *mp, struct vnode *vp,
const void *key, size_t key_len, const void **new_key)
{
struct lfs *fs;
union lfs_dinode *dip;
struct inode *ip;
struct buf *bp;
IFILE *ifp;
struct ulfsmount *ump;
ino_t ino;
daddr_t daddr;
int error, retries;
struct timespec ts;
KASSERT(key_len == sizeof(ino));
memcpy(&ino, key, key_len);
memset(&ts, 0, sizeof ts);
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
mutex_enter(&lfs_lock);
while ((fs->lfs_flags & LFS_NOTYET) && curproc->p_pid != fs->lfs_rfpid)
mtsleep(&fs->lfs_flags, PRIBIO+1, "lfs_notyet", 0,
&lfs_lock);
mutex_exit(&lfs_lock);
KASSERT(ino >= LFS_IFILE_INUM);
if (ino == LFS_IFILE_INUM)
daddr = lfs_sb_getidaddr(fs);
else {
LFS_IENTRY(ifp, fs, ino, bp);
daddr = lfs_if_getdaddr(fs, ifp);
if (lfs_sb_getversion(fs) > 1) {
ts.tv_sec = lfs_if_getatime_sec(fs, ifp);
ts.tv_nsec = lfs_if_getatime_nsec(fs, ifp);
}
brelse(bp, 0);
if (DADDR_IS_BAD(daddr))
return (ENOENT);
}
lfs_init_vnode(ump, ino, vp);
ip = VTOI(vp);
if (curlwp == fs->lfs_cleaner_thread && fs->lfs_cleaner_hint != NULL &&
fs->lfs_cleaner_hint->bi_lbn == LFS_UNUSED_LBN) {
dip = fs->lfs_cleaner_hint->bi_bp;
if (fs->lfs_is64) {
error = copyin(dip, &ip->i_din->u_64,
sizeof(struct lfs64_dinode));
} else {
error = copyin(dip, &ip->i_din->u_32,
sizeof(struct lfs32_dinode));
}
if (error) {
lfs_deinit_vnode(ump, vp);
return error;
}
KASSERT(ip->i_number == ino);
goto out;
}
retries = 0;
again:
error = bread(fs->lfs_devvp, LFS_FSBTODB(fs, daddr),
(lfs_sb_getversion(fs) == 1 ? lfs_sb_getbsize(fs) : lfs_sb_getibsize(fs)),
0, &bp);
if (error) {
lfs_deinit_vnode(ump, vp);
return error;
}
dip = lfs_ifind(fs, ino, bp);
if (dip == NULL) {
brelse(bp, BC_INVAL);
++retries;
if (retries <= LFS_IFIND_RETRIES) {
mutex_enter(&lfs_lock);
if (fs->lfs_iocount) {
DLOG((DLOG_VNODE,
"%s: dinode %d not found, retrying...\n",
__func__, ino));
(void)mtsleep(&fs->lfs_iocount, PRIBIO + 1,
"lfs ifind", 1, &lfs_lock);
} else
retries = LFS_IFIND_RETRIES;
mutex_exit(&lfs_lock);
goto again;
}
#ifdef DEBUG
mutex_enter(&lfs_lock);
if (fs->lfs_seglock > 0) {
struct buf **bpp;
union lfs_dinode *dp;
int i;
for (bpp = fs->lfs_sp->bpp;
bpp != fs->lfs_sp->cbpp; ++bpp) {
if ((*bpp)->b_vp == fs->lfs_ivnode &&
bpp != fs->lfs_sp->bpp) {
printf("%s: block 0x%" PRIx64 ": ",
__func__, (*bpp)->b_blkno);
for (i = 0; i < LFS_INOPB(fs); i++) {
dp = DINO_IN_BLOCK(fs,
(*bpp)->b_data, i);
if (lfs_dino_getinumber(fs, dp))
printf("%ju ",
(uintmax_t)lfs_dino_getinumber(fs, dp));
}
printf("\n");
}
}
}
mutex_exit(&lfs_lock);
#endif
panic("lfs_loadvnode: dinode not found");
}
lfs_copy_dinode(fs, ip->i_din, dip);
brelse(bp, 0);
out:
if (lfs_sb_getversion(fs) > 1) {
lfs_dino_setatime(fs, ip->i_din, ts.tv_sec);
lfs_dino_setatimensec(fs, ip->i_din, ts.tv_nsec);
}
lfs_vinit(mp, &vp);
*new_key = &ip->i_number;
return 0;
}
int
lfs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
struct vattr *vap, kauth_cred_t cred, void *extra,
size_t *key_len, const void **new_key)
{
ino_t ino;
struct inode *ip;
struct ulfsmount *ump;
struct lfs *fs;
int error, mode, gen;
ump = VFSTOULFS(mp);
fs = ump->um_lfs;
KASSERT(dvp != NULL || vap->va_fileid > 0);
KASSERT((fs->lfs_flags & LFS_NOTYET) ||
(dvp != NULL && dvp->v_mount == mp));
KASSERT(vap->va_type != VNON);
*key_len = sizeof(ino);
mode = MAKEIMODE(vap->va_type, vap->va_mode);
if (dvp == NULL) {
ino = vap->va_fileid;
gen = vap->va_gen;
error = lfs_valloc_fixed(fs, ino, gen);
} else {
error = lfs_valloc(dvp, mode, cred, &ino, &gen);
}
if (error)
return error;
lfs_init_vnode(ump, ino, vp);
ip = VTOI(vp);
mutex_enter(&lfs_lock);
LFS_SET_UINO(ip, IN_CHANGE);
mutex_exit(&lfs_lock);
ip->i_lfs_hiblk = -1;
ip->i_gen = gen;
lfs_dino_setgen(fs, ip->i_din, gen);
memset(ip->i_lfs_fragsize, 0,
ULFS_NDADDR * sizeof(*ip->i_lfs_fragsize));
if (cred == NOCRED || cred == FSCRED) {
ip->i_gid = 0;
ip->i_uid = 0;
} else {
ip->i_gid = VTOI(dvp)->i_gid;
ip->i_uid = kauth_cred_geteuid(cred);
}
DIP_ASSIGN(ip, gid, ip->i_gid);
DIP_ASSIGN(ip, uid, ip->i_uid);
#if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
error = lfs_chkiq(ip, 1, cred, 0);
if (error) {
lfs_vfree(dvp, ino, mode);
lfs_deinit_vnode(ump, vp);
return error;
}
#endif
ip->i_flags = 0;
DIP_ASSIGN(ip, flags, 0);
ip->i_mode = mode;
DIP_ASSIGN(ip, mode, mode);
if (vap->va_rdev != VNOVAL) {
if (ump->um_fstype == ULFS1)
ip->i_din->u_32.di_rdev = ulfs_rw32(vap->va_rdev,
ULFS_MPNEEDSWAP(fs));
else
ip->i_din->u_64.di_rdev = ulfs_rw64(vap->va_rdev,
ULFS_MPNEEDSWAP(fs));
}
lfs_vinit(mp, &vp);
*new_key = &ip->i_number;
return 0;
}
int
lfs_fhtovp(struct mount *mp, struct fid *fhp, int lktype, struct vnode **vpp)
{
struct lfid lfh;
struct lfs *fs;
if (fhp->fid_len != sizeof(struct lfid))
return EINVAL;
memcpy(&lfh, fhp, sizeof(lfh));
if (lfh.lfid_ino < LFS_IFILE_INUM)
return ESTALE;
fs = VFSTOULFS(mp)->um_lfs;
if (lfh.lfid_ident != lfs_sb_getident(fs))
return ESTALE;
if (lfh.lfid_ino >
((lfs_dino_getsize(fs, VTOI(fs->lfs_ivnode)->i_din) >> lfs_sb_getbshift(fs)) -
lfs_sb_getcleansz(fs) - lfs_sb_getsegtabsz(fs)) * lfs_sb_getifpb(fs))
return ESTALE;
return (ulfs_fhtovp(mp, &lfh.lfid_ufid, lktype, vpp));
}
int
lfs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
{
struct inode *ip;
struct lfid lfh;
if (*fh_size < sizeof(struct lfid)) {
*fh_size = sizeof(struct lfid);
return E2BIG;
}
*fh_size = sizeof(struct lfid);
ip = VTOI(vp);
memset(&lfh, 0, sizeof(lfh));
lfh.lfid_len = sizeof(struct lfid);
lfh.lfid_ino = ip->i_number;
lfh.lfid_gen = ip->i_gen;
lfh.lfid_ident = lfs_sb_getident(ip->i_lfs);
memcpy(fhp, &lfh, sizeof(lfh));
return (0);
}
static bool
lfs_issequential_hole(const struct lfs *fs,
daddr_t daddr0, daddr_t daddr1)
{
(void)fs;
KASSERT(daddr0 == UNWRITTEN ||
(0 <= daddr0 && daddr0 <= LFS_MAX_DADDR(fs)));
KASSERT(daddr1 == UNWRITTEN ||
(0 <= daddr1 && daddr1 <= LFS_MAX_DADDR(fs)));
if (daddr0 != 0 && daddr1 != 0)
return true;
if (daddr0 == 0 && daddr1 == 0)
return true;
return false;
}
static int
lfs_gop_write(struct vnode *vp, struct vm_page **pgs, int npages,
int flags)
{
int i, error, run, haveeof = 0;
int fs_bshift;
vaddr_t kva;
off_t eof, offset, startoffset = 0;
size_t bytes, iobytes, skipbytes;
bool async = (flags & PGO_SYNCIO) == 0;
daddr_t lbn, blkno;
struct vm_page *pg;
struct buf *mbp, *bp;
struct vnode *devvp = VTOI(vp)->i_devvp;
struct inode *ip = VTOI(vp);
struct lfs *fs = ip->i_lfs;
struct segment *sp = fs->lfs_sp;
SEGSUM *ssp;
UVMHIST_FUNC("lfs_gop_write"); UVMHIST_CALLED(ubchist);
const char * failreason = NULL;
ASSERT_SEGLOCK(fs);
KASSERT(vp != fs->lfs_ivnode);
mutex_enter(vp->v_interlock);
if (LFS_STARVED_FOR_SEGS(fs) && vdead_check(vp, VDEAD_NOWAIT) == 0) {
mutex_exit(vp->v_interlock);
failreason = "Starved for segs and not flushing vp";
goto tryagain;
}
mutex_exit(vp->v_interlock);
if (!LFS_SEGLOCK_HELD(fs)) {
failreason = "Seglock not held";
goto tryagain;
}
if (ip->i_lfs_iflags & LFSI_NO_GOP_WRITE) {
failreason = "Inode with no_gop_write";
goto tryagain;
}
if ((pgs[0]->offset & lfs_sb_getbmask(fs)) != 0) {
failreason = "Bad page offset";
goto tryagain;
}
UVMHIST_LOG(ubchist, "vp %#jx pgs %#jx npages %jd flags 0x%jx",
(uintptr_t)vp, (uintptr_t)pgs, npages, flags);
GOP_SIZE(vp, vp->v_size, &eof, 0);
haveeof = 1;
if (vp->v_type == VREG)
fs_bshift = vp->v_mount->mnt_fs_bshift;
else
fs_bshift = DEV_BSHIFT;
error = 0;
pg = pgs[0];
startoffset = pg->offset;
KASSERT(eof >= 0);
if (startoffset >= eof) {
failreason = "Offset beyond EOF";
goto tryagain;
} else
bytes = MIN(npages << PAGE_SHIFT, eof - startoffset);
skipbytes = 0;
KASSERT(bytes != 0);
for (i = 0; i < npages; i++) {
if (pgs[i]->flags & PG_DELWRI) {
KASSERT(!(pgs[i]->flags & PG_PAGEOUT));
pgs[i]->flags &= ~PG_DELWRI;
pgs[i]->flags |= PG_PAGEOUT;
uvm_pageout_start(1);
rw_enter(vp->v_uobj.vmobjlock, RW_WRITER);
uvm_pagelock(pgs[i]);
uvm_pageunwire(pgs[i]);
uvm_pageunlock(pgs[i]);
rw_exit(vp->v_uobj.vmobjlock);
}
}
if (startoffset & lfs_sb_getbmask(fs))
printf("%" PRId64 " & %" PRIu64 " = %" PRId64 "\n",
startoffset, lfs_sb_getbmask(fs),
startoffset & lfs_sb_getbmask(fs));
KASSERT((startoffset & lfs_sb_getbmask(fs)) == 0);
if (bytes & lfs_sb_getffmask(fs)) {
printf("lfs_gop_write: asked to write %ld bytes\n", (long)bytes);
panic("lfs_gop_write: non-integer blocks");
}
ssp = (SEGSUM *)sp->segsum;
if ((kva = uvm_pagermapin(pgs, npages, UVMPAGER_MAPIN_WRITE |
(lfs_ss_getnfinfo(fs, ssp) < 1 ?
UVMPAGER_MAPIN_WAITOK : 0))) == 0x0) {
DLOG((DLOG_PAGE, "lfs_gop_write: forcing write\n"));
lfs_updatemeta(sp);
lfs_release_finfo(fs);
(void) lfs_writeseg(fs, sp);
lfs_acquire_finfo(fs, ip->i_number, ip->i_gen);
kva = uvm_pagermapin(pgs, npages, UVMPAGER_MAPIN_WRITE |
UVMPAGER_MAPIN_WAITOK);
}
mbp = getiobuf(NULL, true);
UVMHIST_LOG(ubchist, "vp %#jx mbp %#jx num now %jd bytes 0x%jx",
(uintptr_t)vp, (uintptr_t)mbp, vp->v_numoutput, bytes);
mbp->b_bufsize = npages << PAGE_SHIFT;
mbp->b_data = (void *)kva;
mbp->b_resid = mbp->b_bcount = bytes;
mbp->b_cflags |= BC_BUSY|BC_AGE;
mbp->b_iodone = uvm_aio_aiodone;
bp = NULL;
for (offset = startoffset;
bytes > 0;
offset += iobytes, bytes -= iobytes) {
lbn = offset >> fs_bshift;
error = ulfs_bmaparray(vp, lbn, &blkno, NULL, NULL, &run,
lfs_issequential_hole);
if (error) {
UVMHIST_LOG(ubchist, "ulfs_bmaparray() -> %jd",
error,0,0,0);
skipbytes += bytes;
bytes = 0;
break;
}
iobytes = MIN((((off_t)lbn + 1 + run) << fs_bshift) - offset,
bytes);
if (blkno == (daddr_t)-1) {
skipbytes += iobytes;
continue;
}
if (sp->sum_bytes_left < sizeof(int32_t) ||
sp->seg_bytes_left < (1 << lfs_sb_getbshift(fs))) {
int vers;
lfs_updatemeta(sp);
vers = lfs_fi_getversion(fs, sp->fip);
lfs_release_finfo(fs);
(void) lfs_writeseg(fs, sp);
lfs_acquire_finfo(fs, ip->i_number, vers);
}
iobytes = MIN(iobytes, (sp->seg_bytes_left >> fs_bshift)
<< fs_bshift);
iobytes = MIN(iobytes, (sp->sum_bytes_left / sizeof(int32_t))
<< fs_bshift);
KASSERT(iobytes > 0);
if (offset == startoffset && iobytes == bytes) {
bp = mbp;
mutex_enter(vp->v_interlock);
vp->v_numoutput++;
mutex_exit(vp->v_interlock);
} else {
bp = getiobuf(NULL, true);
UVMHIST_LOG(ubchist, "vp %#jx bp %#jx num now %jd",
(uintptr_t)vp, (uintptr_t)bp, vp->v_numoutput, 0);
nestiobuf_setup(mbp, bp, offset - pg->offset, iobytes);
bp->b_flags &= ~B_ASYNC;
}
mutex_enter(&bufcache_lock);
mutex_enter(vp->v_interlock);
bgetvp(vp, bp);
mutex_exit(vp->v_interlock);
mutex_exit(&bufcache_lock);
bp->b_lblkno = lfs_lblkno(fs, offset);
bp->b_private = mbp;
if (devvp->v_type == VBLK) {
bp->b_dev = devvp->v_rdev;
}
VOP_BWRITE(bp->b_vp, bp);
while (lfs_gatherblock(sp, bp, NULL))
continue;
}
nestiobuf_done(mbp, skipbytes, error);
if (skipbytes) {
UVMHIST_LOG(ubchist, "skipbytes %jd", skipbytes, 0,0,0);
}
UVMHIST_LOG(ubchist, "returning 0", 0,0,0,0);
if (!async) {
UVMHIST_LOG(ubchist, "flushing", 0,0,0,0);
mutex_enter(&lfs_lock);
lfs_flush(fs, 0, 1);
mutex_exit(&lfs_lock);
}
if ((sp->seg_flags & SEGM_SINGLE) && lfs_sb_getcurseg(fs) != fs->lfs_startseg)
return EAGAIN;
return (0);
tryagain:
if (failreason != NULL) {
DLOG((DLOG_PAGE, "lfs_gop_write: %s\n", failreason));
}
if (haveeof && startoffset >= eof) {
DLOG((DLOG_PAGE, "lfs_gop_write: ino %d start 0x%" PRIx64
" eof 0x%" PRIx64 " npages=%d\n", VTOI(vp)->i_number,
pgs[0]->offset, eof, npages));
}
rw_enter(vp->v_uobj.vmobjlock, RW_WRITER);
for (i = 0; i < npages; i++) {
pg = pgs[i];
if (pg->flags & PG_PAGEOUT)
uvm_pageout_done(1);
uvm_pagelock(pg);
if (pg->flags & PG_DELWRI) {
uvm_pageunwire(pg);
}
uvm_pageactivate(pg);
uvm_pageunlock(pg);
pg->flags &= ~(PG_DELWRI|PG_PAGEOUT|PG_RELEASED);
uvm_pagemarkdirty(pg, UVM_PAGE_STATUS_DIRTY);
DLOG((DLOG_PAGE, "pg[%d] = %p (vp %p off %" PRIx64 ")\n", i, pg,
vp, pg->offset));
DLOG((DLOG_PAGE, "pg[%d]->flags = %x\n", i, pg->flags));
DLOG((DLOG_PAGE, "pg[%d]->pqflags = %x\n", i, pg->pqflags));
DLOG((DLOG_PAGE, "pg[%d]->uanon = %p\n", i, pg->uanon));
DLOG((DLOG_PAGE, "pg[%d]->uobject = %p\n", i, pg->uobject));
DLOG((DLOG_PAGE, "pg[%d]->wire_count = %d\n", i,
pg->wire_count));
DLOG((DLOG_PAGE, "pg[%d]->loan_count = %d\n", i,
pg->loan_count));
}
uvm_page_unbusy(pgs, npages);
rw_exit(vp->v_uobj.vmobjlock);
return EAGAIN;
}
void
lfs_vinit(struct mount *mp, struct vnode **vpp)
{
struct vnode *vp = *vpp;
struct inode *ip = VTOI(vp);
struct ulfsmount *ump = VFSTOULFS(mp);
struct lfs *fs = ump->um_lfs;
int i;
ip->i_mode = lfs_dino_getmode(fs, ip->i_din);
ip->i_nlink = lfs_dino_getnlink(fs, ip->i_din);
ip->i_lfs_osize = ip->i_size = lfs_dino_getsize(fs, ip->i_din);
ip->i_flags = lfs_dino_getflags(fs, ip->i_din);
ip->i_gen = lfs_dino_getgen(fs, ip->i_din);
ip->i_uid = lfs_dino_getuid(fs, ip->i_din);
ip->i_gid = lfs_dino_getgid(fs, ip->i_din);
ip->i_lfs_effnblks = lfs_dino_getblocks(fs, ip->i_din);
ip->i_lfs_odnlink = lfs_dino_getnlink(fs, ip->i_din);
ulfs_vinit(mp, lfs_specop_p, lfs_fifoop_p, &vp);
ip = VTOI(vp);
memset(ip->i_lfs_fragsize, 0, ULFS_NDADDR * sizeof(*ip->i_lfs_fragsize));
if (vp->v_type != VLNK || ip->i_size >= ip->i_lfs->um_maxsymlinklen) {
#ifdef DEBUG
for (i = (ip->i_size + lfs_sb_getbsize(fs) - 1) >> lfs_sb_getbshift(fs);
i < ULFS_NDADDR; i++) {
if ((vp->v_type == VBLK || vp->v_type == VCHR) &&
i == 0)
continue;
if (lfs_dino_getdb(fs, ip->i_din, i) != 0) {
lfs_dump_dinode(fs, ip->i_din);
panic("inconsistent inode (direct)");
}
}
for ( ; i < ULFS_NDADDR + ULFS_NIADDR; i++) {
if (lfs_dino_getib(fs, ip->i_din, i - ULFS_NDADDR) != 0) {
lfs_dump_dinode(fs, ip->i_din);
panic("inconsistent inode (indirect)");
}
}
#endif
for (i = 0; i < ULFS_NDADDR; i++)
if (lfs_dino_getdb(fs, ip->i_din, i) != 0)
ip->i_lfs_fragsize[i] = lfs_blksize(fs, ip, i);
}
KASSERTMSG((vp->v_type != VNON),
"lfs_vinit: ino %llu is type VNON! (ifmt=%o)\n",
(unsigned long long)ip->i_number,
(ip->i_mode & LFS_IFMT) >> 12);
ip->i_devvp = fs->lfs_devvp;
vref(ip->i_devvp);
#if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
ulfsquota_init(ip);
#endif
genfs_node_init(vp, &lfs_genfsops);
uvm_vnp_setsize(vp, ip->i_size);
ip->i_lfs_hiblk = lfs_lblkno(ip->i_lfs, ip->i_size + lfs_sb_getbsize(ip->i_lfs) - 1) - 1;
*vpp = vp;
}
int
lfs_resize_fs(struct lfs *fs, int newnsegs)
{
SEGUSE *sup;
CLEANERINFO *cip;
struct buf *bp, *obp;
daddr_t olast, nlast, ilast, noff, start, end, dirtysums;
struct vnode *ivp;
struct inode *ip;
int error, badnews, inc, oldnsegs;
int sbbytes, csbbytes, gain, cgain;
int i, nexti;
if (lfs_sb_getversion(fs) < 2)
return EOPNOTSUPP;
oldnsegs = lfs_sb_getnseg(fs);
if (newnsegs == oldnsegs)
return 0;
if (newnsegs <= lfs_dtosn(fs, lfs_sb_getsboff(fs, 1)))
return EINVAL;
ivp = fs->lfs_ivnode;
ip = VTOI(ivp);
error = 0;
lfs_seglock(fs, SEGM_CKP);
if (newnsegs < oldnsegs) {
lfs_rewind(fs, newnsegs);
for (i = newnsegs; i < oldnsegs; i++) {
uint32_t bytes;
LFS_SEGENTRY(sup, fs, i, bp);
bytes = sup->su_nbytes;
brelse(bp, 0);
if (bytes > 0) {
if ((error = lfs_rewrite_segment(fs, i, NULL, NOCRED, curlwp)) != 0)
goto out;
}
lfs_invalidate(fs, i);
}
}
sbbytes = csbbytes = cgain = dirtysums = 0;
for (i = newnsegs; i < oldnsegs; i++) {
LFS_SEGENTRY(sup, fs, i, bp);
badnews = sup->su_nbytes || !(sup->su_flags & SEGUSE_INVAL);
if (sup->su_flags & SEGUSE_SUPERBLOCK)
sbbytes += LFS_SBPAD;
if (!(sup->su_flags & SEGUSE_DIRTY)) {
++cgain;
if (sup->su_flags & SEGUSE_SUPERBLOCK)
csbbytes += LFS_SBPAD;
} else
dirtysums += sup->su_nsums;
brelse(bp, 0);
if (badnews) {
error = EBUSY;
goto out;
}
}
olast = lfs_sb_getcleansz(fs) + lfs_sb_getsegtabsz(fs);
nlast = howmany(newnsegs, lfs_sb_getsepb(fs)) + lfs_sb_getcleansz(fs);
ilast = ivp->v_size >> lfs_sb_getbshift(fs);
noff = nlast - olast;
rw_enter(&fs->lfs_iflock, RW_WRITER);
for (i = 0; i < ilast; i++) {
bread(ivp, i, lfs_sb_getbsize(fs), 0, &bp);
brelse(bp, 0);
}
for (i = ilast; i < ilast + noff; i++) {
if (lfs_balloc(ivp, i * lfs_sb_getbsize(fs),
lfs_sb_getbsize(fs), NOCRED, 0,
&bp) != 0)
panic("balloc extending ifile");
memset(bp->b_data, 0, lfs_sb_getbsize(fs));
VOP_BWRITE(bp->b_vp, bp);
}
if (noff > 0) {
ip->i_size += noff * lfs_sb_getbsize(fs);
lfs_dino_setsize(fs, ip->i_din, ip->i_size);
uvm_vnp_setsize(ivp, ip->i_size);
}
if (noff != 0) {
if (noff < 0) {
start = nlast;
end = ilast + noff;
inc = 1;
} else {
start = ilast + noff - 1;
end = nlast - 1;
inc = -1;
}
for (i = start; i != end; i += inc) {
if (bread(ivp, i, lfs_sb_getbsize(fs),
B_MODIFY, &bp) != 0)
panic("resize: bread dst blk failed");
if (bread(ivp, i - noff, lfs_sb_getbsize(fs),
0, &obp))
panic("resize: bread src blk failed");
memcpy(bp->b_data, obp->b_data, lfs_sb_getbsize(fs));
VOP_BWRITE(bp->b_vp, bp);
brelse(obp, 0);
}
}
for (i = oldnsegs, nexti = 0; i < newnsegs; i = nexti) {
nexti = roundup(i + 1, lfs_sb_getsepb(fs));
if (nexti > newnsegs)
nexti = newnsegs;
if ((error = bread(ivp, i / lfs_sb_getsepb(fs) +
lfs_sb_getcleansz(fs),
lfs_sb_getbsize(fs),
(i > oldnsegs ? B_MODIFY : 0), &bp)) != 0)
panic("lfs: ifile read: %d", error);
sup = &((SEGUSE *)bp->b_data)[i % lfs_sb_getsepb(fs)];
memset(sup, 0, (nexti - i) * sizeof(SEGUSE));
VOP_BWRITE(bp->b_vp, bp);
}
for (i = 2; i < LFS_MAXNUMSB; i++)
if (lfs_dtosn(fs, lfs_sb_getsboff(fs, i)) >= newnsegs)
lfs_sb_setsboff(fs, i, 0x0);
gain = (newnsegs - oldnsegs);
lfs_sb_setnseg(fs, newnsegs);
lfs_sb_setsegtabsz(fs, nlast - lfs_sb_getcleansz(fs));
lfs_sb_addsize(fs, gain * lfs_btofsb(fs, lfs_sb_getssize(fs)));
lfs_sb_adddsize(fs, gain * lfs_btofsb(fs, lfs_sb_getssize(fs))
+ lfs_btofsb(fs, sbbytes));
lfs_sb_addbfree(fs, gain * lfs_btofsb(fs, lfs_sb_getssize(fs))
+ lfs_btofsb(fs, sbbytes)
#if 0
- gain * lfs_btofsb(fs, lfs_sb_getbsize(fs) / 2)
+ dirtysums * lfs_btofsb(fs, lfs_sb_getsumsize(fs))
#endif
);
if (gain > 0) {
lfs_sb_addnclean(fs, gain);
lfs_sb_addavail(fs, gain * lfs_btofsb(fs, lfs_sb_getssize(fs)));
} else {
lfs_sb_subnclean(fs, cgain);
lfs_sb_subavail(fs, cgain * lfs_btofsb(fs, lfs_sb_getssize(fs)) -
lfs_btofsb(fs, csbbytes));
}
if (noff < 0) {
vn_lock(ivp, LK_EXCLUSIVE);
lfs_truncate(ivp, ivp->v_size + (noff << lfs_sb_getbshift(fs)), 0,
NOCRED);
VOP_UNLOCK(ivp);
}
bread(ivp, 0, lfs_sb_getbsize(fs), B_MODIFY, &bp);
cip = bp->b_data;
lfs_ci_setclean(fs, cip, lfs_sb_getnclean(fs));
lfs_ci_setdirty(fs, cip, lfs_sb_getnseg(fs) - lfs_sb_getnclean(fs));
VOP_BWRITE(bp->b_vp, bp);
rw_exit(&fs->lfs_iflock);
out:
lfs_segunlock(fs);
return error;
}
int
lfs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
int attrnamespace, const char *attrname)
{
#ifdef LFS_EXTATTR
struct ulfsmount *ump;
ump = VFSTOULFS(mp);
if (ump->um_fstype == ULFS1) {
return ulfs_extattrctl(mp, cmd, vp, attrnamespace, attrname);
}
#endif
return vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname);
}
