/* object has an underlying thing to poll */
if (realfp->f_op && realfp->f_op->poll) {
mask = (*realfp->f_op->poll)(realfp, pt_p);
}
}
return mask;
}
#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,36)
long
vnode_fop_ioctl(
FILE_T *file_p,
uint cmd,
ulong arg
)
#else /* LINUX_VERSION_CODE > KERNEL_VERSION(2,6,36) */
int
vnode_fop_ioctl(
INODE_T *ino_p,
FILE_T *file_p,
uint cmd,
ulong arg
)
#endif /* LINUX_VERSION_CODE > KERNEL_VERSION(2,6,36) */
{
int err;
int rval; /* unused */
CALL_DATA_T cd;
struct ioctl_ctx ctx;
ASSERT_KERNEL_LOCKED();
mdki_linux_init_call_data(&cd);
ctx.filp = file_p;
ctx.caller_is_32bit = 0; /* unknown as of yet */
#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,36)
err = VOP_IOCTL(ITOV(file_p->f_path.dentry->d_inode), cmd, (void *)arg, 0, &cd, &rval, NULL, &ctx);
#else
err = VOP_IOCTL(ITOV(ino_p), cmd, (void *)arg, 0, &cd, &rval, NULL, &ctx);
#endif
err = mdki_errno_unix_to_linux(err);
mdki_linux_destroy_call_data(&cd);
return err;
}
static int
unionfs_ioctl(void *v)
{
struct vop_ioctl_args *ap = v;
int error;
struct unionfs_node *unp;
struct unionfs_node_status *unsp;
struct vnode *ovp;
UNIONFS_INTERNAL_DEBUG("unionfs_ioctl: enter\n");
vn_lock(ap->a_vp, LK_EXCLUSIVE | LK_RETRY);
unp = VTOUNIONFS(ap->a_vp);
unionfs_get_node_status(unp, &unsp);
ovp = (unsp->uns_upper_opencnt ? unp->un_uppervp : unp->un_lowervp);
unionfs_tryrem_node_status(unp, unsp);
VOP_UNLOCK(ap->a_vp);
if (ovp == NULLVP)
return (EBADF);
error = VOP_IOCTL(ovp, ap->a_command, ap->a_data, ap->a_fflag,
ap->a_cred);
UNIONFS_INTERNAL_DEBUG("unionfs_ioctl: lease (%d)\n", error);
return (error);
}
int
wapbl_getdisksize(struct vnode *vp, uint64_t *numsecp, unsigned int *secsizep)
{
struct partinfo dpart;
unsigned int secsize;
uint64_t numsec;
int error;
error = VOP_IOCTL(vp, DIOCGPART, &dpart, FREAD, NOCRED);
if (error == 0) {
secsize = dpart.disklab->d_secsize;
numsec = dpart.part->p_size;
}
if (error == 0 &&
(secsize == 0 || secsize > MAXBSIZE || !powerof2(secsize) ||
numsec == 0)) {
#ifdef DIAGNOSTIC
printf("%s: vnode %p returns invalid disksize values"
" (secsize = %u, numsec = %llu)\n", __func__, vp,
secsize, numsec);
#endif
error = EINVAL;
}
if (error == 0) {
*secsizep = secsize;
*numsecp = numsec;
}
return error;
}
static void
udf_discstrat_finish_seq(struct udf_strat_args *args)
{
struct udf_mount *ump = args->ump;
struct strat_private *priv = PRIV(ump);
int error;
if (ump == NULL)
return;
/* stop our sheduling thread */
KASSERT(priv->run_thread == 1);
priv->run_thread = 0;
wakeup(priv->queue_lwp);
do {
error = tsleep(&priv->run_thread, PRIBIO+1,
"udfshedfin", hz);
} while (error);
/* kthread should be finished now */
/* set back old device strategy method */
VOP_IOCTL(ump->devvp, DIOCSSTRATEGY, &priv->old_strategy_setting,
FWRITE, NOCRED);
/* destroy our pool */
pool_destroy(&priv->desc_pool);
mutex_destroy(&priv->discstrat_mutex);
cv_destroy(&priv->discstrat_cv);
/* free our private space */
free(ump->strategy_private, M_UDFTEMP);
ump->strategy_private = NULL;
}
static int
cgdioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l)
{
struct cgd_softc *cs;
struct dk_softc *dksc;
int part = DISKPART(dev);
int pmask = 1 << part;
DPRINTF_FOLLOW(("cgdioctl(0x%"PRIx64", %ld, %p, %d, %p)\n",
dev, cmd, data, flag, l));
switch (cmd) {
case CGDIOCGET:
return cgd_ioctl_get(dev, data, l);
case CGDIOCSET:
case CGDIOCCLR:
if ((flag & FWRITE) == 0)
return EBADF;
/* FALLTHROUGH */
default:
GETCGD_SOFTC(cs, dev);
dksc = &cs->sc_dksc;
break;
}
switch (cmd) {
case CGDIOCSET:
if (DK_ATTACHED(dksc))
return EBUSY;
return cgd_ioctl_set(cs, data, l);
case CGDIOCCLR:
if (DK_BUSY(&cs->sc_dksc, pmask))
return EBUSY;
return cgd_ioctl_clr(cs, l);
case DIOCCACHESYNC:
/*
* XXX Do we really need to care about having a writable
* file descriptor here?
*/
if ((flag & FWRITE) == 0)
return (EBADF);
/*
* We pass this call down to the underlying disk.
*/
return VOP_IOCTL(cs->sc_tvn, cmd, data, flag, l->l_cred);
case DIOCGSTRATEGY:
case DIOCSSTRATEGY:
if (!DK_ATTACHED(dksc))
return ENOENT;
/*FALLTHROUGH*/
default:
return dk_ioctl(dksc, dev, cmd, data, flag, l);
case CGDIOCGET:
KASSERT(0);
return EINVAL;
}
}
/*
* Sync underlying disk caches.
*/
int
dm_target_snapshot_orig_sync(dm_table_entry_t * table_en)
{
int cmd;
dm_target_snapshot_origin_config_t *tsoc;
tsoc = table_en->target_config;
cmd = 1;
return VOP_IOCTL(tsoc->tsoc_real_dev->pdev_vnode, DIOCCACHESYNC, &cmd, FREAD|FWRITE, kauth_cred_get());
}
STATIC void
linvfs_unfreeze_fs(
struct super_block *sb)
{
vfs_t *vfsp = LINVFS_GET_VFS(sb);
vnode_t *vp;
int error;
VFS_ROOT(vfsp, &vp, error);
VOP_IOCTL(vp, LINVFS_GET_IP(vp), NULL, 0, XFS_IOC_THAW, 0, error);
VN_RELE(vp);
}
/*
* File table vnode ioctl routine.
*/
int
vn_ioctl(struct file *fp, u_long com, caddr_t data, struct proc *p)
{
struct vnode *vp = ((struct vnode *)fp->f_data);
struct vattr vattr;
int error;
switch (vp->v_type) {
case VREG:
case VDIR:
if (com == FIONREAD) {
error = VOP_GETATTR(vp, &vattr, p->p_ucred, p);
if (error)
return (error);
*(int *)data = vattr.va_size - fp->f_offset;
return (0);
}
if (com == FIBMAP)
return VOP_IOCTL(vp, com, data, fp->f_flag,
p->p_ucred, p);
if (com == FIONBIO || com == FIOASYNC) /* XXX */
return (0); /* XXX */
/* FALLTHROUGH */
default:
return (ENOTTY);
case VFIFO:
case VCHR:
case VBLK:
error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p);
if (error == 0 && com == TIOCSCTTY) {
if (p->p_session->s_ttyvp)
vrele(p->p_session->s_ttyvp);
p->p_session->s_ttyvp = vp;
VREF(vp);
}
return (error);
}
}
/*
* Helper function for findroot():
* Return non-zero if disk device matches bootinfo.
*/
static int
match_bootdisk(device_t dv, struct btinfo_bootdisk *bid)
{
struct vnode *tmpvn;
int error;
struct disklabel label;
int found = 0;
if (device_is_a(dv, "dk")) {
DPRINTF(("%s: dk %s\n", __func__, device_xname(dv)));
return 0;
}
/*
* A disklabel is required here. The boot loader doesn't refuse
* to boot from a disk without a label, but this is normally not
* wanted.
*/
if (bid->labelsector == -1) {
DPRINTF(("%s: no label %s\n", __func__, device_xname(dv)));
return 0;
}
if ((tmpvn = opendisk(dv)) == NULL) {
DPRINTF(("%s: can't open %s\n", __func__, device_xname(dv)));
return 0;
}
error = VOP_IOCTL(tmpvn, DIOCGDINFO, &label, FREAD, NOCRED);
if (error) {
/*
* XXX Can't happen -- open() would have errored out
* or faked one up.
*/
printf("%s: can't get label for dev %s (%d)\n", __func__,
device_xname(dv), error);
goto closeout;
}
/* Compare with our data. */
if (label.d_type == bid->label.type &&
label.d_checksum == bid->label.checksum &&
strncmp(label.d_packname, bid->label.packname, 16) == 0)
found = 1;
DPRINTF(("%s: %s found=%d\n", __func__, device_xname(dv), found));
closeout:
VOP_CLOSE(tmpvn, FREAD, NOCRED);
vput(tmpvn);
return (found);
}
STATIC void
linvfs_freeze_fs(
struct super_block *sb)
{
vfs_t *vfsp = LINVFS_GET_VFS(sb);
vnode_t *vp;
int error;
if (sb->s_flags & MS_RDONLY)
return;
VFS_ROOT(vfsp, &vp, error);
VOP_IOCTL(vp, LINVFS_GET_IP(vp), NULL, 0, XFS_IOC_FREEZE, 0, error);
VN_RELE(vp);
}
int
RUMP_VOP_IOCTL(struct vnode *vp,
u_long command,
void *data,
int fflag,
struct kauth_cred *cred)
{
int error;
rump_schedule();
error = VOP_IOCTL(vp, command, data, fflag, cred);
rump_unschedule();
return error;
}
/*ARGSUSED*/
int
cttyioctl(dev_t dev, u_long cmd, caddr_t addr, int flag, struct proc *p)
{
struct vnode *ttyvp = cttyvp(p);
if (ttyvp == NULL)
return (EIO);
if (cmd == TIOCSCTTY) /* XXX */
return (EINVAL);
if (cmd == TIOCNOTTY) {
if (!SESS_LEADER(p)) {
atomic_clearbits_int(&p->p_flag, P_CONTROLT);
return (0);
} else
return (EINVAL);
}
return (VOP_IOCTL(ttyvp, cmd, addr, flag, NOCRED, p));
}
/*ARGSUSED*/
static int
cttyioctl(struct dev_ioctl_args *ap)
{
struct vnode *ttyvp;
struct proc *p = curproc;
KKASSERT(p);
lwkt_gettoken(&p->p_token);
lwkt_gettoken(&proc_token);
ttyvp = cttyvp(p);
if (ttyvp == NULL) {
lwkt_reltoken(&proc_token);
lwkt_reltoken(&p->p_token);
return (EIO);
}
/*
* Don't allow controlling tty to be set to the controlling tty
* (infinite recursion).
*/
if (ap->a_cmd == TIOCSCTTY) {
lwkt_reltoken(&proc_token);
lwkt_reltoken(&p->p_token);
return EINVAL;
}
if (ap->a_cmd == TIOCNOTTY) {
if (!SESS_LEADER(p)) {
p->p_flags &= ~P_CONTROLT;
lwkt_reltoken(&proc_token);
lwkt_reltoken(&p->p_token);
return (0);
} else {
lwkt_reltoken(&proc_token);
lwkt_reltoken(&p->p_token);
return (EINVAL);
}
}
lwkt_reltoken(&proc_token);
lwkt_reltoken(&p->p_token);
return (VOP_IOCTL(ttyvp, ap->a_cmd, ap->a_data, ap->a_fflag,
ap->a_cred, ap->a_sysmsg));
}
STATIC long
xfs_file_ioctl_invis(
struct file *filp,
unsigned int cmd,
unsigned long arg)
{
struct inode *inode = filp->f_dentry->d_inode;
vnode_t *vp = vn_from_inode(inode);
int error;
VOP_IOCTL(vp, inode, filp, IO_INVIS, cmd, (void __user *)arg, error);
VMODIFY(vp);
/* NOTE: some of the ioctl's return positive #'s as a
* byte count indicating success, such as
* readlink_by_handle. So we don't "sign flip"
* like most other routines. This means true
* errors need to be returned as a negative value.
*/
return error;
}
STATIC int linvfs_ioctl(
struct inode *inode,
struct file *filp,
unsigned int cmd,
unsigned long arg)
{
int error;
vnode_t *vp = LINVFS_GET_VP(inode);
ASSERT(vp);
VOP_IOCTL(vp, inode, filp, cmd, arg, error);
/* NOTE: some of the ioctl's return positive #'s as a
* byte count indicating success, such as
* readlink_by_handle. So we don't "sign flip"
* like most other routines. This means true
* errors need to be returned as a negative value.
*/
return error;
}
/*ARGSUSED*/
static int
cttyioctl(dev_t dev, u_long cmd, void *addr, int flag, struct lwp *l)
{
struct vnode *ttyvp = cttyvp(l->l_proc);
int rv;
if (ttyvp == NULL)
return (EIO);
if (cmd == TIOCSCTTY) /* XXX */
return (EINVAL);
if (cmd == TIOCNOTTY) {
mutex_enter(proc_lock);
if (!SESS_LEADER(l->l_proc)) {
l->l_proc->p_lflag &= ~PL_CONTROLT;
rv = 0;
} else
rv = EINVAL;
mutex_exit(proc_lock);
return (rv);
}
return (VOP_IOCTL(ttyvp, cmd, addr, flag, NOCRED));
}
/*
* Common code for mount and mountroot
*/
static int
iso_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l,
struct iso_args *argp)
{
struct iso_mnt *isomp = (struct iso_mnt *)0;
struct buf *bp = NULL, *pribp = NULL, *supbp = NULL;
dev_t dev = devvp->v_rdev;
int error = EINVAL;
int ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
int iso_bsize;
int iso_blknum;
int joliet_level;
struct iso_volume_descriptor *vdp;
struct iso_supplementary_descriptor *sup;
int sess = 0;
int ext_attr_length;
struct disklabel label;
if (!ronly)
return EROFS;
/* Flush out any old buffers remaining from a previous use. */
if ((error = vinvalbuf(devvp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
return (error);
/* This is the "logical sector size". The standard says this
* should be 2048 or the physical sector size on the device,
* whichever is greater. For now, we'll just use a constant.
*/
iso_bsize = ISO_DEFAULT_BLOCK_SIZE;
error = VOP_IOCTL(devvp, DIOCGDINFO, &label, FREAD, FSCRED);
if (!error) {
/* XXX more sanity checks? */
sess = label.d_partitions[DISKPART(dev)].p_cdsession;
} else {
/* fallback to old method */
error = VOP_IOCTL(devvp, CDIOREADMSADDR, &sess, 0, FSCRED);
if (error)
sess = 0; /* never mind */
}
#ifdef ISO_DEBUG
printf("isofs: session offset (part %"PRId32") %d\n", DISKPART(dev), sess);
#endif
for (iso_blknum = 16; iso_blknum < 100; iso_blknum++) {
if ((error = bread(devvp, (iso_blknum+sess) * btodb(iso_bsize),
iso_bsize, NOCRED, 0, &bp)) != 0)
goto out;
vdp = (struct iso_volume_descriptor *)bp->b_data;
if (memcmp(vdp->id, ISO_STANDARD_ID, sizeof(vdp->id)) != 0) {
error = EINVAL;
goto out;
}
switch (isonum_711(vdp->type)) {
case ISO_VD_PRIMARY:
if (pribp == NULL) {
pribp = bp;
bp = NULL;
}
break;
case ISO_VD_SUPPLEMENTARY:
if (supbp == NULL) {
supbp = bp;
bp = NULL;
}
break;
default:
break;
}
if (isonum_711 (vdp->type) == ISO_VD_END) {
brelse(bp, 0);
bp = NULL;
break;
}
if (bp != NULL) {
brelse(bp, 0);
bp = NULL;
}
}
if (pribp == NULL) {
error = EINVAL;
goto out;
}
isomp = malloc(sizeof *isomp, M_ISOFSMNT, M_WAITOK);
memset(isomp, 0, sizeof *isomp);
if (iso_makemp(isomp, pribp, &ext_attr_length) == -1) {
error = EINVAL;
goto out;
}
isomp->volume_space_size += sess;
brelse(pribp, BC_AGE);
pribp = NULL;
mp->mnt_data = isomp;
mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_CD9660);
mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
mp->mnt_stat.f_namemax = ISO_MAXNAMLEN;
mp->mnt_flag |= MNT_LOCAL;
mp->mnt_iflag |= IMNT_MPSAFE;
mp->mnt_dev_bshift = iso_bsize;
mp->mnt_fs_bshift = isomp->im_bshift;
isomp->im_mountp = mp;
isomp->im_dev = dev;
isomp->im_devvp = devvp;
/* Check the Rock Ridge Extension support */
if (!(argp->flags & ISOFSMNT_NORRIP)) {
struct iso_directory_record *rootp;
if ((error = bread(isomp->im_devvp,
(isomp->root_extent + ext_attr_length) <<
(isomp->im_bshift - DEV_BSHIFT),
isomp->logical_block_size, NOCRED,
0, &bp)) != 0)
goto out;
rootp = (struct iso_directory_record *)bp->b_data;
if ((isomp->rr_skip = cd9660_rrip_offset(rootp,isomp)) < 0) {
argp->flags |= ISOFSMNT_NORRIP;
} else {
argp->flags &= ~ISOFSMNT_GENS;
}
/*
* The contents are valid,
* but they will get reread as part of another vnode, so...
*/
brelse(bp, BC_AGE);
bp = NULL;
}
isomp->im_flags = argp->flags & (ISOFSMNT_NORRIP | ISOFSMNT_GENS |
ISOFSMNT_EXTATT | ISOFSMNT_NOJOLIET | ISOFSMNT_RRCASEINS);
if (isomp->im_flags & ISOFSMNT_GENS)
isomp->iso_ftype = ISO_FTYPE_9660;
else if (isomp->im_flags & ISOFSMNT_NORRIP) {
isomp->iso_ftype = ISO_FTYPE_DEFAULT;
if (argp->flags & ISOFSMNT_NOCASETRANS)
isomp->im_flags |= ISOFSMNT_NOCASETRANS;
} else
isomp->iso_ftype = ISO_FTYPE_RRIP;
/* Check the Joliet Extension support */
if ((argp->flags & ISOFSMNT_NORRIP) != 0 &&
(argp->flags & ISOFSMNT_NOJOLIET) == 0 &&
supbp != NULL) {
joliet_level = 0;
sup = (struct iso_supplementary_descriptor *)supbp->b_data;
if ((isonum_711(sup->flags) & 1) == 0) {
if (memcmp(sup->escape, "%/@", 3) == 0)
joliet_level = 1;
if (memcmp(sup->escape, "%/C", 3) == 0)
joliet_level = 2;
if (memcmp(sup->escape, "%/E", 3) == 0)
joliet_level = 3;
}
if (joliet_level != 0) {
if (iso_makemp(isomp, supbp, NULL) == -1) {
error = EINVAL;
goto out;
}
isomp->im_joliet_level = joliet_level;
}
}
if (supbp != NULL) {
brelse(supbp, 0);
supbp = NULL;
}
spec_node_setmountedfs(devvp, mp);
return 0;
out:
if (bp)
brelse(bp, 0);
if (pribp)
brelse(pribp, 0);
if (supbp)
brelse(supbp, 0);
if (isomp) {
free(isomp, M_ISOFSMNT);
mp->mnt_data = NULL;
}
return error;
}
/* Configure a single disk in the array. */
int
rf_ConfigureDisk(RF_Raid_t *raidPtr, char *buf, RF_RaidDisk_t *diskPtr,
RF_RowCol_t row, RF_RowCol_t col)
{
char *p;
int retcode;
struct partinfo dpart;
struct vnode *vp;
struct vattr va;
struct proc *proc;
int error;
retcode = 0;
p = rf_find_non_white(buf);
if (*buf != '\0' && p[strlen(p) - 1] == '\n') {
/* Strip off the newline. */
p[strlen(p) - 1] = '\0';
}
(void) strlcpy(diskPtr->devname, p, sizeof diskPtr->devname);
proc = raidPtr->engine_thread;
/* Let's start by claiming the component is fine and well... */
diskPtr->status = rf_ds_optimal;
raidPtr->raid_cinfo[row][col].ci_vp = NULL;
raidPtr->raid_cinfo[row][col].ci_dev = 0;
error = raidlookup(diskPtr->devname, curproc, &vp);
if (error) {
printf("raidlookup on device: %s failed !\n", diskPtr->devname);
if (error == ENXIO) {
/* The component isn't there... Must be dead :-( */
diskPtr->status = rf_ds_failed;
} else {
return (error);
}
}
if (diskPtr->status == rf_ds_optimal) {
if ((error = VOP_GETATTR(vp, &va, proc->p_ucred, proc)) != 0) {
return (error);
}
error = VOP_IOCTL(vp, DIOCGPART, (caddr_t) & dpart, FREAD,
proc->p_ucred, proc);
if (error) {
return (error);
}
diskPtr->blockSize = dpart.disklab->d_secsize;
diskPtr->numBlocks = DL_GETPSIZE(dpart.part) - rf_protectedSectors;
diskPtr->partitionSize = DL_GETPSIZE(dpart.part);
raidPtr->raid_cinfo[row][col].ci_vp = vp;
raidPtr->raid_cinfo[row][col].ci_dev = va.va_rdev;
/* This component was not automatically configured. */
diskPtr->auto_configured = 0;
diskPtr->dev = va.va_rdev;
/*
* We allow the user to specify that only a fraction of the
* disks should be used. This is just for debug: it speeds up
* the parity scan.
*/
diskPtr->numBlocks = diskPtr->numBlocks * rf_sizePercentage
/ 100;
}
return (0);
}
/* Do a complete copyback. */
void
rf_CopybackReconstructedData(RF_Raid_t *raidPtr)
{
RF_ComponentLabel_t c_label;
int done, retcode;
RF_CopybackDesc_t *desc;
RF_RowCol_t frow, fcol;
RF_RaidDisk_t *badDisk;
char *databuf;
struct partinfo dpart;
struct vnode *vp;
struct vattr va;
struct proc *proc;
int ac;
done = 0;
fcol = 0;
for (frow = 0; frow < raidPtr->numRow; frow++) {
for (fcol = 0; fcol < raidPtr->numCol; fcol++) {
if (raidPtr->Disks[frow][fcol].status ==
rf_ds_dist_spared ||
raidPtr->Disks[frow][fcol].status ==
rf_ds_spared) {
done = 1;
break;
}
}
if (done)
break;
}
if (frow == raidPtr->numRow) {
printf("COPYBACK: No disks need copyback.\n");
return;
}
badDisk = &raidPtr->Disks[frow][fcol];
proc = raidPtr->engine_thread;
/*
* This device may have been opened successfully the first time.
* Close it before trying to open it again.
*/
if (raidPtr->raid_cinfo[frow][fcol].ci_vp != NULL) {
printf("Close the opened device: %s.\n",
raidPtr->Disks[frow][fcol].devname);
vp = raidPtr->raid_cinfo[frow][fcol].ci_vp;
ac = raidPtr->Disks[frow][fcol].auto_configured;
rf_close_component(raidPtr, vp, ac);
raidPtr->raid_cinfo[frow][fcol].ci_vp = NULL;
}
/* Note that this disk was *not* auto_configured (any longer). */
raidPtr->Disks[frow][fcol].auto_configured = 0;
printf("About to (re-)open the device: %s.\n",
raidPtr->Disks[frow][fcol].devname);
retcode = raidlookup(raidPtr->Disks[frow][fcol].devname, proc, &vp);
if (retcode) {
printf("COPYBACK: raidlookup on device: %s failed: %d !\n",
raidPtr->Disks[frow][fcol].devname, retcode);
/*
* XXX The component isn't responding properly... Must be
* still dead :-(
*/
return;
} else {
/*
* Ok, so we can at least do a lookup...
* How about actually getting a vp for it ?
*/
if ((retcode = VOP_GETATTR(vp, &va, proc->p_ucred, proc)) != 0)
{
return;
}
retcode = VOP_IOCTL(vp, DIOCGPART, (caddr_t) &dpart, FREAD,
proc->p_ucred, proc);
if (retcode) {
return;
}
raidPtr->Disks[frow][fcol].blockSize = dpart.disklab->d_secsize;
raidPtr->Disks[frow][fcol].numBlocks = dpart.part->p_size -
rf_protectedSectors;
raidPtr->raid_cinfo[frow][fcol].ci_vp = vp;
raidPtr->raid_cinfo[frow][fcol].ci_dev = va.va_rdev;
/* XXX Or the above ? */
raidPtr->Disks[frow][fcol].dev = va.va_rdev;
/*
* We allow the user to specify that only a fraction of the
* disks should be used this is just for debug: it speeds up
* the parity scan.
*/
raidPtr->Disks[frow][fcol].numBlocks =
raidPtr->Disks[frow][fcol].numBlocks *
rf_sizePercentage / 100;
}
#if 0
/* This is the way it was done before the CAM stuff was removed. */
if (rf_extract_ids(badDisk->devname, &bus, &targ, &lun)) {
printf("COPYBACK: unable to extract bus, target, lun from"
" devname %s.\n", badDisk->devname);
return;
}
/*
* TUR the disk that's marked as bad to be sure that it's actually
* alive.
*/
rf_SCSI_AllocTUR(&tur_op);
retcode = rf_SCSI_DoTUR(tur_op, bus, targ, lun, badDisk->dev);
rf_SCSI_FreeDiskOp(tur_op, 0);
#endif
if (retcode) {
printf("COPYBACK: target disk failed TUR.\n");
return;
}
/* Get a buffer to hold one SU. */
RF_Malloc(databuf, rf_RaidAddressToByte(raidPtr,
raidPtr->Layout.sectorsPerStripeUnit), (char *));
/* Create a descriptor. */
RF_Malloc(desc, sizeof(*desc), (RF_CopybackDesc_t *));
desc->raidPtr = raidPtr;
desc->status = 0;
desc->frow = frow;
desc->fcol = fcol;
desc->spRow = badDisk->spareRow;
desc->spCol = badDisk->spareCol;
desc->stripeAddr = 0;
desc->sectPerSU = raidPtr->Layout.sectorsPerStripeUnit;
desc->sectPerStripe = raidPtr->Layout.sectorsPerStripeUnit *
raidPtr->Layout.numDataCol;
desc->databuf = databuf;
desc->mcpair = rf_AllocMCPair();
printf("COPYBACK: Quiescing the array.\n");
/*
* Quiesce the array, since we don't want to code support for user
* accs here.
*/
rf_SuspendNewRequestsAndWait(raidPtr);
/* Adjust state of the array and of the disks. */
RF_LOCK_MUTEX(raidPtr->mutex);
raidPtr->Disks[desc->frow][desc->fcol].status = rf_ds_optimal;
raidPtr->status[desc->frow] = rf_rs_optimal;
rf_copyback_in_progress = 1; /* Debug only. */
RF_UNLOCK_MUTEX(raidPtr->mutex);
printf("COPYBACK: Beginning\n");
RF_GETTIME(desc->starttime);
rf_ContinueCopyback(desc);
/*
* Data has been restored.
* Fix up the component label.
* Don't actually need the read here.
*/
raidread_component_label(raidPtr->raid_cinfo[frow][fcol].ci_dev,
raidPtr->raid_cinfo[frow][fcol].ci_vp,
&c_label);
raid_init_component_label(raidPtr, &c_label);
c_label.row = frow;
c_label.column = fcol;
raidwrite_component_label(raidPtr->raid_cinfo[frow][fcol].ci_dev,
raidPtr->raid_cinfo[frow][fcol].ci_vp,
&c_label);
}
/*
* Common code for mount and mountroot
*/
int
ffs_mountfs(struct vnode *devvp, struct mount *mp, struct proc *p)
{
struct ufsmount *ump;
struct buf *bp;
struct fs *fs;
dev_t dev;
struct partinfo dpart;
caddr_t space;
ufs2_daddr_t sbloc;
int error, i, blks, size, ronly;
int32_t *lp;
size_t strsize;
struct ucred *cred;
u_int64_t maxfilesize; /* XXX */
dev = devvp->v_rdev;
cred = p ? p->p_ucred : NOCRED;
/*
* Disallow multiple mounts of the same device.
* Disallow mounting of a device that is currently in use
* (except for root, which might share swap device for miniroot).
* Flush out any old buffers remaining from a previous use.
*/
if ((error = vfs_mountedon(devvp)) != 0)
return (error);
if (vcount(devvp) > 1 && devvp != rootvp)
return (EBUSY);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0);
VOP_UNLOCK(devvp, 0, p);
if (error)
return (error);
ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
if (error)
return (error);
if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, cred, p) != 0)
size = DEV_BSIZE;
else
size = dpart.disklab->d_secsize;
bp = NULL;
ump = NULL;
/*
* Try reading the super-block in each of its possible locations.
*/
for (i = 0; sbtry[i] != -1; i++) {
if (bp != NULL) {
bp->b_flags |= B_NOCACHE;
brelse(bp);
bp = NULL;
}
error = bread(devvp, sbtry[i] / size, SBSIZE, cred, &bp);
if (error)
goto out;
fs = (struct fs *) bp->b_data;
sbloc = sbtry[i];
#if 0
if (fs->fs_magic == FS_UFS2_MAGIC) {
printf("ffs_mountfs(): Sorry, no UFS2 support (yet)\n");
error = EFTYPE;
goto out;
}
#endif
/*
* Do not look for an FFS1 file system at SBLOCK_UFS2. Doing so
* will find the wrong super-block for file systems with 64k
* block size.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && sbloc == SBLOCK_UFS2)
continue;
if (ffs_validate(fs))
break; /* Super block validated */
}
if (sbtry[i] == -1) {
error = EINVAL;
goto out;
}
fs->fs_fmod = 0;
fs->fs_flags &= ~FS_UNCLEAN;
if (fs->fs_clean == 0) {
fs->fs_flags |= FS_UNCLEAN;
#if 0
/*
* It is safe mount unclean file system
* if it was previously mounted with softdep
* but we may loss space and must
* sometimes run fsck manually.
*/
if (fs->fs_flags & FS_DOSOFTDEP)
printf(
"WARNING: %s was not properly unmounted\n",
fs->fs_fsmnt);
else
#endif
if (ronly || (mp->mnt_flag & MNT_FORCE)) {
printf(
"WARNING: %s was not properly unmounted\n",
fs->fs_fsmnt);
} else {
printf(
"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
fs->fs_fsmnt);
error = EROFS;
goto out;
}
}
if (fs->fs_postblformat == FS_42POSTBLFMT && !ronly) {
#ifndef SMALL_KERNEL
printf("ffs_mountfs(): obsolete rotational table format, "
"please use fsck_ffs(8) -c 1\n");
#endif
error = EFTYPE;
goto out;
}
ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
bzero(ump, sizeof *ump);
ump->um_fs = malloc((u_long)fs->fs_sbsize, M_UFSMNT,
M_WAITOK);
if (fs->fs_magic == FS_UFS1_MAGIC)
ump->um_fstype = UM_UFS1;
#ifdef FFS2
else
ump->um_fstype = UM_UFS2;
#endif
bcopy(bp->b_data, ump->um_fs, (u_int)fs->fs_sbsize);
if (fs->fs_sbsize < SBSIZE)
bp->b_flags |= B_INVAL;
brelse(bp);
bp = NULL;
fs = ump->um_fs;
ffs1_compat_read(fs, ump, sbloc);
fs->fs_ronly = ronly;
size = fs->fs_cssize;
blks = howmany(size, fs->fs_fsize);
if (fs->fs_contigsumsize > 0)
size += fs->fs_ncg * sizeof(int32_t);
space = malloc((u_long)size, M_UFSMNT, M_WAITOK);
fs->fs_csp = (struct csum *)space;
for (i = 0; i < blks; i += fs->fs_frag) {
size = fs->fs_bsize;
if (i + fs->fs_frag > blks)
size = (blks - i) * fs->fs_fsize;
error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
cred, &bp);
if (error) {
free(fs->fs_csp, M_UFSMNT);
goto out;
}
bcopy(bp->b_data, space, (u_int)size);
space += size;
brelse(bp);
bp = NULL;
}
if (fs->fs_contigsumsize > 0) {
fs->fs_maxcluster = lp = (int32_t *)space;
for (i = 0; i < fs->fs_ncg; i++)
*lp++ = fs->fs_contigsumsize;
}
mp->mnt_data = (qaddr_t)ump;
mp->mnt_stat.f_fsid.val[0] = (long)dev;
/* Use on-disk fsid if it exists, else fake it */
if (fs->fs_id[0] != 0 && fs->fs_id[1] != 0)
mp->mnt_stat.f_fsid.val[1] = fs->fs_id[1];
else
mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
mp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
mp->mnt_flag |= MNT_LOCAL;
ump->um_mountp = mp;
ump->um_dev = dev;
ump->um_devvp = devvp;
ump->um_nindir = fs->fs_nindir;
ump->um_bptrtodb = fs->fs_fsbtodb;
ump->um_seqinc = fs->fs_frag;
for (i = 0; i < MAXQUOTAS; i++)
ump->um_quotas[i] = NULLVP;
devvp->v_specmountpoint = mp;
ffs_oldfscompat(fs);
if (ronly)
fs->fs_contigdirs = NULL;
else {
fs->fs_contigdirs = (u_int8_t*)malloc((u_long)fs->fs_ncg,
M_UFSMNT, M_WAITOK);
bzero(fs->fs_contigdirs, fs->fs_ncg);
}
/*
* Set FS local "last mounted on" information (NULL pad)
*/
copystr(mp->mnt_stat.f_mntonname, /* mount point*/
fs->fs_fsmnt, /* copy area*/
sizeof(fs->fs_fsmnt) - 1, /* max size*/
&strsize); /* real size*/
bzero(fs->fs_fsmnt + strsize, sizeof(fs->fs_fsmnt) - strsize);
#if 0
if( mp->mnt_flag & MNT_ROOTFS) {
/*
* Root mount; update timestamp in mount structure.
* this will be used by the common root mount code
* to update the system clock.
*/
mp->mnt_time = fs->fs_time;
}
#endif
/*
* XXX
* Limit max file size. Even though ffs can handle files up to 16TB,
* we do limit the max file to 2^31 pages to prevent overflow of
* a 32-bit unsigned int. The buffer cache has its own checks but
* a little added paranoia never hurts.
*/
ump->um_savedmaxfilesize = fs->fs_maxfilesize; /* XXX */
maxfilesize = (u_int64_t)0x80000000 * MIN(PAGE_SIZE, fs->fs_bsize) - 1;
if (fs->fs_maxfilesize > maxfilesize) /* XXX */
fs->fs_maxfilesize = maxfilesize; /* XXX */
if (ronly == 0) {
if ((fs->fs_flags & FS_DOSOFTDEP) &&
(error = softdep_mount(devvp, mp, fs, cred)) != 0) {
free(fs->fs_csp, M_UFSMNT);
free(fs->fs_contigdirs, M_UFSMNT);
goto out;
}
fs->fs_fmod = 1;
fs->fs_clean = 0;
if (mp->mnt_flag & MNT_SOFTDEP)
fs->fs_flags |= FS_DOSOFTDEP;
else
fs->fs_flags &= ~FS_DOSOFTDEP;
(void) ffs_sbupdate(ump, MNT_WAIT);
}
return (0);
out:
devvp->v_specmountpoint = NULL;
if (bp)
brelse(bp);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred, p);
VOP_UNLOCK(devvp, 0, p);
if (ump) {
free(ump->um_fs, M_UFSMNT);
free(ump, M_UFSMNT);
mp->mnt_data = (qaddr_t)0;
}
return (error);
}
/*
* Reload all incore data for a filesystem (used after running fsck on
* the root filesystem and finding things to fix). The filesystem must
* be mounted read-only.
*
* Things to do to update the mount:
* 1) invalidate all cached meta-data.
* 2) re-read superblock from disk.
* 3) re-read summary information from disk.
* 4) invalidate all inactive vnodes.
* 5) invalidate all cached file data.
* 6) re-read inode data for all active vnodes.
*/
int
ffs_reload(struct mount *mountp, struct ucred *cred, struct proc *p)
{
struct vnode *devvp;
caddr_t space;
struct fs *fs, *newfs;
struct partinfo dpart;
int i, blks, size, error;
int32_t *lp;
struct buf *bp = NULL;
struct ffs_reload_args fra;
if ((mountp->mnt_flag & MNT_RDONLY) == 0)
return (EINVAL);
/*
* Step 1: invalidate all cached meta-data.
*/
devvp = VFSTOUFS(mountp)->um_devvp;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
error = vinvalbuf(devvp, 0, cred, p, 0, 0);
VOP_UNLOCK(devvp, 0, p);
if (error)
panic("ffs_reload: dirty1");
/*
* Step 2: re-read superblock from disk.
*/
if (VOP_IOCTL(devvp, DIOCGPART, (caddr_t)&dpart, FREAD, NOCRED, p) != 0)
size = DEV_BSIZE;
else
size = dpart.disklab->d_secsize;
error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
newfs = (struct fs *)bp->b_data;
if (ffs_validate(newfs) == 0) {
brelse(bp);
return (EINVAL);
}
fs = VFSTOUFS(mountp)->um_fs;
/*
* Copy pointer fields back into superblock before copying in XXX
* new superblock. These should really be in the ufsmount. XXX
* Note that important parameters (eg fs_ncg) are unchanged.
*/
newfs->fs_csp = fs->fs_csp;
newfs->fs_maxcluster = fs->fs_maxcluster;
newfs->fs_ronly = fs->fs_ronly;
bcopy(newfs, fs, (u_int)fs->fs_sbsize);
if (fs->fs_sbsize < SBSIZE)
bp->b_flags |= B_INVAL;
brelse(bp);
mountp->mnt_maxsymlinklen = fs->fs_maxsymlinklen;
ffs1_compat_read(fs, VFSTOUFS(mountp), SBOFF);
ffs_oldfscompat(fs);
(void)ffs_statfs(mountp, &mountp->mnt_stat, p);
/*
* Step 3: re-read summary information from disk.
*/
blks = howmany(fs->fs_cssize, fs->fs_fsize);
space = (caddr_t)fs->fs_csp;
for (i = 0; i < blks; i += fs->fs_frag) {
size = fs->fs_bsize;
if (i + fs->fs_frag > blks)
size = (blks - i) * fs->fs_fsize;
error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), size,
NOCRED, &bp);
if (error) {
brelse(bp);
return (error);
}
bcopy(bp->b_data, space, (u_int)size);
space += size;
brelse(bp);
}
if ((fs->fs_flags & FS_DOSOFTDEP))
(void) softdep_mount(devvp, mountp, fs, cred);
/*
* We no longer know anything about clusters per cylinder group.
*/
if (fs->fs_contigsumsize > 0) {
lp = fs->fs_maxcluster;
for (i = 0; i < fs->fs_ncg; i++)
*lp++ = fs->fs_contigsumsize;
}
fra.p = p;
fra.cred = cred;
fra.fs = fs;
fra.devvp = devvp;
error = vfs_mount_foreach_vnode(mountp, ffs_reload_vnode, &fra);
return (error);
}
static void
udf_discstrat_init_seq(struct udf_strat_args *args)
{
struct udf_mount *ump = args->ump;
struct strat_private *priv = PRIV(ump);
struct disk_strategy dkstrat;
uint32_t lb_size;
KASSERT(ump);
KASSERT(ump->logical_vol);
KASSERT(priv == NULL);
lb_size = udf_rw32(ump->logical_vol->lb_size);
KASSERT(lb_size > 0);
/* initialise our memory space */
ump->strategy_private = malloc(sizeof(struct strat_private),
M_UDFTEMP, M_WAITOK);
priv = ump->strategy_private;
memset(priv, 0 , sizeof(struct strat_private));
/* initialise locks */
cv_init(&priv->discstrat_cv, "udfstrat");
mutex_init(&priv->discstrat_mutex, MUTEX_DEFAULT, IPL_NONE);
/*
* Initialise pool for descriptors associated with nodes. This is done
* in lb_size units though currently lb_size is dictated to be
* sector_size.
*/
pool_init(&priv->desc_pool, lb_size, 0, 0, 0, "udf_desc_pool", NULL,
IPL_NONE);
/*
* remember old device strategy method and explicit set method
* `discsort' since we have our own more complex strategy that is not
* implementable on the CD device and other strategies will get in the
* way.
*/
memset(&priv->old_strategy_setting, 0,
sizeof(struct disk_strategy));
VOP_IOCTL(ump->devvp, DIOCGSTRATEGY, &priv->old_strategy_setting,
FREAD | FKIOCTL, NOCRED);
memset(&dkstrat, 0, sizeof(struct disk_strategy));
strcpy(dkstrat.dks_name, "discsort");
VOP_IOCTL(ump->devvp, DIOCSSTRATEGY, &dkstrat, FWRITE | FKIOCTL,
NOCRED);
/* initialise our internal sheduler */
priv->cur_queue = UDF_SHED_READING;
bufq_alloc(&priv->queues[UDF_SHED_READING], "disksort",
BUFQ_SORT_RAWBLOCK);
bufq_alloc(&priv->queues[UDF_SHED_WRITING], "disksort",
BUFQ_SORT_RAWBLOCK);
bufq_alloc(&priv->queues[UDF_SHED_SEQWRITING], "fcfs", 0);
vfs_timestamp(&priv->last_queued[UDF_SHED_READING]);
vfs_timestamp(&priv->last_queued[UDF_SHED_WRITING]);
vfs_timestamp(&priv->last_queued[UDF_SHED_SEQWRITING]);
/* create our disk strategy thread */
priv->run_thread = 1;
if (kthread_create(PRI_NONE, 0 /* KTHREAD_MPSAFE*/, NULL /* cpu_info*/,
udf_discstrat_thread, ump, &priv->queue_lwp,
"%s", "udf_rw")) {
panic("fork udf_rw");
}
}
int
adosfs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
{
struct disklabel dl;
struct partition *parp;
struct adosfsmount *amp;
struct buf *bp;
struct vnode *rvp;
size_t bitmap_sz = 0;
int error, i;
uint64_t numsecs;
unsigned secsize;
unsigned long secsperblk, blksperdisk, resvblks;
amp = NULL;
if ((error = vinvalbuf(devvp, V_SAVE, l->l_cred, l, 0, 0)) != 0)
return (error);
/*
* open blkdev and read boot and root block
*/
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
if ((error = VOP_OPEN(devvp, FREAD, NOCRED)) != 0) {
VOP_UNLOCK(devvp);
return (error);
}
error = getdisksize(devvp, &numsecs, &secsize);
if (error)
goto fail;
amp = kmem_zalloc(sizeof(struct adosfsmount), KM_SLEEP);
/*
* compute filesystem parameters from disklabel
* on arch/amiga the disklabel is computed from the native
* partition tables
* - p_fsize is the filesystem block size
* - p_frag is the number of sectors per filesystem block
* - p_cpg is the number of reserved blocks (boot blocks)
* - p_psize is reduced by the number of preallocated blocks
* at the end of a partition
*
* XXX
* - bsize and secsperblk could be computed from the first sector
* of the root block
* - resvblks (the number of boot blocks) can only be guessed
* by scanning for the root block as its position moves
* with resvblks
*/
error = VOP_IOCTL(devvp, DIOCGDINFO, &dl, FREAD, NOCRED);
VOP_UNLOCK(devvp);
if (error)
goto fail;
parp = &dl.d_partitions[DISKPART(devvp->v_rdev)];
if (dl.d_type == DTYPE_FLOPPY) {
amp->bsize = secsize;
secsperblk = 1;
resvblks = 2;
} else if (parp->p_fsize > 0 && parp->p_frag > 0) {
amp->bsize = parp->p_fsize * parp->p_frag;
secsperblk = parp->p_frag;
resvblks = parp->p_cpg;
} else {
error = EINVAL;
goto fail;
}
blksperdisk = numsecs / secsperblk;
/* The filesytem variant ('dostype') is stored in the boot block */
bp = NULL;
if ((error = bread(devvp, (daddr_t)BBOFF,
amp->bsize, NOCRED, 0, &bp)) != 0) {
goto fail;
}
amp->dostype = adoswordn(bp, 0);
brelse(bp, 0);
/* basic sanity checks */
if (amp->dostype < 0x444f5300 || amp->dostype > 0x444f5305) {
error = EINVAL;
goto fail;
}
amp->rootb = (blksperdisk - 1 + resvblks) / 2;
amp->numblks = blksperdisk - resvblks;
amp->nwords = amp->bsize >> 2;
amp->dbsize = amp->bsize - (IS_FFS(amp) ? 0 : OFS_DATA_OFFSET);
amp->devvp = devvp;
amp->mp = mp;
mp->mnt_data = amp;
mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)devvp->v_rdev;
mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_ADOSFS);
mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
mp->mnt_stat.f_namemax = ADMAXNAMELEN;
mp->mnt_fs_bshift = ffs(amp->bsize) - 1;
mp->mnt_dev_bshift = DEV_BSHIFT;
mp->mnt_flag |= MNT_LOCAL;
/*
* init anode table.
*/
for (i = 0; i < ANODEHASHSZ; i++)
LIST_INIT(&->anodetab[i]);
/*
* get the root anode, if not a valid fs this will fail.
*/
if ((error = VFS_ROOT(mp, &rvp)) != 0)
goto fail;
/* allocate and load bitmap, set free space */
bitmap_sz = ((amp->numblks + 31) / 32) * sizeof(*amp->bitmap);
amp->bitmap = kmem_alloc(bitmap_sz, KM_SLEEP);
if (amp->bitmap)
adosfs_loadbitmap(amp);
if (mp->mnt_flag & MNT_RDONLY && amp->bitmap) {
/*
* Don't need the bitmap any more if it's read-only.
*/
kmem_free(amp->bitmap, bitmap_sz);
amp->bitmap = NULL;
}
vput(rvp);
return(0);
fail:
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
(void) VOP_CLOSE(devvp, FREAD, NOCRED);
VOP_UNLOCK(devvp);
if (amp && amp->bitmap)
kmem_free(amp->bitmap, bitmap_sz);
if (amp)
kmem_free(amp, sizeof(*amp));
return (error);
}
int
ffs_fsync(void *v)
{
struct vop_fsync_args /* {
struct vnode *a_vp;
kauth_cred_t a_cred;
int a_flags;
off_t a_offlo;
off_t a_offhi;
struct lwp *a_l;
} */ *ap = v;
struct buf *bp;
int num, error, i;
struct indir ia[NIADDR + 1];
int bsize;
daddr_t blk_high;
struct vnode *vp;
struct mount *mp;
vp = ap->a_vp;
mp = vp->v_mount;
fstrans_start(mp, FSTRANS_LAZY);
if ((ap->a_offlo == 0 && ap->a_offhi == 0) || (vp->v_type != VREG)) {
error = ffs_full_fsync(vp, ap->a_flags);
goto out;
}
bsize = mp->mnt_stat.f_iosize;
blk_high = ap->a_offhi / bsize;
if (ap->a_offhi % bsize != 0)
blk_high++;
/*
* First, flush all pages in range.
*/
mutex_enter(vp->v_interlock);
error = VOP_PUTPAGES(vp, trunc_page(ap->a_offlo),
round_page(ap->a_offhi), PGO_CLEANIT |
((ap->a_flags & FSYNC_WAIT) ? PGO_SYNCIO : 0));
if (error) {
goto out;
}
#ifdef WAPBL
KASSERT(vp->v_type == VREG);
if (mp->mnt_wapbl) {
/*
* Don't bother writing out metadata if the syncer is
* making the request. We will let the sync vnode
* write it out in a single burst through a call to
* VFS_SYNC().
*/
if ((ap->a_flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0) {
fstrans_done(mp);
return 0;
}
error = 0;
if (vp->v_tag == VT_UFS && VTOI(vp)->i_flag &
(IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFY |
IN_MODIFIED | IN_ACCESSED)) {
error = UFS_WAPBL_BEGIN(mp);
if (error) {
fstrans_done(mp);
return error;
}
error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
((ap->a_flags & FSYNC_WAIT) ? UPDATE_WAIT : 0));
UFS_WAPBL_END(mp);
}
if (error || (ap->a_flags & FSYNC_NOLOG) != 0) {
fstrans_done(mp);
return error;
}
error = wapbl_flush(mp->mnt_wapbl, 0);
fstrans_done(mp);
return error;
}
#endif /* WAPBL */
/*
* Then, flush indirect blocks.
*/
if (blk_high >= NDADDR) {
error = ufs_getlbns(vp, blk_high, ia, &num);
if (error)
goto out;
mutex_enter(&bufcache_lock);
for (i = 0; i < num; i++) {
if ((bp = incore(vp, ia[i].in_lbn)) == NULL)
continue;
if ((bp->b_cflags & BC_BUSY) != 0 ||
(bp->b_oflags & BO_DELWRI) == 0)
continue;
bp->b_cflags |= BC_BUSY | BC_VFLUSH;
mutex_exit(&bufcache_lock);
bawrite(bp);
mutex_enter(&bufcache_lock);
}
mutex_exit(&bufcache_lock);
}
if (ap->a_flags & FSYNC_WAIT) {
mutex_enter(vp->v_interlock);
while (vp->v_numoutput > 0)
cv_wait(&vp->v_cv, vp->v_interlock);
mutex_exit(vp->v_interlock);
}
error = ffs_update(vp, NULL, NULL, UPDATE_CLOSE |
(((ap->a_flags & (FSYNC_WAIT | FSYNC_DATAONLY)) == FSYNC_WAIT)
? UPDATE_WAIT : 0));
if (error == 0 && ap->a_flags & FSYNC_CACHE) {
int l = 0;
VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &l, FWRITE,
curlwp->l_cred);
}
out:
fstrans_done(mp);
return error;
}
/* ARGSUSED */
int
ffs_full_fsync(struct vnode *vp, int flags)
{
int error, i, uflags;
struct mount *mp;
KASSERT(vp->v_tag == VT_UFS);
KASSERT(VTOI(vp) != NULL);
KASSERT(vp->v_type != VCHR && vp->v_type != VBLK);
error = 0;
uflags = UPDATE_CLOSE | ((flags & FSYNC_WAIT) ? UPDATE_WAIT : 0);
mp = vp->v_mount;
/*
* Flush all dirty data associated with the vnode.
*/
if (vp->v_type == VREG) {
int pflags = PGO_ALLPAGES | PGO_CLEANIT;
if ((flags & FSYNC_WAIT))
pflags |= PGO_SYNCIO;
if (fstrans_getstate(mp) == FSTRANS_SUSPENDING)
pflags |= PGO_FREE;
mutex_enter(vp->v_interlock);
error = VOP_PUTPAGES(vp, 0, 0, pflags);
if (error)
return error;
}
#ifdef WAPBL
if (mp && mp->mnt_wapbl) {
/*
* Don't bother writing out metadata if the syncer is
* making the request. We will let the sync vnode
* write it out in a single burst through a call to
* VFS_SYNC().
*/
if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY)) != 0)
return 0;
if ((VTOI(vp)->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE
| IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) != 0) {
error = UFS_WAPBL_BEGIN(mp);
if (error)
return error;
error = ffs_update(vp, NULL, NULL, uflags);
UFS_WAPBL_END(mp);
}
if (error || (flags & FSYNC_NOLOG) != 0)
return error;
/*
* Don't flush the log if the vnode being flushed
* contains no dirty buffers that could be in the log.
*/
if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
error = wapbl_flush(mp->mnt_wapbl, 0);
if (error)
return error;
}
if ((flags & FSYNC_WAIT) != 0) {
mutex_enter(vp->v_interlock);
while (vp->v_numoutput != 0)
cv_wait(&vp->v_cv, vp->v_interlock);
mutex_exit(vp->v_interlock);
}
return error;
}
#endif /* WAPBL */
error = vflushbuf(vp, (flags & FSYNC_WAIT) != 0);
if (error == 0)
error = ffs_update(vp, NULL, NULL, uflags);
if (error == 0 && (flags & FSYNC_CACHE) != 0) {
i = 1;
(void)VOP_IOCTL(VTOI(vp)->i_devvp, DIOCCACHESYNC, &i, FWRITE,
kauth_cred_get());
}
return error;
}
int
ccdinit(struct ccddevice *ccd, char **cpaths, struct proc *p)
{
struct ccd_softc *cs = &ccd_softc[ccd->ccd_unit];
struct ccdcinfo *ci = NULL;
size_t size;
int ix, rpm;
struct vnode *vp;
struct vattr va;
size_t minsize;
int maxsecsize;
struct partinfo dpart;
struct ccdgeom *ccg = &cs->sc_geom;
char tmppath[MAXPATHLEN];
int error;
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT, ("ccdinit: unit %d cflags %b\n",
ccd->ccd_unit, ccd->ccd_flags, CCDF_BITS));
cs->sc_size = 0;
cs->sc_ileave = ccd->ccd_interleave;
cs->sc_nccdisks = ccd->ccd_ndev;
if (snprintf(cs->sc_xname, sizeof(cs->sc_xname), "ccd%d",
ccd->ccd_unit) >= sizeof(cs->sc_xname)) {
printf("ccdinit: device name too long.\n");
return(ENXIO);
}
/* Allocate space for the component info. */
cs->sc_cinfo = malloc(cs->sc_nccdisks * sizeof(struct ccdcinfo),
M_DEVBUF, M_WAITOK);
bzero(cs->sc_cinfo, cs->sc_nccdisks * sizeof(struct ccdcinfo));
/*
* Verify that each component piece exists and record
* relevant information about it.
*/
maxsecsize = 0;
minsize = 0;
rpm = 0;
for (ix = 0; ix < cs->sc_nccdisks; ix++) {
vp = ccd->ccd_vpp[ix];
ci = &cs->sc_cinfo[ix];
ci->ci_vp = vp;
/*
* Copy in the pathname of the component.
*/
bzero(tmppath, sizeof(tmppath)); /* sanity */
error = copyinstr(cpaths[ix], tmppath,
MAXPATHLEN, &ci->ci_pathlen);
if (error) {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: can't copy path, error = %d\n",
cs->sc_xname, error));
free(cs->sc_cinfo, M_DEVBUF);
return (error);
}
ci->ci_path = malloc(ci->ci_pathlen, M_DEVBUF, M_WAITOK);
bcopy(tmppath, ci->ci_path, ci->ci_pathlen);
/*
* XXX: Cache the component's dev_t.
*/
if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p)) != 0) {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: %s: getattr failed error = %d\n",
cs->sc_xname, ci->ci_path, error));
free(ci->ci_path, M_DEVBUF);
free(cs->sc_cinfo, M_DEVBUF);
return (error);
}
ci->ci_dev = va.va_rdev;
/*
* Get partition information for the component.
*/
error = VOP_IOCTL(vp, DIOCGPART, (caddr_t)&dpart,
FREAD, p->p_ucred, p);
if (error) {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: %s: ioctl failed, error = %d\n",
cs->sc_xname, ci->ci_path, error));
free(ci->ci_path, M_DEVBUF);
free(cs->sc_cinfo, M_DEVBUF);
return (error);
}
if (dpart.part->p_fstype == FS_CCD ||
dpart.part->p_fstype == FS_BSDFFS) {
maxsecsize =
((dpart.disklab->d_secsize > maxsecsize) ?
dpart.disklab->d_secsize : maxsecsize);
size = dpart.part->p_size;
} else {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: %s: incorrect partition type\n",
cs->sc_xname, ci->ci_path));
free(ci->ci_path, M_DEVBUF);
free(cs->sc_cinfo, M_DEVBUF);
return (EFTYPE);
}
/*
* Calculate the size, truncating to an interleave
* boundary if necessary.
*/
if (cs->sc_ileave > 1)
size -= size % cs->sc_ileave;
if (size == 0) {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: %s: size == 0\n", cs->sc_xname, ci->ci_path));
free(ci->ci_path, M_DEVBUF);
free(cs->sc_cinfo, M_DEVBUF);
return (ENODEV);
}
if (minsize == 0 || size < minsize)
minsize = size;
ci->ci_size = size;
cs->sc_size += size;
rpm += dpart.disklab->d_rpm;
}
ccg->ccg_rpm = rpm / cs->sc_nccdisks;
/*
* Don't allow the interleave to be smaller than
* the biggest component sector.
*/
if ((cs->sc_ileave > 0) &&
(cs->sc_ileave < (maxsecsize / DEV_BSIZE))) {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: interleave must be at least %d\n",
cs->sc_xname, (maxsecsize / DEV_BSIZE)));
free(ci->ci_path, M_DEVBUF);
free(cs->sc_cinfo, M_DEVBUF);
return (EINVAL);
}
/*
* Mirroring support requires uniform interleave and
* and even number of components.
*/
if (ccd->ccd_flags & CCDF_MIRROR) {
ccd->ccd_flags |= CCDF_UNIFORM;
if (cs->sc_ileave == 0) {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: mirroring requires interleave\n",
cs->sc_xname));
free(ci->ci_path, M_DEVBUF);
free(cs->sc_cinfo, M_DEVBUF);
return (EINVAL);
}
if (cs->sc_nccdisks % 2) {
CCD_DPRINTF(CCDB_FOLLOW | CCDB_INIT,
("%s: mirroring requires even # of components\n",
cs->sc_xname));
free(ci->ci_path, M_DEVBUF);
free(cs->sc_cinfo, M_DEVBUF);
return (EINVAL);
}
}
/*
* If uniform interleave is desired set all sizes to that of
* the smallest component.
*/
ccg->ccg_ntracks = cs->sc_nccunits = cs->sc_nccdisks;
if (ccd->ccd_flags & CCDF_UNIFORM) {
for (ci = cs->sc_cinfo;
ci < &cs->sc_cinfo[cs->sc_nccdisks]; ci++)
ci->ci_size = minsize;
if (ccd->ccd_flags & CCDF_MIRROR)
cs->sc_nccunits = ccg->ccg_ntracks /= 2;
cs->sc_size = ccg->ccg_ntracks * minsize;
}
cs->sc_cflags = ccd->ccd_flags; /* So we can find out later... */
/*
* Construct the interleave table.
*/
ccdinterleave(cs);
/*
* Create pseudo-geometry based on 1MB cylinders. It's
* pretty close.
*/
ccg->ccg_secsize = DEV_BSIZE;
ccg->ccg_nsectors = cs->sc_ileave? cs->sc_ileave :
1024 * (1024 / ccg->ccg_secsize);
ccg->ccg_ncylinders = cs->sc_size / ccg->ccg_ntracks /
ccg->ccg_nsectors;
cs->sc_flags |= CCDF_INITED;
return (0);
}
/*
* Helper function for findroot():
* Return non-zero if disk device matches bootinfo.
*/
static int
match_bootdisk(struct device *dv, struct btinfo_bootdisk *bid)
{
struct vnode *tmpvn;
int error;
struct disklabel label;
int found = 0;
int bmajor;
/*
* A disklabel is required here. The boot loader doesn't refuse
* to boot from a disk without a label, but this is normally not
* wanted.
*/
if (bid->labelsector == -1)
return (0);
/*
* Lookup major number for disk block device.
*/
bmajor = devsw_name2blk(dv->dv_xname, NULL, 0);
if (bmajor == -1)
return (0); /* XXX panic ??? */
/*
* Fake a temporary vnode for the disk, open it, and read
* the disklabel for comparison.
*/
if (bdevvp(MAKEDISKDEV(bmajor, dv->dv_unit, RAW_PART), &tmpvn))
panic("match_bootdisk: can't alloc vnode");
error = VOP_OPEN(tmpvn, FREAD, NOCRED);
if (error) {
#ifndef DEBUG
/*
* Ignore errors caused by missing device, partition,
* or medium.
*/
if (error != ENXIO && error != ENODEV)
#endif
printf("match_bootdisk: can't open dev %s (%d)\n",
dv->dv_xname, error);
vput(tmpvn);
return (0);
}
error = VOP_IOCTL(tmpvn, DIOCGDINFO, &label, FREAD, NOCRED);
if (error) {
/*
* XXX Can't happen -- open() would have errored out
* or faked one up.
*/
printf("match_bootdisk: can't get label for dev %s (%d)\n",
dv->dv_xname, error);
goto closeout;
}
/* Compare with our data. */
if (label.d_type == bid->label.type &&
label.d_checksum == bid->label.checksum &&
strncmp(label.d_packname, bid->label.packname, 16) == 0)
found = 1;
closeout:
VOP_CLOSE(tmpvn, FREAD, NOCRED);
vput(tmpvn);
return (found);
}
/*
* File table vnode ioctl routine.
*/
static int
vn_ioctl(file_t *fp, u_long com, void *data)
{
struct vnode *vp = fp->f_data, *ovp;
struct vattr vattr;
int error;
switch (vp->v_type) {
case VREG:
case VDIR:
if (com == FIONREAD) {
vn_lock(vp, LK_SHARED | LK_RETRY);
error = VOP_GETATTR(vp, &vattr, kauth_cred_get());
VOP_UNLOCK(vp);
if (error)
return (error);
*(int *)data = vattr.va_size - fp->f_offset;
return (0);
}
if ((com == FIONWRITE) || (com == FIONSPACE)) {
/*
* Files don't have send queues, so there never
* are any bytes in them, nor is there any
* open space in them.
*/
*(int *)data = 0;
return (0);
}
if (com == FIOGETBMAP) {
daddr_t *block;
if (*(daddr_t *)data < 0)
return (EINVAL);
block = (daddr_t *)data;
return (VOP_BMAP(vp, *block, NULL, block, NULL));
}
if (com == OFIOGETBMAP) {
daddr_t ibn, obn;
if (*(int32_t *)data < 0)
return (EINVAL);
ibn = (daddr_t)*(int32_t *)data;
error = VOP_BMAP(vp, ibn, NULL, &obn, NULL);
*(int32_t *)data = (int32_t)obn;
return error;
}
if (com == FIONBIO || com == FIOASYNC) /* XXX */
return (0); /* XXX */
/* fall into ... */
case VFIFO:
case VCHR:
case VBLK:
error = VOP_IOCTL(vp, com, data, fp->f_flag,
kauth_cred_get());
if (error == 0 && com == TIOCSCTTY) {
vref(vp);
mutex_enter(proc_lock);
ovp = curproc->p_session->s_ttyvp;
curproc->p_session->s_ttyvp = vp;
mutex_exit(proc_lock);
if (ovp != NULL)
vrele(ovp);
}
return (error);
default:
return (EPASSTHROUGH);
}
}
/*
* MPALMOSTSAFE - acquires mplock
*/
static int
vn_ioctl(struct file *fp, u_long com, caddr_t data, struct ucred *ucred,
struct sysmsg *msg)
{
struct vnode *vp = ((struct vnode *)fp->f_data);
struct vnode *ovp;
struct vattr vattr;
int error;
off_t size;
switch (vp->v_type) {
case VREG:
case VDIR:
if (com == FIONREAD) {
error = VOP_GETATTR(vp, &vattr);
if (error)
break;
size = vattr.va_size;
if ((vp->v_flag & VNOTSEEKABLE) == 0)
size -= vn_poll_fpf_offset(fp);
if (size > 0x7FFFFFFF)
size = 0x7FFFFFFF;
*(int *)data = size;
error = 0;
break;
}
if (com == FIOASYNC) { /* XXX */
error = 0; /* XXX */
break;
}
/* fall into ... */
default:
#if 0
return (ENOTTY);
#endif
case VFIFO:
case VCHR:
case VBLK:
if (com == FIODTYPE) {
if (vp->v_type != VCHR && vp->v_type != VBLK) {
error = ENOTTY;
break;
}
*(int *)data = dev_dflags(vp->v_rdev) & D_TYPEMASK;
error = 0;
break;
}
error = VOP_IOCTL(vp, com, data, fp->f_flag, ucred, msg);
if (error == 0 && com == TIOCSCTTY) {
struct proc *p = curthread->td_proc;
struct session *sess;
if (p == NULL) {
error = ENOTTY;
break;
}
get_mplock();
sess = p->p_session;
/* Do nothing if reassigning same control tty */
if (sess->s_ttyvp == vp) {
error = 0;
rel_mplock();
break;
}
/* Get rid of reference to old control tty */
ovp = sess->s_ttyvp;
vref(vp);
sess->s_ttyvp = vp;
if (ovp)
vrele(ovp);
rel_mplock();
}
break;
}
return (error);
}
static int
nm_ioctl(vnode_t *vp, int cmd, intptr_t arg, int mode, cred_t *cr, int *rvalp,
caller_context_t *ct)
{
return (VOP_IOCTL(VTONM(vp)->nm_filevp, cmd, arg, mode, cr, rvalp, ct));
}
