static int
hammer_setup_device(struct vnode **devvpp, const char *dev_path, int ronly)
{
int error;
struct nlookupdata nd;
/*
* Get the device vnode
*/
if (*devvpp == NULL) {
error = nlookup_init(&nd, dev_path, UIO_SYSSPACE, NLC_FOLLOW);
if (error == 0)
error = nlookup(&nd);
if (error == 0)
error = cache_vref(&nd.nl_nch, nd.nl_cred, devvpp);
nlookup_done(&nd);
} else {
error = 0;
}
if (error == 0) {
if (vn_isdisk(*devvpp, &error)) {
error = vfs_mountedon(*devvpp);
}
}
if (error == 0 && vcount(*devvpp) > 0)
error = EBUSY;
if (error == 0) {
vn_lock(*devvpp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(*devvpp, V_SAVE, 0, 0);
if (error == 0) {
error = VOP_OPEN(*devvpp,
(ronly ? FREAD : FREAD|FWRITE),
FSCRED, NULL);
}
vn_unlock(*devvpp);
}
if (error && *devvpp) {
vrele(*devvpp);
*devvpp = NULL;
}
return (error);
}
static int
vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
uint64_t *ashift)
{
spa_t *spa = vd->vdev_spa;
vdev_disk_t *dvd = vd->vdev_tsd;
vnode_t *devvp = NULLVP;
vfs_context_t context = NULL;
uint64_t blkcnt;
uint32_t blksize;
int fmode = 0;
int error = 0;
int isssd;
/*
* We must have a pathname, and it must be absolute.
*/
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
/*
* Reopen the device if it's not currently open. Otherwise,
* just update the physical size of the device.
*/
if (dvd != NULL) {
if (dvd->vd_offline) {
/*
* If we are opening a device in its offline notify
* context, the LDI handle was just closed. Clean
* up the LDI event callbacks and free vd->vdev_tsd.
*/
vdev_disk_free(vd);
} else {
ASSERT(vd->vdev_reopening);
devvp = dvd->vd_devvp;
goto skip_open;
}
}
/*
* Create vd->vdev_tsd.
*/
vdev_disk_alloc(vd);
dvd = vd->vdev_tsd;
/*
* When opening a disk device, we want to preserve the user's original
* intent. We always want to open the device by the path the user gave
* us, even if it is one of multiple paths to the same device. But we
* also want to be able to survive disks being removed/recabled.
* Therefore the sequence of opening devices is:
*
* 1. Try opening the device by path. For legacy pools without the
* 'whole_disk' property, attempt to fix the path by appending 's0'.
*
* 2. If the devid of the device matches the stored value, return
* success.
*
* 3. Otherwise, the device may have moved. Try opening the device
* by the devid instead.
*/
/* ### APPLE TODO ### */
#ifdef illumos
if (vd->vdev_devid != NULL) {
if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid,
&dvd->vd_minor) != 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
}
#endif
error = EINVAL; /* presume failure */
if (vd->vdev_path != NULL) {
context = vfs_context_create( spl_vfs_context_kernel() );
/* Obtain an opened/referenced vnode for the device. */
if ((error = vnode_open(vd->vdev_path, spa_mode(spa), 0, 0,
&devvp, context))) {
goto out;
}
if (!vnode_isblk(devvp)) {
error = ENOTBLK;
goto out;
}
/*
* ### APPLE TODO ###
* vnode_authorize devvp for KAUTH_VNODE_READ_DATA and
* KAUTH_VNODE_WRITE_DATA
*/
/*
* Disallow opening of a device that is currently in use.
* Flush out any old buffers remaining from a previous use.
*/
if ((error = vfs_mountedon(devvp))) {
goto out;
}
if (VNOP_FSYNC(devvp, MNT_WAIT, context) != 0) {
error = ENOTBLK;
goto out;
}
if ((error = buf_invalidateblks(devvp, BUF_WRITE_DATA, 0, 0))) {
goto out;
}
} else {
goto out;
}
int len = MAXPATHLEN;
if (vn_getpath(devvp, dvd->vd_readlinkname, &len) == 0) {
dprintf("ZFS: '%s' resolved name is '%s'\n",
vd->vdev_path, dvd->vd_readlinkname);
} else {
dvd->vd_readlinkname[0] = 0;
}
skip_open:
/*
* Determine the actual size of the device.
*/
if (VNOP_IOCTL(devvp, DKIOCGETBLOCKSIZE, (caddr_t)&blksize, 0,
context) != 0 ||
VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&blkcnt, 0,
context) != 0) {
error = EINVAL;
goto out;
}
*psize = blkcnt * (uint64_t)blksize;
*max_psize = *psize;
dvd->vd_ashift = highbit(blksize) - 1;
dprintf("vdev_disk: Device %p ashift set to %d\n", devvp,
dvd->vd_ashift);
*ashift = highbit(MAX(blksize, SPA_MINBLOCKSIZE)) - 1;
/*
* ### APPLE TODO ###
*/
#ifdef illumos
if (vd->vdev_wholedisk == 1) {
int wce = 1;
if (error == 0) {
/*
* If we have the capability to expand, we'd have
* found out via success from DKIOCGMEDIAINFO{,EXT}.
* Adjust max_psize upward accordingly since we know
* we own the whole disk now.
*/
*max_psize = capacity * blksz;
}
/*
* Since we own the whole disk, try to enable disk write
* caching. We ignore errors because it's OK if we can't do it.
*/
(void) ldi_ioctl(dvd->vd_lh, DKIOCSETWCE, (intptr_t)&wce,
FKIOCTL, kcred, NULL);
}
#endif
/*
* Clear the nowritecache bit, so that on a vdev_reopen() we will
* try again.
*/
vd->vdev_nowritecache = B_FALSE;
/* Inform the ZIO pipeline that we are non-rotational */
vd->vdev_nonrot = B_FALSE;
if (VNOP_IOCTL(devvp, DKIOCISSOLIDSTATE, (caddr_t)&isssd, 0,
context) == 0) {
if (isssd)
vd->vdev_nonrot = B_TRUE;
}
dprintf("ZFS: vdev_disk(%s) isSSD %d\n", vd->vdev_path ? vd->vdev_path : "",
isssd);
dvd->vd_devvp = devvp;
out:
if (error) {
if (devvp) {
vnode_close(devvp, fmode, context);
dvd->vd_devvp = NULL;
}
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
}
if (context)
(void) vfs_context_rele(context);
if (error) printf("ZFS: vdev_disk_open('%s') failed error %d\n",
vd->vdev_path ? vd->vdev_path : "", error);
return (error);
}
int
msdosfs_mountfs(struct vnode *devvp, struct mount *mp, struct proc *p,
struct msdosfs_args *argp)
{
struct msdosfsmount *pmp;
struct buf *bp;
dev_t dev = devvp->v_rdev;
union bootsector *bsp;
struct byte_bpb33 *b33;
struct byte_bpb50 *b50;
struct byte_bpb710 *b710;
extern struct vnode *rootvp;
u_int8_t SecPerClust;
int ronly, error, bmapsiz;
uint32_t fat_max_clusters;
/*
* 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, p->p_ucred, 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);
bp = NULL; /* both used in error_exit */
pmp = NULL;
/*
* Read the boot sector of the filesystem, and then check the
* boot signature. If not a dos boot sector then error out.
*/
if ((error = bread(devvp, 0, 4096, NOCRED, &bp)) != 0)
goto error_exit;
bp->b_flags |= B_AGE;
bsp = (union bootsector *)bp->b_data;
b33 = (struct byte_bpb33 *)bsp->bs33.bsBPB;
b50 = (struct byte_bpb50 *)bsp->bs50.bsBPB;
b710 = (struct byte_bpb710 *)bsp->bs710.bsPBP;
pmp = malloc(sizeof *pmp, M_MSDOSFSMNT, M_WAITOK | M_ZERO);
pmp->pm_mountp = mp;
/*
* Compute several useful quantities from the bpb in the
* bootsector. Copy in the dos 5 variant of the bpb then fix up
* the fields that are different between dos 5 and dos 3.3.
*/
SecPerClust = b50->bpbSecPerClust;
pmp->pm_BytesPerSec = getushort(b50->bpbBytesPerSec);
pmp->pm_ResSectors = getushort(b50->bpbResSectors);
pmp->pm_FATs = b50->bpbFATs;
pmp->pm_RootDirEnts = getushort(b50->bpbRootDirEnts);
pmp->pm_Sectors = getushort(b50->bpbSectors);
pmp->pm_FATsecs = getushort(b50->bpbFATsecs);
pmp->pm_SecPerTrack = getushort(b50->bpbSecPerTrack);
pmp->pm_Heads = getushort(b50->bpbHeads);
pmp->pm_Media = b50->bpbMedia;
/* Determine the number of DEV_BSIZE blocks in a MSDOSFS sector */
pmp->pm_BlkPerSec = pmp->pm_BytesPerSec / DEV_BSIZE;
if (!pmp->pm_BytesPerSec || !SecPerClust || pmp->pm_SecPerTrack > 64) {
error = EFTYPE;
goto error_exit;
}
if (pmp->pm_Sectors == 0) {
pmp->pm_HiddenSects = getulong(b50->bpbHiddenSecs);
pmp->pm_HugeSectors = getulong(b50->bpbHugeSectors);
} else {
pmp->pm_HiddenSects = getushort(b33->bpbHiddenSecs);
pmp->pm_HugeSectors = pmp->pm_Sectors;
}
if (pmp->pm_RootDirEnts == 0) {
if (pmp->pm_Sectors || pmp->pm_FATsecs ||
getushort(b710->bpbFSVers)) {
error = EINVAL;
goto error_exit;
}
pmp->pm_fatmask = FAT32_MASK;
pmp->pm_fatmult = 4;
pmp->pm_fatdiv = 1;
pmp->pm_FATsecs = getulong(b710->bpbBigFATsecs);
if (getushort(b710->bpbExtFlags) & FATMIRROR)
pmp->pm_curfat = getushort(b710->bpbExtFlags) & FATNUM;
else
pmp->pm_flags |= MSDOSFS_FATMIRROR;
} else
pmp->pm_flags |= MSDOSFS_FATMIRROR;
/*
* More sanity checks:
* MSDOSFS sectors per cluster: >0 && power of 2
* MSDOSFS sector size: >= DEV_BSIZE && power of 2
* HUGE sector count: >0
* FAT sectors: >0
*/
if ((SecPerClust == 0) || (SecPerClust & (SecPerClust - 1)) ||
(pmp->pm_BytesPerSec < DEV_BSIZE) ||
(pmp->pm_BytesPerSec & (pmp->pm_BytesPerSec - 1)) ||
(pmp->pm_HugeSectors == 0) || (pmp->pm_FATsecs == 0)) {
error = EINVAL;
goto error_exit;
}
pmp->pm_HugeSectors *= pmp->pm_BlkPerSec;
pmp->pm_HiddenSects *= pmp->pm_BlkPerSec;
pmp->pm_FATsecs *= pmp->pm_BlkPerSec;
pmp->pm_fatblk = pmp->pm_ResSectors * pmp->pm_BlkPerSec;
SecPerClust *= pmp->pm_BlkPerSec;
if (FAT32(pmp)) {
pmp->pm_rootdirblk = getulong(b710->bpbRootClust);
pmp->pm_firstcluster = pmp->pm_fatblk
+ (pmp->pm_FATs * pmp->pm_FATsecs);
pmp->pm_fsinfo = getushort(b710->bpbFSInfo) * pmp->pm_BlkPerSec;
} else {
pmp->pm_rootdirblk = pmp->pm_fatblk +
(pmp->pm_FATs * pmp->pm_FATsecs);
pmp->pm_rootdirsize = (pmp->pm_RootDirEnts * sizeof(struct direntry)
+ DEV_BSIZE - 1) / DEV_BSIZE;
pmp->pm_firstcluster = pmp->pm_rootdirblk + pmp->pm_rootdirsize;
}
pmp->pm_nmbrofclusters = (pmp->pm_HugeSectors - pmp->pm_firstcluster) /
SecPerClust;
pmp->pm_maxcluster = pmp->pm_nmbrofclusters + 1;
pmp->pm_fatsize = pmp->pm_FATsecs * DEV_BSIZE;
if (pmp->pm_fatmask == 0) {
if (pmp->pm_maxcluster
<= ((CLUST_RSRVD - CLUST_FIRST) & FAT12_MASK)) {
/*
* This will usually be a floppy disk. This size makes
* sure that one fat entry will not be split across
* multiple blocks.
*/
pmp->pm_fatmask = FAT12_MASK;
pmp->pm_fatmult = 3;
pmp->pm_fatdiv = 2;
} else {
pmp->pm_fatmask = FAT16_MASK;
pmp->pm_fatmult = 2;
pmp->pm_fatdiv = 1;
}
}
if (FAT12(pmp))
pmp->pm_fatblocksize = 3 * pmp->pm_BytesPerSec;
else
pmp->pm_fatblocksize = MAXBSIZE;
/*
* We now have the number of sectors in each FAT, so can work
* out how many clusters can be represented in a FAT. Let's
* make sure the file system doesn't claim to have more clusters
* than this.
*
* We perform the calculation like we do to avoid integer overflow.
*
* This will give us a count of clusters. They are numbered
* from 0, so the max cluster value is one less than the value
* we end up with.
*/
fat_max_clusters = pmp->pm_fatsize / pmp->pm_fatmult;
fat_max_clusters *= pmp->pm_fatdiv;
if (pmp->pm_maxcluster >= fat_max_clusters) {
#ifndef SMALL_KERNEL
printf("msdosfs: reducing max cluster to %d from %d "
"due to FAT size\n", fat_max_clusters - 1,
pmp->pm_maxcluster);
#endif
pmp->pm_maxcluster = fat_max_clusters - 1;
}
pmp->pm_fatblocksec = pmp->pm_fatblocksize / DEV_BSIZE;
pmp->pm_bnshift = ffs(DEV_BSIZE) - 1;
/*
* Compute mask and shift value for isolating cluster relative byte
* offsets and cluster numbers from a file offset.
*/
pmp->pm_bpcluster = SecPerClust * DEV_BSIZE;
pmp->pm_crbomask = pmp->pm_bpcluster - 1;
pmp->pm_cnshift = ffs(pmp->pm_bpcluster) - 1;
/*
* Check for valid cluster size
* must be a power of 2
*/
if (pmp->pm_bpcluster ^ (1 << pmp->pm_cnshift)) {
error = EFTYPE;
goto error_exit;
}
/*
* Release the bootsector buffer.
*/
brelse(bp);
bp = NULL;
/*
* Check FSInfo
*/
if (pmp->pm_fsinfo) {
struct fsinfo *fp;
if ((error = bread(devvp, pmp->pm_fsinfo, fsi_size(pmp),
NOCRED, &bp)) != 0)
goto error_exit;
fp = (struct fsinfo *)bp->b_data;
if (!bcmp(fp->fsisig1, "RRaA", 4)
&& !bcmp(fp->fsisig2, "rrAa", 4)
&& !bcmp(fp->fsisig3, "\0\0\125\252", 4)
&& !bcmp(fp->fsisig4, "\0\0\125\252", 4))
/* Valid FSInfo. */
;
else
pmp->pm_fsinfo = 0;
/* XXX make sure this tiny buf doesn't come back in fillinusemap! */
SET(bp->b_flags, B_INVAL);
brelse(bp);
bp = NULL;
}
/*
* Check and validate (or perhaps invalidate?) the fsinfo structure? XXX
*/
/*
* Allocate memory for the bitmap of allocated clusters, and then
* fill it in.
*/
bmapsiz = (pmp->pm_maxcluster + N_INUSEBITS - 1) / N_INUSEBITS;
if (bmapsiz == 0 || SIZE_MAX / bmapsiz < sizeof(*pmp->pm_inusemap)) {
/* detect multiplicative integer overflow */
error = EINVAL;
goto error_exit;
}
pmp->pm_inusemap = malloc(bmapsiz * sizeof(*pmp->pm_inusemap),
M_MSDOSFSFAT, M_WAITOK | M_CANFAIL);
if (pmp->pm_inusemap == NULL) {
error = EINVAL;
goto error_exit;
}
/*
* fillinusemap() needs pm_devvp.
*/
pmp->pm_dev = dev;
pmp->pm_devvp = devvp;
/*
* Have the inuse map filled in.
*/
if ((error = fillinusemap(pmp)) != 0)
goto error_exit;
/*
* If they want fat updates to be synchronous then let them suffer
* the performance degradation in exchange for the on disk copy of
* the fat being correct just about all the time. I suppose this
* would be a good thing to turn on if the kernel is still flakey.
*/
if (mp->mnt_flag & MNT_SYNCHRONOUS)
pmp->pm_flags |= MSDOSFSMNT_WAITONFAT;
/*
* Finish up.
*/
if (ronly)
pmp->pm_flags |= MSDOSFSMNT_RONLY;
else
pmp->pm_fmod = 1;
mp->mnt_data = (qaddr_t)pmp;
mp->mnt_stat.f_fsid.val[0] = (long)dev;
mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
#ifdef QUOTA
/*
* If we ever do quotas for DOS filesystems this would be a place
* to fill in the info in the msdosfsmount structure. You dolt,
* quotas on dos filesystems make no sense because files have no
* owners on dos filesystems. of course there is some empty space
* in the directory entry where we could put uid's and gid's.
*/
#endif
devvp->v_specmountpoint = mp;
return (0);
error_exit:
devvp->v_specmountpoint = NULL;
if (bp)
brelse(bp);
vn_lock(devvp, LK_EXCLUSIVE|LK_RETRY, p);
(void) VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NOCRED, p);
VOP_UNLOCK(devvp, 0, p);
if (pmp) {
if (pmp->pm_inusemap)
free(pmp->pm_inusemap, M_MSDOSFSFAT);
free(pmp, M_MSDOSFSMNT);
mp->mnt_data = (qaddr_t)0;
}
return (error);
}
/*
* Open a special file.
*/
int
spec_open(void *v)
{
struct vop_open_args *ap = v;
struct proc *p = ap->a_p;
struct vnode *vp = ap->a_vp;
struct vnode *bvp;
dev_t bdev;
dev_t dev = (dev_t)vp->v_rdev;
int maj = major(dev);
int error;
/*
* Don't allow open if fs is mounted -nodev.
*/
if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
return (ENXIO);
switch (vp->v_type) {
case VCHR:
if ((u_int)maj >= nchrdev)
return (ENXIO);
if (ap->a_cred != FSCRED && (ap->a_mode & FWRITE)) {
/*
* When running in very secure mode, do not allow
* opens for writing of any disk character devices.
*/
if (securelevel >= 2 && cdevsw[maj].d_type == D_DISK)
return (EPERM);
/*
* When running in secure mode, do not allow opens
* for writing of /dev/mem, /dev/kmem, or character
* devices whose corresponding block devices are
* currently mounted.
*/
if (securelevel >= 1) {
if ((bdev = chrtoblk(dev)) != NODEV &&
vfinddev(bdev, VBLK, &bvp) &&
bvp->v_usecount > 0 &&
(error = vfs_mountedon(bvp)))
return (error);
if (iskmemdev(dev))
return (EPERM);
}
}
if (cdevsw[maj].d_type == D_TTY)
vp->v_flag |= VISTTY;
if (cdevsw[maj].d_flags & D_CLONE)
return (spec_open_clone(ap));
VOP_UNLOCK(vp, 0, p);
error = (*cdevsw[maj].d_open)(dev, ap->a_mode, S_IFCHR, p);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
return (error);
case VBLK:
if ((u_int)maj >= nblkdev)
return (ENXIO);
/*
* When running in very secure mode, do not allow
* opens for writing of any disk block devices.
*/
if (securelevel >= 2 && ap->a_cred != FSCRED &&
(ap->a_mode & FWRITE) && bdevsw[maj].d_type == D_DISK)
return (EPERM);
/*
* Do not allow opens of block devices that are
* currently mounted.
*/
if ((error = vfs_mountedon(vp)) != 0)
return (error);
return ((*bdevsw[maj].d_open)(dev, ap->a_mode, S_IFBLK, p));
case VNON:
case VLNK:
case VDIR:
case VREG:
case VBAD:
case VFIFO:
case VSOCK:
break;
}
return (0);
}
static int
vdev_disk_open(vdev_t *vd, uint64_t *size, uint64_t *max_size, uint64_t *ashift)
{
vdev_disk_t *dvd = NULL;
vnode_t *devvp = NULLVP;
vfs_context_t context = NULL;
uint64_t blkcnt;
uint32_t blksize;
int fmode = 0;
int error = 0;
/*
* We must have a pathname, and it must be absolute.
*/
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (EINVAL);
}
dvd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP);
if (dvd == NULL)
return ENOMEM;
/*
* When opening a disk device, we want to preserve the user's original
* intent. We always want to open the device by the path the user gave
* us, even if it is one of multiple paths to the save device. But we
* also want to be able to survive disks being removed/recabled.
* Therefore the sequence of opening devices is:
*
* 1. Try opening the device by path. For legacy pools without the
* 'whole_disk' property, attempt to fix the path by appending 's0'.
*
* 2. If the devid of the device matches the stored value, return
* success.
*
* 3. Otherwise, the device may have moved. Try opening the device
* by the devid instead.
*
*/
/* ### APPLE TODO ### */
/* ddi_devid_str_decode */
context = vfs_context_create((vfs_context_t)0);
/* Obtain an opened/referenced vnode for the device. */
error = vnode_open(vd->vdev_path, spa_mode(vd->vdev_spa), 0, 0, &devvp, context);
if (error) {
goto out;
}
if (!vnode_isblk(devvp)) {
error = ENOTBLK;
goto out;
}
/* ### APPLE TODO ### */
/* vnode_authorize devvp for KAUTH_VNODE_READ_DATA and
* KAUTH_VNODE_WRITE_DATA
*/
/*
* Disallow opening of a device that is currently in use.
* Flush out any old buffers remaining from a previous use.
*/
if ((error = vfs_mountedon(devvp))) {
goto out;
}
if (VNOP_FSYNC(devvp, MNT_WAIT, context) != 0) {
error = ENOTBLK;
goto out;
}
if ((error = buf_invalidateblks(devvp, BUF_WRITE_DATA, 0, 0))) {
goto out;
}
/*
* Determine the actual size of the device.
*/
if (VNOP_IOCTL(devvp, DKIOCGETBLOCKSIZE, (caddr_t)&blksize, 0, context)
!= 0 || VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&blkcnt,
0, context) != 0) {
error = EINVAL;
goto out;
}
*size = blkcnt * (uint64_t)blksize;
/*
* ### APPLE TODO ###
* If we own the whole disk, try to enable disk write caching.
*/
/*
* Take the device's minimum transfer size into account.
*/
*ashift = highbit(MAX(blksize, SPA_MINBLOCKSIZE)) - 1;
/*
* Setting the vdev_ashift did in fact break the pool for import
* on ZEVO. This puts the logic into question. It appears that vdev_top
* will also then change. It then panics in space_map from metaslab_alloc
*/
//vd->vdev_ashift = *ashift;
dvd->vd_ashift = *ashift;
/*
* Clear the nowritecache bit, so that on a vdev_reopen() we will
* try again.
*/
vd->vdev_nowritecache = B_FALSE;
vd->vdev_tsd = dvd;
dvd->vd_devvp = devvp;
out:
if (error) {
if (devvp)
vnode_close(devvp, fmode, context);
if (dvd)
kmem_free(dvd, sizeof (vdev_disk_t));
/*
* Since the open has failed, vd->vdev_tsd should
* be NULL when we get here, signaling to the
* rest of the spa not to try and reopen or close this device
*/
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
}
if (context) {
(void) vfs_context_rele(context);
}
return (error);
}
int
udf_mountfs(struct vnode *devvp, struct mount *mp, uint32_t lb, struct proc *p)
{
struct buf *bp = NULL;
struct anchor_vdp avdp;
struct umount *ump = NULL;
struct part_desc *pd;
struct logvol_desc *lvd;
struct fileset_desc *fsd;
struct extfile_entry *xfentry;
struct file_entry *fentry;
uint32_t sector, size, mvds_start, mvds_end;
uint32_t fsd_offset = 0;
uint16_t part_num = 0, fsd_part = 0;
int error = EINVAL;
int logvol_found = 0, part_found = 0, fsd_found = 0;
int bsize;
/*
* 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)))
return (error);
if (vcount(devvp) > 1 && devvp != rootvp)
return (EBUSY);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
error = vinvalbuf(devvp, V_SAVE, p->p_ucred, p, 0, 0);
VOP_UNLOCK(devvp, 0, p);
if (error)
return (error);
error = VOP_OPEN(devvp, FREAD, FSCRED, p);
if (error)
return (error);
ump = malloc(sizeof(*ump), M_UDFMOUNT, M_WAITOK | M_ZERO);
mp->mnt_data = (qaddr_t) ump;
mp->mnt_stat.f_fsid.val[0] = devvp->v_rdev;
mp->mnt_stat.f_fsid.val[1] = makefstype(MOUNT_UDF);
mp->mnt_flag |= MNT_LOCAL;
ump->um_mountp = mp;
ump->um_dev = devvp->v_rdev;
ump->um_devvp = devvp;
bsize = 2048; /* Should probe the media for its size. */
/*
* Get the Anchor Volume Descriptor Pointer from sector 256.
* Should also check sector n - 256, n, and 512.
*/
sector = 256;
if ((error = bread(devvp, sector * btodb(bsize), bsize, NOCRED,
&bp)) != 0)
goto bail;
if ((error = udf_checktag((struct desc_tag *)bp->b_data, TAGID_ANCHOR)))
goto bail;
bcopy(bp->b_data, &avdp, sizeof(struct anchor_vdp));
brelse(bp);
bp = NULL;
/*
* Extract the Partition Descriptor and Logical Volume Descriptor
* from the Volume Descriptor Sequence.
* Should we care about the partition type right now?
* What about multiple partitions?
*/
mvds_start = letoh32(avdp.main_vds_ex.loc);
mvds_end = mvds_start + (letoh32(avdp.main_vds_ex.len) - 1) / bsize;
for (sector = mvds_start; sector < mvds_end; sector++) {
if ((error = bread(devvp, sector * btodb(bsize), bsize,
NOCRED, &bp)) != 0) {
printf("Can't read sector %d of VDS\n", sector);
goto bail;
}
lvd = (struct logvol_desc *)bp->b_data;
if (!udf_checktag(&lvd->tag, TAGID_LOGVOL)) {
ump->um_bsize = letoh32(lvd->lb_size);
ump->um_bmask = ump->um_bsize - 1;
ump->um_bshift = ffs(ump->um_bsize) - 1;
fsd_part = letoh16(lvd->_lvd_use.fsd_loc.loc.part_num);
fsd_offset = letoh32(lvd->_lvd_use.fsd_loc.loc.lb_num);
if (udf_find_partmaps(ump, lvd))
break;
logvol_found = 1;
}
pd = (struct part_desc *)bp->b_data;
if (!udf_checktag(&pd->tag, TAGID_PARTITION)) {
part_found = 1;
part_num = letoh16(pd->part_num);
ump->um_len = ump->um_reallen = letoh32(pd->part_len);
ump->um_start = ump->um_realstart = letoh32(pd->start_loc);
}
brelse(bp);
bp = NULL;
if ((part_found) && (logvol_found))
break;
}
if (!part_found || !logvol_found) {
error = EINVAL;
goto bail;
}
if (ISSET(ump->um_flags, UDF_MNT_USES_META)) {
/* Read Metadata File 'File Entry' to find Metadata file. */
struct long_ad *la;
sector = ump->um_start + ump->um_meta_start; /* Set in udf_get_mpartmap() */
if ((error = RDSECTOR(devvp, sector, ump->um_bsize, &bp)) != 0) {
printf("Cannot read sector %d for Metadata File Entry\n", sector);
error = EINVAL;
goto bail;
}
xfentry = (struct extfile_entry *)bp->b_data;
fentry = (struct file_entry *)bp->b_data;
if (udf_checktag(&xfentry->tag, TAGID_EXTFENTRY) == 0)
la = (struct long_ad *)&xfentry->data[letoh32(xfentry->l_ea)];
else if (udf_checktag(&fentry->tag, TAGID_FENTRY) == 0)
la = (struct long_ad *)&fentry->data[letoh32(fentry->l_ea)];
else {
printf("Invalid Metadata File FE @ sector %d! (tag.id %d)\n",
sector, fentry->tag.id);
error = EINVAL;
goto bail;
}
ump->um_meta_start = letoh32(la->loc.lb_num);
ump->um_meta_len = letoh32(la->len);
if (bp != NULL) {
brelse(bp);
bp = NULL;
}
} else if (fsd_part != part_num) {
printf("FSD does not lie within the partition!\n");
error = EINVAL;
goto bail;
}
mtx_init(&ump->um_hashmtx, IPL_NONE);
ump->um_hashtbl = hashinit(UDF_HASHTBLSIZE, M_UDFMOUNT, M_WAITOK,
&ump->um_hashsz);
/* Get the VAT, if needed */
if (ump->um_flags & UDF_MNT_FIND_VAT) {
error = udf_vat_get(ump, lb);
if (error)
goto bail;
}
/*
* Grab the Fileset Descriptor
* Thanks to Chuck McCrobie <*****@*****.**> for pointing
* me in the right direction here.
*/
if (ISSET(ump->um_flags, UDF_MNT_USES_META))
sector = ump->um_meta_start;
else
sector = fsd_offset;
udf_vat_map(ump, §or);
if ((error = RDSECTOR(devvp, sector, ump->um_bsize, &bp)) != 0) {
printf("Cannot read sector %d of FSD\n", sector);
goto bail;
}
fsd = (struct fileset_desc *)bp->b_data;
if (!udf_checktag(&fsd->tag, TAGID_FSD)) {
fsd_found = 1;
bcopy(&fsd->rootdir_icb, &ump->um_root_icb,
sizeof(struct long_ad));
if (ISSET(ump->um_flags, UDF_MNT_USES_META)) {
ump->um_root_icb.loc.lb_num += ump->um_meta_start;
ump->um_root_icb.loc.part_num = part_num;
}
}
brelse(bp);
bp = NULL;
if (!fsd_found) {
printf("Couldn't find the fsd\n");
error = EINVAL;
goto bail;
}
/*
* Find the file entry for the root directory.
*/
sector = letoh32(ump->um_root_icb.loc.lb_num);
size = letoh32(ump->um_root_icb.len);
udf_vat_map(ump, §or);
if ((error = udf_readlblks(ump, sector, size, &bp)) != 0) {
printf("Cannot read sector %d\n", sector);
goto bail;
}
xfentry = (struct extfile_entry *)bp->b_data;
fentry = (struct file_entry *)bp->b_data;
error = udf_checktag(&xfentry->tag, TAGID_EXTFENTRY);
if (error) {
error = udf_checktag(&fentry->tag, TAGID_FENTRY);
if (error) {
printf("Invalid root file entry!\n");
goto bail;
}
}
brelse(bp);
bp = NULL;
devvp->v_specmountpoint = mp;
return (0);
bail:
if (ump->um_hashtbl != NULL)
free(ump->um_hashtbl, M_UDFMOUNT);
if (ump != NULL) {
free(ump, M_UDFMOUNT);
mp->mnt_data = NULL;
mp->mnt_flag &= ~MNT_LOCAL;
}
if (bp != NULL)
brelse(bp);
vn_lock(devvp, LK_EXCLUSIVE|LK_RETRY, p);
VOP_CLOSE(devvp, FREAD, FSCRED, p);
VOP_UNLOCK(devvp, 0, p);
return (error);
}
/*
* 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);
}
/*
* Common code for mount and mountroot
*/
int
ext2fs_mountfs(struct vnode *devvp, struct mount *mp, struct proc *p)
{
struct ufsmount *ump;
struct buf *bp;
struct ext2fs *fs;
struct m_ext2fs *m_fs;
dev_t dev;
int error, i, ronly;
struct ucred *cred;
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);
if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
return (error);
ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED);
if (error)
return (error);
bp = NULL;
ump = NULL;
#ifdef DEBUG_EXT2
printf("ext2 sb size: %d\n", sizeof(struct ext2fs));
#endif
error = bread(devvp, (daddr_t)(SBOFF / DEV_BSIZE), SBSIZE, &bp);
if (error)
goto out;
fs = (struct ext2fs *)bp->b_data;
error = ext2fs_checksb(fs, ronly);
if (error)
goto out;
ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK | M_ZERO);
ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT,
M_WAITOK | M_ZERO);
e2fs_sbload((struct ext2fs*)bp->b_data, &ump->um_e2fs->e2fs);
brelse(bp);
bp = NULL;
m_fs = ump->um_e2fs;
m_fs->e2fs_ronly = ronly;
ump->um_fstype = UM_EXT2FS;
#ifdef DEBUG_EXT2
printf("ext2 ino size %d\n", EXT2_DINODE_SIZE(m_fs));
#endif
if (ronly == 0) {
if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
m_fs->e2fs.e2fs_state = 0;
else
m_fs->e2fs.e2fs_state = E2FS_ERRORS;
m_fs->e2fs_fmod = 1;
}
/* compute dynamic sb infos */
m_fs->e2fs_ncg =
howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
m_fs->e2fs.e2fs_bpg);
/* XXX assume hw bsize = 512 */
m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + 1;
m_fs->e2fs_bsize = 1024 << m_fs->e2fs.e2fs_log_bsize;
m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg,
m_fs->e2fs_bsize / sizeof(struct ext2_gd));
m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE(m_fs);
m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg/m_fs->e2fs_ipb;
m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
M_UFSMNT, M_WAITOK);
for (i=0; i < m_fs->e2fs_ngdb; i++) {
error = bread(devvp ,
fsbtodb(m_fs, ((m_fs->e2fs_bsize>1024)? 0 : 1) + i + 1),
m_fs->e2fs_bsize, &bp);
if (error) {
free(m_fs->e2fs_gd, M_UFSMNT);
goto out;
}
e2fs_cgload((struct ext2_gd*)bp->b_data,
&m_fs->e2fs_gd[i * m_fs->e2fs_bsize
/ sizeof(struct ext2_gd)],
m_fs->e2fs_bsize);
brelse(bp);
bp = NULL;
}
mp->mnt_data = (qaddr_t)ump;
mp->mnt_stat.f_fsid.val[0] = (long)dev;
mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
mp->mnt_flag |= MNT_LOCAL;
ump->um_mountp = mp;
ump->um_dev = dev;
ump->um_devvp = devvp;
ump->um_nindir = NINDIR(m_fs);
ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
ump->um_seqinc = 1; /* no frags */
devvp->v_specmountpoint = mp;
return (0);
out:
if (bp)
brelse(bp);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY, p);
(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, cred);
VOP_UNLOCK(devvp, 0);
if (ump) {
free(ump->um_e2fs, M_UFSMNT);
free(ump, M_UFSMNT);
mp->mnt_data = (qaddr_t)0;
}
return (error);
}
/*
* Open a special file.
*/
int
spec_open(struct vnop_open_args *ap)
{
struct proc *p = vfs_context_proc(ap->a_context);
kauth_cred_t cred = vfs_context_ucred(ap->a_context);
struct vnode *vp = ap->a_vp;
dev_t bdev, dev = (dev_t)vp->v_rdev;
int maj = major(dev);
int error;
/*
* Don't allow open if fs is mounted -nodev.
*/
if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
return (ENXIO);
switch (vp->v_type) {
case VCHR:
if ((u_int)maj >= (u_int)nchrdev)
return (ENXIO);
if (cred != FSCRED && (ap->a_mode & FWRITE)) {
/*
* When running in very secure mode, do not allow
* opens for writing of any disk character devices.
*/
if (securelevel >= 2 && isdisk(dev, VCHR))
return (EPERM);
/*
* When running in secure mode, do not allow opens
* for writing of /dev/mem, /dev/kmem, or character
* devices whose corresponding block devices are
* currently mounted.
*/
if (securelevel >= 1) {
if ((bdev = chrtoblk(dev)) != NODEV && check_mountedon(bdev, VBLK, &error))
return (error);
if (iskmemdev(dev))
return (EPERM);
}
}
if (cdevsw[maj].d_type == D_TTY) {
vnode_lock(vp);
vp->v_flag |= VISTTY;
vnode_unlock(vp);
}
error = (*cdevsw[maj].d_open)(dev, ap->a_mode, S_IFCHR, p);
return (error);
case VBLK:
if ((u_int)maj >= (u_int)nblkdev)
return (ENXIO);
/*
* When running in very secure mode, do not allow
* opens for writing of any disk block devices.
*/
if (securelevel >= 2 && cred != FSCRED &&
(ap->a_mode & FWRITE) && bdevsw[maj].d_type == D_DISK)
return (EPERM);
/*
* Do not allow opens of block devices that are
* currently mounted.
*/
if ( (error = vfs_mountedon(vp)) )
return (error);
error = (*bdevsw[maj].d_open)(dev, ap->a_mode, S_IFBLK, p);
if (!error) {
u_int64_t blkcnt;
u_int32_t blksize;
int setsize = 0;
u_int32_t size512 = 512;
if (!VNOP_IOCTL(vp, DKIOCGETBLOCKSIZE, (caddr_t)&blksize, 0, ap->a_context)) {
/* Switch to 512 byte sectors (temporarily) */
if (!VNOP_IOCTL(vp, DKIOCSETBLOCKSIZE, (caddr_t)&size512, FWRITE, ap->a_context)) {
/* Get the number of 512 byte physical blocks. */
if (!VNOP_IOCTL(vp, DKIOCGETBLOCKCOUNT, (caddr_t)&blkcnt, 0, ap->a_context)) {
setsize = 1;
}
}
/* If it doesn't set back, we can't recover */
if (VNOP_IOCTL(vp, DKIOCSETBLOCKSIZE, (caddr_t)&blksize, FWRITE, ap->a_context))
error = ENXIO;
}
vnode_lock(vp);
set_blocksize(vp, dev);
/*
* Cache the size in bytes of the block device for later
* use by spec_write().
*/
if (setsize)
vp->v_specdevsize = blkcnt * (u_int64_t)size512;
else
vp->v_specdevsize = (u_int64_t)0; /* Default: Can't get */
vnode_unlock(vp);
}
return(error);
default:
panic("spec_open type");
}
return (0);
}
/*
* Common code for mount and mountroot
*/
int
ntfs_mountfs(struct vnode *devvp, struct mount *mp, struct ntfs_args *argsp,
struct ucred *cred)
{
struct buf *bp;
struct ntfsmount *ntmp;
cdev_t dev;
int error, ronly, ncount, i;
struct vnode *vp;
char cs_local[ICONV_CSNMAXLEN];
char cs_ntfs[ICONV_CSNMAXLEN];
/*
* 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.
*/
error = vfs_mountedon(devvp);
if (error)
return (error);
ncount = vcount(devvp);
if (devvp->v_object)
ncount -= 1;
if (ncount > 1)
return (EBUSY);
VN_LOCK(devvp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(devvp, V_SAVE, 0, 0);
VOP__UNLOCK(devvp, 0);
if (error)
return (error);
ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
VN_LOCK(devvp, LK_EXCLUSIVE | LK_RETRY);
error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, NULL);
VOP__UNLOCK(devvp, 0);
if (error)
return (error);
dev = devvp->v_rdev;
bp = NULL;
error = bread(devvp, BBLOCK, BBSIZE, &bp);
if (error)
goto out;
ntmp = kmalloc(sizeof *ntmp, M_NTFSMNT, M_WAITOK | M_ZERO);
bcopy( bp->b_data, &ntmp->ntm_bootfile, sizeof(struct bootfile) );
/*
* We must not cache the boot block if its size is not exactly
* one cluster in order to avoid confusing the buffer cache when
* the boot file is read later by ntfs_readntvattr_plain(), which
* reads a cluster at a time.
*/
if (ntfs_cntob(1) != BBSIZE)
bp->b_flags |= B_NOCACHE;
brelse( bp );
bp = NULL;
if (strncmp(ntmp->ntm_bootfile.bf_sysid, NTFS_BBID, NTFS_BBIDLEN)) {
error = EINVAL;
dprintf(("ntfs_mountfs: invalid boot block\n"));
goto out;
}
{
int8_t cpr = ntmp->ntm_mftrecsz;
if( cpr > 0 )
ntmp->ntm_bpmftrec = ntmp->ntm_spc * cpr;
else
ntmp->ntm_bpmftrec = (1 << (-cpr)) / ntmp->ntm_bps;
}
dprintf(("ntfs_mountfs(): bps: %d, spc: %d, media: %x, mftrecsz: %d (%d sects)\n",
ntmp->ntm_bps,ntmp->ntm_spc,ntmp->ntm_bootfile.bf_media,
ntmp->ntm_mftrecsz,ntmp->ntm_bpmftrec));
dprintf(("ntfs_mountfs(): mftcn: 0x%x|0x%x\n",
(u_int32_t)ntmp->ntm_mftcn,(u_int32_t)ntmp->ntm_mftmirrcn));
ntmp->ntm_mountp = mp;
ntmp->ntm_dev = dev;
ntmp->ntm_devvp = devvp;
ntmp->ntm_uid = argsp->uid;
ntmp->ntm_gid = argsp->gid;
ntmp->ntm_mode = argsp->mode;
ntmp->ntm_flag = argsp->flag;
if (argsp->flag & NTFS_MFLAG_KICONV && ntfs_iconv) {
bcopy(argsp->cs_local, cs_local, sizeof(cs_local));
bcopy(argsp->cs_ntfs, cs_ntfs, sizeof(cs_ntfs));
ntfs_82u_init(ntmp, cs_local, cs_ntfs);
ntfs_u28_init(ntmp, NULL, cs_local, cs_ntfs);
} else {
ntfs_82u_init(ntmp, NULL, NULL);
ntfs_u28_init(ntmp, ntmp->ntm_82u, NULL, NULL);
}
mp->mnt_data = (qaddr_t)ntmp;
dprintf(("ntfs_mountfs(): case-%s,%s uid: %d, gid: %d, mode: %o\n",
(ntmp->ntm_flag & NTFS_MFLAG_CASEINS)?"insens.":"sens.",
(ntmp->ntm_flag & NTFS_MFLAG_ALLNAMES)?" allnames,":"",
ntmp->ntm_uid, ntmp->ntm_gid, ntmp->ntm_mode));
vfs_add_vnodeops(mp, &ntfs_vnode_vops, &mp->mnt_vn_norm_ops);
/*
* We read in some system nodes to do not allow
* reclaim them and to have everytime access to them.
*/
{
int pi[3] = { NTFS_MFTINO, NTFS_ROOTINO, NTFS_BITMAPINO };
for (i=0; i<3; i++) {
error = VFS_VGET(mp, NULL,
pi[i], &(ntmp->ntm_sysvn[pi[i]]));
if(error)
goto out1;
vsetflags(ntmp->ntm_sysvn[pi[i]], VSYSTEM);
vref(ntmp->ntm_sysvn[pi[i]]);
vput(ntmp->ntm_sysvn[pi[i]]);
}
}
/* read the Unicode lowercase --> uppercase translation table,
* if necessary */
if ((error = ntfs_toupper_use(mp, ntmp)))
goto out1;
/*
* Scan $BitMap and count free clusters
*/
error = ntfs_calccfree(ntmp, &ntmp->ntm_cfree);
if(error)
goto out1;
/*
* Read and translate to internal format attribute
* definition file.
*/
{
int num,j;
struct attrdef ad;
/* Open $AttrDef */
error = VFS_VGET(mp, NULL, NTFS_ATTRDEFINO, &vp);
if(error)
goto out1;
/* Count valid entries */
for(num=0;;num++) {
error = ntfs_readattr(ntmp, VTONT(vp),
NTFS_A_DATA, NULL,
num * sizeof(ad), sizeof(ad),
&ad, NULL);
if (error)
goto out1;
if (ad.ad_name[0] == 0)
break;
}
/* Alloc memory for attribute definitions */
ntmp->ntm_ad = kmalloc(num * sizeof(struct ntvattrdef),
M_NTFSMNT, M_WAITOK);
ntmp->ntm_adnum = num;
/* Read them and translate */
for(i=0;i<num;i++){
error = ntfs_readattr(ntmp, VTONT(vp),
NTFS_A_DATA, NULL,
i * sizeof(ad), sizeof(ad),
&ad, NULL);
if (error)
goto out1;
j = 0;
do {
ntmp->ntm_ad[i].ad_name[j] = ad.ad_name[j];
} while(ad.ad_name[j++]);
ntmp->ntm_ad[i].ad_namelen = j - 1;
ntmp->ntm_ad[i].ad_type = ad.ad_type;
}
vput(vp);
}
mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
mp->mnt_maxsymlinklen = 0;
mp->mnt_flag |= MNT_LOCAL;
dev->si_mountpoint = mp;
return (0);
out1:
for(i=0;i<NTFS_SYSNODESNUM;i++)
if(ntmp->ntm_sysvn[i]) vrele(ntmp->ntm_sysvn[i]);
if (vflush(mp, 0, 0))
dprintf(("ntfs_mountfs: vflush failed\n"));
out:
dev->si_mountpoint = NULL;
if (bp)
brelse(bp);
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
(void)VOP_CLOSE(devvp, ronly ? FREAD : FREAD|FWRITE, NULL);
vn_unlock(devvp);
return (error);
}
/************************************************************************
* VOLUMES *
************************************************************************
*
* Load a HAMMER volume by name. Returns 0 on success or a positive error
* code on failure. Volumes must be loaded at mount time, get_volume() will
* not load a new volume.
*
* The passed devvp is vref()'d but not locked. This function consumes the
* ref (typically by associating it with the volume structure).
*
* Calls made to hammer_load_volume() or single-threaded
*/
int
hammer_install_volume(struct hammer_mount *hmp, const char *volname,
struct vnode *devvp)
{
struct mount *mp;
hammer_volume_t volume;
struct hammer_volume_ondisk *ondisk;
struct nlookupdata nd;
struct buf *bp = NULL;
int error;
int ronly;
int setmp = 0;
mp = hmp->mp;
ronly = ((mp->mnt_flag & MNT_RDONLY) ? 1 : 0);
/*
* Allocate a volume structure
*/
++hammer_count_volumes;
volume = kmalloc(sizeof(*volume), hmp->m_misc, M_WAITOK|M_ZERO);
volume->vol_name = kstrdup(volname, hmp->m_misc);
volume->io.hmp = hmp; /* bootstrap */
hammer_io_init(&volume->io, volume, HAMMER_STRUCTURE_VOLUME);
volume->io.offset = 0LL;
volume->io.bytes = HAMMER_BUFSIZE;
/*
* Get the device vnode
*/
if (devvp == NULL) {
error = nlookup_init(&nd, volume->vol_name, UIO_SYSSPACE, NLC_FOLLOW);
if (error == 0)
error = nlookup(&nd);
if (error == 0)
error = cache_vref(&nd.nl_nch, nd.nl_cred, &volume->devvp);
nlookup_done(&nd);
} else {
error = 0;
volume->devvp = devvp;
}
if (error == 0) {
if (vn_isdisk(volume->devvp, &error)) {
error = vfs_mountedon(volume->devvp);
}
}
if (error == 0 && vcount(volume->devvp) > 0)
error = EBUSY;
if (error == 0) {
vn_lock(volume->devvp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(volume->devvp, V_SAVE, 0, 0);
if (error == 0) {
error = VOP_OPEN(volume->devvp,
(ronly ? FREAD : FREAD|FWRITE),
FSCRED, NULL);
}
vn_unlock(volume->devvp);
}
if (error) {
hammer_free_volume(volume);
return(error);
}
volume->devvp->v_rdev->si_mountpoint = mp;
setmp = 1;
/*
* Extract the volume number from the volume header and do various
* sanity checks.
*/
error = bread(volume->devvp, 0LL, HAMMER_BUFSIZE, &bp);
if (error)
goto late_failure;
ondisk = (void *)bp->b_data;
if (ondisk->vol_signature != HAMMER_FSBUF_VOLUME) {
kprintf("hammer_mount: volume %s has an invalid header\n",
volume->vol_name);
error = EFTYPE;
goto late_failure;
}
volume->vol_no = ondisk->vol_no;
volume->buffer_base = ondisk->vol_buf_beg;
volume->vol_flags = ondisk->vol_flags;
volume->nblocks = ondisk->vol_nblocks;
volume->maxbuf_off = HAMMER_ENCODE_RAW_BUFFER(volume->vol_no,
ondisk->vol_buf_end - ondisk->vol_buf_beg);
volume->maxraw_off = ondisk->vol_buf_end;
if (RB_EMPTY(&hmp->rb_vols_root)) {
hmp->fsid = ondisk->vol_fsid;
} else if (bcmp(&hmp->fsid, &ondisk->vol_fsid, sizeof(uuid_t))) {
kprintf("hammer_mount: volume %s's fsid does not match "
"other volumes\n", volume->vol_name);
error = EFTYPE;
goto late_failure;
}
/*
* Insert the volume structure into the red-black tree.
*/
if (RB_INSERT(hammer_vol_rb_tree, &hmp->rb_vols_root, volume)) {
kprintf("hammer_mount: volume %s has a duplicate vol_no %d\n",
volume->vol_name, volume->vol_no);
error = EEXIST;
}
/*
* Set the root volume . HAMMER special cases rootvol the structure.
* We do not hold a ref because this would prevent related I/O
* from being flushed.
*/
if (error == 0 && ondisk->vol_rootvol == ondisk->vol_no) {
hmp->rootvol = volume;
hmp->nvolumes = ondisk->vol_count;
if (bp) {
brelse(bp);
bp = NULL;
}
hmp->mp->mnt_stat.f_blocks += ondisk->vol0_stat_bigblocks *
(HAMMER_LARGEBLOCK_SIZE / HAMMER_BUFSIZE);
hmp->mp->mnt_vstat.f_blocks += ondisk->vol0_stat_bigblocks *
(HAMMER_LARGEBLOCK_SIZE / HAMMER_BUFSIZE);
}
late_failure:
if (bp)
brelse(bp);
if (error) {
/*vinvalbuf(volume->devvp, V_SAVE, 0, 0);*/
if (setmp)
volume->devvp->v_rdev->si_mountpoint = NULL;
vn_lock(volume->devvp, LK_EXCLUSIVE | LK_RETRY);
VOP_CLOSE(volume->devvp, ronly ? FREAD : FREAD|FWRITE);
vn_unlock(volume->devvp);
hammer_free_volume(volume);
}
return (error);
}
/*
* Common code for mount and mountroot
*/
int
ext2fs_mountfs(struct vnode *devvp, struct mount *mp, struct proc *p)
{
struct ufsmount *ump;
struct buf *bp;
struct ext2fs *fs;
dev_t dev;
int error, ronly;
struct ucred *cred;
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);
if ((error = vinvalbuf(devvp, V_SAVE, cred, p, 0, 0)) != 0)
return (error);
ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
error = VOP_OPEN(devvp, ronly ? FREAD : FREAD|FWRITE, FSCRED, p);
if (error)
return (error);
bp = NULL;
ump = NULL;
/*
* Read the superblock from disk.
*/
error = bread(devvp, (daddr_t)(SBOFF / DEV_BSIZE), SBSIZE, &bp);
if (error)
goto out;
fs = (struct ext2fs *)bp->b_data;
error = e2fs_sbcheck(fs, ronly);
if (error)
goto out;
ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK | M_ZERO);
ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT,
M_WAITOK | M_ZERO);
/*
* Copy in the superblock, compute in-memory values
* and load group descriptors.
*/
e2fs_sbload(fs, &ump->um_e2fs->e2fs);
if ((error = e2fs_sbfill(devvp, ump->um_e2fs)) != 0)
goto out;
brelse(bp);
bp = NULL;
fs = &ump->um_e2fs->e2fs;
ump->um_e2fs->e2fs_ronly = ronly;
ump->um_fstype = UM_EXT2FS;
if (ronly == 0) {
if (fs->e2fs_state == E2FS_ISCLEAN)
fs->e2fs_state = 0;
else
fs->e2fs_state = E2FS_ERRORS;
ump->um_e2fs->e2fs_fmod = 1;
}
mp->mnt_data = (qaddr_t)ump;
mp->mnt_stat.f_fsid.val[0] = (long)dev;
mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
mp->mnt_stat.f_namemax = MAXNAMLEN;
mp->mnt_flag |= MNT_LOCAL;
ump->um_mountp = mp;
ump->um_dev = dev;
ump->um_devvp = devvp;
ump->um_nindir = NINDIR(ump->um_e2fs);
ump->um_bptrtodb = ump->um_e2fs->e2fs_fsbtodb;
ump->um_seqinc = 1; /* no frags */
ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
devvp->v_specmountpoint = mp;
return (0);
out:
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, p);
if (ump) {
free(ump->um_e2fs, M_UFSMNT, sizeof *ump->um_e2fs);
free(ump, M_UFSMNT, sizeof *ump);
mp->mnt_data = NULL;
}
return (error);
}
static int
devfs_spec_open(struct vop_open_args *ap)
{
struct vnode *vp = ap->a_vp;
struct vnode *orig_vp = NULL;
struct devfs_node *node = DEVFS_NODE(vp);
struct devfs_node *newnode;
cdev_t dev, ndev = NULL;
int error = 0;
if (node) {
if (node->d_dev == NULL)
return ENXIO;
if (!devfs_node_is_accessible(node))
return ENOENT;
}
if ((dev = vp->v_rdev) == NULL)
return ENXIO;
vn_lock(vp, LK_UPGRADE | LK_RETRY);
if (node && ap->a_fp) {
devfs_debug(DEVFS_DEBUG_DEBUG, "devfs_spec_open: -1.1-\n");
lockmgr(&devfs_lock, LK_EXCLUSIVE);
ndev = devfs_clone(dev, node->d_dir.d_name,
node->d_dir.d_namlen,
ap->a_mode, ap->a_cred);
if (ndev != NULL) {
newnode = devfs_create_device_node(
DEVFS_MNTDATA(vp->v_mount)->root_node,
ndev, NULL, NULL);
/* XXX: possibly destroy device if this happens */
if (newnode != NULL) {
dev = ndev;
devfs_link_dev(dev);
devfs_debug(DEVFS_DEBUG_DEBUG,
"parent here is: %s, node is: |%s|\n",
((node->parent->node_type == Nroot) ?
"ROOT!" : node->parent->d_dir.d_name),
newnode->d_dir.d_name);
devfs_debug(DEVFS_DEBUG_DEBUG,
"test: %s\n",
((struct devfs_node *)(TAILQ_LAST(DEVFS_DENODE_HEAD(node->parent), devfs_node_head)))->d_dir.d_name);
/*
* orig_vp is set to the original vp if we cloned.
*/
/* node->flags |= DEVFS_CLONED; */
devfs_allocv(&vp, newnode);
orig_vp = ap->a_vp;
ap->a_vp = vp;
}
}
lockmgr(&devfs_lock, LK_RELEASE);
}
devfs_debug(DEVFS_DEBUG_DEBUG,
"devfs_spec_open() called on %s! \n",
dev->si_name);
/*
* Make this field valid before any I/O in ->d_open
*/
if (!dev->si_iosize_max)
/* XXX: old DFLTPHYS == 64KB dependency */
dev->si_iosize_max = min(MAXPHYS,64*1024);
if (dev_dflags(dev) & D_TTY)
vsetflags(vp, VISTTY);
/*
* Open underlying device
*/
vn_unlock(vp);
error = dev_dopen(dev, ap->a_mode, S_IFCHR, ap->a_cred, ap->a_fp);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
/*
* Clean up any cloned vp if we error out.
*/
if (error) {
if (orig_vp) {
vput(vp);
ap->a_vp = orig_vp;
/* orig_vp = NULL; */
}
return error;
}
/*
* This checks if the disk device is going to be opened for writing.
* It will be only allowed in the cases where securelevel permits it
* and it's not mounted R/W.
*/
if ((dev_dflags(dev) & D_DISK) && (ap->a_mode & FWRITE) &&
(ap->a_cred != FSCRED)) {
/* Very secure mode. No open for writing allowed */
if (securelevel >= 2)
return EPERM;
/*
* If it is mounted R/W, do not allow to open for writing.
* In the case it's mounted read-only but securelevel
* is >= 1, then do not allow opening for writing either.
*/
if (vfs_mountedon(vp)) {
if (!(dev->si_mountpoint->mnt_flag & MNT_RDONLY))
return EBUSY;
else if (securelevel >= 1)
return EPERM;
}
}
if (dev_dflags(dev) & D_TTY) {
if (dev->si_tty) {
struct tty *tp;
tp = dev->si_tty;
if (!tp->t_stop) {
devfs_debug(DEVFS_DEBUG_DEBUG,
"devfs: no t_stop\n");
tp->t_stop = nottystop;
}
}
}
if (vn_isdisk(vp, NULL)) {
if (!dev->si_bsize_phys)
dev->si_bsize_phys = DEV_BSIZE;
vinitvmio(vp, IDX_TO_OFF(INT_MAX), PAGE_SIZE, -1);
}
vop_stdopen(ap);
#if 0
if (node)
nanotime(&node->atime);
#endif
/*
* If we replaced the vp the vop_stdopen() call will have loaded
* it into fp->f_data and vref()d the vp, giving us two refs. So
* instead of just unlocking it here we have to vput() it.
*/
if (orig_vp)
vput(vp);
/* Ugly pty magic, to make pty devices appear once they are opened */
if (node && (node->flags & DEVFS_PTY) == DEVFS_PTY)
node->flags &= ~DEVFS_INVISIBLE;
if (ap->a_fp) {
KKASSERT(ap->a_fp->f_type == DTYPE_VNODE);
KKASSERT((ap->a_fp->f_flag & FMASK) == (ap->a_mode & FMASK));
ap->a_fp->f_ops = &devfs_dev_fileops;
KKASSERT(ap->a_fp->f_data == (void *)vp);
}
return 0;
}
