X-Git-Url: https://git.camperquake.de/gitweb.cgi?a=blobdiff_plain;f=module%2Fzfs%2Fzfs_vnops.c;h=876d44b3563dae38e3989c632c356d2b7b9637b1;hb=refs%2Fheads%2Frertzinger%2Ffeature-zpool-get--p;hp=a0720079cf4679583c3dd088d3f2604e4e0d86ef;hpb=572e285762521df27fe5b026f409ba1a21abb7ac;p=zfs.git diff --git a/module/zfs/zfs_vnops.c b/module/zfs/zfs_vnops.c index a072007..876d44b 100644 --- a/module/zfs/zfs_vnops.c +++ b/module/zfs/zfs_vnops.c @@ -20,11 +20,13 @@ */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2012 by Delphix. All rights reserved. */ /* Portions Copyright 2007 Jeremy Teo */ /* Portions Copyright 2010 Robert Milkowski */ + #include #include #include @@ -33,7 +35,6 @@ #include #include #include -#include #include #include #include @@ -41,13 +42,7 @@ #include #include #include -#include -#include #include -#include -#include -#include -#include #include #include #include @@ -66,18 +61,20 @@ #include #include #include -#include #include +#include #include "fs/fs_subr.h" #include #include #include +#include #include #include #include #include #include #include +#include /* * Programming rules. @@ -91,12 +88,12 @@ * to freed memory. The example below illustrates the following Big Rules: * * (1) A check must be made in each zfs thread for a mounted file system. - * This is done avoiding races using ZFS_ENTER(zfsvfs). - * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes + * This is done avoiding races using ZFS_ENTER(zsb). + * A ZFS_EXIT(zsb) is needed before all returns. Any znodes * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros * can return EIO from the calling function. * - * (2) VN_RELE() should always be the last thing except for zil_commit() + * (2) iput() should always be the last thing except for zil_commit() * (if necessary) and ZFS_EXIT(). This is for 3 reasons: * First, if it's the last reference, the vnode/znode * can be freed, so the zp may point to freed memory. Second, the last @@ -104,7 +101,7 @@ * pushing cached pages (which acquires range locks) and syncing out * cached atime changes. Third, zfs_zinactive() may require a new tx, * which could deadlock the system if you were already holding one. - * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC(). + * If you must call iput() within a tx then use iput_ASYNC(). * * (3) All range locks must be grabbed before calling dmu_tx_assign(), * as they can span dmu_tx_assign() calls. @@ -120,7 +117,7 @@ * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() * forever, because the previous txg can't quiesce until B's tx commits. * - * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, + * If dmu_tx_assign() returns ERESTART and zsb->z_assign is TXG_NOWAIT, * then drop all locks, call dmu_tx_wait(), and try again. * * (5) If the operation succeeded, generate the intent log entry for it @@ -137,9 +134,9 @@ * * In general, this is how things should be ordered in each vnode op: * - * ZFS_ENTER(zfsvfs); // exit if unmounted + * ZFS_ENTER(zsb); // exit if unmounted * top: - * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD()) + * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) * rw_enter(...); // grab any other locks you need * tx = dmu_tx_create(...); // get DMU tx * dmu_tx_hold_*(); // hold each object you might modify @@ -147,14 +144,14 @@ * if (error) { * rw_exit(...); // drop locks * zfs_dirent_unlock(dl); // unlock directory entry - * VN_RELE(...); // release held vnodes + * iput(...); // release held vnodes * if (error == ERESTART) { * dmu_tx_wait(tx); * dmu_tx_abort(tx); * goto top; * } * dmu_tx_abort(tx); // abort DMU tx - * ZFS_EXIT(zfsvfs); // finished in zfs + * ZFS_EXIT(zsb); // finished in zfs * return (error); // really out of space * } * error = do_real_work(); // do whatever this VOP does @@ -163,83 +160,95 @@ * dmu_tx_commit(tx); // commit DMU tx -- error or not * rw_exit(...); // drop locks * zfs_dirent_unlock(dl); // unlock directory entry - * VN_RELE(...); // release held vnodes + * iput(...); // release held vnodes * zil_commit(zilog, foid); // synchronous when necessary - * ZFS_EXIT(zfsvfs); // finished in zfs + * ZFS_EXIT(zsb); // finished in zfs * return (error); // done, report error */ -/* ARGSUSED */ +/* + * Virus scanning is unsupported. It would be possible to add a hook + * here to performance the required virus scan. This could be done + * entirely in the kernel or potentially as an update to invoke a + * scanning utility. + */ static int -zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct) +zfs_vscan(struct inode *ip, cred_t *cr, int async) +{ + return (0); +} + +/* ARGSUSED */ +int +zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) { - znode_t *zp = VTOZ(*vpp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) && - ((flag & FAPPEND) == 0)) { - ZFS_EXIT(zfsvfs); + /* Honor ZFS_APPENDONLY file attribute */ + if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && + ((flag & O_APPEND) == 0)) { + ZFS_EXIT(zsb); return (EPERM); } - if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && - ZTOV(zp)->v_type == VREG && + /* Virus scan eligible files on open */ + if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { - if (fs_vscan(*vpp, cr, 0) != 0) { - ZFS_EXIT(zfsvfs); + if (zfs_vscan(ip, cr, 0) != 0) { + ZFS_EXIT(zsb); return (EACCES); } } /* Keep a count of the synchronous opens in the znode */ - if (flag & (FSYNC | FDSYNC)) + if (flag & O_SYNC) atomic_inc_32(&zp->z_sync_cnt); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (0); } +EXPORT_SYMBOL(zfs_open); /* ARGSUSED */ -static int -zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr, - caller_context_t *ct) +int +zfs_close(struct inode *ip, int flag, cred_t *cr) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); - /* - * Clean up any locks held by this process on the vp. - */ - cleanlocks(vp, ddi_get_pid(), 0); - cleanshares(vp, ddi_get_pid()); - - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - /* Decrement the synchronous opens in the znode */ - if ((flag & (FSYNC | FDSYNC)) && (count == 1)) - atomic_dec_32(&zp->z_sync_cnt); + /* + * Zero the synchronous opens in the znode. Under Linux the + * zfs_close() hook is not symmetric with zfs_open(), it is + * only called once when the last reference is dropped. + */ + if (flag & O_SYNC) + zp->z_sync_cnt = 0; - if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && - ZTOV(zp)->v_type == VREG && + if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) - VERIFY(fs_vscan(vp, cr, 1) == 0); + VERIFY(zfs_vscan(ip, cr, 1) == 0); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (0); } +EXPORT_SYMBOL(zfs_close); +#if defined(SEEK_HOLE) && defined(SEEK_DATA) /* - * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and - * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter. + * Lseek support for finding holes (cmd == SEEK_HOLE) and + * data (cmd == SEEK_DATA). "off" is an in/out parameter. */ static int -zfs_holey(vnode_t *vp, int cmd, offset_t *off) +zfs_holey_common(struct inode *ip, int cmd, loff_t *off) { - znode_t *zp = VTOZ(vp); + znode_t *zp = ITOZ(ip); uint64_t noff = (uint64_t)*off; /* new offset */ uint64_t file_sz; int error; @@ -250,12 +259,12 @@ zfs_holey(vnode_t *vp, int cmd, offset_t *off) return (ENXIO); } - if (cmd == _FIO_SEEK_HOLE) + if (cmd == SEEK_HOLE) hole = B_TRUE; else hole = B_FALSE; - error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff); + error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); /* end of file? */ if ((error == ESRCH) || (noff > file_sz)) { @@ -275,75 +284,25 @@ zfs_holey(vnode_t *vp, int cmd, offset_t *off) return (error); } -/* ARGSUSED */ -static int -zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred, - int *rvalp, caller_context_t *ct) +int +zfs_holey(struct inode *ip, int cmd, loff_t *off) { - offset_t off; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); int error; - zfsvfs_t *zfsvfs; - znode_t *zp; - - switch (com) { - case _FIOFFS: - return (zfs_sync(vp->v_vfsp, 0, cred)); - - /* - * The following two ioctls are used by bfu. Faking out, - * necessary to avoid bfu errors. - */ - case _FIOGDIO: - case _FIOSDIO: - return (0); - - case _FIO_SEEK_DATA: - case _FIO_SEEK_HOLE: - if (ddi_copyin((void *)data, &off, sizeof (off), flag)) - return (EFAULT); - - zp = VTOZ(vp); - zfsvfs = zp->z_zfsvfs; - ZFS_ENTER(zfsvfs); - ZFS_VERIFY_ZP(zp); - /* offset parameter is in/out */ - error = zfs_holey(vp, com, &off); - ZFS_EXIT(zfsvfs); - if (error) - return (error); - if (ddi_copyout(&off, (void *)data, sizeof (off), flag)) - return (EFAULT); - return (0); - } - return (ENOTTY); -} + ZFS_ENTER(zsb); + ZFS_VERIFY_ZP(zp); -/* - * Utility functions to map and unmap a single physical page. These - * are used to manage the mappable copies of ZFS file data, and therefore - * do not update ref/mod bits. - */ -caddr_t -zfs_map_page(page_t *pp, enum seg_rw rw) -{ - if (kpm_enable) - return (hat_kpm_mapin(pp, 0)); - ASSERT(rw == S_READ || rw == S_WRITE); - return (ppmapin(pp, PROT_READ | ((rw == S_WRITE) ? PROT_WRITE : 0), - (caddr_t)-1)); -} + error = zfs_holey_common(ip, cmd, off); -void -zfs_unmap_page(page_t *pp, caddr_t addr) -{ - if (kpm_enable) { - hat_kpm_mapout(pp, 0, addr); - } else { - ppmapout(addr); - } + ZFS_EXIT(zsb); + return (error); } +EXPORT_SYMBOL(zfs_holey); +#endif /* SEEK_HOLE && SEEK_DATA */ +#if defined(_KERNEL) /* * When a file is memory mapped, we must keep the IO data synchronized * between the DMU cache and the memory mapped pages. What this means: @@ -352,24 +311,39 @@ zfs_unmap_page(page_t *pp, caddr_t addr) * the page and the dmu buffer. */ static void -update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid) +update_pages(struct inode *ip, int64_t start, int len, + objset_t *os, uint64_t oid) { + struct address_space *mp = ip->i_mapping; + struct page *pp; + uint64_t nbytes; int64_t off; + void *pb; - off = start & PAGEOFFSET; - for (start &= PAGEMASK; len > 0; start += PAGESIZE) { - page_t *pp; - uint64_t nbytes = MIN(PAGESIZE - off, len); + off = start & (PAGE_CACHE_SIZE-1); + for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { + nbytes = MIN(PAGE_CACHE_SIZE - off, len); - if (pp = page_lookup(vp, start, SE_SHARED)) { - caddr_t va; + pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); + if (pp) { + if (mapping_writably_mapped(mp)) + flush_dcache_page(pp); - va = zfs_map_page(pp, S_WRITE); - (void) dmu_read(os, oid, start+off, nbytes, va+off, + pb = kmap(pp); + (void) dmu_read(os, oid, start+off, nbytes, pb+off, DMU_READ_PREFETCH); - zfs_unmap_page(pp, va); - page_unlock(pp); + kunmap(pp); + + if (mapping_writably_mapped(mp)) + flush_dcache_page(pp); + + mark_page_accessed(pp); + SetPageUptodate(pp); + ClearPageError(pp); + unlock_page(pp); + page_cache_release(pp); } + len -= nbytes; off = 0; } @@ -386,30 +360,41 @@ update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid) * the file is memory mapped. */ static int -mappedread(vnode_t *vp, int nbytes, uio_t *uio) +mappedread(struct inode *ip, int nbytes, uio_t *uio) { - znode_t *zp = VTOZ(vp); - objset_t *os = zp->z_zfsvfs->z_os; + struct address_space *mp = ip->i_mapping; + struct page *pp; + znode_t *zp = ITOZ(ip); + objset_t *os = ITOZSB(ip)->z_os; int64_t start, off; + uint64_t bytes; int len = nbytes; int error = 0; + void *pb; start = uio->uio_loffset; - off = start & PAGEOFFSET; - for (start &= PAGEMASK; len > 0; start += PAGESIZE) { - page_t *pp; - uint64_t bytes = MIN(PAGESIZE - off, len); - - if (pp = page_lookup(vp, start, SE_SHARED)) { - caddr_t va; - - va = zfs_map_page(pp, S_READ); - error = uiomove(va + off, bytes, UIO_READ, uio); - zfs_unmap_page(pp, va); - page_unlock(pp); + off = start & (PAGE_CACHE_SIZE-1); + for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { + bytes = MIN(PAGE_CACHE_SIZE - off, len); + + pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); + if (pp) { + ASSERT(PageUptodate(pp)); + + pb = kmap(pp); + error = uiomove(pb + off, bytes, UIO_READ, uio); + kunmap(pp); + + if (mapping_writably_mapped(mp)) + flush_dcache_page(pp); + + mark_page_accessed(pp); + unlock_page(pp); + page_cache_release(pp); } else { error = dmu_read_uio(os, zp->z_id, uio, bytes); } + len -= bytes; off = 0; if (error) @@ -417,18 +402,19 @@ mappedread(vnode_t *vp, int nbytes, uio_t *uio) } return (error); } +#endif /* _KERNEL */ -offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */ +unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ /* * Read bytes from specified file into supplied buffer. * - * IN: vp - vnode of file to be read from. + * IN: ip - inode of file to be read from. * uio - structure supplying read location, range info, * and return buffer. - * ioflag - SYNC flags; used to provide FRSYNC semantics. + * ioflag - FSYNC flags; used to provide FRSYNC semantics. + * O_DIRECT flag; used to bypass page cache. * cr - credentials of caller. - * ct - caller context * * OUT: uio - updated offset and range, buffer filled. * @@ -436,26 +422,28 @@ offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */ * error code if failure * * Side Effects: - * vp - atime updated if byte count > 0 + * inode - atime updated if byte count > 0 */ /* ARGSUSED */ -static int -zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) +int +zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); objset_t *os; ssize_t n, nbytes; - int error; + int error = 0; rl_t *rl; +#ifdef HAVE_UIO_ZEROCOPY xuio_t *xuio = NULL; +#endif /* HAVE_UIO_ZEROCOPY */ - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - os = zfsvfs->z_os; + os = zsb->z_os; if (zp->z_pflags & ZFS_AV_QUARANTINED) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EACCES); } @@ -463,7 +451,7 @@ zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) * Validate file offset */ if (uio->uio_loffset < (offset_t)0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EINVAL); } @@ -471,26 +459,24 @@ zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) * Fasttrack empty reads */ if (uio->uio_resid == 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (0); } /* * Check for mandatory locks */ - if (MANDMODE(zp->z_mode)) { - if (error = chklock(vp, FREAD, - uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) { - ZFS_EXIT(zfsvfs); - return (error); - } + if (mandatory_lock(ip) && + !lock_may_read(ip, uio->uio_loffset, uio->uio_resid)) { + ZFS_EXIT(zsb); + return (EAGAIN); } /* * If we're in FRSYNC mode, sync out this znode before reading it. */ - if (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zfsvfs->z_log, zp->z_id); + if (ioflag & FRSYNC || zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) + zil_commit(zsb->z_log, zp->z_id); /* * Lock the range against changes. @@ -509,6 +495,7 @@ zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) ASSERT(uio->uio_loffset < zp->z_size); n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); +#ifdef HAVE_UIO_ZEROCOPY if ((uio->uio_extflg == UIO_XUIO) && (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { int nblk; @@ -525,7 +512,7 @@ zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) } (void) dmu_xuio_init(xuio, nblk); - if (vn_has_cached_data(vp)) { + if (vn_has_cached_data(ip)) { /* * For simplicity, we always allocate a full buffer * even if we only expect to read a portion of a block. @@ -537,15 +524,17 @@ zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) } } } +#endif /* HAVE_UIO_ZEROCOPY */ while (n > 0) { nbytes = MIN(n, zfs_read_chunk_size - P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); - if (vn_has_cached_data(vp)) - error = mappedread(vp, nbytes, uio); + if (zp->z_is_mapped && !(ioflag & O_DIRECT)) + error = mappedread(ip, nbytes, uio); else error = dmu_read_uio(os, zp->z_id, uio, nbytes); + if (error) { /* convert checksum errors into IO errors */ if (error == ECKSUM) @@ -558,20 +547,22 @@ zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) out: zfs_range_unlock(rl); - ZFS_ACCESSTIME_STAMP(zfsvfs, zp); - ZFS_EXIT(zfsvfs); + ZFS_ACCESSTIME_STAMP(zsb, zp); + zfs_inode_update(zp); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_read); /* * Write the bytes to a file. * - * IN: vp - vnode of file to be written to. + * IN: ip - inode of file to be written to. * uio - structure supplying write location, range info, * and data buffer. * ioflag - FAPPEND flag set if in append mode. + * O_DIRECT flag; used to bypass page cache. * cr - credentials of caller. - * ct - caller context (NFS/CIFS fem monitor only) * * OUT: uio - updated offset and range. * @@ -579,36 +570,36 @@ out: * error code if failure * * Timestamps: - * vp - ctime|mtime updated if byte count > 0 + * ip - ctime|mtime updated if byte count > 0 */ /* ARGSUSED */ -static int -zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) +int +zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) { - znode_t *zp = VTOZ(vp); - rlim64_t limit = uio->uio_llimit; + znode_t *zp = ITOZ(ip); + rlim64_t limit = uio->uio_limit; ssize_t start_resid = uio->uio_resid; ssize_t tx_bytes; uint64_t end_size; dmu_tx_t *tx; - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + zfs_sb_t *zsb = ZTOZSB(zp); zilog_t *zilog; offset_t woff; ssize_t n, nbytes; rl_t *rl; - int max_blksz = zfsvfs->z_max_blksz; - int error; + int max_blksz = zsb->z_max_blksz; + int error = 0; arc_buf_t *abuf; - iovec_t *aiov; + iovec_t *aiov = NULL; xuio_t *xuio = NULL; int i_iov = 0; - int iovcnt = uio->uio_iovcnt; iovec_t *iovp = uio->uio_iov; int write_eof; int count = 0; sa_bulk_attr_t bulk[4]; uint64_t mtime[2], ctime[2]; + ASSERTV(int iovcnt = uio->uio_iovcnt); /* * Fasttrack empty write @@ -620,14 +611,13 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) limit = MAXOFFSET_T; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, - &zp->z_size, 8); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, 8); /* @@ -636,18 +626,18 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && (uio->uio_loffset < zp->z_size))) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EPERM); } - zilog = zfsvfs->z_log; + zilog = zsb->z_log; /* * Validate file offset */ woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; if (woff < 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EINVAL); } @@ -655,12 +645,12 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) * Check for mandatory locks before calling zfs_range_lock() * in order to prevent a deadlock with locks set via fcntl(). */ - if (MANDMODE((mode_t)zp->z_mode) && - (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) { - ZFS_EXIT(zfsvfs); - return (error); + if (mandatory_lock(ip) && !lock_may_write(ip, woff, n)) { + ZFS_EXIT(zsb); + return (EAGAIN); } +#ifdef HAVE_UIO_ZEROCOPY /* * Pre-fault the pages to ensure slow (eg NFS) pages * don't hold up txg. @@ -671,6 +661,7 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) xuio = (xuio_t *)uio; else uio_prefaultpages(MIN(n, max_blksz), uio); +#endif /* HAVE_UIO_ZEROCOPY */ /* * If in append mode, set the io offset pointer to eof. @@ -702,7 +693,7 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) if (woff >= limit) { zfs_range_unlock(rl); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EFBIG); } @@ -723,8 +714,8 @@ zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) abuf = NULL; woff = uio->uio_loffset; again: - if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || - zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { + if (zfs_owner_overquota(zsb, zp, B_FALSE) || + zfs_owner_overquota(zsb, zp, B_TRUE)) { if (abuf != NULL) dmu_return_arcbuf(abuf); error = EDQUOT; @@ -736,8 +727,6 @@ again: aiov = &iovp[i_iov]; abuf = dmu_xuio_arcbuf(xuio, i_iov); dmu_xuio_clear(xuio, i_iov); - DTRACE_PROBE3(zfs_cp_write, int, i_iov, - iovec_t *, aiov, arc_buf_t *, abuf); ASSERT((aiov->iov_base == abuf->b_data) || ((char *)aiov->iov_base - (char *)abuf->b_data + aiov->iov_len == arc_buf_size(abuf))); @@ -759,8 +748,8 @@ again: max_blksz); ASSERT(abuf != NULL); ASSERT(arc_buf_size(abuf) == max_blksz); - if (error = uiocopy(abuf->b_data, max_blksz, - UIO_WRITE, uio, &cbytes)) { + if ((error = uiocopy(abuf->b_data, max_blksz, + UIO_WRITE, uio, &cbytes))) { dmu_return_arcbuf(abuf); break; } @@ -770,7 +759,7 @@ again: /* * Start a transaction. */ - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); zfs_sa_upgrade_txholds(tx, zp); @@ -829,7 +818,7 @@ again: if (tx_bytes < max_blksz && (!write_eof || aiov->iov_base != abuf->b_data)) { ASSERT(xuio); - dmu_write(zfsvfs->z_os, zp->z_id, woff, + dmu_write(zsb->z_os, zp->z_id, woff, aiov->iov_len, aiov->iov_base, tx); dmu_return_arcbuf(abuf); xuio_stat_wbuf_copied(); @@ -841,17 +830,16 @@ again: ASSERT(tx_bytes <= uio->uio_resid); uioskip(uio, tx_bytes); } - if (tx_bytes && vn_has_cached_data(vp)) { - update_pages(vp, woff, - tx_bytes, zfsvfs->z_os, zp->z_id); - } + + if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) + update_pages(ip, woff, tx_bytes, zsb->z_os, zp->z_id); /* * If we made no progress, we're done. If we made even * partial progress, update the znode and ZIL accordingly. */ if (tx_bytes == 0) { - (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), + (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), (void *)&zp->z_size, sizeof (uint64_t), tx); dmu_tx_commit(tx); ASSERT(error != 0); @@ -878,7 +866,7 @@ again: uint64_t newmode; zp->z_mode &= ~(S_ISUID | S_ISGID); newmode = zp->z_mode; - (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), + (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zsb), (void *)&newmode, sizeof (uint64_t), tx); } mutex_exit(&zp->z_acl_lock); @@ -900,8 +888,8 @@ again: * the file size to the specified eof. Note, there's no * concurrency during replay. */ - if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) - zp->z_size = zfsvfs->z_replay_eof; + if (zsb->z_replay && zsb->z_replay_eof != 0) + zp->z_size = zsb->z_replay_eof; error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); @@ -923,24 +911,36 @@ again: * If we're in replay mode, or we made no progress, return error. * Otherwise, it's at least a partial write, so it's successful. */ - if (zfsvfs->z_replay || uio->uio_resid == start_resid) { - ZFS_EXIT(zfsvfs); + if (zsb->z_replay || uio->uio_resid == start_resid) { + ZFS_EXIT(zsb); return (error); } if (ioflag & (FSYNC | FDSYNC) || - zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, zp->z_id); - ZFS_EXIT(zfsvfs); + zfs_inode_update(zp); + ZFS_EXIT(zsb); return (0); } +EXPORT_SYMBOL(zfs_write); + +static void +iput_async(struct inode *ip, taskq_t *taskq) +{ + ASSERT(atomic_read(&ip->i_count) > 0); + if (atomic_read(&ip->i_count) == 1) + taskq_dispatch(taskq, (task_func_t *)iput, ip, TQ_PUSHPAGE); + else + iput(ip); +} void zfs_get_done(zgd_t *zgd, int error) { znode_t *zp = zgd->zgd_private; - objset_t *os = zp->z_zfsvfs->z_os; + objset_t *os = ZTOZSB(zp)->z_os; if (zgd->zgd_db) dmu_buf_rele(zgd->zgd_db, zgd); @@ -951,7 +951,7 @@ zfs_get_done(zgd_t *zgd, int error) * Release the vnode asynchronously as we currently have the * txg stopped from syncing. */ - VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os))); + iput_async(ZTOI(zp), dsl_pool_iput_taskq(dmu_objset_pool(os))); if (error == 0 && zgd->zgd_bp) zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); @@ -969,8 +969,8 @@ static int zil_fault_io = 0; int zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) { - zfsvfs_t *zfsvfs = arg; - objset_t *os = zfsvfs->z_os; + zfs_sb_t *zsb = arg; + objset_t *os = zsb->z_os; znode_t *zp; uint64_t object = lr->lr_foid; uint64_t offset = lr->lr_offset; @@ -986,20 +986,19 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) /* * Nothing to do if the file has been removed */ - if (zfs_zget(zfsvfs, object, &zp) != 0) + if (zfs_zget(zsb, object, &zp) != 0) return (ENOENT); if (zp->z_unlinked) { /* * Release the vnode asynchronously as we currently have the * txg stopped from syncing. */ - VN_RELE_ASYNC(ZTOV(zp), - dsl_pool_vnrele_taskq(dmu_objset_pool(os))); + iput_async(ZTOI(zp), dsl_pool_iput_taskq(dmu_objset_pool(os))); return (ENOENT); } - zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); - zgd->zgd_zilog = zfsvfs->z_log; + zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_PUSHPAGE); + zgd->zgd_zilog = zsb->z_log; zgd->zgd_private = zp; /* @@ -1084,15 +1083,14 @@ zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) } /*ARGSUSED*/ -static int -zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr, - caller_context_t *ct) +int +zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); int error; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); if (flag & V_ACE_MASK) @@ -1100,46 +1098,23 @@ zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr, else error = zfs_zaccess_rwx(zp, mode, flag, cr); - ZFS_EXIT(zfsvfs); - return (error); -} - -/* - * If vnode is for a device return a specfs vnode instead. - */ -static int -specvp_check(vnode_t **vpp, cred_t *cr) -{ - int error = 0; - - if (IS_DEVVP(*vpp)) { - struct vnode *svp; - - svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); - VN_RELE(*vpp); - if (svp == NULL) - error = ENOSYS; - *vpp = svp; - } + ZFS_EXIT(zsb); return (error); } - +EXPORT_SYMBOL(zfs_access); /* * Lookup an entry in a directory, or an extended attribute directory. - * If it exists, return a held vnode reference for it. + * If it exists, return a held inode reference for it. * - * IN: dvp - vnode of directory to search. + * IN: dip - inode of directory to search. * nm - name of entry to lookup. - * pnp - full pathname to lookup [UNUSED]. * flags - LOOKUP_XATTR set if looking for an attribute. - * rdir - root directory vnode [UNUSED]. * cr - credentials of caller. - * ct - caller context * direntflags - directory lookup flags * realpnp - returned pathname. * - * OUT: vpp - vnode of located entry, NULL if not found. + * OUT: ipp - inode of located entry, NULL if not found. * * RETURN: 0 if success * error code if failure @@ -1148,19 +1123,18 @@ specvp_check(vnode_t **vpp, cred_t *cr) * NA */ /* ARGSUSED */ -static int -zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, - int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct, - int *direntflags, pathname_t *realpnp) +int +zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, + cred_t *cr, int *direntflags, pathname_t *realpnp) { - znode_t *zdp = VTOZ(dvp); - zfsvfs_t *zfsvfs = zdp->z_zfsvfs; - int error = 0; + znode_t *zdp = ITOZ(dip); + zfs_sb_t *zsb = ITOZSB(dip); + int error = 0; /* fast path */ if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { - if (dvp->v_type != VDIR) { + if (!S_ISDIR(dip->i_mode)) { return (ENOTDIR); } else if (zdp->z_sa_hdl == NULL) { return (EIO); @@ -1169,58 +1143,50 @@ zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { error = zfs_fastaccesschk_execute(zdp, cr); if (!error) { - *vpp = dvp; - VN_HOLD(*vpp); + *ipp = dip; + igrab(*ipp); return (0); } return (error); +#ifdef HAVE_DNLC } else { vnode_t *tvp = dnlc_lookup(dvp, nm); if (tvp) { error = zfs_fastaccesschk_execute(zdp, cr); if (error) { - VN_RELE(tvp); + iput(tvp); return (error); } if (tvp == DNLC_NO_VNODE) { - VN_RELE(tvp); + iput(tvp); return (ENOENT); } else { *vpp = tvp; return (specvp_check(vpp, cr)); } } +#endif /* HAVE_DNLC */ } } - DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm); - - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zdp); - *vpp = NULL; + *ipp = NULL; if (flags & LOOKUP_XATTR) { /* - * If the xattr property is off, refuse the lookup request. - */ - if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) { - ZFS_EXIT(zfsvfs); - return (EINVAL); - } - - /* * We don't allow recursive attributes.. * Maybe someday we will. */ if (zdp->z_pflags & ZFS_XATTR) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EINVAL); } - if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) { - ZFS_EXIT(zfsvfs); + if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { + ZFS_EXIT(zsb); return (error); } @@ -1228,18 +1194,18 @@ zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, * Do we have permission to get into attribute directory? */ - if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0, - B_FALSE, cr)) { - VN_RELE(*vpp); - *vpp = NULL; + if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, + B_FALSE, cr))) { + iput(*ipp); + *ipp = NULL; } - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } - if (dvp->v_type != VDIR) { - ZFS_EXIT(zfsvfs); + if (!S_ISDIR(dip->i_mode)) { + ZFS_EXIT(zsb); return (ENOTDIR); } @@ -1247,66 +1213,64 @@ zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, * Check accessibility of directory. */ - if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) { - ZFS_EXIT(zfsvfs); + if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { + ZFS_EXIT(zsb); return (error); } - if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), + if (zsb->z_utf8 && u8_validate(nm, strlen(nm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EILSEQ); } - error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp); - if (error == 0) - error = specvp_check(vpp, cr); + error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); + if ((error == 0) && (*ipp)) + zfs_inode_update(ITOZ(*ipp)); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_lookup); /* * Attempt to create a new entry in a directory. If the entry * already exists, truncate the file if permissible, else return - * an error. Return the vp of the created or trunc'd file. + * an error. Return the ip of the created or trunc'd file. * - * IN: dvp - vnode of directory to put new file entry in. + * IN: dip - inode of directory to put new file entry in. * name - name of new file entry. * vap - attributes of new file. * excl - flag indicating exclusive or non-exclusive mode. * mode - mode to open file with. * cr - credentials of caller. * flag - large file flag [UNUSED]. - * ct - caller context - * vsecp - ACL to be set + * vsecp - ACL to be set * - * OUT: vpp - vnode of created or trunc'd entry. + * OUT: ipp - inode of created or trunc'd entry. * * RETURN: 0 if success * error code if failure * * Timestamps: - * dvp - ctime|mtime updated if new entry created - * vp - ctime|mtime always, atime if new + * dip - ctime|mtime updated if new entry created + * ip - ctime|mtime always, atime if new */ /* ARGSUSED */ -static int -zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl, - int mode, vnode_t **vpp, cred_t *cr, int flag, caller_context_t *ct, - vsecattr_t *vsecp) +int +zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, + int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) { - znode_t *zp, *dzp = VTOZ(dvp); - zfsvfs_t *zfsvfs = dzp->z_zfsvfs; + znode_t *zp, *dzp = ITOZ(dip); + zfs_sb_t *zsb = ITOZSB(dip); zilog_t *zilog; objset_t *os; zfs_dirlock_t *dl; dmu_tx_t *tx; int error; - ksid_t *ksid; uid_t uid; - gid_t gid = crgetgid(cr); + gid_t gid; zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; boolean_t have_acl = B_FALSE; @@ -1316,51 +1280,44 @@ zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl, * make sure file system is at proper version */ - ksid = crgetsid(cr, KSID_OWNER); - if (ksid) - uid = ksid_getid(ksid); - else - uid = crgetuid(cr); + gid = crgetgid(cr); + uid = crgetuid(cr); - if (zfsvfs->z_use_fuids == B_FALSE && - (vsecp || (vap->va_mask & AT_XVATTR) || - IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) + if (zsb->z_use_fuids == B_FALSE && + (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) return (EINVAL); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(dzp); - os = zfsvfs->z_os; - zilog = zfsvfs->z_log; + os = zsb->z_os; + zilog = zsb->z_log; - if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), + if (zsb->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EILSEQ); } - if (vap->va_mask & AT_XVATTR) { + if (vap->va_mask & ATTR_XVATTR) { if ((error = secpolicy_xvattr((xvattr_t *)vap, - crgetuid(cr), cr, vap->va_type)) != 0) { - ZFS_EXIT(zfsvfs); + crgetuid(cr), cr, vap->va_mode)) != 0) { + ZFS_EXIT(zsb); return (error); } } -top: - *vpp = NULL; - - if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr)) - vap->va_mode &= ~VSVTX; +top: + *ipp = NULL; if (*name == '\0') { /* * Null component name refers to the directory itself. */ - VN_HOLD(dvp); + igrab(dip); zp = dzp; dl = NULL; error = 0; } else { - /* possible VN_HOLD(zp) */ + /* possible igrab(zp) */ int zflg = 0; if (flag & FIGNORECASE) @@ -1373,7 +1330,7 @@ top: zfs_acl_ids_free(&acl_ids); if (strcmp(name, "..") == 0) error = EISDIR; - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } } @@ -1385,7 +1342,7 @@ top: * Create a new file object and update the directory * to reference it. */ - if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { + if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { if (have_acl) zfs_acl_ids_free(&acl_ids); goto out; @@ -1396,8 +1353,7 @@ top: * extended attribute directories. */ - if ((dzp->z_pflags & ZFS_XATTR) && - (vap->va_type != VREG)) { + if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { if (have_acl) zfs_acl_ids_free(&acl_ids); error = EINVAL; @@ -1409,7 +1365,7 @@ top: goto out; have_acl = B_TRUE; - if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { + if (zfs_acl_ids_overquota(zsb, &acl_ids)) { zfs_acl_ids_free(&acl_ids); error = EDQUOT; goto out; @@ -1420,12 +1376,12 @@ top: dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE); - fuid_dirtied = zfsvfs->z_fuid_dirty; + fuid_dirtied = zsb->z_fuid_dirty; if (fuid_dirtied) - zfs_fuid_txhold(zfsvfs, tx); + zfs_fuid_txhold(zsb, tx); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); - if (!zfsvfs->z_use_sa && + if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); @@ -1440,13 +1396,13 @@ top: } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); if (fuid_dirtied) - zfs_fuid_sync(zfsvfs, tx); + zfs_fuid_sync(zsb, tx); (void) zfs_link_create(dl, zp, tx, ZNEW); txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); @@ -1469,14 +1425,14 @@ top: /* * Can't truncate an existing file if in exclusive mode. */ - if (excl == EXCL) { + if (excl) { error = EEXIST; goto out; } /* * Can't open a directory for writing. */ - if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) { + if (S_ISDIR(ZTOI(zp)->i_mode)) { error = EISDIR; goto out; } @@ -1494,15 +1450,12 @@ top: /* * Truncate regular files if requested. */ - if ((ZTOV(zp)->v_type == VREG) && - (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) { + if (S_ISREG(ZTOI(zp)->i_mode) && + (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { /* we can't hold any locks when calling zfs_freesp() */ zfs_dirent_unlock(dl); dl = NULL; error = zfs_freesp(zp, 0, 0, mode, TRUE); - if (error == 0) { - vnevent_create(ZTOV(zp), ct); - } } } out: @@ -1512,70 +1465,72 @@ out: if (error) { if (zp) - VN_RELE(ZTOV(zp)); + iput(ZTOI(zp)); } else { - *vpp = ZTOV(zp); - error = specvp_check(vpp, cr); + zfs_inode_update(dzp); + zfs_inode_update(zp); + *ipp = ZTOI(zp); } - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_create); /* * Remove an entry from a directory. * - * IN: dvp - vnode of directory to remove entry from. + * IN: dip - inode of directory to remove entry from. * name - name of entry to remove. * cr - credentials of caller. - * ct - caller context - * flags - case flags * * RETURN: 0 if success * error code if failure * * Timestamps: - * dvp - ctime|mtime - * vp - ctime (if nlink > 0) + * dip - ctime|mtime + * ip - ctime (if nlink > 0) */ uint64_t null_xattr = 0; /*ARGSUSED*/ -static int -zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct, - int flags) +int +zfs_remove(struct inode *dip, char *name, cred_t *cr) { - znode_t *zp, *dzp = VTOZ(dvp); + znode_t *zp, *dzp = ITOZ(dip); znode_t *xzp; - vnode_t *vp; - zfsvfs_t *zfsvfs = dzp->z_zfsvfs; + struct inode *ip; + zfs_sb_t *zsb = ITOZSB(dip); zilog_t *zilog; - uint64_t acl_obj, xattr_obj; - uint64_t xattr_obj_unlinked = 0; + uint64_t xattr_obj; + uint64_t xattr_obj_unlinked = 0; uint64_t obj = 0; zfs_dirlock_t *dl; dmu_tx_t *tx; - boolean_t may_delete_now, delete_now = FALSE; - boolean_t unlinked, toobig = FALSE; + boolean_t unlinked; uint64_t txtype; pathname_t *realnmp = NULL; +#ifdef HAVE_PN_UTILS pathname_t realnm; +#endif /* HAVE_PN_UTILS */ int error; int zflg = ZEXISTS; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(dzp); - zilog = zfsvfs->z_log; + zilog = zsb->z_log; +#ifdef HAVE_PN_UTILS if (flags & FIGNORECASE) { zflg |= ZCILOOK; pn_alloc(&realnm); realnmp = &realnm; } +#endif /* HAVE_PN_UTILS */ top: xattr_obj = 0; @@ -1583,92 +1538,79 @@ top: /* * Attempt to lock directory; fail if entry doesn't exist. */ - if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, - NULL, realnmp)) { + if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, + NULL, realnmp))) { +#ifdef HAVE_PN_UTILS if (realnmp) pn_free(realnmp); - ZFS_EXIT(zfsvfs); +#endif /* HAVE_PN_UTILS */ + ZFS_EXIT(zsb); return (error); } - vp = ZTOV(zp); + ip = ZTOI(zp); - if (error = zfs_zaccess_delete(dzp, zp, cr)) { + if ((error = zfs_zaccess_delete(dzp, zp, cr))) { goto out; } /* * Need to use rmdir for removing directories. */ - if (vp->v_type == VDIR) { + if (S_ISDIR(ip->i_mode)) { error = EPERM; goto out; } - vnevent_remove(vp, dvp, name, ct); - +#ifdef HAVE_DNLC if (realnmp) dnlc_remove(dvp, realnmp->pn_buf); else dnlc_remove(dvp, name); - - mutex_enter(&vp->v_lock); - may_delete_now = vp->v_count == 1 && !vn_has_cached_data(vp); - mutex_exit(&vp->v_lock); +#endif /* HAVE_DNLC */ /* - * We may delete the znode now, or we may put it in the unlinked set; - * it depends on whether we're the last link, and on whether there are - * other holds on the vnode. So we dmu_tx_hold() the right things to - * allow for either case. + * We never delete the znode and always place it in the unlinked + * set. The dentry cache will always hold the last reference and + * is responsible for safely freeing the znode. */ obj = zp->z_id; - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); zfs_sa_upgrade_txholds(tx, dzp); - if (may_delete_now) { - toobig = - zp->z_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT; - /* if the file is too big, only hold_free a token amount */ - dmu_tx_hold_free(tx, zp->z_id, 0, - (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); - } /* are there any extended attributes? */ - error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), + error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), &xattr_obj, sizeof (xattr_obj)); if (error == 0 && xattr_obj) { - error = zfs_zget(zfsvfs, xattr_obj, &xzp); - ASSERT3U(error, ==, 0); + error = zfs_zget(zsb, xattr_obj, &xzp); + ASSERT0(error); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); } - mutex_enter(&zp->z_lock); - if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) - dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); - mutex_exit(&zp->z_lock); - /* charge as an update -- would be nice not to charge at all */ - dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); + dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); error = dmu_tx_assign(tx, TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); - VN_RELE(vp); + iput(ip); if (xzp) - VN_RELE(ZTOV(xzp)); + iput(ZTOI(xzp)); if (error == ERESTART) { dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } +#ifdef HAVE_PN_UTILS if (realnmp) pn_free(realnmp); +#endif /* HAVE_PN_UTILS */ dmu_tx_abort(tx); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } @@ -1683,163 +1625,127 @@ top: } if (unlinked) { - /* * Hold z_lock so that we can make sure that the ACL obj * hasn't changed. Could have been deleted due to * zfs_sa_upgrade(). */ mutex_enter(&zp->z_lock); - mutex_enter(&vp->v_lock); - (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), + (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); - delete_now = may_delete_now && !toobig && - vp->v_count == 1 && !vn_has_cached_data(vp) && - xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == - acl_obj; - mutex_exit(&vp->v_lock); - } - - if (delete_now) { - if (xattr_obj_unlinked) { - ASSERT3U(xzp->z_links, ==, 2); - mutex_enter(&xzp->z_lock); - xzp->z_unlinked = 1; - xzp->z_links = 0; - error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), - &xzp->z_links, sizeof (xzp->z_links), tx); - ASSERT3U(error, ==, 0); - mutex_exit(&xzp->z_lock); - zfs_unlinked_add(xzp, tx); - - if (zp->z_is_sa) - error = sa_remove(zp->z_sa_hdl, - SA_ZPL_XATTR(zfsvfs), tx); - else - error = sa_update(zp->z_sa_hdl, - SA_ZPL_XATTR(zfsvfs), &null_xattr, - sizeof (uint64_t), tx); - ASSERT3U(error, ==, 0); - } - mutex_enter(&vp->v_lock); - vp->v_count--; - ASSERT3U(vp->v_count, ==, 0); - mutex_exit(&vp->v_lock); - mutex_exit(&zp->z_lock); - zfs_znode_delete(zp, tx); - } else if (unlinked) { mutex_exit(&zp->z_lock); zfs_unlinked_add(zp, tx); } txtype = TX_REMOVE; +#ifdef HAVE_PN_UTILS if (flags & FIGNORECASE) txtype |= TX_CI; +#endif /* HAVE_PN_UTILS */ zfs_log_remove(zilog, tx, txtype, dzp, name, obj); dmu_tx_commit(tx); out: +#ifdef HAVE_PN_UTILS if (realnmp) pn_free(realnmp); +#endif /* HAVE_PN_UTILS */ zfs_dirent_unlock(dl); + zfs_inode_update(dzp); + zfs_inode_update(zp); + if (xzp) + zfs_inode_update(xzp); - if (!delete_now) - VN_RELE(vp); + iput(ip); if (xzp) - VN_RELE(ZTOV(xzp)); + iput(ZTOI(xzp)); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_remove); /* - * Create a new directory and insert it into dvp using the name + * Create a new directory and insert it into dip using the name * provided. Return a pointer to the inserted directory. * - * IN: dvp - vnode of directory to add subdir to. + * IN: dip - inode of directory to add subdir to. * dirname - name of new directory. * vap - attributes of new directory. * cr - credentials of caller. - * ct - caller context * vsecp - ACL to be set * - * OUT: vpp - vnode of created directory. + * OUT: ipp - inode of created directory. * * RETURN: 0 if success * error code if failure * * Timestamps: - * dvp - ctime|mtime updated - * vp - ctime|mtime|atime updated + * dip - ctime|mtime updated + * ipp - ctime|mtime|atime updated */ /*ARGSUSED*/ -static int -zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr, - caller_context_t *ct, int flags, vsecattr_t *vsecp) +int +zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, + cred_t *cr, int flags, vsecattr_t *vsecp) { - znode_t *zp, *dzp = VTOZ(dvp); - zfsvfs_t *zfsvfs = dzp->z_zfsvfs; + znode_t *zp, *dzp = ITOZ(dip); + zfs_sb_t *zsb = ITOZSB(dip); zilog_t *zilog; zfs_dirlock_t *dl; uint64_t txtype; dmu_tx_t *tx; int error; int zf = ZNEW; - ksid_t *ksid; uid_t uid; gid_t gid = crgetgid(cr); zfs_acl_ids_t acl_ids; boolean_t fuid_dirtied; - ASSERT(vap->va_type == VDIR); + ASSERT(S_ISDIR(vap->va_mode)); /* * If we have an ephemeral id, ACL, or XVATTR then * make sure file system is at proper version */ - ksid = crgetsid(cr, KSID_OWNER); - if (ksid) - uid = ksid_getid(ksid); - else - uid = crgetuid(cr); - if (zfsvfs->z_use_fuids == B_FALSE && - (vsecp || (vap->va_mask & AT_XVATTR) || - IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) + uid = crgetuid(cr); + if (zsb->z_use_fuids == B_FALSE && + (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) return (EINVAL); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(dzp); - zilog = zfsvfs->z_log; + zilog = zsb->z_log; if (dzp->z_pflags & ZFS_XATTR) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EINVAL); } - if (zfsvfs->z_utf8 && u8_validate(dirname, + if (zsb->z_utf8 && u8_validate(dirname, strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EILSEQ); } if (flags & FIGNORECASE) zf |= ZCILOOK; - if (vap->va_mask & AT_XVATTR) { + if (vap->va_mask & ATTR_XVATTR) { if ((error = secpolicy_xvattr((xvattr_t *)vap, - crgetuid(cr), cr, vap->va_type)) != 0) { - ZFS_EXIT(zfsvfs); + crgetuid(cr), cr, vap->va_mode)) != 0) { + ZFS_EXIT(zsb); return (error); } } if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, &acl_ids)) != 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } /* @@ -1850,39 +1756,39 @@ zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr, * to fail. */ top: - *vpp = NULL; + *ipp = NULL; - if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, - NULL, NULL)) { + if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, + NULL, NULL))) { zfs_acl_ids_free(&acl_ids); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } - if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) { + if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } - if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { + if (zfs_acl_ids_overquota(zsb, &acl_ids)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EDQUOT); } /* * Add a new entry to the directory. */ - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); - fuid_dirtied = zfsvfs->z_fuid_dirty; + fuid_dirtied = zsb->z_fuid_dirty; if (fuid_dirtied) - zfs_fuid_txhold(zfsvfs, tx); - if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { + zfs_fuid_txhold(zsb, tx); + if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } @@ -1900,7 +1806,7 @@ top: } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } @@ -1910,14 +1816,14 @@ top: zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); if (fuid_dirtied) - zfs_fuid_sync(zfsvfs, tx); + zfs_fuid_sync(zsb, tx); /* * Now put new name in parent dir. */ (void) zfs_link_create(dl, zp, tx, ZNEW); - *vpp = ZTOV(zp); + *ipp = ZTOI(zp); txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); if (flags & FIGNORECASE) @@ -1931,49 +1837,51 @@ top: zfs_dirent_unlock(dl); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); + zfs_inode_update(dzp); + zfs_inode_update(zp); + ZFS_EXIT(zsb); return (0); } +EXPORT_SYMBOL(zfs_mkdir); /* * Remove a directory subdir entry. If the current working * directory is the same as the subdir to be removed, the * remove will fail. * - * IN: dvp - vnode of directory to remove from. + * IN: dip - inode of directory to remove from. * name - name of directory to be removed. - * cwd - vnode of current working directory. + * cwd - inode of current working directory. * cr - credentials of caller. - * ct - caller context * flags - case flags * * RETURN: 0 if success * error code if failure * * Timestamps: - * dvp - ctime|mtime updated + * dip - ctime|mtime updated */ /*ARGSUSED*/ -static int -zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr, - caller_context_t *ct, int flags) +int +zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, + int flags) { - znode_t *dzp = VTOZ(dvp); + znode_t *dzp = ITOZ(dip); znode_t *zp; - vnode_t *vp; - zfsvfs_t *zfsvfs = dzp->z_zfsvfs; + struct inode *ip; + zfs_sb_t *zsb = ITOZSB(dip); zilog_t *zilog; zfs_dirlock_t *dl; dmu_tx_t *tx; int error; int zflg = ZEXISTS; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(dzp); - zilog = zfsvfs->z_log; + zilog = zsb->z_log; if (flags & FIGNORECASE) zflg |= ZCILOOK; @@ -1983,30 +1891,28 @@ top: /* * Attempt to lock directory; fail if entry doesn't exist. */ - if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, - NULL, NULL)) { - ZFS_EXIT(zfsvfs); + if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, + NULL, NULL))) { + ZFS_EXIT(zsb); return (error); } - vp = ZTOV(zp); + ip = ZTOI(zp); - if (error = zfs_zaccess_delete(dzp, zp, cr)) { + if ((error = zfs_zaccess_delete(dzp, zp, cr))) { goto out; } - if (vp->v_type != VDIR) { + if (!S_ISDIR(ip->i_mode)) { error = ENOTDIR; goto out; } - if (vp == cwd) { + if (ip == cwd) { error = EINVAL; goto out; } - vnevent_rmdir(vp, dvp, name, ct); - /* * Grab a lock on the directory to make sure that noone is * trying to add (or lookup) entries while we are removing it. @@ -2019,10 +1925,10 @@ top: */ rw_enter(&zp->z_parent_lock, RW_WRITER); - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); - dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); + dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); zfs_sa_upgrade_txholds(tx, zp); zfs_sa_upgrade_txholds(tx, dzp); error = dmu_tx_assign(tx, TXG_NOWAIT); @@ -2030,14 +1936,14 @@ top: rw_exit(&zp->z_parent_lock); rw_exit(&zp->z_name_lock); zfs_dirent_unlock(dl); - VN_RELE(vp); + iput(ip); if (error == ERESTART) { dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } dmu_tx_abort(tx); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } @@ -2057,35 +1963,32 @@ top: out: zfs_dirent_unlock(dl); - VN_RELE(vp); + zfs_inode_update(dzp); + zfs_inode_update(zp); + iput(ip); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_rmdir); /* * Read as many directory entries as will fit into the provided - * buffer from the given directory cursor position (specified in - * the uio structure. + * dirent buffer from the given directory cursor position. * - * IN: vp - vnode of directory to read. - * uio - structure supplying read location, range info, - * and return buffer. - * cr - credentials of caller. - * ct - caller context - * flags - case flags + * IN: ip - inode of directory to read. + * dirent - buffer for directory entries. * - * OUT: uio - updated offset and range, buffer filled. - * eofp - set to true if end-of-file detected. + * OUT: dirent - filler buffer of directory entries. * * RETURN: 0 if success * error code if failure * * Timestamps: - * vp - atime updated + * ip - atime updated * * Note that the low 4 bits of the cookie returned by zap is always zero. * This allows us to use the low range for "special" directory entries: @@ -2093,70 +1996,42 @@ out: * we use the offset 2 for the '.zfs' directory. */ /* ARGSUSED */ -static int -zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, - caller_context_t *ct, int flags) +int +zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) { - znode_t *zp = VTOZ(vp); - iovec_t *iovp; - edirent_t *eodp; - dirent64_t *odp; - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); objset_t *os; - caddr_t outbuf; - size_t bufsize; zap_cursor_t zc; zap_attribute_t zap; - uint_t bytes_wanted; - uint64_t offset; /* must be unsigned; checks for < 1 */ - uint64_t parent; - int local_eof; int outcount; int error; uint8_t prefetch; - boolean_t check_sysattrs; + int done = 0; + uint64_t parent; + loff_t *pos = &(ctx->pos); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), - &parent, sizeof (parent))) != 0) { - ZFS_EXIT(zfsvfs); - return (error); - } - - /* - * If we are not given an eof variable, - * use a local one. - */ - if (eofp == NULL) - eofp = &local_eof; - - /* - * Check for valid iov_len. - */ - if (uio->uio_iov->iov_len <= 0) { - ZFS_EXIT(zfsvfs); - return (EINVAL); - } + if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zsb), + &parent, sizeof (parent))) != 0) + goto out; /* * Quit if directory has been removed (posix) */ - if ((*eofp = zp->z_unlinked) != 0) { - ZFS_EXIT(zfsvfs); - return (0); - } - error = 0; - os = zfsvfs->z_os; - offset = uio->uio_loffset; + if (zp->z_unlinked) + goto out; + + os = zsb->z_os; prefetch = zp->z_zn_prefetch; /* * Initialize the iterator cursor. */ - if (offset <= 3) { + if (*pos <= 3) { /* * Start iteration from the beginning of the directory. */ @@ -2165,55 +2040,28 @@ zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, /* * The offset is a serialized cursor. */ - zap_cursor_init_serialized(&zc, os, zp->z_id, offset); - } - - /* - * Get space to change directory entries into fs independent format. - */ - iovp = uio->uio_iov; - bytes_wanted = iovp->iov_len; - if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) { - bufsize = bytes_wanted; - outbuf = kmem_alloc(bufsize, KM_SLEEP); - odp = (struct dirent64 *)outbuf; - } else { - bufsize = bytes_wanted; - odp = (struct dirent64 *)iovp->iov_base; + zap_cursor_init_serialized(&zc, os, zp->z_id, *pos); } - eodp = (struct edirent *)odp; - - /* - * If this VFS supports the system attribute view interface; and - * we're looking at an extended attribute directory; and we care - * about normalization conflicts on this vfs; then we must check - * for normalization conflicts with the sysattr name space. - */ - check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) && - (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm && - (flags & V_RDDIR_ENTFLAGS); /* * Transform to file-system independent format */ outcount = 0; - while (outcount < bytes_wanted) { - ino64_t objnum; - ushort_t reclen; - off64_t *next = NULL; + while (!done) { + uint64_t objnum; /* * Special case `.', `..', and `.zfs'. */ - if (offset == 0) { + if (*pos == 0) { (void) strcpy(zap.za_name, "."); zap.za_normalization_conflict = 0; objnum = zp->z_id; - } else if (offset == 1) { + } else if (*pos == 1) { (void) strcpy(zap.za_name, ".."); zap.za_normalization_conflict = 0; objnum = parent; - } else if (offset == 2 && zfs_show_ctldir(zp)) { + } else if (*pos == 2 && zfs_show_ctldir(zp)) { (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); zap.za_normalization_conflict = 0; objnum = ZFSCTL_INO_ROOT; @@ -2221,194 +2069,111 @@ zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, /* * Grab next entry. */ - if (error = zap_cursor_retrieve(&zc, &zap)) { - if ((*eofp = (error == ENOENT)) != 0) + if ((error = zap_cursor_retrieve(&zc, &zap))) { + if (error == ENOENT) break; else goto update; } + /* + * Allow multiple entries provided the first entry is + * the object id. Non-zpl consumers may safely make + * use of the additional space. + * + * XXX: This should be a feature flag for compatibility + */ if (zap.za_integer_length != 8 || - zap.za_num_integers != 1) { + zap.za_num_integers == 0) { cmn_err(CE_WARN, "zap_readdir: bad directory " - "entry, obj = %lld, offset = %lld\n", + "entry, obj = %lld, offset = %lld, " + "length = %d, num = %lld\n", (u_longlong_t)zp->z_id, - (u_longlong_t)offset); + (u_longlong_t)*pos, + zap.za_integer_length, + (u_longlong_t)zap.za_num_integers); error = ENXIO; goto update; } objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); - /* - * MacOS X can extract the object type here such as: - * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer); - */ - - if (check_sysattrs && !zap.za_normalization_conflict) { - zap.za_normalization_conflict = - xattr_sysattr_casechk(zap.za_name); - } } - if (flags & V_RDDIR_ACCFILTER) { - /* - * If we have no access at all, don't include - * this entry in the returned information - */ - znode_t *ezp; - if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0) - goto skip_entry; - if (!zfs_has_access(ezp, cr)) { - VN_RELE(ZTOV(ezp)); - goto skip_entry; - } - VN_RELE(ZTOV(ezp)); - } - - if (flags & V_RDDIR_ENTFLAGS) - reclen = EDIRENT_RECLEN(strlen(zap.za_name)); - else - reclen = DIRENT64_RECLEN(strlen(zap.za_name)); - - /* - * Will this entry fit in the buffer? - */ - if (outcount + reclen > bufsize) { - /* - * Did we manage to fit anything in the buffer? - */ - if (!outcount) { - error = EINVAL; - goto update; - } + done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), + objnum, ZFS_DIRENT_TYPE(zap.za_first_integer)); + if (done) break; - } - if (flags & V_RDDIR_ENTFLAGS) { - /* - * Add extended flag entry: - */ - eodp->ed_ino = objnum; - eodp->ed_reclen = reclen; - /* NOTE: ed_off is the offset for the *next* entry */ - next = &(eodp->ed_off); - eodp->ed_eflags = zap.za_normalization_conflict ? - ED_CASE_CONFLICT : 0; - (void) strncpy(eodp->ed_name, zap.za_name, - EDIRENT_NAMELEN(reclen)); - eodp = (edirent_t *)((intptr_t)eodp + reclen); - } else { - /* - * Add normal entry: - */ - odp->d_ino = objnum; - odp->d_reclen = reclen; - /* NOTE: d_off is the offset for the *next* entry */ - next = &(odp->d_off); - (void) strncpy(odp->d_name, zap.za_name, - DIRENT64_NAMELEN(reclen)); - odp = (dirent64_t *)((intptr_t)odp + reclen); - } - outcount += reclen; - - ASSERT(outcount <= bufsize); /* Prefetch znode */ - if (prefetch) + if (prefetch) { dmu_prefetch(os, objnum, 0, 0); + } - skip_entry: - /* - * Move to the next entry, fill in the previous offset. - */ - if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { + if (*pos > 2 || (*pos == 2 && !zfs_show_ctldir(zp))) { zap_cursor_advance(&zc); - offset = zap_cursor_serialize(&zc); + *pos = zap_cursor_serialize(&zc); } else { - offset += 1; + (*pos)++; } - if (next) - *next = offset; } zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ - if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) { - iovp->iov_base += outcount; - iovp->iov_len -= outcount; - uio->uio_resid -= outcount; - } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) { - /* - * Reset the pointer. - */ - offset = uio->uio_loffset; - } - update: zap_cursor_fini(&zc); - if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) - kmem_free(outbuf, bufsize); - if (error == ENOENT) error = 0; - ZFS_ACCESSTIME_STAMP(zfsvfs, zp); + ZFS_ACCESSTIME_STAMP(zsb, zp); + zfs_inode_update(zp); + +out: + ZFS_EXIT(zsb); - uio->uio_loffset = offset; - ZFS_EXIT(zfsvfs); return (error); } +EXPORT_SYMBOL(zfs_readdir); ulong_t zfs_fsync_sync_cnt = 4; -static int -zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct) +int +zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - - /* - * Regardless of whether this is required for standards conformance, - * this is the logical behavior when fsync() is called on a file with - * dirty pages. We use B_ASYNC since the ZIL transactions are already - * going to be pushed out as part of the zil_commit(). - */ - if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) && - (vp->v_type == VREG) && !(IS_SWAPVP(vp))) - (void) VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, B_ASYNC, cr, ct); + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); - if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { - ZFS_ENTER(zfsvfs); + if (zsb->z_os->os_sync != ZFS_SYNC_DISABLED) { + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - zil_commit(zfsvfs->z_log, zp->z_id); - ZFS_EXIT(zfsvfs); + zil_commit(zsb->z_log, zp->z_id); + ZFS_EXIT(zsb); } return (0); } +EXPORT_SYMBOL(zfs_fsync); /* * Get the requested file attributes and place them in the provided * vattr structure. * - * IN: vp - vnode of file. + * IN: ip - inode of file. * vap - va_mask identifies requested attributes. - * If AT_XVATTR set, then optional attrs are requested + * If ATTR_XVATTR set, then optional attrs are requested * flags - ATTR_NOACLCHECK (CIFS server context) * cr - credentials of caller. - * ct - caller context * * OUT: vap - attribute values. * * RETURN: 0 (always succeeds) */ /* ARGSUSED */ -static int -zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, - caller_context_t *ct) +int +zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); int error = 0; uint64_t links; uint64_t mtime[2], ctime[2]; @@ -2418,16 +2183,16 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, sa_bulk_attr_t bulk[2]; int count = 0; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } @@ -2438,9 +2203,9 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, */ if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && (vap->va_uid != crgetuid(cr))) { - if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, - skipaclchk, cr)) { - ZFS_EXIT(zfsvfs); + if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, + skipaclchk, cr))) { + ZFS_EXIT(zsb); return (error); } } @@ -2451,24 +2216,24 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, */ mutex_enter(&zp->z_lock); - vap->va_type = vp->v_type; - vap->va_mode = zp->z_mode & MODEMASK; - vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev; + vap->va_type = vn_mode_to_vtype(zp->z_mode); + vap->va_mode = zp->z_mode; + vap->va_fsid = ZTOI(zp)->i_sb->s_dev; vap->va_nodeid = zp->z_id; - if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp)) + if ((zp->z_id == zsb->z_root) && zfs_show_ctldir(zp)) links = zp->z_links + 1; else links = zp->z_links; - vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */ - vap->va_size = zp->z_size; - vap->va_rdev = vp->v_rdev; - vap->va_seq = zp->z_seq; + vap->va_nlink = MIN(links, ZFS_LINK_MAX); + vap->va_size = i_size_read(ip); + vap->va_rdev = ip->i_rdev; + vap->va_seq = ip->i_generation; /* * Add in any requested optional attributes and the create time. * Also set the corresponding bits in the returned attribute bitmap. */ - if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { + if ((xoap = xva_getxoptattr(xvap)) != NULL && zsb->z_use_fuids) { if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { xoap->xoa_archive = ((zp->z_pflags & ZFS_ARCHIVE) != 0); @@ -2536,14 +2301,14 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && - vp->v_type == VREG) { + S_ISREG(ip->i_mode)) { zfs_sa_get_scanstamp(zp, xvap); } if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { uint64_t times[2]; - (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), + (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zsb), times, sizeof (times)); ZFS_TIME_DECODE(&xoap->xoa_createtime, times); XVA_SET_RTN(xvap, XAT_CREATETIME); @@ -2583,42 +2348,84 @@ zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, /* * Block size hasn't been set; suggest maximal I/O transfers. */ - vap->va_blksize = zfsvfs->z_max_blksz; + vap->va_blksize = zsb->z_max_blksz; + } + + ZFS_EXIT(zsb); + return (0); +} +EXPORT_SYMBOL(zfs_getattr); + +/* + * Get the basic file attributes and place them in the provided kstat + * structure. The inode is assumed to be the authoritative source + * for most of the attributes. However, the znode currently has the + * authoritative atime, blksize, and block count. + * + * IN: ip - inode of file. + * + * OUT: sp - kstat values. + * + * RETURN: 0 (always succeeds) + */ +/* ARGSUSED */ +int +zfs_getattr_fast(struct inode *ip, struct kstat *sp) +{ + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); + + ZFS_ENTER(zsb); + ZFS_VERIFY_ZP(zp); + + mutex_enter(&zp->z_lock); + + generic_fillattr(ip, sp); + ZFS_TIME_DECODE(&sp->atime, zp->z_atime); + + sa_object_size(zp->z_sa_hdl, (uint32_t *)&sp->blksize, &sp->blocks); + if (unlikely(zp->z_blksz == 0)) { + /* + * Block size hasn't been set; suggest maximal I/O transfers. + */ + sp->blksize = zsb->z_max_blksz; } - ZFS_EXIT(zfsvfs); + mutex_exit(&zp->z_lock); + + ZFS_EXIT(zsb); + return (0); } +EXPORT_SYMBOL(zfs_getattr_fast); /* * Set the file attributes to the values contained in the * vattr structure. * - * IN: vp - vnode of file to be modified. + * IN: ip - inode of file to be modified. * vap - new attribute values. - * If AT_XVATTR set, then optional attrs are being set + * If ATTR_XVATTR set, then optional attrs are being set * flags - ATTR_UTIME set if non-default time values provided. * - ATTR_NOACLCHECK (CIFS context only). * cr - credentials of caller. - * ct - caller context * * RETURN: 0 if success * error code if failure * * Timestamps: - * vp - ctime updated, mtime updated if size changed. + * ip - ctime updated, mtime updated if size changed. */ /* ARGSUSED */ -static int -zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, - caller_context_t *ct) +int +zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); zilog_t *zilog; dmu_tx_t *tx; vattr_t oldva; - xvattr_t tmpxvattr; + xvattr_t *tmpxvattr; uint_t mask = vap->va_mask; uint_t saved_mask; int trim_mask = 0; @@ -2635,40 +2442,37 @@ zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, zfs_acl_t *aclp; boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; boolean_t fuid_dirtied = B_FALSE; - sa_bulk_attr_t bulk[7], xattr_bulk[7]; + sa_bulk_attr_t *bulk, *xattr_bulk; int count = 0, xattr_count = 0; if (mask == 0) return (0); - if (mask & AT_NOSET) - return (EINVAL); - - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - zilog = zfsvfs->z_log; + zilog = zsb->z_log; /* * Make sure that if we have ephemeral uid/gid or xvattr specified * that file system is at proper version level */ - if (zfsvfs->z_use_fuids == B_FALSE && - (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) || - ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) || - (mask & AT_XVATTR))) { - ZFS_EXIT(zfsvfs); + if (zsb->z_use_fuids == B_FALSE && + (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || + ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || + (mask & ATTR_XVATTR))) { + ZFS_EXIT(zsb); return (EINVAL); } - if (mask & AT_SIZE && vp->v_type == VDIR) { - ZFS_EXIT(zfsvfs); + if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { + ZFS_EXIT(zsb); return (EISDIR); } - if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) { - ZFS_EXIT(zfsvfs); + if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { + ZFS_EXIT(zsb); return (EINVAL); } @@ -2678,21 +2482,25 @@ zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, */ xoap = xva_getxoptattr(xvap); - xva_init(&tmpxvattr); + tmpxvattr = kmem_alloc(sizeof(xvattr_t), KM_SLEEP); + xva_init(tmpxvattr); + + bulk = kmem_alloc(sizeof(sa_bulk_attr_t) * 7, KM_SLEEP); + xattr_bulk = kmem_alloc(sizeof(sa_bulk_attr_t) * 7, KM_SLEEP); /* * Immutable files can only alter immutable bit and atime */ if ((zp->z_pflags & ZFS_IMMUTABLE) && - ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) || - ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { - ZFS_EXIT(zfsvfs); - return (EPERM); + ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || + ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { + err = EPERM; + goto out3; } - if ((mask & AT_SIZE) && (zp->z_pflags & ZFS_READONLY)) { - ZFS_EXIT(zfsvfs); - return (EPERM); + if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { + err = EPERM; + goto out3; } /* @@ -2701,11 +2509,11 @@ zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, * handle times greater than 2039. This check should be removed * once large timestamps are fully supported. */ - if (mask & (AT_ATIME | AT_MTIME)) { - if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || - ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { - ZFS_EXIT(zfsvfs); - return (EOVERFLOW); + if (mask & (ATTR_ATIME | ATTR_MTIME)) { + if (((mask & ATTR_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || + ((mask & ATTR_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { + err = EOVERFLOW; + goto out3; } } @@ -2714,21 +2522,22 @@ top: aclp = NULL; /* Can this be moved to before the top label? */ - if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { - ZFS_EXIT(zfsvfs); - return (EROFS); + if (zfs_is_readonly(zsb)) { + err = EROFS; + goto out3; } /* * First validate permissions */ - if (mask & AT_SIZE) { + if (mask & ATTR_SIZE) { err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); - if (err) { - ZFS_EXIT(zfsvfs); - return (err); - } + if (err) + goto out3; + + truncate_setsize(ip, vap->va_size); + /* * XXX - Note, we are not providing any open * mode flags here (like FNDELAY), so we may @@ -2737,14 +2546,12 @@ top: */ /* XXX - would it be OK to generate a log record here? */ err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); - if (err) { - ZFS_EXIT(zfsvfs); - return (err); - } + if (err) + goto out3; } - if (mask & (AT_ATIME|AT_MTIME) || - ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || + if (mask & (ATTR_ATIME|ATTR_MTIME) || + ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || XVA_ISSET_REQ(xvap, XAT_READONLY) || XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || XVA_ISSET_REQ(xvap, XAT_OFFLINE) || @@ -2755,8 +2562,8 @@ top: skipaclchk, cr); } - if (mask & (AT_UID|AT_GID)) { - int idmask = (mask & (AT_UID|AT_GID)); + if (mask & (ATTR_UID|ATTR_GID)) { + int idmask = (mask & (ATTR_UID|ATTR_GID)); int take_owner; int take_group; @@ -2765,19 +2572,19 @@ top: * we may clear S_ISUID/S_ISGID bits. */ - if (!(mask & AT_MODE)) + if (!(mask & ATTR_MODE)) vap->va_mode = zp->z_mode; /* * Take ownership or chgrp to group we are a member of */ - take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr)); - take_group = (mask & AT_GID) && - zfs_groupmember(zfsvfs, vap->va_gid, cr); + take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); + take_group = (mask & ATTR_GID) && + zfs_groupmember(zsb, vap->va_gid, cr); /* - * If both AT_UID and AT_GID are set then take_owner and + * If both ATTR_UID and ATTR_GID are set then take_owner and * take_group must both be set in order to allow taking * ownership. * @@ -2785,16 +2592,17 @@ top: * */ - if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) || - ((idmask == AT_UID) && take_owner) || - ((idmask == AT_GID) && take_group)) { + if (((idmask == (ATTR_UID|ATTR_GID)) && + take_owner && take_group) || + ((idmask == ATTR_UID) && take_owner) || + ((idmask == ATTR_GID) && take_group)) { if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, skipaclchk, cr) == 0) { /* * Remove setuid/setgid for non-privileged users */ - secpolicy_setid_clear(vap, cr); - trim_mask = (mask & (AT_UID|AT_GID)); + (void) secpolicy_setid_clear(vap, cr); + trim_mask = (mask & (ATTR_UID|ATTR_GID)); } else { need_policy = TRUE; } @@ -2806,7 +2614,7 @@ top: mutex_enter(&zp->z_lock); oldva.va_mode = zp->z_mode; zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); - if (mask & AT_XVATTR) { + if (mask & ATTR_XVATTR) { /* * Update xvattr mask to include only those attributes * that are actually changing. @@ -2820,7 +2628,7 @@ top: need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_APPENDONLY); - XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY); + XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); } } @@ -2830,7 +2638,7 @@ top: need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_NOUNLINK); - XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK); + XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); } } @@ -2840,7 +2648,7 @@ top: need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_IMMUTABLE); - XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE); + XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); } } @@ -2850,7 +2658,7 @@ top: need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_NODUMP); - XVA_SET_REQ(&tmpxvattr, XAT_NODUMP); + XVA_SET_REQ(tmpxvattr, XAT_NODUMP); } } @@ -2860,26 +2668,26 @@ top: need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); - XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED); + XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); } } if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { - if ((vp->v_type != VREG && + if ((!S_ISREG(ip->i_mode) && xoap->xoa_av_quarantined) || xoap->xoa_av_quarantined != ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { need_policy = TRUE; } else { XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); - XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED); + XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); } } if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { mutex_exit(&zp->z_lock); - ZFS_EXIT(zfsvfs); - return (EPERM); + err = EPERM; + goto out3; } if (need_policy == FALSE && @@ -2891,15 +2699,14 @@ top: mutex_exit(&zp->z_lock); - if (mask & AT_MODE) { + if (mask & ATTR_MODE) { if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { - err = secpolicy_setid_setsticky_clear(vp, vap, + err = secpolicy_setid_setsticky_clear(ip, vap, &oldva, cr); - if (err) { - ZFS_EXIT(zfsvfs); - return (err); - } - trim_mask |= AT_MODE; + if (err) + goto out3; + + trim_mask |= ATTR_MODE; } else { need_policy = TRUE; } @@ -2918,12 +2725,10 @@ top: saved_mask = vap->va_mask; vap->va_mask &= ~trim_mask; } - err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, + err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); - if (err) { - ZFS_EXIT(zfsvfs); - return (err); - } + if (err) + goto out3; if (trim_mask) vap->va_mask |= saved_mask; @@ -2935,42 +2740,42 @@ top: */ mask = vap->va_mask; - if ((mask & (AT_UID | AT_GID))) { - err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), + if ((mask & (ATTR_UID | ATTR_GID))) { + err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), &xattr_obj, sizeof (xattr_obj)); if (err == 0 && xattr_obj) { - err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp); + err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); if (err) goto out2; } - if (mask & AT_UID) { - new_uid = zfs_fuid_create(zfsvfs, + if (mask & ATTR_UID) { + new_uid = zfs_fuid_create(zsb, (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); if (new_uid != zp->z_uid && - zfs_fuid_overquota(zfsvfs, B_FALSE, new_uid)) { + zfs_fuid_overquota(zsb, B_FALSE, new_uid)) { if (attrzp) - VN_RELE(ZTOV(attrzp)); + iput(ZTOI(attrzp)); err = EDQUOT; goto out2; } } - if (mask & AT_GID) { - new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid, + if (mask & ATTR_GID) { + new_gid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); if (new_gid != zp->z_gid && - zfs_fuid_overquota(zfsvfs, B_TRUE, new_gid)) { + zfs_fuid_overquota(zsb, B_TRUE, new_gid)) { if (attrzp) - VN_RELE(ZTOV(attrzp)); + iput(ZTOI(attrzp)); err = EDQUOT; goto out2; } } } - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(zsb->z_os); - if (mask & AT_MODE) { + if (mask & ATTR_MODE) { uint64_t pmode = zp->z_mode; uint64_t acl_obj; new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); @@ -2983,7 +2788,7 @@ top: * Are we upgrading ACL from old V0 format * to V1 format? */ - if (zfsvfs->z_version >= ZPL_VERSION_FUID && + if (zsb->z_version >= ZPL_VERSION_FUID && zfs_znode_acl_version(zp) == ZFS_ACL_VERSION_INITIAL) { dmu_tx_hold_free(tx, acl_obj, 0, @@ -3001,7 +2806,7 @@ top: mutex_exit(&zp->z_lock); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); } else { - if ((mask & AT_XVATTR) && + if ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); else @@ -3012,9 +2817,9 @@ top: dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); } - fuid_dirtied = zfsvfs->z_fuid_dirty; + fuid_dirtied = zsb->z_fuid_dirty; if (fuid_dirtied) - zfs_fuid_txhold(zfsvfs, tx); + zfs_fuid_txhold(zsb, tx); zfs_sa_upgrade_txholds(tx, zp); @@ -3035,49 +2840,49 @@ top: */ - if (mask & (AT_UID|AT_GID|AT_MODE)) + if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_enter(&zp->z_acl_lock); mutex_enter(&zp->z_lock); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, sizeof (zp->z_pflags)); if (attrzp) { - if (mask & (AT_UID|AT_GID|AT_MODE)) + if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_enter(&attrzp->z_acl_lock); mutex_enter(&attrzp->z_lock); SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, - SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, + SA_ZPL_FLAGS(zsb), NULL, &attrzp->z_pflags, sizeof (attrzp->z_pflags)); } - if (mask & (AT_UID|AT_GID)) { + if (mask & (ATTR_UID|ATTR_GID)) { - if (mask & AT_UID) { - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, + if (mask & ATTR_UID) { + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &new_uid, sizeof (new_uid)); zp->z_uid = new_uid; if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, - SA_ZPL_UID(zfsvfs), NULL, &new_uid, + SA_ZPL_UID(zsb), NULL, &new_uid, sizeof (new_uid)); attrzp->z_uid = new_uid; } } - if (mask & AT_GID) { - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), + if (mask & ATTR_GID) { + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &new_gid, sizeof (new_gid)); zp->z_gid = new_gid; if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, - SA_ZPL_GID(zfsvfs), NULL, &new_gid, + SA_ZPL_GID(zsb), NULL, &new_gid, sizeof (new_gid)); attrzp->z_gid = new_gid; } } - if (!(mask & AT_MODE)) { - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), + if (!(mask & ATTR_MODE)) { + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &new_mode, sizeof (new_mode)); new_mode = zp->z_mode; } @@ -3089,13 +2894,13 @@ top: } } - if (mask & AT_MODE) { - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, + if (mask & ATTR_MODE) { + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &new_mode, sizeof (new_mode)); zp->z_mode = new_mode; - ASSERT3U((uintptr_t)aclp, !=, NULL); + ASSERT3P(aclp, !=, NULL); err = zfs_aclset_common(zp, aclp, cr, tx); - ASSERT3U(err, ==, 0); + ASSERT0(err); if (zp->z_acl_cached) zfs_acl_free(zp->z_acl_cached); zp->z_acl_cached = aclp; @@ -3103,34 +2908,34 @@ top: } - if (mask & AT_ATIME) { + if (mask & ATTR_ATIME) { ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, &zp->z_atime, sizeof (zp->z_atime)); } - if (mask & AT_MTIME) { + if (mask & ATTR_MTIME) { ZFS_TIME_ENCODE(&vap->va_mtime, mtime); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, sizeof (mtime)); } /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ - if (mask & AT_SIZE && !(mask & AT_MTIME)) { - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), + if (mask & ATTR_SIZE && !(mask & ATTR_MTIME)) { + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, sizeof (mtime)); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, sizeof (ctime)); zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); } else if (mask != 0) { - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, sizeof (ctime)); zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, B_TRUE); if (attrzp) { SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, - SA_ZPL_CTIME(zfsvfs), NULL, + SA_ZPL_CTIME(zsb), NULL, &ctime, sizeof (ctime)); zfs_tstamp_update_setup(attrzp, STATE_CHANGED, mtime, ctime, B_TRUE); @@ -3141,50 +2946,50 @@ top: * update from toggling bit */ - if (xoap && (mask & AT_XVATTR)) { + if (xoap && (mask & ATTR_XVATTR)) { /* * restore trimmed off masks * so that return masks can be set for caller. */ - if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) { + if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { XVA_SET_REQ(xvap, XAT_APPENDONLY); } - if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) { + if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { XVA_SET_REQ(xvap, XAT_NOUNLINK); } - if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) { + if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { XVA_SET_REQ(xvap, XAT_IMMUTABLE); } - if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) { + if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { XVA_SET_REQ(xvap, XAT_NODUMP); } - if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) { + if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { XVA_SET_REQ(xvap, XAT_AV_MODIFIED); } - if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) { + if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); } if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) - ASSERT(vp->v_type == VREG); + ASSERT(S_ISREG(ip->i_mode)); zfs_xvattr_set(zp, xvap, tx); } if (fuid_dirtied) - zfs_fuid_sync(zfsvfs, tx); + zfs_fuid_sync(zsb, tx); if (mask != 0) zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); mutex_exit(&zp->z_lock); - if (mask & (AT_UID|AT_GID|AT_MODE)) + if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_exit(&zp->z_acl_lock); if (attrzp) { - if (mask & (AT_UID|AT_GID|AT_MODE)) + if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) mutex_exit(&attrzp->z_acl_lock); mutex_exit(&attrzp->z_lock); } @@ -3196,7 +3001,7 @@ out: } if (attrzp) - VN_RELE(ZTOV(attrzp)); + iput(ZTOI(attrzp)); if (aclp) zfs_acl_free(aclp); @@ -3212,15 +3017,21 @@ out: } else { err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); dmu_tx_commit(tx); + zfs_inode_update(zp); } out2: - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); +out3: + kmem_free(xattr_bulk, sizeof(sa_bulk_attr_t) * 7); + kmem_free(bulk, sizeof(sa_bulk_attr_t) * 7); + kmem_free(tmpxvattr, sizeof(xvattr_t)); + ZFS_EXIT(zsb); return (err); } +EXPORT_SYMBOL(zfs_setattr); typedef struct zfs_zlock { krwlock_t *zl_rwlock; /* lock we acquired */ @@ -3238,7 +3049,7 @@ zfs_rename_unlock(zfs_zlock_t **zlpp) while ((zl = *zlpp) != NULL) { if (zl->zl_znode != NULL) - VN_RELE(ZTOV(zl->zl_znode)); + iput(ZTOI(zl->zl_znode)); rw_exit(zl->zl_rwlock); *zlpp = zl->zl_next; kmem_free(zl, sizeof (*zl)); @@ -3256,7 +3067,7 @@ zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) { zfs_zlock_t *zl; znode_t *zp = tdzp; - uint64_t rootid = zp->z_zfsvfs->z_root; + uint64_t rootid = ZTOZSB(zp)->z_root; uint64_t oidp = zp->z_id; krwlock_t *rwlp = &szp->z_parent_lock; krw_t rw = RW_WRITER; @@ -3304,12 +3115,12 @@ zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) return (0); if (rw == RW_READER) { /* i.e. not the first pass */ - int error = zfs_zget(zp->z_zfsvfs, oidp, &zp); + int error = zfs_zget(ZTOZSB(zp), oidp, &zp); if (error) return (error); zl->zl_znode = zp; } - (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zp->z_zfsvfs), + (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), &oidp, sizeof (oidp)); rwlp = &zp->z_parent_lock; rw = RW_READER; @@ -3323,30 +3134,28 @@ zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) * Move an entry from the provided source directory to the target * directory. Change the entry name as indicated. * - * IN: sdvp - Source directory containing the "old entry". + * IN: sdip - Source directory containing the "old entry". * snm - Old entry name. - * tdvp - Target directory to contain the "new entry". + * tdip - Target directory to contain the "new entry". * tnm - New entry name. * cr - credentials of caller. - * ct - caller context * flags - case flags * * RETURN: 0 if success * error code if failure * * Timestamps: - * sdvp,tdvp - ctime|mtime updated + * sdip,tdip - ctime|mtime updated */ /*ARGSUSED*/ -static int -zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr, - caller_context_t *ct, int flags) +int +zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, + cred_t *cr, int flags) { znode_t *tdzp, *szp, *tzp; - znode_t *sdzp = VTOZ(sdvp); - zfsvfs_t *zfsvfs = sdzp->z_zfsvfs; + znode_t *sdzp = ITOZ(sdip); + zfs_sb_t *zsb = ITOZSB(sdip); zilog_t *zilog; - vnode_t *realvp; zfs_dirlock_t *sdl, *tdl; dmu_tx_t *tx; zfs_zlock_t *zl; @@ -3354,26 +3163,20 @@ zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr, int error = 0; int zflg = 0; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(sdzp); - zilog = zfsvfs->z_log; + zilog = zsb->z_log; - /* - * Make sure we have the real vp for the target directory. - */ - if (VOP_REALVP(tdvp, &realvp, ct) == 0) - tdvp = realvp; - - if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) { - ZFS_EXIT(zfsvfs); + if (tdip->i_sb != sdip->i_sb) { + ZFS_EXIT(zsb); return (EXDEV); } - tdzp = VTOZ(tdvp); + tdzp = ITOZ(tdip); ZFS_VERIFY_ZP(tdzp); - if (zfsvfs->z_utf8 && u8_validate(tnm, + if (zsb->z_utf8 && u8_validate(tnm, strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EILSEQ); } @@ -3391,7 +3194,7 @@ top: * See the comment in zfs_link() for why this is considered bad. */ if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EINVAL); } @@ -3410,10 +3213,10 @@ top: * First compare the two name arguments without * considering any case folding. */ - int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); + int nofold = (zsb->z_norm & ~U8_TEXTPREP_TOUPPER); cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); - ASSERT(error == 0 || !zfsvfs->z_utf8); + ASSERT(error == 0 || !zsb->z_utf8); if (cmp == 0) { /* * POSIX: "If the old argument and the new argument @@ -3421,7 +3224,7 @@ top: * the rename() function shall return successfully * and perform no other action." */ - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (0); } /* @@ -3442,10 +3245,10 @@ top: * is an exact match, we will allow this to proceed as * a name-change request. */ - if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || - (zfsvfs->z_case == ZFS_CASE_MIXED && + if ((zsb->z_case == ZFS_CASE_INSENSITIVE || + (zsb->z_case == ZFS_CASE_MIXED && flags & FIGNORECASE)) && - u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, + u8_strcmp(snm, tnm, 0, zsb->z_norm, U8_UNICODE_LATEST, &error) == 0) { /* * case preserving rename request, require exact @@ -3485,7 +3288,7 @@ top: if (!terr) { zfs_dirent_unlock(tdl); if (tzp) - VN_RELE(ZTOV(tzp)); + iput(ZTOI(tzp)); } if (sdzp == tdzp) @@ -3493,19 +3296,19 @@ top: if (strcmp(snm, "..") == 0) serr = EINVAL; - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (serr); } if (terr) { zfs_dirent_unlock(sdl); - VN_RELE(ZTOV(szp)); + iput(ZTOI(szp)); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); if (strcmp(tnm, "..") == 0) terr = EINVAL; - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (terr); } @@ -3516,15 +3319,15 @@ top: * done in a single check. */ - if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)) + if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) goto out; - if (ZTOV(szp)->v_type == VDIR) { + if (S_ISDIR(ZTOI(szp)->i_mode)) { /* * Check to make sure rename is valid. * Can't do a move like this: /usr/a/b to /usr/a/b/c/d */ - if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl)) + if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) goto out; } @@ -3535,13 +3338,13 @@ top: /* * Source and target must be the same type. */ - if (ZTOV(szp)->v_type == VDIR) { - if (ZTOV(tzp)->v_type != VDIR) { + if (S_ISDIR(ZTOI(szp)->i_mode)) { + if (!S_ISDIR(ZTOI(tzp)->i_mode)) { error = ENOTDIR; goto out; } } else { - if (ZTOV(tzp)->v_type == VDIR) { + if (S_ISDIR(ZTOI(tzp)->i_mode)) { error = EISDIR; goto out; } @@ -3557,19 +3360,7 @@ top: } } - vnevent_rename_src(ZTOV(szp), sdvp, snm, ct); - if (tzp) - vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct); - - /* - * notify the target directory if it is not the same - * as source directory. - */ - if (tdvp != sdvp) { - vnevent_rename_dest_dir(tdvp, ct); - } - - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); @@ -3584,7 +3375,7 @@ top: } zfs_sa_upgrade_txholds(tx, szp); - dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); + dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); error = dmu_tx_assign(tx, TXG_NOWAIT); if (error) { if (zl != NULL) @@ -3595,16 +3386,16 @@ top: if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); - VN_RELE(ZTOV(szp)); + iput(ZTOI(szp)); if (tzp) - VN_RELE(ZTOV(tzp)); + iput(ZTOI(tzp)); if (error == ERESTART) { dmu_tx_wait(tx); dmu_tx_abort(tx); goto top; } dmu_tx_abort(tx); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } @@ -3616,21 +3407,15 @@ top: if (error == 0) { szp->z_pflags |= ZFS_AV_MODIFIED; - error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), + error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zsb), (void *)&szp->z_pflags, sizeof (uint64_t), tx); - ASSERT3U(error, ==, 0); + ASSERT0(error); error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); if (error == 0) { zfs_log_rename(zilog, tx, TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0), sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp); - - /* - * Update path information for the target vnode - */ - vn_renamepath(tdvp, ZTOV(szp), tnm, - strlen(tnm)); } else { /* * At this point, we have successfully created @@ -3658,47 +3443,54 @@ out: zfs_dirent_unlock(sdl); zfs_dirent_unlock(tdl); + zfs_inode_update(sdzp); if (sdzp == tdzp) rw_exit(&sdzp->z_name_lock); + if (sdzp != tdzp) + zfs_inode_update(tdzp); - VN_RELE(ZTOV(szp)); - if (tzp) - VN_RELE(ZTOV(tzp)); + zfs_inode_update(szp); + iput(ZTOI(szp)); + if (tzp) { + zfs_inode_update(tzp); + iput(ZTOI(tzp)); + } - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_rename); /* * Insert the indicated symbolic reference entry into the directory. * - * IN: dvp - Directory to contain new symbolic link. + * IN: dip - Directory to contain new symbolic link. * link - Name for new symlink entry. * vap - Attributes of new entry. * target - Target path of new symlink. + * * cr - credentials of caller. - * ct - caller context * flags - case flags * * RETURN: 0 if success * error code if failure * * Timestamps: - * dvp - ctime|mtime updated + * dip - ctime|mtime updated */ /*ARGSUSED*/ -static int -zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr, - caller_context_t *ct, int flags) +int +zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, + struct inode **ipp, cred_t *cr, int flags) { - znode_t *zp, *dzp = VTOZ(dvp); + znode_t *zp, *dzp = ITOZ(dip); zfs_dirlock_t *dl; dmu_tx_t *tx; - zfsvfs_t *zfsvfs = dzp->z_zfsvfs; + zfs_sb_t *zsb = ITOZSB(dip); zilog_t *zilog; uint64_t len = strlen(link); int error; @@ -3707,67 +3499,69 @@ zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr, boolean_t fuid_dirtied; uint64_t txtype = TX_SYMLINK; - ASSERT(vap->va_type == VLNK); + ASSERT(S_ISLNK(vap->va_mode)); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(dzp); - zilog = zfsvfs->z_log; + zilog = zsb->z_log; - if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), + if (zsb->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EILSEQ); } if (flags & FIGNORECASE) zflg |= ZCILOOK; if (len > MAXPATHLEN) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (ENAMETOOLONG); } if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids)) != 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } top: + *ipp = NULL; + /* * Attempt to lock directory; fail if entry already exists. */ error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); if (error) { zfs_acl_ids_free(&acl_ids); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } - if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { + if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } - if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { + if (zfs_acl_ids_overquota(zsb, &acl_ids)) { zfs_acl_ids_free(&acl_ids); zfs_dirent_unlock(dl); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EDQUOT); } - tx = dmu_tx_create(zfsvfs->z_os); - fuid_dirtied = zfsvfs->z_fuid_dirty; + tx = dmu_tx_create(zsb->z_os); + fuid_dirtied = zsb->z_fuid_dirty; dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + ZFS_SA_BASE_ATTR_SIZE + len); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); - if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { + if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, acl_ids.z_aclp->z_acl_bytes); } if (fuid_dirtied) - zfs_fuid_txhold(zfsvfs, tx); + zfs_fuid_txhold(zsb, tx); error = dmu_tx_assign(tx, TXG_NOWAIT); if (error) { zfs_dirent_unlock(dl); @@ -3778,7 +3572,7 @@ top: } zfs_acl_ids_free(&acl_ids); dmu_tx_abort(tx); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } @@ -3789,18 +3583,18 @@ top: zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); if (fuid_dirtied) - zfs_fuid_sync(zfsvfs, tx); + zfs_fuid_sync(zsb, tx); mutex_enter(&zp->z_lock); if (zp->z_is_sa) - error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), + error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zsb), link, len, tx); else zfs_sa_symlink(zp, link, len, tx); mutex_exit(&zp->z_lock); zp->z_size = len; - (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), + (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), &zp->z_size, sizeof (zp->z_size), tx); /* * Insert the new object into the directory. @@ -3811,141 +3605,139 @@ top: txtype |= TX_CI; zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); + zfs_inode_update(dzp); + zfs_inode_update(zp); + zfs_acl_ids_free(&acl_ids); dmu_tx_commit(tx); zfs_dirent_unlock(dl); - VN_RELE(ZTOV(zp)); + *ipp = ZTOI(zp); - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_symlink); /* * Return, in the buffer contained in the provided uio structure, - * the symbolic path referred to by vp. + * the symbolic path referred to by ip. * - * IN: vp - vnode of symbolic link. - * uoip - structure to contain the link path. + * IN: ip - inode of symbolic link + * uio - structure to contain the link path. * cr - credentials of caller. - * ct - caller context - * - * OUT: uio - structure to contain the link path. * * RETURN: 0 if success * error code if failure * * Timestamps: - * vp - atime updated + * ip - atime updated */ /* ARGSUSED */ -static int -zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct) +int +zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); int error; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); mutex_enter(&zp->z_lock); if (zp->z_is_sa) error = sa_lookup_uio(zp->z_sa_hdl, - SA_ZPL_SYMLINK(zfsvfs), uio); + SA_ZPL_SYMLINK(zsb), uio); else error = zfs_sa_readlink(zp, uio); mutex_exit(&zp->z_lock); - ZFS_ACCESSTIME_STAMP(zfsvfs, zp); - - ZFS_EXIT(zfsvfs); + ZFS_ACCESSTIME_STAMP(zsb, zp); + zfs_inode_update(zp); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_readlink); /* - * Insert a new entry into directory tdvp referencing svp. + * Insert a new entry into directory tdip referencing sip. * - * IN: tdvp - Directory to contain new entry. - * svp - vnode of new entry. + * IN: tdip - Directory to contain new entry. + * sip - inode of new entry. * name - name of new entry. * cr - credentials of caller. - * ct - caller context * * RETURN: 0 if success * error code if failure * * Timestamps: - * tdvp - ctime|mtime updated - * svp - ctime updated + * tdip - ctime|mtime updated + * sip - ctime updated */ /* ARGSUSED */ -static int -zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr, - caller_context_t *ct, int flags) +int +zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr) { - znode_t *dzp = VTOZ(tdvp); + znode_t *dzp = ITOZ(tdip); znode_t *tzp, *szp; - zfsvfs_t *zfsvfs = dzp->z_zfsvfs; + zfs_sb_t *zsb = ITOZSB(tdip); zilog_t *zilog; zfs_dirlock_t *dl; dmu_tx_t *tx; - vnode_t *realvp; int error; int zf = ZNEW; uint64_t parent; uid_t owner; - ASSERT(tdvp->v_type == VDIR); + ASSERT(S_ISDIR(tdip->i_mode)); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(dzp); - zilog = zfsvfs->z_log; - - if (VOP_REALVP(svp, &realvp, ct) == 0) - svp = realvp; + zilog = zsb->z_log; /* * POSIX dictates that we return EPERM here. * Better choices include ENOTSUP or EISDIR. */ - if (svp->v_type == VDIR) { - ZFS_EXIT(zfsvfs); + if (S_ISDIR(sip->i_mode)) { + ZFS_EXIT(zsb); return (EPERM); } - if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) { - ZFS_EXIT(zfsvfs); + if (sip->i_sb != tdip->i_sb) { + ZFS_EXIT(zsb); return (EXDEV); } - szp = VTOZ(svp); + szp = ITOZ(sip); ZFS_VERIFY_ZP(szp); /* Prevent links to .zfs/shares files */ - if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), + if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zsb), &parent, sizeof (uint64_t))) != 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } - if (parent == zfsvfs->z_shares_dir) { - ZFS_EXIT(zfsvfs); + if (parent == zsb->z_shares_dir) { + ZFS_EXIT(zsb); return (EPERM); } - if (zfsvfs->z_utf8 && u8_validate(name, + if (zsb->z_utf8 && u8_validate(name, strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EILSEQ); } +#ifdef HAVE_PN_UTILS if (flags & FIGNORECASE) zf |= ZCILOOK; +#endif /* HAVE_PN_UTILS */ /* * We do not support links between attributes and non-attributes @@ -3954,19 +3746,18 @@ zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr, * imposed in attribute space. */ if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EINVAL); } - - owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER); + owner = zfs_fuid_map_id(zsb, szp->z_uid, cr, ZFS_OWNER); if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EPERM); } - if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { - ZFS_EXIT(zfsvfs); + if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { + ZFS_EXIT(zsb); return (error); } @@ -3976,11 +3767,11 @@ top: */ error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); if (error) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } - tx = dmu_tx_create(zfsvfs->z_os); + tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); zfs_sa_upgrade_txholds(tx, szp); @@ -3994,7 +3785,7 @@ top: goto top; } dmu_tx_abort(tx); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } @@ -4002,8 +3793,10 @@ top: if (error == 0) { uint64_t txtype = TX_LINK; +#ifdef HAVE_PN_UTILS if (flags & FIGNORECASE) txtype |= TX_CI; +#endif /* HAVE_PN_UTILS */ zfs_log_link(zilog, tx, txtype, dzp, szp, name); } @@ -4011,294 +3804,235 @@ top: zfs_dirent_unlock(dl); - if (error == 0) { - vnevent_link(svp, ct); - } - - if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) + if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) zil_commit(zilog, 0); - ZFS_EXIT(zfsvfs); + zfs_inode_update(dzp); + zfs_inode_update(szp); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_link); -/* - * zfs_null_putapage() is used when the file system has been force - * unmounted. It just drops the pages. - */ -/* ARGSUSED */ -static int -zfs_null_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, - size_t *lenp, int flags, cred_t *cr) +static void +zfs_putpage_commit_cb(void *arg, int error) { - pvn_write_done(pp, B_INVAL|B_FORCE|B_ERROR); - return (0); + struct page *pp = arg; + + if (error) { + __set_page_dirty_nobuffers(pp); + + if (error != ECANCELED) + SetPageError(pp); + } else { + ClearPageError(pp); + } + + end_page_writeback(pp); } /* - * Push a page out to disk, klustering if possible. + * Push a page out to disk, once the page is on stable storage the + * registered commit callback will be run as notification of completion. * - * IN: vp - file to push page to. - * pp - page to push. - * flags - additional flags. - * cr - credentials of caller. - * - * OUT: offp - start of range pushed. - * lenp - len of range pushed. + * IN: ip - page mapped for inode. + * pp - page to push (page is locked) + * wbc - writeback control data * * RETURN: 0 if success * error code if failure * - * NOTE: callers must have locked the page to be pushed. On - * exit, the page (and all other pages in the kluster) must be - * unlocked. + * Timestamps: + * ip - ctime|mtime updated */ /* ARGSUSED */ -static int -zfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, - size_t *lenp, int flags, cred_t *cr) +int +zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); + loff_t offset; + loff_t pgoff; + unsigned int pglen; + rl_t *rl; dmu_tx_t *tx; - u_offset_t off, koff; - size_t len, klen; - int err; + caddr_t va; + int err = 0; + uint64_t mtime[2], ctime[2]; + sa_bulk_attr_t bulk[3]; + int cnt = 0; + int sync; - off = pp->p_offset; - len = PAGESIZE; - /* - * If our blocksize is bigger than the page size, try to kluster - * multiple pages so that we write a full block (thus avoiding - * a read-modify-write). - */ - if (off < zp->z_size && zp->z_blksz > PAGESIZE) { - klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE); - koff = ISP2(klen) ? P2ALIGN(off, (u_offset_t)klen) : 0; - ASSERT(koff <= zp->z_size); - if (koff + klen > zp->z_size) - klen = P2ROUNDUP(zp->z_size - koff, (uint64_t)PAGESIZE); - pp = pvn_write_kluster(vp, pp, &off, &len, koff, klen, flags); - } - ASSERT3U(btop(len), ==, btopr(len)); + ZFS_ENTER(zsb); + ZFS_VERIFY_ZP(zp); - /* - * Can't push pages past end-of-file. - */ - if (off >= zp->z_size) { - /* ignore all pages */ - err = 0; - goto out; - } else if (off + len > zp->z_size) { - int npages = btopr(zp->z_size - off); - page_t *trunc; + ASSERT(PageLocked(pp)); + + pgoff = page_offset(pp); /* Page byte-offset in file */ + offset = i_size_read(ip); /* File length in bytes */ + pglen = MIN(PAGE_CACHE_SIZE, /* Page length in bytes */ + P2ROUNDUP(offset, PAGE_CACHE_SIZE)-pgoff); - page_list_break(&pp, &trunc, npages); - /* ignore pages past end of file */ - if (trunc) - pvn_write_done(trunc, flags); - len = zp->z_size - off; + /* Page is beyond end of file */ + if (pgoff >= offset) { + unlock_page(pp); + ZFS_EXIT(zsb); + return (0); } - if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || - zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { + /* Truncate page length to end of file */ + if (pgoff + pglen > offset) + pglen = offset - pgoff; + +#if 0 + /* + * FIXME: Allow mmap writes past its quota. The correct fix + * is to register a page_mkwrite() handler to count the page + * against its quota when it is about to be dirtied. + */ + if (zfs_owner_overquota(zsb, zp, B_FALSE) || + zfs_owner_overquota(zsb, zp, B_TRUE)) { err = EDQUOT; - goto out; } -top: - tx = dmu_tx_create(zfsvfs->z_os); - dmu_tx_hold_write(tx, zp->z_id, off, len); +#endif + + set_page_writeback(pp); + unlock_page(pp); + + rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER); + tx = dmu_tx_create(zsb->z_os); + + sync = ((zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) || + (wbc->sync_mode == WB_SYNC_ALL)); + if (!sync) + dmu_tx_callback_register(tx, zfs_putpage_commit_cb, pp); + + dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); err = dmu_tx_assign(tx, TXG_NOWAIT); if (err != 0) { - if (err == ERESTART) { + if (err == ERESTART) dmu_tx_wait(tx); - dmu_tx_abort(tx); - goto top; - } + + /* Will call all registered commit callbacks */ dmu_tx_abort(tx); - goto out; - } - if (zp->z_blksz <= PAGESIZE) { - caddr_t va = zfs_map_page(pp, S_READ); - ASSERT3U(len, <=, PAGESIZE); - dmu_write(zfsvfs->z_os, zp->z_id, off, len, va, tx); - zfs_unmap_page(pp, va); - } else { - err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, pp, tx); + /* + * For the synchronous case the commit callback must be + * explicitly called because there is no registered callback. + */ + if (sync) + zfs_putpage_commit_cb(pp, ECANCELED); + + zfs_range_unlock(rl); + ZFS_EXIT(zsb); + return (err); } - if (err == 0) { - uint64_t mtime[2], ctime[2]; - sa_bulk_attr_t bulk[3]; - int count = 0; + va = kmap(pp); + ASSERT3U(pglen, <=, PAGE_CACHE_SIZE); + dmu_write(zsb->z_os, zp->z_id, pgoff, pglen, va, tx); + kunmap(pp); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, - &mtime, 16); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, - &ctime, 16); - SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, - &zp->z_pflags, 8); - zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, - B_TRUE); - zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, len, 0); - } + SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); + SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); + SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, 8); + + /* Preserve the mtime and ctime provided by the inode */ + ZFS_TIME_ENCODE(&ip->i_mtime, mtime); + ZFS_TIME_ENCODE(&ip->i_ctime, ctime); + zp->z_atime_dirty = 0; + zp->z_seq++; + + err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); + + zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0); dmu_tx_commit(tx); -out: - pvn_write_done(pp, (err ? B_ERROR : 0) | flags); - if (offp) - *offp = off; - if (lenp) - *lenp = len; + zfs_range_unlock(rl); + + if (sync) { + zil_commit(zsb->z_log, zp->z_id); + zfs_putpage_commit_cb(pp, err); + } + ZFS_EXIT(zsb); return (err); } /* - * Copy the portion of the file indicated from pages into the file. - * The pages are stored in a page list attached to the files vnode. - * - * IN: vp - vnode of file to push page data to. - * off - position in file to put data. - * len - amount of data to write. - * flags - flags to control the operation. - * cr - credentials of caller. - * ct - caller context. - * - * RETURN: 0 if success - * error code if failure - * - * Timestamps: - * vp - ctime|mtime updated + * Update the system attributes when the inode has been dirtied. For the + * moment we're conservative and only update the atime, mtime, and ctime. */ -/*ARGSUSED*/ -static int -zfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr, - caller_context_t *ct) +int +zfs_dirty_inode(struct inode *ip, int flags) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - page_t *pp; - size_t io_len; - u_offset_t io_off; - uint_t blksz; - rl_t *rl; - int error = 0; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); + dmu_tx_t *tx; + uint64_t atime[2], mtime[2], ctime[2]; + sa_bulk_attr_t bulk[3]; + int error; + int cnt = 0; - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - /* - * Align this request to the file block size in case we kluster. - * XXX - this can result in pretty aggresive locking, which can - * impact simultanious read/write access. One option might be - * to break up long requests (len == 0) into block-by-block - * operations to get narrower locking. - */ - blksz = zp->z_blksz; - if (ISP2(blksz)) - io_off = P2ALIGN_TYPED(off, blksz, u_offset_t); - else - io_off = 0; - if (len > 0 && ISP2(blksz)) - io_len = P2ROUNDUP_TYPED(len + (off - io_off), blksz, size_t); - else - io_len = 0; + tx = dmu_tx_create(zsb->z_os); - if (io_len == 0) { - /* - * Search the entire vp list for pages >= io_off. - */ - rl = zfs_range_lock(zp, io_off, UINT64_MAX, RL_WRITER); - error = pvn_vplist_dirty(vp, io_off, zfs_putapage, flags, cr); - goto out; - } - rl = zfs_range_lock(zp, io_off, io_len, RL_WRITER); + dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); + zfs_sa_upgrade_txholds(tx, zp); - if (off > zp->z_size) { - /* past end of file */ - zfs_range_unlock(rl); - ZFS_EXIT(zfsvfs); - return (0); + error = dmu_tx_assign(tx, TXG_WAIT); + if (error) { + dmu_tx_abort(tx); + goto out; } - len = MIN(io_len, P2ROUNDUP(zp->z_size, PAGESIZE) - io_off); + mutex_enter(&zp->z_lock); + SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zsb), NULL, &atime, 16); + SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); + SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); - for (off = io_off; io_off < off + len; io_off += io_len) { - if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) { - pp = page_lookup(vp, io_off, - (flags & (B_INVAL | B_FREE)) ? SE_EXCL : SE_SHARED); - } else { - pp = page_lookup_nowait(vp, io_off, - (flags & B_FREE) ? SE_EXCL : SE_SHARED); - } + /* Preserve the mtime and ctime provided by the inode */ + ZFS_TIME_ENCODE(&ip->i_atime, atime); + ZFS_TIME_ENCODE(&ip->i_mtime, mtime); + ZFS_TIME_ENCODE(&ip->i_ctime, ctime); + zp->z_atime_dirty = 0; - if (pp != NULL && pvn_getdirty(pp, flags)) { - int err; + error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); + mutex_exit(&zp->z_lock); - /* - * Found a dirty page to push - */ - err = zfs_putapage(vp, pp, &io_off, &io_len, flags, cr); - if (err) - error = err; - } else { - io_len = PAGESIZE; - } - } + dmu_tx_commit(tx); out: - zfs_range_unlock(rl); - if ((flags & B_ASYNC) == 0 || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) - zil_commit(zfsvfs->z_log, zp->z_id); - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (error); } +EXPORT_SYMBOL(zfs_dirty_inode); /*ARGSUSED*/ void -zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) +zfs_inactive(struct inode *ip) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); int error; - rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); - if (zp->z_sa_hdl == NULL) { - /* - * The fs has been unmounted, or we did a - * suspend/resume and this file no longer exists. - */ - if (vn_has_cached_data(vp)) { - (void) pvn_vplist_dirty(vp, 0, zfs_null_putapage, - B_INVAL, cr); - } - - mutex_enter(&zp->z_lock); - mutex_enter(&vp->v_lock); - ASSERT(vp->v_count == 1); - vp->v_count = 0; - mutex_exit(&vp->v_lock); - mutex_exit(&zp->z_lock); - rw_exit(&zfsvfs->z_teardown_inactive_lock); - zfs_znode_free(zp); + if (zfsctl_is_node(ip)) { + zfsctl_inode_inactive(ip); return; } - /* - * Attempt to push any data in the page cache. If this fails - * we will get kicked out later in zfs_zinactive(). - */ - if (vn_has_cached_data(vp)) { - (void) pvn_vplist_dirty(vp, 0, zfs_putapage, B_INVAL|B_ASYNC, - cr); + rw_enter(&zsb->z_teardown_inactive_lock, RW_READER); + if (zp->z_sa_hdl == NULL) { + rw_exit(&zsb->z_teardown_inactive_lock); + return; } if (zp->z_atime_dirty && zp->z_unlinked == 0) { - dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); + dmu_tx_t *tx = dmu_tx_create(zsb->z_os); dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); zfs_sa_upgrade_txholds(tx, zp); @@ -4307,7 +4041,7 @@ zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) dmu_tx_abort(tx); } else { mutex_enter(&zp->z_lock); - (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), + (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), (void *)&zp->z_atime, sizeof (zp->z_atime), tx); zp->z_atime_dirty = 0; mutex_exit(&zp->z_lock); @@ -4316,13 +4050,14 @@ zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) } zfs_zinactive(zp); - rw_exit(&zfsvfs->z_teardown_inactive_lock); + rw_exit(&zsb->z_teardown_inactive_lock); } +EXPORT_SYMBOL(zfs_inactive); /* * Bounds-check the seek operation. * - * IN: vp - vnode seeking within + * IN: ip - inode seeking within * ooff - old file offset * noffp - pointer to new file offset * ct - caller context @@ -4331,138 +4066,69 @@ zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) * EINVAL if new offset invalid */ /* ARGSUSED */ -static int -zfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp, - caller_context_t *ct) +int +zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) { - if (vp->v_type == VDIR) + if (S_ISDIR(ip->i_mode)) return (0); return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); } +EXPORT_SYMBOL(zfs_seek); /* - * Pre-filter the generic locking function to trap attempts to place - * a mandatory lock on a memory mapped file. + * Fill pages with data from the disk. */ static int -zfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset, - flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct) +zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - - ZFS_ENTER(zfsvfs); - ZFS_VERIFY_ZP(zp); + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); + objset_t *os; + struct page *cur_pp; + u_offset_t io_off, total; + size_t io_len; + loff_t i_size; + unsigned page_idx; + int err; + + os = zsb->z_os; + io_len = nr_pages << PAGE_CACHE_SHIFT; + i_size = i_size_read(ip); + io_off = page_offset(pl[0]); + + if (io_off + io_len > i_size) + io_len = i_size - io_off; /* - * We are following the UFS semantics with respect to mapcnt - * here: If we see that the file is mapped already, then we will - * return an error, but we don't worry about races between this - * function and zfs_map(). + * Iterate over list of pages and read each page individually. */ - if (zp->z_mapcnt > 0 && MANDMODE(zp->z_mode)) { - ZFS_EXIT(zfsvfs); - return (EAGAIN); - } - ZFS_EXIT(zfsvfs); - return (fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr, ct)); -} - -/* - * If we can't find a page in the cache, we will create a new page - * and fill it with file data. For efficiency, we may try to fill - * multiple pages at once (klustering) to fill up the supplied page - * list. Note that the pages to be filled are held with an exclusive - * lock to prevent access by other threads while they are being filled. - */ -static int -zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg, - caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw) -{ - znode_t *zp = VTOZ(vp); - page_t *pp, *cur_pp; - objset_t *os = zp->z_zfsvfs->z_os; - u_offset_t io_off, total; - size_t io_len; - int err; - - if (plsz == PAGESIZE || zp->z_blksz <= PAGESIZE) { - /* - * We only have a single page, don't bother klustering - */ - io_off = off; - io_len = PAGESIZE; - pp = page_create_va(vp, io_off, io_len, - PG_EXCL | PG_WAIT, seg, addr); - } else { - /* - * Try to find enough pages to fill the page list - */ - pp = pvn_read_kluster(vp, off, seg, addr, &io_off, - &io_len, off, plsz, 0); - } - if (pp == NULL) { - /* - * The page already exists, nothing to do here. - */ - *pl = NULL; - return (0); - } - - /* - * Fill the pages in the kluster. - */ - cur_pp = pp; + page_idx = 0; + cur_pp = pl[0]; for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { caddr_t va; - ASSERT3U(io_off, ==, cur_pp->p_offset); - va = zfs_map_page(cur_pp, S_WRITE); + va = kmap(cur_pp); err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, DMU_READ_PREFETCH); - zfs_unmap_page(cur_pp, va); + kunmap(cur_pp); if (err) { - /* On error, toss the entire kluster */ - pvn_read_done(pp, B_ERROR); /* convert checksum errors into IO errors */ if (err == ECKSUM) err = EIO; return (err); } - cur_pp = cur_pp->p_next; + cur_pp = pl[++page_idx]; } - /* - * Fill in the page list array from the kluster starting - * from the desired offset `off'. - * NOTE: the page list will always be null terminated. - */ - pvn_plist_init(pp, pl, plsz, off, io_len, rw); - ASSERT(pl == NULL || (*pl)->p_offset == off); - return (0); } /* - * Return pointers to the pages for the file region [off, off + len] - * in the pl array. If plsz is greater than len, this function may - * also return page pointers from after the specified region - * (i.e. the region [off, off + plsz]). These additional pages are - * only returned if they are already in the cache, or were created as - * part of a klustered read. + * Uses zfs_fillpage to read data from the file and fill the pages. * - * IN: vp - vnode of file to get data from. - * off - position in file to get data from. - * len - amount of data to retrieve. - * plsz - length of provided page list. - * seg - segment to obtain pages for. - * addr - virtual address of fault. - * rw - mode of created pages. - * cr - credentials of caller. - * ct - caller context. - * - * OUT: protp - protection mode of created pages. - * pl - list of pages created. + * IN: ip - inode of file to get data from. + * pl - list of pages to read + * nr_pages - number of pages to read * * RETURN: 0 if success * error code if failure @@ -4471,217 +4137,119 @@ zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg, * vp - atime updated */ /* ARGSUSED */ -static int -zfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp, - page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr, - enum seg_rw rw, cred_t *cr, caller_context_t *ct) +int +zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - page_t **pl0 = pl; - int err = 0; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); + int err; - /* we do our own caching, faultahead is unnecessary */ if (pl == NULL) return (0); - else if (len > plsz) - len = plsz; - else - len = P2ROUNDUP(len, PAGESIZE); - ASSERT(plsz >= len); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - if (protp) - *protp = PROT_ALL; + err = zfs_fillpage(ip, pl, nr_pages); - /* - * Loop through the requested range [off, off + len) looking - * for pages. If we don't find a page, we will need to create - * a new page and fill it with data from the file. - */ - while (len > 0) { - if (*pl = page_lookup(vp, off, SE_SHARED)) - *(pl+1) = NULL; - else if (err = zfs_fillpage(vp, off, seg, addr, pl, plsz, rw)) - goto out; - while (*pl) { - ASSERT3U((*pl)->p_offset, ==, off); - off += PAGESIZE; - addr += PAGESIZE; - if (len > 0) { - ASSERT3U(len, >=, PAGESIZE); - len -= PAGESIZE; - } - ASSERT3U(plsz, >=, PAGESIZE); - plsz -= PAGESIZE; - pl++; - } - } + if (!err) + ZFS_ACCESSTIME_STAMP(zsb, zp); - /* - * Fill out the page array with any pages already in the cache. - */ - while (plsz > 0 && - (*pl++ = page_lookup_nowait(vp, off, SE_SHARED))) { - off += PAGESIZE; - plsz -= PAGESIZE; - } -out: - if (err) { - /* - * Release any pages we have previously locked. - */ - while (pl > pl0) - page_unlock(*--pl); - } else { - ZFS_ACCESSTIME_STAMP(zfsvfs, zp); - } - - *pl = NULL; - - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (err); } +EXPORT_SYMBOL(zfs_getpage); /* - * Request a memory map for a section of a file. This code interacts - * with common code and the VM system as follows: - * - * common code calls mmap(), which ends up in smmap_common() + * Check ZFS specific permissions to memory map a section of a file. * - * this calls VOP_MAP(), which takes you into (say) zfs + * IN: ip - inode of the file to mmap + * off - file offset + * addrp - start address in memory region + * len - length of memory region + * vm_flags- address flags * - * zfs_map() calls as_map(), passing segvn_create() as the callback - * - * segvn_create() creates the new segment and calls VOP_ADDMAP() - * - * zfs_addmap() updates z_mapcnt + * RETURN: 0 if success + * error code if failure */ /*ARGSUSED*/ -static int -zfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp, - size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, - caller_context_t *ct) +int +zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, + unsigned long vm_flags) { - znode_t *zp = VTOZ(vp); - zfsvfs_t *zfsvfs = zp->z_zfsvfs; - segvn_crargs_t vn_a; - int error; + znode_t *zp = ITOZ(ip); + zfs_sb_t *zsb = ITOZSB(ip); - ZFS_ENTER(zfsvfs); + ZFS_ENTER(zsb); ZFS_VERIFY_ZP(zp); - if ((prot & PROT_WRITE) && (zp->z_pflags & + if ((vm_flags & VM_WRITE) && (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EPERM); } - if ((prot & (PROT_READ | PROT_EXEC)) && + if ((vm_flags & (VM_READ | VM_EXEC)) && (zp->z_pflags & ZFS_AV_QUARANTINED)) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (EACCES); } - if (vp->v_flag & VNOMAP) { - ZFS_EXIT(zfsvfs); - return (ENOSYS); - } - if (off < 0 || len > MAXOFFSET_T - off) { - ZFS_EXIT(zfsvfs); + ZFS_EXIT(zsb); return (ENXIO); } - if (vp->v_type != VREG) { - ZFS_EXIT(zfsvfs); - return (ENODEV); - } - - /* - * If file is locked, disallow mapping. - */ - if (MANDMODE(zp->z_mode) && vn_has_flocks(vp)) { - ZFS_EXIT(zfsvfs); - return (EAGAIN); - } - - as_rangelock(as); - error = choose_addr(as, addrp, len, off, ADDR_VACALIGN, flags); - if (error != 0) { - as_rangeunlock(as); - ZFS_EXIT(zfsvfs); - return (error); - } - - vn_a.vp = vp; - vn_a.offset = (u_offset_t)off; - vn_a.type = flags & MAP_TYPE; - vn_a.prot = prot; - vn_a.maxprot = maxprot; - vn_a.cred = cr; - vn_a.amp = NULL; - vn_a.flags = flags & ~MAP_TYPE; - vn_a.szc = 0; - vn_a.lgrp_mem_policy_flags = 0; - - error = as_map(as, *addrp, len, segvn_create, &vn_a); - - as_rangeunlock(as); - ZFS_EXIT(zfsvfs); - return (error); -} - -/* ARGSUSED */ -static int -zfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, - size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, - caller_context_t *ct) -{ - uint64_t pages = btopr(len); - - atomic_add_64(&VTOZ(vp)->z_mapcnt, pages); + ZFS_EXIT(zsb); return (0); } +EXPORT_SYMBOL(zfs_map); /* - * The reason we push dirty pages as part of zfs_delmap() is so that we get a - * more accurate mtime for the associated file. Since we don't have a way of - * detecting when the data was actually modified, we have to resort to - * heuristics. If an explicit msync() is done, then we mark the mtime when the - * last page is pushed. The problem occurs when the msync() call is omitted, - * which by far the most common case: - * - * open() - * mmap() - * - * munmap() - * close() - *