2 * Copyright (c) 1999, Boris Popov
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by Boris Popov.
16 * 4. Neither the name of the author nor the names of any co-contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/kernel.h>
40 #include <sys/mount.h>
41 #include <sys/namei.h>
42 #include <sys/vnode.h>
43 #include <sys/dirent.h>
44 #include <sys/sysctl.h>
47 #include <vm/vm_page.h>
48 #include <vm/vm_extern.h>
49 #include <vm/vm_object.h>
50 #include <vm/vm_pager.h>
51 #include <vm/vnode_pager.h>
53 #include <netncp/ncp.h>
54 #include <netncp/ncp_conn.h>
55 #include <netncp/ncp_subr.h>
56 #include <netncp/ncp_ncp.h>
58 #include <fs/nwfs/nwfs.h>
59 #include <fs/nwfs/nwfs_node.h>
60 #include <fs/nwfs/nwfs_subr.h>
62 static int nwfs_fastlookup = 1;
64 SYSCTL_DECL(_vfs_nwfs);
65 SYSCTL_INT(_vfs_nwfs, OID_AUTO, fastlookup, CTLFLAG_RW, &nwfs_fastlookup, 0, "");
68 extern int nwfs_pbuf_freecnt;
70 #define DE_SIZE (sizeof(struct dirent))
74 nwfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred) {
75 struct nwmount *nmp = VTONWFS(vp);
78 struct nwnode *np = VTONW(vp);
79 struct nw_entry_info fattr;
81 struct componentname cn;
85 NCPVNDEBUG("dirname='%s'\n",np->n_name);
86 if (uio->uio_resid < DE_SIZE || (uio->uio_offset < 0))
90 i = uio->uio_offset / DE_SIZE; /* offset in directory */
92 error = ncp_initsearch(vp, uio->uio_td, cred);
94 NCPVNDEBUG("cannot initialize search, error=%d",error);
99 for (; uio->uio_resid >= DE_SIZE; i++) {
100 bzero((char *) &dp, DE_SIZE);
101 dp.d_reclen = DE_SIZE;
105 dp.d_fileno = (i == 0) ? np->n_fid.f_id : np->n_parent.f_id;
106 if (!dp.d_fileno) dp.d_fileno = NWFS_ROOT_INO;
110 dp.d_name[i + 1] = '\0';
114 error = ncp_search_for_file_or_subdir(nmp, &np->n_seq, &fattr, uio->uio_td, cred);
115 if (error && error < 0x80) break;
116 dp.d_fileno = fattr.dirEntNum;
117 dp.d_type = (fattr.attributes & aDIR) ? DT_DIR : DT_REG;
118 dp.d_namlen = fattr.nameLen;
119 bcopy(fattr.entryName, dp.d_name, dp.d_namlen);
120 dp.d_name[dp.d_namlen] = '\0';
122 if (error && eofflag) {
129 if (nwfs_fastlookup && !error && i > 1) {
130 fid.f_id = fattr.dirEntNum;
131 fid.f_parent = np->n_fid.f_id;
132 error = nwfs_nget(vp->v_mount, fid, &fattr, vp, &newvp);
134 VTONW(newvp)->n_ctime = VTONW(newvp)->n_vattr.va_ctime.tv_sec;
135 cn.cn_nameptr = dp.d_name;
136 cn.cn_namelen = dp.d_namlen;
137 cache_enter(vp, newvp, &cn);
146 if ((error = uiomove(&dp, DE_SIZE, uio)))
150 uio->uio_offset = i * DE_SIZE;
155 nwfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred) {
156 struct nwmount *nmp = VFSTONWFS(vp->v_mount);
157 struct nwnode *np = VTONW(vp);
162 if (vp->v_type != VREG && vp->v_type != VDIR) {
163 printf("%s: vn types other than VREG or VDIR are unsupported !\n",__func__);
166 if (uiop->uio_resid == 0) return 0;
167 if (uiop->uio_offset < 0) return EINVAL;
168 /* if (uiop->uio_offset + uiop->uio_resid > nmp->nm_maxfilesize)
171 if (vp->v_type == VDIR) {
172 error = nwfs_readvdir(vp, uiop, cred);
175 biosize = NWFSTOCONN(nmp)->buffer_size;
176 if (np->n_flag & NMODIFIED) {
177 nwfs_attr_cacheremove(vp);
178 error = VOP_GETATTR(vp, &vattr, cred);
179 if (error) return (error);
180 np->n_mtime = vattr.va_mtime.tv_sec;
182 error = VOP_GETATTR(vp, &vattr, cred);
183 if (error) return (error);
184 if (np->n_mtime != vattr.va_mtime.tv_sec) {
185 error = nwfs_vinvalbuf(vp, td);
186 if (error) return (error);
187 np->n_mtime = vattr.va_mtime.tv_sec;
190 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
195 nwfs_writevnode(vp, uiop, cred, ioflag)
201 struct nwmount *nmp = VTONWFS(vp);
202 struct nwnode *np = VTONW(vp);
204 /* struct vattr vattr;*/
207 if (vp->v_type != VREG) {
208 printf("%s: vn types other than VREG unsupported !\n",__func__);
211 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
212 if (uiop->uio_offset < 0) return EINVAL;
213 /* if (uiop->uio_offset + uiop->uio_resid > nmp->nm_maxfilesize)
216 if (ioflag & (IO_APPEND | IO_SYNC)) {
217 if (np->n_flag & NMODIFIED) {
218 nwfs_attr_cacheremove(vp);
219 error = nwfs_vinvalbuf(vp, td);
220 if (error) return (error);
222 if (ioflag & IO_APPEND) {
223 /* We can relay only on local information about file size,
224 * because until file is closed NetWare will not return
225 * the correct size. */
227 nwfs_attr_cacheremove(vp);
228 error = VOP_GETATTR(vp, &vattr, cred);
229 if (error) return (error);
231 uiop->uio_offset = np->n_size;
234 if (uiop->uio_resid == 0) return 0;
236 if (vn_rlimit_fsize(vp, uiop, td))
239 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
240 NCPVNDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
242 if (uiop->uio_offset > np->n_size) {
243 np->n_vattr.va_size = np->n_size = uiop->uio_offset;
244 vnode_pager_setsize(vp, np->n_size);
251 * Do an I/O operation to/from a cache block.
254 nwfs_doio(vp, bp, cr, td)
268 nmp = VFSTONWFS(vp->v_mount);
271 uiop->uio_iovcnt = 1;
272 uiop->uio_segflg = UIO_SYSSPACE;
274 if (bp->b_iocmd == BIO_READ) {
275 io.iov_len = uiop->uio_resid = bp->b_bcount;
276 io.iov_base = bp->b_data;
277 uiop->uio_rw = UIO_READ;
278 switch (vp->v_type) {
280 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
281 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
284 if (uiop->uio_resid) {
285 int left = uiop->uio_resid;
286 int nread = bp->b_bcount - left;
288 bzero((char *)bp->b_data + nread, left);
292 nfsstats.readdir_bios++;
293 uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * NFS_DIRBLKSIZ;
294 if (nmp->nm_flag & NFSMNT_RDIRPLUS) {
295 error = nfs_readdirplusrpc(vp, uiop, cr);
296 if (error == NFSERR_NOTSUPP)
297 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
299 if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0)
300 error = nfs_readdirrpc(vp, uiop, cr);
301 if (error == 0 && uiop->uio_resid == bp->b_bcount)
302 bp->b_flags |= B_INVAL;
306 printf("nwfs_doio: type %x unexpected\n",vp->v_type);
310 bp->b_ioflags |= BIO_ERROR;
314 if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size)
315 bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE);
317 if (bp->b_dirtyend > bp->b_dirtyoff) {
318 io.iov_len = uiop->uio_resid = bp->b_dirtyend - bp->b_dirtyoff;
319 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
320 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
321 uiop->uio_rw = UIO_WRITE;
322 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
325 * For an interrupted write, the buffer is still valid
326 * and the write hasn't been pushed to the server yet,
327 * so we can't set BIO_ERROR and report the interruption
328 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
329 * is not relevant, so the rpc attempt is essentially
330 * a noop. For the case of a V3 write rpc not being
331 * committed to stable storage, the block is still
332 * dirty and requires either a commit rpc or another
333 * write rpc with iomode == NFSV3WRITE_FILESYNC before
334 * the block is reused. This is indicated by setting
335 * the B_DELWRI and B_NEEDCOMMIT flags.
338 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
342 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
343 if ((bp->b_flags & B_ASYNC) == 0)
344 bp->b_flags |= B_EINTR;
345 if ((bp->b_flags & B_PAGING) == 0) {
347 bp->b_flags &= ~B_DONE;
349 if ((bp->b_flags & B_ASYNC) == 0)
350 bp->b_flags |= B_EINTR;
354 bp->b_ioflags |= BIO_ERROR;
355 bp->b_error /*= np->n_error */= error;
356 /* np->n_flag |= NWRITEERR;*/
358 bp->b_dirtyoff = bp->b_dirtyend = 0;
366 bp->b_resid = uiop->uio_resid;
372 * Vnode op for VM getpages.
373 * Wish wish .... get rid from multiple IO routines
377 struct vop_getpages_args /* {
382 vm_ooffset_t a_offset;
386 return vop_stdgetpages(ap);(ap->a_vp, ap->a_m, ap->a_count,
388 int i, error, nextoff, size, toff, npages, count;
402 td = curthread; /* XXX */
403 cred = td->td_ucred; /* XXX */
405 nmp = VFSTONWFS(vp->v_mount);
409 if ((object = vp->v_object) == NULL) {
410 printf("nwfs_getpages: called with non-merged cache vnode??\n");
411 return VM_PAGER_ERROR;
414 bp = getpbuf(&nwfs_pbuf_freecnt);
415 npages = btoc(count);
416 kva = (vm_offset_t) bp->b_data;
417 pmap_qenter(kva, pages, npages);
419 iov.iov_base = (caddr_t) kva;
423 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
424 uio.uio_resid = count;
425 uio.uio_segflg = UIO_SYSSPACE;
426 uio.uio_rw = UIO_READ;
429 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, &uio,cred);
430 pmap_qremove(kva, npages);
432 relpbuf(bp, &nwfs_pbuf_freecnt);
434 VM_OBJECT_LOCK(object);
435 if (error && (uio.uio_resid == count)) {
436 printf("nwfs_getpages: error %d\n",error);
437 vm_page_lock_queues();
438 for (i = 0; i < npages; i++) {
439 if (ap->a_reqpage != i)
440 vm_page_free(pages[i]);
442 vm_page_unlock_queues();
443 VM_OBJECT_UNLOCK(object);
444 return VM_PAGER_ERROR;
447 size = count - uio.uio_resid;
449 vm_page_lock_queues();
450 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
452 nextoff = toff + PAGE_SIZE;
455 if (nextoff <= size) {
456 m->valid = VM_PAGE_BITS_ALL;
457 KASSERT(m->dirty == 0,
458 ("nwfs_getpages: page %p is dirty", m));
460 int nvalid = ((size + DEV_BSIZE - 1) - toff) & ~(DEV_BSIZE - 1);
461 vm_page_set_valid(m, 0, nvalid);
462 KASSERT((m->dirty & vm_page_bits(0, nvalid)) == 0,
463 ("nwfs_getpages: page %p is dirty", m));
466 if (i != ap->a_reqpage) {
468 * Whether or not to leave the page activated is up in
469 * the air, but we should put the page on a page queue
470 * somewhere (it already is in the object). Result:
471 * It appears that emperical results show that
472 * deactivating pages is best.
476 * Just in case someone was asking for this page we
477 * now tell them that it is ok to use.
480 if (m->oflags & VPO_WANTED)
483 vm_page_deactivate(m);
490 vm_page_unlock_queues();
491 VM_OBJECT_UNLOCK(object);
493 #endif /* NWFS_RWCACHE */
497 * Vnode op for VM putpages.
498 * possible bug: all IO done in sync mode
499 * Note that vop_close always invalidate pages before close, so it's
500 * not necessary to open vnode.
504 struct vop_putpages_args /* {
510 vm_ooffset_t a_offset;
514 struct vnode *vp = ap->a_vp;
519 td = curthread; /* XXX */
520 cred = td->td_ucred; /* XXX */
521 VOP_OPEN(vp, FWRITE, cred, td, NULL);
522 error = vop_stdputpages(ap);
523 VOP_CLOSE(vp, FWRITE, cred, td);
530 int i, npages, count;
536 td = curthread; /* XXX */
537 cred = td->td_ucred; /* XXX */
538 /* VOP_OPEN(vp, FWRITE, cred, td, NULL);*/
540 nmp = VFSTONWFS(vp->v_mount);
543 rtvals = ap->a_rtvals;
544 npages = btoc(count);
546 for (i = 0; i < npages; i++) {
547 rtvals[i] = VM_PAGER_ERROR;
550 bp = getpbuf(&nwfs_pbuf_freecnt);
551 kva = (vm_offset_t) bp->b_data;
552 pmap_qenter(kva, pages, npages);
554 iov.iov_base = (caddr_t) kva;
558 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
559 uio.uio_resid = count;
560 uio.uio_segflg = UIO_SYSSPACE;
561 uio.uio_rw = UIO_WRITE;
563 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
565 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, &uio, cred);
566 /* VOP_CLOSE(vp, FWRITE, cred, td);*/
567 NCPVNDEBUG("paged write done: %d\n", error);
569 pmap_qremove(kva, npages);
570 relpbuf(bp, &nwfs_pbuf_freecnt);
573 vnode_pager_undirty_pages(pages, rtvals, count - uio.uio_resid);
575 #endif /* NWFS_RWCACHE */
578 * Flush and invalidate all dirty buffers. If another process is already
579 * doing the flush, just wait for completion.
582 nwfs_vinvalbuf(vp, td)
586 struct nwnode *np = VTONW(vp);
587 /* struct nwmount *nmp = VTONWFS(vp);*/
590 if (vp->v_iflag & VI_DOOMED)
593 while (np->n_flag & NFLUSHINPROG) {
594 np->n_flag |= NFLUSHWANT;
595 error = tsleep(&np->n_flag, PRIBIO + 2, "nwfsvinv", 2 * hz);
596 error = ncp_chkintr(NWFSTOCONN(VTONWFS(vp)), td);
600 np->n_flag |= NFLUSHINPROG;
602 if (vp->v_bufobj.bo_object != NULL) {
603 VM_OBJECT_LOCK(vp->v_bufobj.bo_object);
604 vm_object_page_clean(vp->v_bufobj.bo_object, 0, 0, OBJPC_SYNC);
605 VM_OBJECT_UNLOCK(vp->v_bufobj.bo_object);
608 error = vinvalbuf(vp, V_SAVE, PCATCH, 0);
610 if (error == ERESTART || error == EINTR) {
611 np->n_flag &= ~NFLUSHINPROG;
612 if (np->n_flag & NFLUSHWANT) {
613 np->n_flag &= ~NFLUSHWANT;
618 error = vinvalbuf(vp, V_SAVE, PCATCH, 0);
620 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
621 if (np->n_flag & NFLUSHWANT) {
622 np->n_flag &= ~NFLUSHWANT;