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/resourcevar.h> /* defines plimit structure in proc struct */
38 #include <sys/kernel.h>
42 #include <sys/mount.h>
43 #include <sys/namei.h>
44 #include <sys/vnode.h>
45 #include <sys/dirent.h>
46 #include <sys/signalvar.h>
47 #include <sys/sysctl.h>
50 #include <vm/vm_page.h>
51 #include <vm/vm_extern.h>
52 #include <vm/vm_object.h>
53 #include <vm/vm_pager.h>
54 #include <vm/vnode_pager.h>
56 #include <netncp/ncp.h>
57 #include <netncp/ncp_conn.h>
58 #include <netncp/ncp_subr.h>
59 #include <netncp/ncp_ncp.h>
61 #include <fs/nwfs/nwfs.h>
62 #include <fs/nwfs/nwfs_node.h>
63 #include <fs/nwfs/nwfs_subr.h>
65 static int nwfs_fastlookup = 1;
67 SYSCTL_DECL(_vfs_nwfs);
68 SYSCTL_INT(_vfs_nwfs, OID_AUTO, fastlookup, CTLFLAG_RW, &nwfs_fastlookup, 0, "");
71 extern int nwfs_pbuf_freecnt;
73 #define DE_SIZE (sizeof(struct dirent))
77 nwfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred) {
78 struct nwmount *nmp = VTONWFS(vp);
81 struct nwnode *np = VTONW(vp);
82 struct nw_entry_info fattr;
84 struct componentname cn;
88 NCPVNDEBUG("dirname='%s'\n",np->n_name);
89 if (uio->uio_resid < DE_SIZE || (uio->uio_offset < 0))
93 i = uio->uio_offset / DE_SIZE; /* offset in directory */
95 error = ncp_initsearch(vp, uio->uio_td, cred);
97 NCPVNDEBUG("cannot initialize search, error=%d",error);
102 for (; uio->uio_resid >= DE_SIZE; i++) {
103 bzero((char *) &dp, DE_SIZE);
104 dp.d_reclen = DE_SIZE;
108 dp.d_fileno = (i == 0) ? np->n_fid.f_id : np->n_parent.f_id;
109 if (!dp.d_fileno) dp.d_fileno = NWFS_ROOT_INO;
113 dp.d_name[i + 1] = '\0';
117 error = ncp_search_for_file_or_subdir(nmp, &np->n_seq, &fattr, uio->uio_td, cred);
118 if (error && error < 0x80) break;
119 dp.d_fileno = fattr.dirEntNum;
120 dp.d_type = (fattr.attributes & aDIR) ? DT_DIR : DT_REG;
121 dp.d_namlen = fattr.nameLen;
122 bcopy(fattr.entryName, dp.d_name, dp.d_namlen);
123 dp.d_name[dp.d_namlen] = '\0';
125 if (error && eofflag) {
132 if (nwfs_fastlookup && !error && i > 1) {
133 fid.f_id = fattr.dirEntNum;
134 fid.f_parent = np->n_fid.f_id;
135 error = nwfs_nget(vp->v_mount, fid, &fattr, vp, &newvp);
137 VTONW(newvp)->n_ctime = VTONW(newvp)->n_vattr.va_ctime.tv_sec;
138 cn.cn_nameptr = dp.d_name;
139 cn.cn_namelen = dp.d_namlen;
140 cache_enter(vp, newvp, &cn);
149 if ((error = uiomove(&dp, DE_SIZE, uio)))
153 uio->uio_offset = i * DE_SIZE;
158 nwfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred) {
159 struct nwmount *nmp = VFSTONWFS(vp->v_mount);
160 struct nwnode *np = VTONW(vp);
165 if (vp->v_type != VREG && vp->v_type != VDIR) {
166 printf("%s: vn types other than VREG or VDIR are unsupported !\n",__func__);
169 if (uiop->uio_resid == 0) return 0;
170 if (uiop->uio_offset < 0) return EINVAL;
171 /* if (uiop->uio_offset + uiop->uio_resid > nmp->nm_maxfilesize)
174 if (vp->v_type == VDIR) {
175 error = nwfs_readvdir(vp, uiop, cred);
178 biosize = NWFSTOCONN(nmp)->buffer_size;
179 if (np->n_flag & NMODIFIED) {
180 nwfs_attr_cacheremove(vp);
181 error = VOP_GETATTR(vp, &vattr, cred, td);
182 if (error) return (error);
183 np->n_mtime = vattr.va_mtime.tv_sec;
185 error = VOP_GETATTR(vp, &vattr, cred, td);
186 if (error) return (error);
187 if (np->n_mtime != vattr.va_mtime.tv_sec) {
188 error = nwfs_vinvalbuf(vp, td);
189 if (error) return (error);
190 np->n_mtime = vattr.va_mtime.tv_sec;
193 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
198 nwfs_writevnode(vp, uiop, cred, ioflag)
204 struct nwmount *nmp = VTONWFS(vp);
205 struct nwnode *np = VTONW(vp);
207 /* struct vattr vattr;*/
210 if (vp->v_type != VREG) {
211 printf("%s: vn types other than VREG unsupported !\n",__func__);
214 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
215 if (uiop->uio_offset < 0) return EINVAL;
216 /* if (uiop->uio_offset + uiop->uio_resid > nmp->nm_maxfilesize)
219 if (ioflag & (IO_APPEND | IO_SYNC)) {
220 if (np->n_flag & NMODIFIED) {
221 nwfs_attr_cacheremove(vp);
222 error = nwfs_vinvalbuf(vp, td);
223 if (error) return (error);
225 if (ioflag & IO_APPEND) {
226 /* We can relay only on local information about file size,
227 * because until file is closed NetWare will not return
228 * the correct size. */
230 nwfs_attr_cacheremove(vp);
231 error = VOP_GETATTR(vp, &vattr, cred, td);
232 if (error) return (error);
234 uiop->uio_offset = np->n_size;
237 if (uiop->uio_resid == 0) return 0;
239 PROC_LOCK(td->td_proc);
240 if (uiop->uio_offset + uiop->uio_resid >
241 lim_cur(td->td_proc, RLIMIT_FSIZE)) {
242 psignal(td->td_proc, SIGXFSZ);
243 PROC_UNLOCK(td->td_proc);
246 PROC_UNLOCK(td->td_proc);
248 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
249 NCPVNDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
251 if (uiop->uio_offset > np->n_size) {
252 np->n_vattr.va_size = np->n_size = uiop->uio_offset;
253 vnode_pager_setsize(vp, np->n_size);
260 * Do an I/O operation to/from a cache block.
263 nwfs_doio(vp, bp, cr, td)
277 nmp = VFSTONWFS(vp->v_mount);
280 uiop->uio_iovcnt = 1;
281 uiop->uio_segflg = UIO_SYSSPACE;
283 if (bp->b_iocmd == BIO_READ) {
284 io.iov_len = uiop->uio_resid = bp->b_bcount;
285 io.iov_base = bp->b_data;
286 uiop->uio_rw = UIO_READ;
287 switch (vp->v_type) {
289 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
290 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
293 if (uiop->uio_resid) {
294 int left = uiop->uio_resid;
295 int nread = bp->b_bcount - left;
297 bzero((char *)bp->b_data + nread, left);
301 nfsstats.readdir_bios++;
302 uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * NFS_DIRBLKSIZ;
303 if (nmp->nm_flag & NFSMNT_RDIRPLUS) {
304 error = nfs_readdirplusrpc(vp, uiop, cr);
305 if (error == NFSERR_NOTSUPP)
306 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
308 if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0)
309 error = nfs_readdirrpc(vp, uiop, cr);
310 if (error == 0 && uiop->uio_resid == bp->b_bcount)
311 bp->b_flags |= B_INVAL;
315 printf("nwfs_doio: type %x unexpected\n",vp->v_type);
319 bp->b_ioflags |= BIO_ERROR;
323 if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size)
324 bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE);
326 if (bp->b_dirtyend > bp->b_dirtyoff) {
327 io.iov_len = uiop->uio_resid = bp->b_dirtyend - bp->b_dirtyoff;
328 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
329 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
330 uiop->uio_rw = UIO_WRITE;
331 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
334 * For an interrupted write, the buffer is still valid
335 * and the write hasn't been pushed to the server yet,
336 * so we can't set BIO_ERROR and report the interruption
337 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
338 * is not relevant, so the rpc attempt is essentially
339 * a noop. For the case of a V3 write rpc not being
340 * committed to stable storage, the block is still
341 * dirty and requires either a commit rpc or another
342 * write rpc with iomode == NFSV3WRITE_FILESYNC before
343 * the block is reused. This is indicated by setting
344 * the B_DELWRI and B_NEEDCOMMIT flags.
347 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
351 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
352 if ((bp->b_flags & B_ASYNC) == 0)
353 bp->b_flags |= B_EINTR;
354 if ((bp->b_flags & B_PAGING) == 0) {
356 bp->b_flags &= ~B_DONE;
358 if ((bp->b_flags & B_ASYNC) == 0)
359 bp->b_flags |= B_EINTR;
363 bp->b_ioflags |= BIO_ERROR;
364 bp->b_error /*= np->n_error */= error;
365 /* np->n_flag |= NWRITEERR;*/
367 bp->b_dirtyoff = bp->b_dirtyend = 0;
375 bp->b_resid = uiop->uio_resid;
381 * Vnode op for VM getpages.
382 * Wish wish .... get rid from multiple IO routines
386 struct vop_getpages_args /* {
391 vm_ooffset_t a_offset;
395 return vop_stdgetpages(ap);(ap->a_vp, ap->a_m, ap->a_count,
397 int i, error, nextoff, size, toff, npages, count;
411 td = curthread; /* XXX */
412 cred = td->td_ucred; /* XXX */
414 nmp = VFSTONWFS(vp->v_mount);
418 if ((object = vp->v_object) == NULL) {
419 printf("nwfs_getpages: called with non-merged cache vnode??\n");
420 return VM_PAGER_ERROR;
423 bp = getpbuf(&nwfs_pbuf_freecnt);
424 npages = btoc(count);
425 kva = (vm_offset_t) bp->b_data;
426 pmap_qenter(kva, pages, npages);
428 iov.iov_base = (caddr_t) kva;
432 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
433 uio.uio_resid = count;
434 uio.uio_segflg = UIO_SYSSPACE;
435 uio.uio_rw = UIO_READ;
438 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, &uio,cred);
439 pmap_qremove(kva, npages);
441 relpbuf(bp, &nwfs_pbuf_freecnt);
443 VM_OBJECT_LOCK(object);
444 if (error && (uio.uio_resid == count)) {
445 printf("nwfs_getpages: error %d\n",error);
446 vm_page_lock_queues();
447 for (i = 0; i < npages; i++) {
448 if (ap->a_reqpage != i)
449 vm_page_free(pages[i]);
451 vm_page_unlock_queues();
452 VM_OBJECT_UNLOCK(object);
453 return VM_PAGER_ERROR;
456 size = count - uio.uio_resid;
458 vm_page_lock_queues();
459 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
461 nextoff = toff + PAGE_SIZE;
464 if (nextoff <= size) {
465 m->valid = VM_PAGE_BITS_ALL;
468 int nvalid = ((size + DEV_BSIZE - 1) - toff) & ~(DEV_BSIZE - 1);
469 vm_page_set_validclean(m, 0, nvalid);
472 if (i != ap->a_reqpage) {
474 * Whether or not to leave the page activated is up in
475 * the air, but we should put the page on a page queue
476 * somewhere (it already is in the object). Result:
477 * It appears that emperical results show that
478 * deactivating pages is best.
482 * Just in case someone was asking for this page we
483 * now tell them that it is ok to use.
486 if (m->flags & PG_WANTED)
489 vm_page_deactivate(m);
496 vm_page_unlock_queues();
497 VM_OBJECT_UNLOCK(object);
499 #endif /* NWFS_RWCACHE */
503 * Vnode op for VM putpages.
504 * possible bug: all IO done in sync mode
505 * Note that vop_close always invalidate pages before close, so it's
506 * not necessary to open vnode.
510 struct vop_putpages_args /* {
516 vm_ooffset_t a_offset;
520 struct vnode *vp = ap->a_vp;
525 td = curthread; /* XXX */
526 cred = td->td_ucred; /* XXX */
527 VOP_OPEN(vp, FWRITE, cred, td, -1);
528 error = vop_stdputpages(ap);
529 VOP_CLOSE(vp, FWRITE, cred, td);
536 int i, npages, count;
542 td = curthread; /* XXX */
543 cred = td->td_ucred; /* XXX */
544 /* VOP_OPEN(vp, FWRITE, cred, td, -1);*/
546 nmp = VFSTONWFS(vp->v_mount);
549 rtvals = ap->a_rtvals;
550 npages = btoc(count);
552 for (i = 0; i < npages; i++) {
553 rtvals[i] = VM_PAGER_AGAIN;
556 bp = getpbuf(&nwfs_pbuf_freecnt);
557 kva = (vm_offset_t) bp->b_data;
558 pmap_qenter(kva, pages, npages);
560 iov.iov_base = (caddr_t) kva;
564 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
565 uio.uio_resid = count;
566 uio.uio_segflg = UIO_SYSSPACE;
567 uio.uio_rw = UIO_WRITE;
569 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
571 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, &uio, cred);
572 /* VOP_CLOSE(vp, FWRITE, cred, td);*/
573 NCPVNDEBUG("paged write done: %d\n", error);
575 pmap_qremove(kva, npages);
576 relpbuf(bp, &nwfs_pbuf_freecnt);
579 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
580 vm_page_lock_queues();
581 for (i = 0; i < nwritten; i++) {
582 rtvals[i] = VM_PAGER_OK;
583 vm_page_undirty(pages[i]);
585 vm_page_unlock_queues();
588 #endif /* NWFS_RWCACHE */
591 * Flush and invalidate all dirty buffers. If another process is already
592 * doing the flush, just wait for completion.
595 nwfs_vinvalbuf(vp, td)
599 struct nwnode *np = VTONW(vp);
600 /* struct nwmount *nmp = VTONWFS(vp);*/
603 if (vp->v_iflag & VI_XLOCK)
606 while (np->n_flag & NFLUSHINPROG) {
607 np->n_flag |= NFLUSHWANT;
608 error = tsleep(&np->n_flag, PRIBIO + 2, "nwfsvinv", 2 * hz);
609 error = ncp_chkintr(NWFSTOCONN(VTONWFS(vp)), td);
613 np->n_flag |= NFLUSHINPROG;
614 error = vinvalbuf(vp, V_SAVE, td, PCATCH, 0);
616 if (error == ERESTART || error == EINTR) {
617 np->n_flag &= ~NFLUSHINPROG;
618 if (np->n_flag & NFLUSHWANT) {
619 np->n_flag &= ~NFLUSHWANT;
624 error = vinvalbuf(vp, V_SAVE, td, PCATCH, 0);
626 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
627 if (np->n_flag & NFLUSHWANT) {
628 np->n_flag &= ~NFLUSHWANT;