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
32 * $FreeBSD: src/sys/nwfs/nwfs_io.c,v 1.6.2.1 2000/10/25 02:11:10 bp Exp $
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>
41 #include <sys/mount.h>
42 #include <sys/namei.h>
43 #include <sys/vnode.h>
44 #include <sys/dirent.h>
45 #include <sys/signalvar.h>
46 #include <sys/sysctl.h>
49 #include <vm/vm_page.h>
50 #include <vm/vm_extern.h>
51 #include <vm/vm_object.h>
52 #include <vm/vm_pager.h>
53 #include <vm/vnode_pager.h>
55 #include <netncp/ncp.h>
56 #include <netncp/ncp_conn.h>
57 #include <netncp/ncp_subr.h>
59 #include <nwfs/nwfs.h>
60 #include <nwfs/nwfs_node.h>
61 #include <nwfs/nwfs_subr.h>
63 static int nwfs_fastlookup = 1;
65 extern struct linker_set sysctl_vfs_nwfs;
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_procp, 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_procp, 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((caddr_t)&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",__FUNCTION__);
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, p);
182 if (error) return (error);
183 np->n_mtime = vattr.va_mtime.tv_sec;
185 error = VOP_GETATTR(vp, &vattr, cred, p);
186 if (error) return (error);
187 if (np->n_mtime != vattr.va_mtime.tv_sec) {
188 error = nwfs_vinvalbuf(vp, V_SAVE, cred, p, 1);
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",__FUNCTION__);
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, V_SAVE, cred, p, 1);
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, p);
232 if (error) return (error);
234 uiop->uio_offset = np->n_size;
237 if (uiop->uio_resid == 0) return 0;
238 if (p && uiop->uio_offset + uiop->uio_resid > p->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
242 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
243 NCPVNDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
245 if (uiop->uio_offset > np->n_size) {
246 np->n_vattr.va_size = np->n_size = uiop->uio_offset;
247 vnode_pager_setsize(vp, np->n_size);
254 * Do an I/O operation to/from a cache block.
272 nmp = VFSTONWFS(vp->v_mount);
275 uiop->uio_iovcnt = 1;
276 uiop->uio_segflg = UIO_SYSSPACE;
278 if (bp->b_flags & B_READ) {
279 io.iov_len = uiop->uio_resid = bp->b_bcount;
280 io.iov_base = bp->b_data;
281 uiop->uio_rw = UIO_READ;
282 switch (vp->v_type) {
284 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
285 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
288 if (uiop->uio_resid) {
289 int left = uiop->uio_resid;
290 int nread = bp->b_bcount - left;
292 bzero((char *)bp->b_data + nread, left);
296 nfsstats.readdir_bios++;
297 uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * NFS_DIRBLKSIZ;
298 if (nmp->nm_flag & NFSMNT_RDIRPLUS) {
299 error = nfs_readdirplusrpc(vp, uiop, cr);
300 if (error == NFSERR_NOTSUPP)
301 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
303 if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0)
304 error = nfs_readdirrpc(vp, uiop, cr);
305 if (error == 0 && uiop->uio_resid == bp->b_bcount)
306 bp->b_flags |= B_INVAL;
310 printf("nwfs_doio: type %x unexpected\n",vp->v_type);
314 bp->b_flags |= B_ERROR;
318 if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size)
319 bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE);
321 if (bp->b_dirtyend > bp->b_dirtyoff) {
322 io.iov_len = uiop->uio_resid = bp->b_dirtyend - bp->b_dirtyoff;
323 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
324 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
325 uiop->uio_rw = UIO_WRITE;
326 bp->b_flags |= B_WRITEINPROG;
327 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
328 bp->b_flags &= ~B_WRITEINPROG;
331 * For an interrupted write, the buffer is still valid
332 * and the write hasn't been pushed to the server yet,
333 * so we can't set B_ERROR and report the interruption
334 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
335 * is not relevant, so the rpc attempt is essentially
336 * a noop. For the case of a V3 write rpc not being
337 * committed to stable storage, the block is still
338 * dirty and requires either a commit rpc or another
339 * write rpc with iomode == NFSV3WRITE_FILESYNC before
340 * the block is reused. This is indicated by setting
341 * the B_DELWRI and B_NEEDCOMMIT flags.
344 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
348 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
349 if ((bp->b_flags & B_ASYNC) == 0)
350 bp->b_flags |= B_EINTR;
351 if ((bp->b_flags & B_PAGING) == 0) {
353 bp->b_flags &= ~B_DONE;
355 if ((bp->b_flags & B_ASYNC) == 0)
356 bp->b_flags |= B_EINTR;
360 bp->b_flags |= B_ERROR;
361 bp->b_error /*= np->n_error */= error;
362 /* np->n_flag |= NWRITEERR;*/
364 bp->b_dirtyoff = bp->b_dirtyend = 0;
372 bp->b_resid = uiop->uio_resid;
378 * Vnode op for VM getpages.
379 * Wish wish .... get rid from multiple IO routines
383 struct vop_getpages_args /* {
388 vm_ooffset_t a_offset;
392 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
395 int i, error, nextoff, size, toff, npages, count;
408 p = curproc; /* XXX */
409 cred = curproc->p_ucred; /* XXX */
411 nmp = VFSTONWFS(vp->v_mount);
415 if (vp->v_object == NULL) {
416 printf("nwfs_getpages: called with non-merged cache vnode??\n");
417 return VM_PAGER_ERROR;
420 bp = getpbuf(&nwfs_pbuf_freecnt);
421 npages = btoc(count);
422 kva = (vm_offset_t) bp->b_data;
423 pmap_qenter(kva, pages, npages);
425 iov.iov_base = (caddr_t) kva;
429 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
430 uio.uio_resid = count;
431 uio.uio_segflg = UIO_SYSSPACE;
432 uio.uio_rw = UIO_READ;
435 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, &uio,cred);
436 pmap_qremove(kva, npages);
438 relpbuf(bp, &nwfs_pbuf_freecnt);
440 if (error && (uio.uio_resid == count)) {
441 printf("nwfs_getpages: error %d\n",error);
442 for (i = 0; i < npages; i++) {
443 if (ap->a_reqpage != i)
444 vnode_pager_freepage(pages[i]);
446 return VM_PAGER_ERROR;
449 size = count - uio.uio_resid;
451 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
453 nextoff = toff + PAGE_SIZE;
456 m->flags &= ~PG_ZERO;
458 if (nextoff <= size) {
459 m->valid = VM_PAGE_BITS_ALL;
462 int nvalid = ((size + DEV_BSIZE - 1) - toff) & ~(DEV_BSIZE - 1);
463 vm_page_set_validclean(m, 0, nvalid);
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->flags & PG_WANTED)
483 vm_page_deactivate(m);
486 vnode_pager_freepage(m);
491 #endif /* NWFS_RWCACHE */
495 * Vnode op for VM putpages.
496 * possible bug: all IO done in sync mode
497 * Note that vop_close always invalidate pages before close, so it's
498 * not necessary to open vnode.
502 struct vop_putpages_args /* {
508 vm_ooffset_t a_offset;
512 struct vnode *vp = ap->a_vp;
517 p = curproc; /* XXX */
518 cred = p->p_ucred; /* XXX */
519 VOP_OPEN(vp, FWRITE, cred, p);
520 error = vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
521 ap->a_sync, ap->a_rtvals);
522 VOP_CLOSE(vp, FWRITE, cred, p);
529 int i, npages, count;
535 p = curproc; /* XXX */
536 cred = p->p_ucred; /* XXX */
537 /* VOP_OPEN(vp, FWRITE, cred, p);*/
539 nmp = VFSTONWFS(vp->v_mount);
542 rtvals = ap->a_rtvals;
543 npages = btoc(count);
545 for (i = 0; i < npages; i++) {
546 rtvals[i] = VM_PAGER_AGAIN;
549 bp = getpbuf(&nwfs_pbuf_freecnt);
550 kva = (vm_offset_t) bp->b_data;
551 pmap_qenter(kva, pages, npages);
553 iov.iov_base = (caddr_t) kva;
557 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
558 uio.uio_resid = count;
559 uio.uio_segflg = UIO_SYSSPACE;
560 uio.uio_rw = UIO_WRITE;
562 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
564 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, &uio, cred);
565 /* VOP_CLOSE(vp, FWRITE, cred, p);*/
566 NCPVNDEBUG("paged write done: %d\n", error);
568 pmap_qremove(kva, npages);
569 relpbuf(bp, &nwfs_pbuf_freecnt);
572 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
573 for (i = 0; i < nwritten; i++) {
574 rtvals[i] = VM_PAGER_OK;
579 #endif /* NWFS_RWCACHE */
582 * Flush and invalidate all dirty buffers. If another process is already
583 * doing the flush, just wait for completion.
586 nwfs_vinvalbuf(vp, flags, cred, p, intrflg)
593 struct nwnode *np = VTONW(vp);
594 /* struct nwmount *nmp = VTONWFS(vp);*/
595 int error = 0, slpflag, slptimeo;
597 if (vp->v_flag & VXLOCK) {
607 while (np->n_flag & NFLUSHINPROG) {
608 np->n_flag |= NFLUSHWANT;
609 error = tsleep((caddr_t)&np->n_flag, PRIBIO + 2, "nwfsvinv", slptimeo);
610 error = ncp_chkintr(NWFSTOCONN(VTONWFS(vp)), p);
611 if (error == EINTR && intrflg)
614 np->n_flag |= NFLUSHINPROG;
615 error = vinvalbuf(vp, flags, cred, p, slpflag, 0);
617 if (intrflg && (error == ERESTART || error == EINTR)) {
618 np->n_flag &= ~NFLUSHINPROG;
619 if (np->n_flag & NFLUSHWANT) {
620 np->n_flag &= ~NFLUSHWANT;
621 wakeup((caddr_t)&np->n_flag);
625 error = vinvalbuf(vp, flags, cred, p, slpflag, 0);
627 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
628 if (np->n_flag & NFLUSHWANT) {
629 np->n_flag &= ~NFLUSHWANT;
630 wakeup((caddr_t)&np->n_flag);