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 $
33 * $DragonFly: src/sys/vfs/nwfs/nwfs_io.c,v 1.10 2004/04/22 17:56:44 cpressey Exp $
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/resourcevar.h> /* defines plimit structure in proc struct */
39 #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 <netproto/ncp/ncp.h>
57 #include <netproto/ncp/ncp_conn.h>
58 #include <netproto/ncp/ncp_subr.h>
61 #include "nwfs_node.h"
62 #include "nwfs_subr.h"
64 static int nwfs_fastlookup = 1;
66 SYSCTL_DECL(_vfs_nwfs);
67 SYSCTL_INT(_vfs_nwfs, OID_AUTO, fastlookup, CTLFLAG_RW, &nwfs_fastlookup, 0, "");
70 extern int nwfs_pbuf_freecnt;
72 #define DE_SIZE (sizeof(struct dirent))
76 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, NCPNULL, 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)
160 struct nwmount *nmp = VFSTONWFS(vp->v_mount);
161 struct nwnode *np = VTONW(vp);
166 if (vp->v_type != VREG && vp->v_type != VDIR) {
167 printf("%s: vn types other than VREG or VDIR are unsupported !\n",__FUNCTION__);
170 if (uiop->uio_resid == 0) return 0;
171 if (uiop->uio_offset < 0) return EINVAL;
172 /* if (uiop->uio_offset + uiop->uio_resid > nmp->nm_maxfilesize)
175 if (vp->v_type == VDIR) {
176 error = nwfs_readvdir(vp, uiop, cred);
179 biosize = NWFSTOCONN(nmp)->buffer_size;
180 if (np->n_flag & NMODIFIED) {
181 nwfs_attr_cacheremove(vp);
182 error = VOP_GETATTR(vp, &vattr, td);
183 if (error) return (error);
184 np->n_mtime = vattr.va_mtime.tv_sec;
186 error = VOP_GETATTR(vp, &vattr, td);
187 if (error) return (error);
188 if (np->n_mtime != vattr.va_mtime.tv_sec) {
189 error = nwfs_vinvalbuf(vp, V_SAVE, td, 1);
190 if (error) return (error);
191 np->n_mtime = vattr.va_mtime.tv_sec;
194 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop,cred);
199 nwfs_writevnode(struct vnode *vp, struct uio *uiop, struct ucred *cred,
202 struct nwmount *nmp = VTONWFS(vp);
203 struct nwnode *np = VTONW(vp);
205 /* struct vattr vattr;*/
208 if (vp->v_type != VREG) {
209 printf("%s: vn types other than VREG unsupported !\n",__FUNCTION__);
212 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
213 if (uiop->uio_offset < 0) return EINVAL;
214 /* if (uiop->uio_offset + uiop->uio_resid > nmp->nm_maxfilesize)
217 if (ioflag & (IO_APPEND | IO_SYNC)) {
218 if (np->n_flag & NMODIFIED) {
219 nwfs_attr_cacheremove(vp);
220 error = nwfs_vinvalbuf(vp, V_SAVE, td, 1);
221 if (error) return (error);
223 if (ioflag & IO_APPEND) {
224 /* We can relay only on local information about file size,
225 * because until file is closed NetWare will not return
226 * the correct size. */
228 nwfs_attr_cacheremove(vp);
229 error = VOP_GETATTR(vp, &vattr, td);
230 if (error) return (error);
232 uiop->uio_offset = np->n_size;
235 if (uiop->uio_resid == 0) return 0;
236 if (td->td_proc && uiop->uio_offset + uiop->uio_resid >
237 td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
238 psignal(td->td_proc, SIGXFSZ);
241 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
242 NCPVNDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
244 if (uiop->uio_offset > np->n_size) {
245 np->n_vattr.va_size = np->n_size = uiop->uio_offset;
246 vnode_pager_setsize(vp, np->n_size);
253 * Do an I/O operation to/from a cache block.
256 nwfs_doio(struct buf *bp, struct ucred *cr, struct thread *td)
268 nmp = VFSTONWFS(vp->v_mount);
271 uiop->uio_iovcnt = 1;
272 uiop->uio_segflg = UIO_SYSSPACE;
274 if (bp->b_flags & B_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_flags |= B_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 bp->b_flags |= B_WRITEINPROG;
323 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
324 bp->b_flags &= ~B_WRITEINPROG;
327 * For an interrupted write, the buffer is still valid
328 * and the write hasn't been pushed to the server yet,
329 * so we can't set B_ERROR and report the interruption
330 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
331 * is not relevant, so the rpc attempt is essentially
332 * a noop. For the case of a V3 write rpc not being
333 * committed to stable storage, the block is still
334 * dirty and requires either a commit rpc or another
335 * write rpc with iomode == NFSV3WRITE_FILESYNC before
336 * the block is reused. This is indicated by setting
337 * the B_DELWRI and B_NEEDCOMMIT flags.
340 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
344 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
345 if ((bp->b_flags & B_ASYNC) == 0)
346 bp->b_flags |= B_EINTR;
347 if ((bp->b_flags & B_PAGING) == 0) {
349 bp->b_flags &= ~B_DONE;
351 if ((bp->b_flags & B_ASYNC) == 0)
352 bp->b_flags |= B_EINTR;
356 bp->b_flags |= B_ERROR;
357 bp->b_error /*= np->n_error */= error;
358 /* np->n_flag |= NWRITEERR;*/
360 bp->b_dirtyoff = bp->b_dirtyend = 0;
368 bp->b_resid = uiop->uio_resid;
374 * Vnode op for VM getpages.
375 * Wish wish .... get rid from multiple IO routines
377 * nwfs_getpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
378 * int a_reqpage, vm_ooffset_t a_offset)
381 nwfs_getpages(struct vop_getpages_args *ap)
384 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
387 int i, error, nextoff, size, toff, npages, count;
393 struct thread *td = curthread; /* XXX */
399 KKASSERT(td->td_proc);
400 cred = td->td_proc->p_ucred;
404 nmp = VFSTONWFS(vp->v_mount);
408 if (vp->v_object == NULL) {
409 printf("nwfs_getpages: called with non-merged cache vnode??\n");
410 return VM_PAGER_ERROR;
413 bp = getpbuf(&nwfs_pbuf_freecnt);
414 npages = btoc(count);
415 kva = (vm_offset_t) bp->b_data;
416 pmap_qenter(kva, pages, npages);
418 iov.iov_base = (caddr_t) kva;
422 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
423 uio.uio_resid = count;
424 uio.uio_segflg = UIO_SYSSPACE;
425 uio.uio_rw = UIO_READ;
428 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, &uio,cred);
429 pmap_qremove(kva, npages);
431 relpbuf(bp, &nwfs_pbuf_freecnt);
433 if (error && (uio.uio_resid == count)) {
434 printf("nwfs_getpages: error %d\n",error);
435 for (i = 0; i < npages; i++) {
436 if (ap->a_reqpage != i)
437 vnode_pager_freepage(pages[i]);
439 return VM_PAGER_ERROR;
442 size = count - uio.uio_resid;
444 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
446 nextoff = toff + PAGE_SIZE;
449 m->flags &= ~PG_ZERO;
451 if (nextoff <= size) {
452 m->valid = VM_PAGE_BITS_ALL;
455 int nvalid = ((size + DEV_BSIZE - 1) - toff) & ~(DEV_BSIZE - 1);
456 vm_page_set_validclean(m, 0, nvalid);
459 if (i != ap->a_reqpage) {
461 * Whether or not to leave the page activated is up in
462 * the air, but we should put the page on a page queue
463 * somewhere (it already is in the object). Result:
464 * It appears that emperical results show that
465 * deactivating pages is best.
469 * Just in case someone was asking for this page we
470 * now tell them that it is ok to use.
473 if (m->flags & PG_WANTED)
476 vm_page_deactivate(m);
479 vnode_pager_freepage(m);
484 #endif /* NWFS_RWCACHE */
488 * Vnode op for VM putpages.
489 * possible bug: all IO done in sync mode
490 * Note that vop_close always invalidate pages before close, so it's
491 * not necessary to open vnode.
493 * nwfs_putpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
494 * int a_sync, int *a_rtvals, vm_ooffset_t a_offset)
497 nwfs_putpages(struct vop_putpages_args *ap)
500 struct thread *td = curthread; /* XXX */
501 struct vnode *vp = ap->a_vp;
505 KKASSERT(td->td_proc);
506 cred = td->td_proc->p_ucred; /* XXX */
507 VOP_OPEN(vp, FWRITE, cred, td);
508 error = vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
509 ap->a_sync, ap->a_rtvals);
510 VOP_CLOSE(vp, FWRITE, cred, td);
517 int i, npages, count;
523 KKASSERT(td->td_proc);
524 cred = td->td_proc->p_ucred; /* XXX */
526 /* VOP_OPEN(vp, FWRITE, cred, td);*/
528 nmp = VFSTONWFS(vp->v_mount);
531 rtvals = ap->a_rtvals;
532 npages = btoc(count);
534 for (i = 0; i < npages; i++) {
535 rtvals[i] = VM_PAGER_AGAIN;
538 bp = getpbuf(&nwfs_pbuf_freecnt);
539 kva = (vm_offset_t) bp->b_data;
540 pmap_qenter(kva, pages, npages);
542 iov.iov_base = (caddr_t) kva;
546 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
547 uio.uio_resid = count;
548 uio.uio_segflg = UIO_SYSSPACE;
549 uio.uio_rw = UIO_WRITE;
551 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
553 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, &uio, cred);
554 /* VOP_CLOSE(vp, FWRITE, cred, td);*/
555 NCPVNDEBUG("paged write done: %d\n", error);
557 pmap_qremove(kva, npages);
558 relpbuf(bp, &nwfs_pbuf_freecnt);
561 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
562 for (i = 0; i < nwritten; i++) {
563 rtvals[i] = VM_PAGER_OK;
568 #endif /* NWFS_RWCACHE */
571 * Flush and invalidate all dirty buffers. If another process is already
572 * doing the flush, just wait for completion.
575 nwfs_vinvalbuf(struct vnode *vp, int flags, struct thread *td, int intrflg)
577 struct nwnode *np = VTONW(vp);
578 /* struct nwmount *nmp = VTONWFS(vp);*/
579 int error = 0, slpflag, slptimeo;
581 if (vp->v_flag & VXLOCK) {
591 while (np->n_flag & NFLUSHINPROG) {
592 np->n_flag |= NFLUSHWANT;
593 error = tsleep((caddr_t)&np->n_flag, 0, "nwfsvinv", slptimeo);
594 error = ncp_chkintr(NWFSTOCONN(VTONWFS(vp)), td);
595 if (error == EINTR && intrflg)
598 np->n_flag |= NFLUSHINPROG;
599 error = vinvalbuf(vp, flags, td, slpflag, 0);
601 if (intrflg && (error == ERESTART || error == EINTR)) {
602 np->n_flag &= ~NFLUSHINPROG;
603 if (np->n_flag & NFLUSHWANT) {
604 np->n_flag &= ~NFLUSHWANT;
605 wakeup((caddr_t)&np->n_flag);
609 error = vinvalbuf(vp, flags, td, slpflag, 0);
611 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
612 if (np->n_flag & NFLUSHWANT) {
613 np->n_flag &= ~NFLUSHWANT;
614 wakeup((caddr_t)&np->n_flag);