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 1.15 2005/04/15 19:08:24 dillon 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;
87 NCPVNDEBUG("dirname='%s'\n",np->n_name);
88 if (uio->uio_resid < DE_SIZE || (uio->uio_offset < 0))
92 i = uio->uio_offset / DE_SIZE; /* offset in directory */
94 error = ncp_initsearch(vp, uio->uio_td, cred);
96 NCPVNDEBUG("cannot initialize search, error=%d",error);
101 for (; uio->uio_resid >= DE_SIZE; i++) {
102 bzero((char *) &dp, DE_SIZE);
103 dp.d_reclen = DE_SIZE;
107 dp.d_fileno = (i == 0) ? np->n_fid.f_id : np->n_parent.f_id;
108 if (!dp.d_fileno) dp.d_fileno = NWFS_ROOT_INO;
112 dp.d_name[i + 1] = '\0';
116 error = ncp_search_for_file_or_subdir(nmp, &np->n_seq, &fattr, uio->uio_td, cred);
117 if (error && error < 0x80) break;
118 dp.d_fileno = fattr.dirEntNum;
119 dp.d_type = (fattr.attributes & aDIR) ? DT_DIR : DT_REG;
120 dp.d_namlen = fattr.nameLen;
121 bcopy(fattr.entryName, dp.d_name, dp.d_namlen);
122 dp.d_name[dp.d_namlen] = '\0';
124 if (error && eofflag) {
131 if (nwfs_fastlookup && !error && i > 1) {
132 fid.f_id = fattr.dirEntNum;
133 fid.f_parent = np->n_fid.f_id;
134 error = nwfs_nget(vp->v_mount, fid, &fattr, vp, &newvp);
136 VTONW(newvp)->n_ctime = VTONW(newvp)->n_vattr.va_ctime.tv_sec;
145 if ((error = uiomove((caddr_t)&dp, DE_SIZE, uio)))
149 uio->uio_offset = i * DE_SIZE;
154 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, td);
179 if (error) return (error);
180 np->n_mtime = vattr.va_mtime.tv_sec;
182 error = VOP_GETATTR(vp, &vattr, td);
183 if (error) return (error);
184 if (np->n_mtime != vattr.va_mtime.tv_sec) {
185 error = nwfs_vinvalbuf(vp, V_SAVE, td, 1);
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(struct vnode *vp, struct uio *uiop, struct ucred *cred,
198 struct nwmount *nmp = VTONWFS(vp);
199 struct nwnode *np = VTONW(vp);
201 /* struct vattr vattr;*/
204 if (vp->v_type != VREG) {
205 printf("%s: vn types other than VREG unsupported !\n",__func__);
208 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
209 if (uiop->uio_offset < 0) return EINVAL;
210 /* if (uiop->uio_offset + uiop->uio_resid > nmp->nm_maxfilesize)
213 if (ioflag & (IO_APPEND | IO_SYNC)) {
214 if (np->n_flag & NMODIFIED) {
215 nwfs_attr_cacheremove(vp);
216 error = nwfs_vinvalbuf(vp, V_SAVE, td, 1);
217 if (error) return (error);
219 if (ioflag & IO_APPEND) {
220 /* We can relay only on local information about file size,
221 * because until file is closed NetWare will not return
222 * the correct size. */
224 nwfs_attr_cacheremove(vp);
225 error = VOP_GETATTR(vp, &vattr, td);
226 if (error) return (error);
228 uiop->uio_offset = np->n_size;
231 if (uiop->uio_resid == 0) return 0;
232 if (td->td_proc && uiop->uio_offset + uiop->uio_resid >
233 td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
234 psignal(td->td_proc, SIGXFSZ);
237 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
238 NCPVNDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
240 if (uiop->uio_offset > np->n_size) {
241 np->n_vattr.va_size = np->n_size = uiop->uio_offset;
242 vnode_pager_setsize(vp, np->n_size);
249 * Do an I/O operation to/from a cache block.
252 nwfs_doio(struct buf *bp, struct ucred *cr, struct thread *td)
264 nmp = VFSTONWFS(vp->v_mount);
267 uiop->uio_iovcnt = 1;
268 uiop->uio_segflg = UIO_SYSSPACE;
270 if (bp->b_flags & B_READ) {
271 io.iov_len = uiop->uio_resid = bp->b_bcount;
272 io.iov_base = bp->b_data;
273 uiop->uio_rw = UIO_READ;
274 switch (vp->v_type) {
276 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
277 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
280 if (uiop->uio_resid) {
281 int left = uiop->uio_resid;
282 int nread = bp->b_bcount - left;
284 bzero((char *)bp->b_data + nread, left);
288 nfsstats.readdir_bios++;
289 uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * NFS_DIRBLKSIZ;
290 if (nmp->nm_flag & NFSMNT_RDIRPLUS) {
291 error = nfs_readdirplusrpc(vp, uiop, cr);
292 if (error == NFSERR_NOTSUPP)
293 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
295 if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0)
296 error = nfs_readdirrpc(vp, uiop, cr);
297 if (error == 0 && uiop->uio_resid == bp->b_bcount)
298 bp->b_flags |= B_INVAL;
302 printf("nwfs_doio: type %x unexpected\n",vp->v_type);
306 bp->b_flags |= B_ERROR;
310 if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size)
311 bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE);
313 if (bp->b_dirtyend > bp->b_dirtyoff) {
314 io.iov_len = uiop->uio_resid = bp->b_dirtyend - bp->b_dirtyoff;
315 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
316 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
317 uiop->uio_rw = UIO_WRITE;
318 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
321 * For an interrupted write, the buffer is still valid
322 * and the write hasn't been pushed to the server yet,
323 * so we can't set B_ERROR and report the interruption
324 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
325 * is not relevant, so the rpc attempt is essentially
326 * a noop. For the case of a V3 write rpc not being
327 * committed to stable storage, the block is still
328 * dirty and requires either a commit rpc or another
329 * write rpc with iomode == NFSV3WRITE_FILESYNC before
330 * the block is reused. This is indicated by setting
331 * the B_DELWRI and B_NEEDCOMMIT flags.
334 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
338 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
339 if ((bp->b_flags & B_ASYNC) == 0)
340 bp->b_flags |= B_EINTR;
341 if ((bp->b_flags & B_PAGING) == 0) {
343 bp->b_flags &= ~B_DONE;
345 if ((bp->b_flags & B_ASYNC) == 0)
346 bp->b_flags |= B_EINTR;
350 bp->b_flags |= B_ERROR;
351 bp->b_error /*= np->n_error */= error;
352 /* np->n_flag |= NWRITEERR;*/
354 bp->b_dirtyoff = bp->b_dirtyend = 0;
362 bp->b_resid = uiop->uio_resid;
368 * Vnode op for VM getpages.
369 * Wish wish .... get rid from multiple IO routines
371 * nwfs_getpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
372 * int a_reqpage, vm_ooffset_t a_offset)
375 nwfs_getpages(struct vop_getpages_args *ap)
378 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
381 int i, error, nextoff, size, toff, npages, count;
387 struct thread *td = curthread; /* XXX */
393 KKASSERT(td->td_proc);
394 cred = td->td_proc->p_ucred;
398 nmp = VFSTONWFS(vp->v_mount);
402 if (vp->v_object == NULL) {
403 printf("nwfs_getpages: called with non-merged cache vnode??\n");
404 return VM_PAGER_ERROR;
407 bp = getpbuf(&nwfs_pbuf_freecnt);
408 npages = btoc(count);
409 kva = (vm_offset_t) bp->b_data;
410 pmap_qenter(kva, pages, npages);
412 iov.iov_base = (caddr_t) kva;
416 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
417 uio.uio_resid = count;
418 uio.uio_segflg = UIO_SYSSPACE;
419 uio.uio_rw = UIO_READ;
422 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, &uio,cred);
423 pmap_qremove(kva, npages);
425 relpbuf(bp, &nwfs_pbuf_freecnt);
427 if (error && (uio.uio_resid == count)) {
428 printf("nwfs_getpages: error %d\n",error);
429 for (i = 0; i < npages; i++) {
430 if (ap->a_reqpage != i)
431 vnode_pager_freepage(pages[i]);
433 return VM_PAGER_ERROR;
436 size = count - uio.uio_resid;
438 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
440 nextoff = toff + PAGE_SIZE;
443 m->flags &= ~PG_ZERO;
445 if (nextoff <= size) {
446 m->valid = VM_PAGE_BITS_ALL;
449 int nvalid = ((size + DEV_BSIZE - 1) - toff) & ~(DEV_BSIZE - 1);
450 vm_page_set_validclean(m, 0, nvalid);
453 if (i != ap->a_reqpage) {
455 * Whether or not to leave the page activated is up in
456 * the air, but we should put the page on a page queue
457 * somewhere (it already is in the object). Result:
458 * It appears that emperical results show that
459 * deactivating pages is best.
463 * Just in case someone was asking for this page we
464 * now tell them that it is ok to use.
467 if (m->flags & PG_WANTED)
470 vm_page_deactivate(m);
473 vnode_pager_freepage(m);
478 #endif /* NWFS_RWCACHE */
482 * Vnode op for VM putpages.
483 * possible bug: all IO done in sync mode
484 * Note that vop_close always invalidate pages before close, so it's
485 * not necessary to open vnode.
487 * nwfs_putpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
488 * int a_sync, int *a_rtvals, vm_ooffset_t a_offset)
491 nwfs_putpages(struct vop_putpages_args *ap)
494 struct thread *td = curthread; /* XXX */
495 struct vnode *vp = ap->a_vp;
499 KKASSERT(td->td_proc);
500 cred = td->td_proc->p_ucred; /* XXX */
501 VOP_OPEN(vp, FWRITE, cred, NULL, td);
502 error = vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
503 ap->a_sync, ap->a_rtvals);
504 VOP_CLOSE(vp, FWRITE, cred, td);
511 int i, npages, count;
517 KKASSERT(td->td_proc);
518 cred = td->td_proc->p_ucred; /* XXX */
520 /* VOP_OPEN(vp, FWRITE, cred, td);*/
522 nmp = VFSTONWFS(vp->v_mount);
525 rtvals = ap->a_rtvals;
526 npages = btoc(count);
528 for (i = 0; i < npages; i++) {
529 rtvals[i] = VM_PAGER_AGAIN;
532 bp = getpbuf(&nwfs_pbuf_freecnt);
533 kva = (vm_offset_t) bp->b_data;
534 pmap_qenter(kva, pages, npages);
536 iov.iov_base = (caddr_t) kva;
540 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
541 uio.uio_resid = count;
542 uio.uio_segflg = UIO_SYSSPACE;
543 uio.uio_rw = UIO_WRITE;
545 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
547 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, &uio, cred);
548 /* VOP_CLOSE(vp, FWRITE, cred, td);*/
549 NCPVNDEBUG("paged write done: %d\n", error);
551 pmap_qremove(kva, npages);
552 relpbuf(bp, &nwfs_pbuf_freecnt);
555 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
556 for (i = 0; i < nwritten; i++) {
557 rtvals[i] = VM_PAGER_OK;
562 #endif /* NWFS_RWCACHE */
565 * Flush and invalidate all dirty buffers. If another process is already
566 * doing the flush, just wait for completion.
569 nwfs_vinvalbuf(struct vnode *vp, int flags, struct thread *td, int intrflg)
571 struct nwnode *np = VTONW(vp);
572 /* struct nwmount *nmp = VTONWFS(vp);*/
573 int error = 0, slpflag, slptimeo;
575 if (vp->v_flag & VRECLAIMED) {
585 while (np->n_flag & NFLUSHINPROG) {
586 np->n_flag |= NFLUSHWANT;
587 error = tsleep((caddr_t)&np->n_flag, 0, "nwfsvinv", slptimeo);
588 error = ncp_chkintr(NWFSTOCONN(VTONWFS(vp)), td);
589 if (error == EINTR && intrflg)
592 np->n_flag |= NFLUSHINPROG;
593 error = vinvalbuf(vp, flags, td, slpflag, 0);
595 if (intrflg && (error == ERESTART || error == EINTR)) {
596 np->n_flag &= ~NFLUSHINPROG;
597 if (np->n_flag & NFLUSHWANT) {
598 np->n_flag &= ~NFLUSHWANT;
599 wakeup((caddr_t)&np->n_flag);
603 error = vinvalbuf(vp, flags, td, slpflag, 0);
605 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
606 if (np->n_flag & NFLUSHWANT) {
607 np->n_flag &= ~NFLUSHWANT;
608 wakeup((caddr_t)&np->n_flag);