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_page2.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 <netproto/ncp/ncp.h>
56 #include <netproto/ncp/ncp_conn.h>
57 #include <netproto/ncp/ncp_subr.h>
59 #include <sys/thread2.h>
61 #include <machine/limits.h>
64 #include "nwfs_node.h"
65 #include "nwfs_subr.h"
67 static int nwfs_fastlookup = 1;
69 SYSCTL_DECL(_vfs_nwfs);
70 SYSCTL_INT(_vfs_nwfs, OID_AUTO, fastlookup, CTLFLAG_RW, &nwfs_fastlookup, 0, "");
73 extern int nwfs_pbuf_freecnt;
78 nwfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred)
80 struct nwmount *nmp = VTONWFS(vp);
83 struct nw_entry_info fattr;
92 NCPVNDEBUG("dirname='%s'\n",np->n_name);
93 if (uio->uio_offset < 0 || uio->uio_offset > INT_MAX)
96 i = (int)uio->uio_offset; /* offset in directory */
98 error = ncp_initsearch(vp, uio->uio_td, cred);
100 NCPVNDEBUG("cannot initialize search, error=%d",error);
105 for (; !error && uio->uio_resid > 0; i++) {
108 d_ino = np->n_fid.f_id;
110 d_ino = NWFS_ROOT_INO;
116 d_ino = np->n_parent.f_id;
118 d_ino = NWFS_ROOT_INO;
124 error = ncp_search_for_file_or_subdir(nmp, &np->n_seq, &fattr, uio->uio_td, cred);
125 if (error && error < 0x80)
127 d_ino = fattr.dirEntNum;
128 d_type = (fattr.attributes & aDIR) ? DT_DIR : DT_REG;
129 d_namlen = fattr.nameLen;
130 d_name = fattr.entryName;
132 if (error && eofflag) {
139 if (nwfs_fastlookup && !error && i > 1) {
140 fid.f_id = fattr.dirEntNum;
141 fid.f_parent = np->n_fid.f_id;
142 error = nwfs_nget(vp->v_mount, fid, &fattr, vp, &newvp);
144 VTONW(newvp)->n_ctime = VTONW(newvp)->n_vattr.va_ctime.tv_sec;
153 if (vop_write_dirent(&error, uio, d_ino, d_type, d_namlen, d_name))
162 nwfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred)
164 struct nwmount *nmp = VFSTONWFS(vp->v_mount);
165 struct nwnode *np = VTONW(vp);
169 if (vp->v_type != VREG && vp->v_type != VDIR) {
170 kprintf("%s: vn types other than VREG or VDIR are unsupported !\n",__func__);
173 if (uiop->uio_resid == 0) return 0;
174 if (uiop->uio_offset < 0) return EINVAL;
175 if (vp->v_type == VDIR) {
176 error = nwfs_readvdir(vp, uiop, cred);
179 if (np->n_flag & NMODIFIED) {
180 nwfs_attr_cacheremove(vp);
181 error = VOP_GETATTR(vp, &vattr);
182 if (error) return (error);
183 np->n_mtime = vattr.va_mtime.tv_sec;
185 error = VOP_GETATTR(vp, &vattr);
186 if (error) return (error);
187 if (np->n_mtime != vattr.va_mtime.tv_sec) {
188 error = nwfs_vinvalbuf(vp, V_SAVE, 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(struct vnode *vp, struct uio *uiop, struct ucred *cred,
201 struct nwmount *nmp = VTONWFS(vp);
202 struct nwnode *np = VTONW(vp);
204 /* struct vattr vattr;*/
207 if (vp->v_type != VREG) {
208 kprintf("%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;
214 if (ioflag & (IO_APPEND | IO_SYNC)) {
215 if (np->n_flag & NMODIFIED) {
216 nwfs_attr_cacheremove(vp);
217 error = nwfs_vinvalbuf(vp, V_SAVE, 1);
218 if (error) return (error);
220 if (ioflag & IO_APPEND) {
221 /* We can relay only on local information about file size,
222 * because until file is closed NetWare will not return
223 * the correct size. */
225 nwfs_attr_cacheremove(vp);
226 error = VOP_GETATTR(vp, &vattr);
227 if (error) return (error);
229 uiop->uio_offset = np->n_size;
232 if (uiop->uio_resid == 0) return 0;
233 if (td->td_proc && uiop->uio_offset + uiop->uio_resid >
234 td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
235 lwpsignal(td->td_proc, td->td_lwp, SIGXFSZ);
238 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cred);
239 NCPVNDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
241 if (uiop->uio_offset > np->n_size) {
242 np->n_vattr.va_size = np->n_size = uiop->uio_offset;
243 vnode_pager_setsize(vp, np->n_size);
250 * Do an I/O operation to/from a cache block.
253 nwfs_doio(struct vnode *vp, struct bio *bio, struct ucred *cr, struct thread *td)
255 struct buf *bp = bio->bio_buf;
264 nmp = VFSTONWFS(vp->v_mount);
267 uiop->uio_iovcnt = 1;
268 uiop->uio_segflg = UIO_SYSSPACE;
271 if (bp->b_cmd == BUF_CMD_READ) {
272 io.iov_len = uiop->uio_resid = (size_t)bp->b_bcount;
273 io.iov_base = bp->b_data;
274 uiop->uio_rw = UIO_READ;
275 switch (vp->v_type) {
277 uiop->uio_offset = bio->bio_offset;
278 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
281 if (uiop->uio_resid) {
282 size_t left = uiop->uio_resid;
283 size_t nread = bp->b_bcount - left;
285 bzero((char *)bp->b_data + nread, left);
289 nfsstats.readdir_bios++;
290 uiop->uio_offset = bio->bio_offset;
291 if (nmp->nm_flag & NFSMNT_RDIRPLUS) {
292 error = nfs_readdirplusrpc(vp, uiop, cr);
293 if (error == NFSERR_NOTSUPP)
294 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
296 if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0)
297 error = nfs_readdirrpc(vp, uiop, cr);
298 if (error == 0 && uiop->uio_resid == (size_t)bp->b_bcount)
299 bp->b_flags |= B_INVAL;
303 kprintf("nwfs_doio: type %x unexpected\n",vp->v_type);
307 bp->b_flags |= B_ERROR;
311 KKASSERT(bp->b_cmd == BUF_CMD_WRITE);
312 if (bio->bio_offset + bp->b_dirtyend > np->n_size)
313 bp->b_dirtyend = np->n_size - bio->bio_offset;
315 if (bp->b_dirtyend > bp->b_dirtyoff) {
316 io.iov_len = uiop->uio_resid =
317 (size_t)(bp->b_dirtyend - bp->b_dirtyoff);
318 uiop->uio_offset = bio->bio_offset + bp->b_dirtyoff;
319 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
320 uiop->uio_rw = UIO_WRITE;
321 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
324 * For an interrupted write, the buffer is still valid
325 * and the write hasn't been pushed to the server yet,
326 * so we can't set B_ERROR and report the interruption
327 * by setting B_EINTR. For the async case, B_EINTR
328 * is not relevant, so the rpc attempt is essentially
329 * a noop. For the case of a V3 write rpc not being
330 * committed to stable storage, the block is still
331 * dirty and requires either a commit rpc or another
332 * write rpc with iomode == NFSV3WRITE_FILESYNC before
333 * the block is reused. This is indicated by setting
334 * the B_DELWRI and B_NEEDCOMMIT flags.
337 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
340 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
341 if ((bp->b_flags & B_PAGING) == 0)
343 bp->b_flags |= B_EINTR;
347 bp->b_flags |= B_ERROR;
348 bp->b_error /*= np->n_error */= error;
349 /* np->n_flag |= NWRITEERR;*/
351 bp->b_dirtyoff = bp->b_dirtyend = 0;
359 bp->b_resid = (int)uiop->uio_resid;
365 * Vnode op for VM getpages.
366 * Wish wish .... get rid from multiple IO routines
368 * nwfs_getpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
369 * int a_reqpage, vm_ooffset_t a_offset)
372 nwfs_getpages(struct vop_getpages_args *ap)
375 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
376 ap->a_reqpage, ap->a_seqaccess);
378 int i, error, npages;
379 size_t nextoff, toff;
387 struct thread *td = curthread; /* XXX */
393 KKASSERT(td->td_proc);
394 cred = td->td_proc->p_ucred;
398 nmp = VFSTONWFS(vp->v_mount);
400 count = (size_t)ap->a_count;
402 if (vp->v_object == NULL) {
403 kprintf("nwfs_getpages: called with non-merged cache vnode??\n");
404 return VM_PAGER_ERROR;
407 bp = getpbuf_kva(&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 kprintf("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;
446 * NOTE: pmap dirty bit should have already been cleared.
447 * We do not clear it here.
449 if (nextoff <= size) {
450 m->valid = VM_PAGE_BITS_ALL;
453 int nvalid = ((size + DEV_BSIZE - 1) - toff) &
455 vm_page_set_validclean(m, 0, nvalid);
458 if (i != ap->a_reqpage) {
460 * Whether or not to leave the page activated is up in
461 * the air, but we should put the page on a page queue
462 * somewhere (it already is in the object). Result:
463 * It appears that emperical results show that
464 * deactivating pages is best.
468 * Just in case someone was asking for this page we
469 * now tell them that it is ok to use.
472 if (m->flags & PG_REFERENCED)
475 vm_page_deactivate(m);
478 vnode_pager_freepage(m);
483 #endif /* NWFS_RWCACHE */
487 * Vnode op for VM putpages.
488 * possible bug: all IO done in sync mode
489 * Note that vop_close always invalidate pages before close, so it's
490 * not necessary to open vnode.
492 * nwfs_putpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
493 * int a_sync, int *a_rtvals, vm_ooffset_t a_offset)
496 nwfs_putpages(struct vop_putpages_args *ap)
499 struct thread *td = curthread; /* XXX */
500 struct vnode *vp = ap->a_vp;
504 KKASSERT(td->td_proc);
505 cred = td->td_proc->p_ucred; /* XXX */
506 VOP_OPEN(vp, FWRITE, cred, NULL);
507 error = vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
508 ap->a_sync, ap->a_rtvals);
509 VOP_CLOSE(vp, FWRITE, cred);
516 int i, npages, count;
522 KKASSERT(td->td_proc);
523 cred = td->td_proc->p_ucred; /* XXX */
525 /* VOP_OPEN(vp, FWRITE, cred, NULL);*/
527 nmp = VFSTONWFS(vp->v_mount);
530 rtvals = ap->a_rtvals;
531 npages = btoc(count);
533 for (i = 0; i < npages; i++) {
534 rtvals[i] = VM_PAGER_AGAIN;
537 bp = getpbuf_kva(&nwfs_pbuf_freecnt);
538 kva = (vm_offset_t) bp->b_data;
539 pmap_qenter(kva, pages, npages);
541 iov.iov_base = (caddr_t) kva;
545 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
546 uio.uio_resid = count;
547 uio.uio_segflg = UIO_SYSSPACE;
548 uio.uio_rw = UIO_WRITE;
550 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
552 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, &uio, cred);
553 /* VOP_CLOSE(vp, FWRITE, cred);*/
554 NCPVNDEBUG("paged write done: %d\n", error);
556 pmap_qremove(kva, npages);
557 relpbuf(bp, &nwfs_pbuf_freecnt);
560 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
561 for (i = 0; i < nwritten; i++) {
562 rtvals[i] = VM_PAGER_OK;
563 vm_page_undirty(pages[i]);
567 #endif /* NWFS_RWCACHE */
570 * Flush and invalidate all dirty buffers. If another process is already
571 * doing the flush, just wait for completion.
574 nwfs_vinvalbuf(struct vnode *vp, int flags, int intrflg)
576 struct nwnode *np = VTONW(vp);
577 /* struct nwmount *nmp = VTONWFS(vp);*/
578 int error = 0, slpflag, slptimeo;
580 if (vp->v_flag & VRECLAIMED) {
590 while (np->n_flag & NFLUSHINPROG) {
591 np->n_flag |= NFLUSHWANT;
592 error = tsleep((caddr_t)&np->n_flag, 0, "nwfsvinv", slptimeo);
593 error = ncp_chkintr(NWFSTOCONN(VTONWFS(vp)), curthread);
594 if (error == EINTR && intrflg)
597 np->n_flag |= NFLUSHINPROG;
598 error = vinvalbuf(vp, flags, slpflag, 0);
600 if (intrflg && (error == ERESTART || error == EINTR)) {
601 np->n_flag &= ~NFLUSHINPROG;
602 if (np->n_flag & NFLUSHWANT) {
603 np->n_flag &= ~NFLUSHWANT;
604 wakeup((caddr_t)&np->n_flag);
608 error = vinvalbuf(vp, flags, slpflag, 0);
610 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
611 if (np->n_flag & NFLUSHWANT) {
612 np->n_flag &= ~NFLUSHWANT;
613 wakeup((caddr_t)&np->n_flag);