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 1.24 2007/02/22 15:50:50 corecode 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>
60 #include <sys/thread2.h>
62 #include <machine/limits.h>
65 #include "nwfs_node.h"
66 #include "nwfs_subr.h"
68 static int nwfs_fastlookup = 1;
70 SYSCTL_DECL(_vfs_nwfs);
71 SYSCTL_INT(_vfs_nwfs, OID_AUTO, fastlookup, CTLFLAG_RW, &nwfs_fastlookup, 0, "");
74 extern int nwfs_pbuf_freecnt;
79 nwfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred)
81 struct nwmount *nmp = VTONWFS(vp);
84 struct nw_entry_info fattr;
93 NCPVNDEBUG("dirname='%s'\n",np->n_name);
94 if (uio->uio_offset < 0 || uio->uio_offset > INT_MAX)
97 i = (int)uio->uio_offset; /* offset in directory */
99 error = ncp_initsearch(vp, uio->uio_td, cred);
101 NCPVNDEBUG("cannot initialize search, error=%d",error);
106 for (; !error && uio->uio_resid > 0; i++) {
109 d_ino = np->n_fid.f_id;
111 d_ino = NWFS_ROOT_INO;
117 d_ino = np->n_parent.f_id;
119 d_ino = NWFS_ROOT_INO;
125 error = ncp_search_for_file_or_subdir(nmp, &np->n_seq, &fattr, uio->uio_td, cred);
126 if (error && error < 0x80)
128 d_ino = fattr.dirEntNum;
129 d_type = (fattr.attributes & aDIR) ? DT_DIR : DT_REG;
130 d_namlen = fattr.nameLen;
131 d_name = fattr.entryName;
133 if (error && eofflag) {
140 if (nwfs_fastlookup && !error && i > 1) {
141 fid.f_id = fattr.dirEntNum;
142 fid.f_parent = np->n_fid.f_id;
143 error = nwfs_nget(vp->v_mount, fid, &fattr, vp, &newvp);
145 VTONW(newvp)->n_ctime = VTONW(newvp)->n_vattr.va_ctime.tv_sec;
154 if (vop_write_dirent(&error, uio, d_ino, d_type, d_namlen, d_name))
163 nwfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred)
165 struct nwmount *nmp = VFSTONWFS(vp->v_mount);
166 struct nwnode *np = VTONW(vp);
171 if (vp->v_type != VREG && vp->v_type != VDIR) {
172 kprintf("%s: vn types other than VREG or VDIR are unsupported !\n",__func__);
175 if (uiop->uio_resid == 0) return 0;
176 if (uiop->uio_offset < 0) return EINVAL;
178 if (vp->v_type == VDIR) {
179 error = nwfs_readvdir(vp, uiop, cred);
182 biosize = NWFSTOCONN(nmp)->buffer_size;
183 if (np->n_flag & NMODIFIED) {
184 nwfs_attr_cacheremove(vp);
185 error = VOP_GETATTR(vp, &vattr);
186 if (error) return (error);
187 np->n_mtime = vattr.va_mtime.tv_sec;
189 error = VOP_GETATTR(vp, &vattr);
190 if (error) return (error);
191 if (np->n_mtime != vattr.va_mtime.tv_sec) {
192 error = nwfs_vinvalbuf(vp, V_SAVE, 1);
193 if (error) return (error);
194 np->n_mtime = vattr.va_mtime.tv_sec;
197 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop,cred);
202 nwfs_writevnode(struct vnode *vp, struct uio *uiop, struct ucred *cred,
205 struct nwmount *nmp = VTONWFS(vp);
206 struct nwnode *np = VTONW(vp);
208 /* struct vattr vattr;*/
211 if (vp->v_type != VREG) {
212 kprintf("%s: vn types other than VREG unsupported !\n",__func__);
215 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
216 if (uiop->uio_offset < 0) return EINVAL;
218 if (ioflag & (IO_APPEND | IO_SYNC)) {
219 if (np->n_flag & NMODIFIED) {
220 nwfs_attr_cacheremove(vp);
221 error = nwfs_vinvalbuf(vp, V_SAVE, 1);
222 if (error) return (error);
224 if (ioflag & IO_APPEND) {
225 /* We can relay only on local information about file size,
226 * because until file is closed NetWare will not return
227 * the correct size. */
229 nwfs_attr_cacheremove(vp);
230 error = VOP_GETATTR(vp, &vattr);
231 if (error) return (error);
233 uiop->uio_offset = np->n_size;
236 if (uiop->uio_resid == 0) return 0;
237 if (td->td_proc && uiop->uio_offset + uiop->uio_resid >
238 td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
239 lwpsignal(td->td_proc, td->td_lwp, SIGXFSZ);
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.
257 nwfs_doio(struct vnode *vp, struct bio *bio, struct ucred *cr, struct thread *td)
259 struct buf *bp = bio->bio_buf;
268 nmp = VFSTONWFS(vp->v_mount);
271 uiop->uio_iovcnt = 1;
272 uiop->uio_segflg = UIO_SYSSPACE;
275 if (bp->b_cmd == BUF_CMD_READ) {
276 io.iov_len = uiop->uio_resid = (size_t)bp->b_bcount;
277 io.iov_base = bp->b_data;
278 uiop->uio_rw = UIO_READ;
279 switch (vp->v_type) {
281 uiop->uio_offset = bio->bio_offset;
282 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
285 if (uiop->uio_resid) {
286 size_t left = uiop->uio_resid;
287 size_t nread = bp->b_bcount - left;
289 bzero((char *)bp->b_data + nread, left);
293 nfsstats.readdir_bios++;
294 uiop->uio_offset = bio->bio_offset;
295 if (nmp->nm_flag & NFSMNT_RDIRPLUS) {
296 error = nfs_readdirplusrpc(vp, uiop, cr);
297 if (error == NFSERR_NOTSUPP)
298 nmp->nm_flag &= ~NFSMNT_RDIRPLUS;
300 if ((nmp->nm_flag & NFSMNT_RDIRPLUS) == 0)
301 error = nfs_readdirrpc(vp, uiop, cr);
302 if (error == 0 && uiop->uio_resid == (size_t)bp->b_bcount)
303 bp->b_flags |= B_INVAL;
307 kprintf("nwfs_doio: type %x unexpected\n",vp->v_type);
311 bp->b_flags |= B_ERROR;
315 KKASSERT(bp->b_cmd == BUF_CMD_WRITE);
316 if (bio->bio_offset + bp->b_dirtyend > np->n_size)
317 bp->b_dirtyend = np->n_size - bio->bio_offset;
319 if (bp->b_dirtyend > bp->b_dirtyoff) {
320 io.iov_len = uiop->uio_resid =
321 (size_t)(bp->b_dirtyend - bp->b_dirtyoff);
322 uiop->uio_offset = bio->bio_offset + bp->b_dirtyoff;
323 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
324 uiop->uio_rw = UIO_WRITE;
325 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, uiop, cr);
328 * For an interrupted write, the buffer is still valid
329 * and the write hasn't been pushed to the server yet,
330 * so we can't set B_ERROR and report the interruption
331 * by setting B_EINTR. For the async case, B_EINTR
332 * is not relevant, so the rpc attempt is essentially
333 * a noop. For the case of a V3 write rpc not being
334 * committed to stable storage, the block is still
335 * dirty and requires either a commit rpc or another
336 * write rpc with iomode == NFSV3WRITE_FILESYNC before
337 * the block is reused. This is indicated by setting
338 * the B_DELWRI and B_NEEDCOMMIT flags.
341 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
344 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
345 if ((bp->b_flags & B_PAGING) == 0)
347 bp->b_flags |= B_EINTR;
351 bp->b_flags |= B_ERROR;
352 bp->b_error /*= np->n_error */= error;
353 /* np->n_flag |= NWRITEERR;*/
355 bp->b_dirtyoff = bp->b_dirtyend = 0;
363 bp->b_resid = (int)uiop->uio_resid;
369 * Vnode op for VM getpages.
370 * Wish wish .... get rid from multiple IO routines
372 * nwfs_getpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
373 * int a_reqpage, vm_ooffset_t a_offset)
376 nwfs_getpages(struct vop_getpages_args *ap)
379 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
380 ap->a_reqpage, ap->a_seqaccess);
382 int i, error, npages;
383 size_t nextoff, toff;
391 struct thread *td = curthread; /* XXX */
397 KKASSERT(td->td_proc);
398 cred = td->td_proc->p_ucred;
402 nmp = VFSTONWFS(vp->v_mount);
404 count = (size_t)ap->a_count;
406 if (vp->v_object == NULL) {
407 kprintf("nwfs_getpages: called with non-merged cache vnode??\n");
408 return VM_PAGER_ERROR;
411 bp = getpbuf_kva(&nwfs_pbuf_freecnt);
412 npages = btoc(count);
413 kva = (vm_offset_t) bp->b_data;
414 pmap_qenter(kva, pages, npages);
416 iov.iov_base = (caddr_t) kva;
420 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
421 uio.uio_resid = count;
422 uio.uio_segflg = UIO_SYSSPACE;
423 uio.uio_rw = UIO_READ;
426 error = ncp_read(NWFSTOCONN(nmp), &np->n_fh, &uio,cred);
427 pmap_qremove(kva, npages);
429 relpbuf(bp, &nwfs_pbuf_freecnt);
431 if (error && (uio.uio_resid == count)) {
432 kprintf("nwfs_getpages: error %d\n",error);
433 for (i = 0; i < npages; i++) {
434 if (ap->a_reqpage != i)
435 vnode_pager_freepage(pages[i]);
437 return VM_PAGER_ERROR;
440 size = count - uio.uio_resid;
442 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
444 nextoff = toff + PAGE_SIZE;
447 m->flags &= ~PG_ZERO;
450 * NOTE: pmap dirty bit should have already been cleared.
451 * We do not clear it here.
453 if (nextoff <= size) {
454 m->valid = VM_PAGE_BITS_ALL;
457 int nvalid = ((size + DEV_BSIZE - 1) - toff) &
459 vm_page_set_validclean(m, 0, nvalid);
462 if (i != ap->a_reqpage) {
464 * Whether or not to leave the page activated is up in
465 * the air, but we should put the page on a page queue
466 * somewhere (it already is in the object). Result:
467 * It appears that emperical results show that
468 * deactivating pages is best.
472 * Just in case someone was asking for this page we
473 * now tell them that it is ok to use.
476 if (m->flags & PG_WANTED)
479 vm_page_deactivate(m);
482 vnode_pager_freepage(m);
487 #endif /* NWFS_RWCACHE */
491 * Vnode op for VM putpages.
492 * possible bug: all IO done in sync mode
493 * Note that vop_close always invalidate pages before close, so it's
494 * not necessary to open vnode.
496 * nwfs_putpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
497 * int a_sync, int *a_rtvals, vm_ooffset_t a_offset)
500 nwfs_putpages(struct vop_putpages_args *ap)
503 struct thread *td = curthread; /* XXX */
504 struct vnode *vp = ap->a_vp;
508 KKASSERT(td->td_proc);
509 cred = td->td_proc->p_ucred; /* XXX */
510 VOP_OPEN(vp, FWRITE, cred, NULL);
511 error = vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
512 ap->a_sync, ap->a_rtvals);
513 VOP_CLOSE(vp, FWRITE, cred);
520 int i, npages, count;
526 KKASSERT(td->td_proc);
527 cred = td->td_proc->p_ucred; /* XXX */
529 /* VOP_OPEN(vp, FWRITE, cred, NULL);*/
531 nmp = VFSTONWFS(vp->v_mount);
534 rtvals = ap->a_rtvals;
535 npages = btoc(count);
537 for (i = 0; i < npages; i++) {
538 rtvals[i] = VM_PAGER_AGAIN;
541 bp = getpbuf_kva(&nwfs_pbuf_freecnt);
542 kva = (vm_offset_t) bp->b_data;
543 pmap_qenter(kva, pages, npages);
545 iov.iov_base = (caddr_t) kva;
549 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
550 uio.uio_resid = count;
551 uio.uio_segflg = UIO_SYSSPACE;
552 uio.uio_rw = UIO_WRITE;
554 NCPVNDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
556 error = ncp_write(NWFSTOCONN(nmp), &np->n_fh, &uio, cred);
557 /* VOP_CLOSE(vp, FWRITE, cred);*/
558 NCPVNDEBUG("paged write done: %d\n", error);
560 pmap_qremove(kva, npages);
561 relpbuf(bp, &nwfs_pbuf_freecnt);
564 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
565 for (i = 0; i < nwritten; i++) {
566 rtvals[i] = VM_PAGER_OK;
567 vm_page_undirty(pages[i]);
571 #endif /* NWFS_RWCACHE */
574 * Flush and invalidate all dirty buffers. If another process is already
575 * doing the flush, just wait for completion.
578 nwfs_vinvalbuf(struct vnode *vp, int flags, int intrflg)
580 struct nwnode *np = VTONW(vp);
581 /* struct nwmount *nmp = VTONWFS(vp);*/
582 int error = 0, slpflag, slptimeo;
584 if (vp->v_flag & VRECLAIMED) {
594 while (np->n_flag & NFLUSHINPROG) {
595 np->n_flag |= NFLUSHWANT;
596 error = tsleep((caddr_t)&np->n_flag, 0, "nwfsvinv", slptimeo);
597 error = ncp_chkintr(NWFSTOCONN(VTONWFS(vp)), curthread);
598 if (error == EINTR && intrflg)
601 np->n_flag |= NFLUSHINPROG;
602 error = vinvalbuf(vp, flags, slpflag, 0);
604 if (intrflg && (error == ERESTART || error == EINTR)) {
605 np->n_flag &= ~NFLUSHINPROG;
606 if (np->n_flag & NFLUSHWANT) {
607 np->n_flag &= ~NFLUSHWANT;
608 wakeup((caddr_t)&np->n_flag);
612 error = vinvalbuf(vp, flags, slpflag, 0);
614 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
615 if (np->n_flag & NFLUSHWANT) {
616 np->n_flag &= ~NFLUSHWANT;
617 wakeup((caddr_t)&np->n_flag);