2 * Copyright (c) 2000-2001, 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
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>
40 #include <sys/fcntl.h>
43 #include <sys/mount.h>
44 #include <sys/namei.h>
45 #include <sys/vnode.h>
46 #include <sys/dirent.h>
47 #include <sys/signalvar.h>
48 #include <sys/sysctl.h>
49 #include <sys/vmmeter.h>
52 #include <vm/vm_page.h>
53 #include <vm/vm_extern.h>
54 #include <vm/vm_object.h>
55 #include <vm/vm_pager.h>
56 #include <vm/vnode_pager.h>
58 #include <sys/ioccom.h>
60 #include <netsmb/smb.h>
61 #include <netsmb/smb_conn.h>
62 #include <netsmb/smb_subr.h>
64 #include <fs/smbfs/smbfs.h>
65 #include <fs/smbfs/smbfs_node.h>
66 #include <fs/smbfs/smbfs_subr.h>
68 /*#define SMBFS_RWGENERIC*/
70 extern int smbfs_pbuf_freecnt;
72 static int smbfs_fastlookup = 1;
74 SYSCTL_DECL(_vfs_smbfs);
75 SYSCTL_INT(_vfs_smbfs, OID_AUTO, fastlookup, CTLFLAG_RW, &smbfs_fastlookup, 0, "");
78 #define DE_SIZE (sizeof(struct dirent))
81 smbfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred)
84 struct componentname cn;
85 struct smb_cred scred;
86 struct smbfs_fctx *ctx;
88 struct smbnode *np = VTOSMB(vp);
89 int error/*, *eofflag = ap->a_eofflag*/;
93 SMBVDEBUG("dirname='%s'\n", np->n_name);
94 smb_makescred(&scred, uio->uio_td, cred);
95 offset = uio->uio_offset / DE_SIZE; /* offset in the directory */
96 limit = uio->uio_resid / DE_SIZE;
97 if (uio->uio_resid < DE_SIZE || uio->uio_offset < 0)
99 while (limit && offset < 2) {
101 bzero((caddr_t)&de, DE_SIZE);
102 de.d_reclen = DE_SIZE;
103 de.d_fileno = (offset == 0) ? np->n_ino :
104 (np->n_parent ? VTOSMB(np->n_parent)->n_ino : 2);
105 if (de.d_fileno == 0)
106 de.d_fileno = 0x7ffffffd + offset;
107 de.d_namlen = offset + 1;
110 de.d_name[offset + 1] = '\0';
112 error = uiomove(&de, DE_SIZE, uio);
116 uio->uio_offset += DE_SIZE;
120 if (offset != np->n_dirofs || np->n_dirseq == NULL) {
121 SMBVDEBUG("Reopening search %ld:%ld\n", offset, np->n_dirofs);
123 smbfs_findclose(np->n_dirseq, &scred);
127 error = smbfs_findopen(np, "*", 1,
128 SMB_FA_SYSTEM | SMB_FA_HIDDEN | SMB_FA_DIR,
131 SMBVDEBUG("can not open search, error = %d", error);
137 while (np->n_dirofs < offset) {
138 error = smbfs_findnext(ctx, offset - np->n_dirofs++, &scred);
140 smbfs_findclose(np->n_dirseq, &scred);
142 return error == ENOENT ? 0 : error;
146 for (; limit; limit--, offset++) {
147 error = smbfs_findnext(ctx, limit, &scred);
151 bzero((caddr_t)&de, DE_SIZE);
152 de.d_reclen = DE_SIZE;
153 de.d_fileno = ctx->f_attr.fa_ino;
154 de.d_type = (ctx->f_attr.fa_attr & SMB_FA_DIR) ? DT_DIR : DT_REG;
155 de.d_namlen = ctx->f_nmlen;
156 bcopy(ctx->f_name, de.d_name, de.d_namlen);
157 de.d_name[de.d_namlen] = '\0';
158 if (smbfs_fastlookup) {
159 error = smbfs_nget(vp->v_mount, vp, ctx->f_name,
160 ctx->f_nmlen, &ctx->f_attr, &newvp);
162 cn.cn_nameptr = de.d_name;
163 cn.cn_namelen = de.d_namlen;
164 cache_enter(vp, newvp, &cn);
168 error = uiomove(&de, DE_SIZE, uio);
174 uio->uio_offset = offset * DE_SIZE;
179 smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred)
181 struct smbmount *smp = VFSTOSMBFS(vp->v_mount);
182 struct smbnode *np = VTOSMB(vp);
185 struct smb_cred scred;
189 * Protect against method which is not supported for now
191 if (uiop->uio_segflg == UIO_NOCOPY)
194 if (vp->v_type != VREG && vp->v_type != VDIR) {
195 SMBFSERR("vn types other than VREG or VDIR are unsupported !\n");
198 if (uiop->uio_resid == 0)
200 if (uiop->uio_offset < 0)
202 /* if (uiop->uio_offset + uiop->uio_resid > smp->nm_maxfilesize)
205 if (vp->v_type == VDIR) {
206 lks = LK_EXCLUSIVE; /* lockstatus(vp->v_vnlock); */
207 if (lks == LK_SHARED)
208 vn_lock(vp, LK_UPGRADE | LK_RETRY);
209 error = smbfs_readvdir(vp, uiop, cred);
210 if (lks == LK_SHARED)
211 vn_lock(vp, LK_DOWNGRADE | LK_RETRY);
215 /* biosize = SSTOCN(smp->sm_share)->sc_txmax;*/
216 if (np->n_flag & NMODIFIED) {
217 smbfs_attr_cacheremove(vp);
218 error = VOP_GETATTR(vp, &vattr, cred);
221 np->n_mtime.tv_sec = vattr.va_mtime.tv_sec;
223 error = VOP_GETATTR(vp, &vattr, cred);
226 if (np->n_mtime.tv_sec != vattr.va_mtime.tv_sec) {
227 error = smbfs_vinvalbuf(vp, td);
230 np->n_mtime.tv_sec = vattr.va_mtime.tv_sec;
233 smb_makescred(&scred, td, cred);
234 return smb_read(smp->sm_share, np->n_fid, uiop, &scred);
238 smbfs_writevnode(struct vnode *vp, struct uio *uiop,
239 struct ucred *cred, int ioflag)
241 struct smbmount *smp = VTOSMBFS(vp);
242 struct smbnode *np = VTOSMB(vp);
243 struct smb_cred scred;
248 if (vp->v_type != VREG) {
249 SMBERROR("vn types other than VREG unsupported !\n");
252 SMBVDEBUG("ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
253 if (uiop->uio_offset < 0)
255 /* if (uiop->uio_offset + uiop->uio_resid > smp->nm_maxfilesize)
259 if (ioflag & (IO_APPEND | IO_SYNC)) {
260 if (np->n_flag & NMODIFIED) {
261 smbfs_attr_cacheremove(vp);
262 error = smbfs_vinvalbuf(vp, td);
266 if (ioflag & IO_APPEND) {
269 * File size can be changed by another client
271 smbfs_attr_cacheremove(vp);
272 error = VOP_GETATTR(vp, &vattr, cred);
273 if (error) return (error);
275 uiop->uio_offset = np->n_size;
278 if (uiop->uio_resid == 0)
282 if (uiop->uio_offset + uiop->uio_resid >
283 lim_cur(p, RLIMIT_FSIZE)) {
290 smb_makescred(&scred, td, cred);
291 error = smb_write(smp->sm_share, np->n_fid, uiop, &scred);
292 SMBVDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
294 if (uiop->uio_offset > np->n_size) {
295 np->n_size = uiop->uio_offset;
296 vnode_pager_setsize(vp, np->n_size);
303 * Do an I/O operation to/from a cache block.
306 smbfs_doio(struct vnode *vp, struct buf *bp, struct ucred *cr, struct thread *td)
308 struct smbmount *smp = VFSTOSMBFS(vp->v_mount);
309 struct smbnode *np = VTOSMB(vp);
310 struct uio uio, *uiop = &uio;
312 struct smb_cred scred;
316 uiop->uio_iovcnt = 1;
317 uiop->uio_segflg = UIO_SYSSPACE;
320 smb_makescred(&scred, td, cr);
322 if (bp->b_iocmd == BIO_READ) {
323 io.iov_len = uiop->uio_resid = bp->b_bcount;
324 io.iov_base = bp->b_data;
325 uiop->uio_rw = UIO_READ;
326 switch (vp->v_type) {
328 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
329 error = smb_read(smp->sm_share, np->n_fid, uiop, &scred);
332 if (uiop->uio_resid) {
333 int left = uiop->uio_resid;
334 int nread = bp->b_bcount - left;
336 bzero((char *)bp->b_data + nread, left);
340 printf("smbfs_doio: type %x unexpected\n",vp->v_type);
345 bp->b_ioflags |= BIO_ERROR;
348 if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size)
349 bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE);
351 if (bp->b_dirtyend > bp->b_dirtyoff) {
352 io.iov_len = uiop->uio_resid = bp->b_dirtyend - bp->b_dirtyoff;
353 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
354 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
355 uiop->uio_rw = UIO_WRITE;
356 error = smb_write(smp->sm_share, np->n_fid, uiop, &scred);
359 * For an interrupted write, the buffer is still valid
360 * and the write hasn't been pushed to the server yet,
361 * so we can't set BIO_ERROR and report the interruption
362 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
363 * is not relevant, so the rpc attempt is essentially
364 * a noop. For the case of a V3 write rpc not being
365 * committed to stable storage, the block is still
366 * dirty and requires either a commit rpc or another
367 * write rpc with iomode == NFSV3WRITE_FILESYNC before
368 * the block is reused. This is indicated by setting
369 * the B_DELWRI and B_NEEDCOMMIT flags.
372 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
376 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
377 if ((bp->b_flags & B_ASYNC) == 0)
378 bp->b_flags |= B_EINTR;
379 if ((bp->b_flags & B_PAGING) == 0) {
381 bp->b_flags &= ~B_DONE;
383 if ((bp->b_flags & B_ASYNC) == 0)
384 bp->b_flags |= B_EINTR;
388 bp->b_ioflags |= BIO_ERROR;
391 bp->b_dirtyoff = bp->b_dirtyend = 0;
399 bp->b_resid = uiop->uio_resid;
405 * Vnode op for VM getpages.
406 * Wish wish .... get rid from multiple IO routines
410 struct vop_getpages_args /* {
415 vm_ooffset_t a_offset;
418 #ifdef SMBFS_RWGENERIC
419 return vop_stdgetpages(ap);
421 int i, error, nextoff, size, toff, npages, count, reqpage;
429 struct smbmount *smp;
431 struct smb_cred scred;
436 if ((object = vp->v_object) == NULL) {
437 printf("smbfs_getpages: called with non-merged cache vnode??\n");
438 return VM_PAGER_ERROR;
441 td = curthread; /* XXX */
442 cred = td->td_ucred; /* XXX */
444 smp = VFSTOSMBFS(vp->v_mount);
447 npages = btoc(count);
448 reqpage = ap->a_reqpage;
451 * If the requested page is partially valid, just return it and
452 * allow the pager to zero-out the blanks. Partially valid pages
453 * can only occur at the file EOF.
457 VM_OBJECT_LOCK(object);
459 vm_page_lock_queues();
460 for (i = 0; i < npages; ++i) {
462 vm_page_free(pages[i]);
464 vm_page_unlock_queues();
465 VM_OBJECT_UNLOCK(object);
468 VM_OBJECT_UNLOCK(object);
470 smb_makescred(&scred, td, cred);
472 bp = getpbuf(&smbfs_pbuf_freecnt);
474 kva = (vm_offset_t) bp->b_data;
475 pmap_qenter(kva, pages, npages);
476 PCPU_INC(cnt.v_vnodein);
477 PCPU_ADD(cnt.v_vnodepgsin, npages);
479 iov.iov_base = (caddr_t) kva;
483 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
484 uio.uio_resid = count;
485 uio.uio_segflg = UIO_SYSSPACE;
486 uio.uio_rw = UIO_READ;
489 error = smb_read(smp->sm_share, np->n_fid, &uio, &scred);
490 pmap_qremove(kva, npages);
492 relpbuf(bp, &smbfs_pbuf_freecnt);
494 VM_OBJECT_LOCK(object);
495 if (error && (uio.uio_resid == count)) {
496 printf("smbfs_getpages: error %d\n",error);
497 vm_page_lock_queues();
498 for (i = 0; i < npages; i++) {
500 vm_page_free(pages[i]);
502 vm_page_unlock_queues();
503 VM_OBJECT_UNLOCK(object);
504 return VM_PAGER_ERROR;
507 size = count - uio.uio_resid;
509 vm_page_lock_queues();
510 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
512 nextoff = toff + PAGE_SIZE;
515 if (nextoff <= size) {
517 * Read operation filled an entire page
519 m->valid = VM_PAGE_BITS_ALL;
520 KASSERT(m->dirty == 0,
521 ("smbfs_getpages: page %p is dirty", m));
522 } else if (size > toff) {
524 * Read operation filled a partial page.
527 vm_page_set_valid(m, 0, size - toff);
528 KASSERT(m->dirty == 0,
529 ("smbfs_getpages: page %p is dirty", m));
532 * Read operation was short. If no error occured
533 * we may have hit a zero-fill section. We simply
534 * leave valid set to 0.
541 * Whether or not to leave the page activated is up in
542 * the air, but we should put the page on a page queue
543 * somewhere (it already is in the object). Result:
544 * It appears that emperical results show that
545 * deactivating pages is best.
549 * Just in case someone was asking for this page we
550 * now tell them that it is ok to use.
553 if (m->oflags & VPO_WANTED)
556 vm_page_deactivate(m);
563 vm_page_unlock_queues();
564 VM_OBJECT_UNLOCK(object);
566 #endif /* SMBFS_RWGENERIC */
570 * Vnode op for VM putpages.
571 * possible bug: all IO done in sync mode
572 * Note that vop_close always invalidate pages before close, so it's
573 * not necessary to open vnode.
577 struct vop_putpages_args /* {
583 vm_ooffset_t a_offset;
587 struct vnode *vp = ap->a_vp;
591 #ifdef SMBFS_RWGENERIC
592 td = curthread; /* XXX */
593 cred = td->td_ucred; /* XXX */
594 VOP_OPEN(vp, FWRITE, cred, td, NULL);
595 error = vop_stdputpages(ap);
596 VOP_CLOSE(vp, FWRITE, cred, td);
603 int i, npages, count;
605 struct smbmount *smp;
607 struct smb_cred scred;
610 td = curthread; /* XXX */
611 cred = td->td_ucred; /* XXX */
612 /* VOP_OPEN(vp, FWRITE, cred, td, NULL);*/
614 smp = VFSTOSMBFS(vp->v_mount);
617 rtvals = ap->a_rtvals;
618 npages = btoc(count);
620 for (i = 0; i < npages; i++) {
621 rtvals[i] = VM_PAGER_AGAIN;
624 bp = getpbuf(&smbfs_pbuf_freecnt);
626 kva = (vm_offset_t) bp->b_data;
627 pmap_qenter(kva, pages, npages);
628 PCPU_INC(cnt.v_vnodeout);
629 PCPU_ADD(cnt.v_vnodepgsout, count);
631 iov.iov_base = (caddr_t) kva;
635 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
636 uio.uio_resid = count;
637 uio.uio_segflg = UIO_SYSSPACE;
638 uio.uio_rw = UIO_WRITE;
640 SMBVDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
642 smb_makescred(&scred, td, cred);
643 error = smb_write(smp->sm_share, np->n_fid, &uio, &scred);
644 /* VOP_CLOSE(vp, FWRITE, cred, td);*/
645 SMBVDEBUG("paged write done: %d\n", error);
647 pmap_qremove(kva, npages);
649 relpbuf(bp, &smbfs_pbuf_freecnt);
652 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
653 vm_page_lock_queues();
654 for (i = 0; i < nwritten; i++) {
655 rtvals[i] = VM_PAGER_OK;
656 vm_page_undirty(pages[i]);
658 vm_page_unlock_queues();
661 #endif /* SMBFS_RWGENERIC */
665 * Flush and invalidate all dirty buffers. If another process is already
666 * doing the flush, just wait for completion.
669 smbfs_vinvalbuf(struct vnode *vp, struct thread *td)
671 struct smbnode *np = VTOSMB(vp);
674 if (vp->v_iflag & VI_DOOMED)
677 while (np->n_flag & NFLUSHINPROG) {
678 np->n_flag |= NFLUSHWANT;
679 error = tsleep(&np->n_flag, PRIBIO + 2, "smfsvinv", 2 * hz);
680 error = smb_td_intr(td);
684 np->n_flag |= NFLUSHINPROG;
686 if (vp->v_bufobj.bo_object != NULL) {
687 VM_OBJECT_LOCK(vp->v_bufobj.bo_object);
688 vm_object_page_clean(vp->v_bufobj.bo_object, 0, 0, OBJPC_SYNC);
689 VM_OBJECT_UNLOCK(vp->v_bufobj.bo_object);
692 error = vinvalbuf(vp, V_SAVE, PCATCH, 0);
694 if (error == ERESTART || error == EINTR) {
695 np->n_flag &= ~NFLUSHINPROG;
696 if (np->n_flag & NFLUSHWANT) {
697 np->n_flag &= ~NFLUSHWANT;
702 error = vinvalbuf(vp, V_SAVE, PCATCH, 0);
704 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
705 if (np->n_flag & NFLUSHWANT) {
706 np->n_flag &= ~NFLUSHWANT;