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
32 * $FreeBSD: src/sys/fs/smbfs/smbfs_io.c,v 1.3.2.3 2003/01/17 08:20:26 tjr Exp $
33 * $DragonFly: src/sys/vfs/smbfs/smbfs_io.c,v 1.15 2005/01/08 18:57:48 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>
41 #include <sys/fcntl.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 <sys/ioccom.h>
58 #include <netproto/smb/smb.h>
59 #include <netproto/smb/smb_conn.h>
60 #include <netproto/smb/smb_subr.h>
63 #include "smbfs_node.h"
64 #include "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 smb_cred scred;
85 struct smbfs_fctx *ctx;
87 struct smbnode *np = VTOSMB(vp);
88 int error/*, *eofflag = ap->a_eofflag*/;
92 SMBVDEBUG("dirname='%s'\n", np->n_name);
93 smb_makescred(&scred, uio->uio_td, cred);
94 offset = uio->uio_offset / DE_SIZE; /* offset in the directory */
95 limit = uio->uio_resid / DE_SIZE;
96 if (uio->uio_resid < DE_SIZE || uio->uio_offset < 0)
98 while (limit && offset < 2) {
100 bzero((caddr_t)&de, DE_SIZE);
101 de.d_reclen = DE_SIZE;
102 de.d_fileno = (offset == 0) ? np->n_ino :
103 (np->n_parent ? VTOSMB(np->n_parent)->n_ino : 2);
104 if (de.d_fileno == 0)
105 de.d_fileno = 0x7ffffffd + offset;
106 de.d_namlen = offset + 1;
109 de.d_name[offset + 1] = '\0';
111 error = uiomove((caddr_t)&de, DE_SIZE, uio);
115 uio->uio_offset += DE_SIZE;
119 if (offset != np->n_dirofs || np->n_dirseq == NULL) {
120 SMBVDEBUG("Reopening search %ld:%ld\n", offset, np->n_dirofs);
122 smbfs_findclose(np->n_dirseq, &scred);
126 error = smbfs_findopen(np, "*", 1,
127 SMB_FA_SYSTEM | SMB_FA_HIDDEN | SMB_FA_DIR,
130 SMBVDEBUG("can not open search, error = %d", error);
136 while (np->n_dirofs < offset) {
137 error = smbfs_findnext(ctx, offset - np->n_dirofs++, &scred);
139 smbfs_findclose(np->n_dirseq, &scred);
141 return error == ENOENT ? 0 : error;
145 for (; limit; limit--, offset++) {
146 error = smbfs_findnext(ctx, limit, &scred);
150 bzero((caddr_t)&de, DE_SIZE);
151 de.d_reclen = DE_SIZE;
152 de.d_fileno = ctx->f_attr.fa_ino;
153 de.d_type = (ctx->f_attr.fa_attr & SMB_FA_DIR) ? DT_DIR : DT_REG;
154 de.d_namlen = ctx->f_nmlen;
155 bcopy(ctx->f_name, de.d_name, de.d_namlen);
156 de.d_name[de.d_namlen] = '\0';
157 if (smbfs_fastlookup) {
158 error = smbfs_nget(vp->v_mount, vp, ctx->f_name,
159 ctx->f_nmlen, &ctx->f_attr, &newvp);
163 error = uiomove((caddr_t)&de, DE_SIZE, uio);
169 uio->uio_offset = offset * DE_SIZE;
174 smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred)
177 struct smbmount *smp = VFSTOSMBFS(vp->v_mount);
178 struct smbnode *np = VTOSMB(vp);
180 struct smb_cred scred;
184 * Protect against method which is not supported for now
186 if (uiop->uio_segflg == UIO_NOCOPY)
189 if (vp->v_type != VREG && vp->v_type != VDIR) {
190 SMBFSERR("vn types other than VREG or VDIR are unsupported !\n");
193 if (uiop->uio_resid == 0)
195 if (uiop->uio_offset < 0)
197 /* if (uiop->uio_offset + uiop->uio_resid > smp->nm_maxfilesize)
200 if (vp->v_type == VDIR) {
201 lks = LK_EXCLUSIVE;/*lockstatus(&vp->v_lock, td);*/
202 if (lks == LK_SHARED)
203 vn_lock(vp, LK_UPGRADE | LK_RETRY, td);
204 error = smbfs_readvdir(vp, uiop, cred);
205 if (lks == LK_SHARED)
206 vn_lock(vp, LK_DOWNGRADE | LK_RETRY, td);
210 /* biosize = SSTOCN(smp->sm_share)->sc_txmax;*/
211 if (np->n_flag & NMODIFIED) {
212 smbfs_attr_cacheremove(vp);
213 error = VOP_GETATTR(vp, &vattr, td);
216 np->n_mtime.tv_sec = vattr.va_mtime.tv_sec;
218 error = VOP_GETATTR(vp, &vattr, td);
221 if (np->n_mtime.tv_sec != vattr.va_mtime.tv_sec) {
222 error = smbfs_vinvalbuf(vp, V_SAVE, td, 1);
225 np->n_mtime.tv_sec = vattr.va_mtime.tv_sec;
228 smb_makescred(&scred, td, cred);
229 return smb_read(smp->sm_share, np->n_fid, uiop, &scred);
233 smbfs_writevnode(struct vnode *vp, struct uio *uiop,
234 struct ucred *cred, int ioflag)
237 struct smbmount *smp = VTOSMBFS(vp);
238 struct smbnode *np = VTOSMB(vp);
239 struct smb_cred scred;
242 if (vp->v_type != VREG) {
243 SMBERROR("vn types other than VREG unsupported !\n");
246 SMBVDEBUG("ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
247 if (uiop->uio_offset < 0)
249 /* if (uiop->uio_offset + uiop->uio_resid > smp->nm_maxfilesize)
252 if (ioflag & (IO_APPEND | IO_SYNC)) {
253 if (np->n_flag & NMODIFIED) {
254 smbfs_attr_cacheremove(vp);
255 error = smbfs_vinvalbuf(vp, V_SAVE, td, 1);
259 if (ioflag & IO_APPEND) {
262 * File size can be changed by another client
264 smbfs_attr_cacheremove(vp);
265 error = VOP_GETATTR(vp, &vattr, td);
266 if (error) return (error);
268 uiop->uio_offset = np->n_size;
271 if (uiop->uio_resid == 0)
274 uiop->uio_offset + uiop->uio_resid >
275 td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
276 psignal(td->td_proc, SIGXFSZ);
279 smb_makescred(&scred, td, cred);
280 error = smb_write(smp->sm_share, np->n_fid, uiop, &scred);
281 SMBVDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
283 if (uiop->uio_offset > np->n_size) {
284 np->n_size = uiop->uio_offset;
285 vnode_pager_setsize(vp, np->n_size);
292 * Do an I/O operation to/from a cache block.
295 smbfs_doio(struct buf *bp, struct ucred *cr, struct thread *td)
297 struct vnode *vp = bp->b_vp;
298 struct smbmount *smp = VFSTOSMBFS(vp->v_mount);
299 struct smbnode *np = VTOSMB(vp);
300 struct uio uio, *uiop = &uio;
302 struct smb_cred scred;
306 uiop->uio_iovcnt = 1;
307 uiop->uio_segflg = UIO_SYSSPACE;
310 smb_makescred(&scred, td, cr);
312 if (bp->b_flags & B_READ) {
313 io.iov_len = uiop->uio_resid = bp->b_bcount;
314 io.iov_base = bp->b_data;
315 uiop->uio_rw = UIO_READ;
316 switch (vp->v_type) {
318 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
319 error = smb_read(smp->sm_share, np->n_fid, uiop, &scred);
322 if (uiop->uio_resid) {
323 int left = uiop->uio_resid;
324 int nread = bp->b_bcount - left;
326 bzero((char *)bp->b_data + nread, left);
330 printf("smbfs_doio: type %x unexpected\n",vp->v_type);
335 bp->b_flags |= B_ERROR;
338 if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size)
339 bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE);
341 if (bp->b_dirtyend > bp->b_dirtyoff) {
342 io.iov_len = uiop->uio_resid = bp->b_dirtyend - bp->b_dirtyoff;
343 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
344 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
345 uiop->uio_rw = UIO_WRITE;
346 bp->b_flags |= B_WRITEINPROG;
347 error = smb_write(smp->sm_share, np->n_fid, uiop, &scred);
348 bp->b_flags &= ~B_WRITEINPROG;
351 * For an interrupted write, the buffer is still valid
352 * and the write hasn't been pushed to the server yet,
353 * so we can't set BIO_ERROR and report the interruption
354 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
355 * is not relevant, so the rpc attempt is essentially
356 * a noop. For the case of a V3 write rpc not being
357 * committed to stable storage, the block is still
358 * dirty and requires either a commit rpc or another
359 * write rpc with iomode == NFSV3WRITE_FILESYNC before
360 * the block is reused. This is indicated by setting
361 * the B_DELWRI and B_NEEDCOMMIT flags.
364 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
368 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
369 if ((bp->b_flags & B_ASYNC) == 0)
370 bp->b_flags |= B_EINTR;
371 if ((bp->b_flags & B_PAGING) == 0) {
373 bp->b_flags &= ~B_DONE;
375 if ((bp->b_flags & B_ASYNC) == 0)
376 bp->b_flags |= B_EINTR;
380 bp->b_flags |= B_ERROR;
383 bp->b_dirtyoff = bp->b_dirtyend = 0;
391 bp->b_resid = uiop->uio_resid;
397 * Vnode op for VM getpages.
398 * Wish wish .... get rid from multiple IO routines
400 * smbfs_getpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
401 * int a_reqpage, vm_ooffset_t a_offset)
404 smbfs_getpages(struct vop_getpages_args *ap)
406 #ifdef SMBFS_RWGENERIC
407 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
410 int i, error, nextoff, size, toff, npages, count;
417 struct thread *td = curthread; /* XXX */
419 struct smbmount *smp;
421 struct smb_cred scred;
424 KKASSERT(td->td_proc);
427 cred = td->td_proc->p_ucred;
429 smp = VFSTOSMBFS(vp->v_mount);
433 if (vp->v_object == NULL) {
434 printf("smbfs_getpages: called with non-merged cache vnode??\n");
435 return VM_PAGER_ERROR;
437 smb_makescred(&scred, td, cred);
439 bp = getpbuf(&smbfs_pbuf_freecnt);
440 npages = btoc(count);
441 kva = (vm_offset_t) bp->b_data;
442 pmap_qenter(kva, pages, npages);
444 iov.iov_base = (caddr_t) kva;
448 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
449 uio.uio_resid = count;
450 uio.uio_segflg = UIO_SYSSPACE;
451 uio.uio_rw = UIO_READ;
455 * This is kinda nasty. Since smbfs is physically closing the
456 * fid on close(), we have to reopen it if necessary. There are
457 * other races here too, such as if another process opens the same
458 * file while we are blocked in read. XXX
462 if (np->n_opencount == 0) {
463 error = smbfs_smb_open(np, SMB_AM_OPENREAD, &scred);
468 error = smb_read(smp->sm_share, np->n_fid, &uio, &scred);
470 smbfs_smb_close(smp->sm_share, np->n_fid, NULL, &scred);
471 pmap_qremove(kva, npages);
473 relpbuf(bp, &smbfs_pbuf_freecnt);
475 if (error && (uio.uio_resid == count)) {
476 printf("smbfs_getpages: error %d\n",error);
477 for (i = 0; i < npages; i++) {
478 if (ap->a_reqpage != i)
479 vnode_pager_freepage(pages[i]);
481 return VM_PAGER_ERROR;
484 size = count - uio.uio_resid;
486 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
488 nextoff = toff + PAGE_SIZE;
491 m->flags &= ~PG_ZERO;
493 if (nextoff <= size) {
494 m->valid = VM_PAGE_BITS_ALL;
497 int nvalid = ((size + DEV_BSIZE - 1) - toff) & ~(DEV_BSIZE - 1);
498 vm_page_set_validclean(m, 0, nvalid);
501 if (i != ap->a_reqpage) {
503 * Whether or not to leave the page activated is up in
504 * the air, but we should put the page on a page queue
505 * somewhere (it already is in the object). Result:
506 * It appears that emperical results show that
507 * deactivating pages is best.
511 * Just in case someone was asking for this page we
512 * now tell them that it is ok to use.
515 if (m->flags & PG_WANTED)
518 vm_page_deactivate(m);
521 vnode_pager_freepage(m);
526 #endif /* SMBFS_RWGENERIC */
530 * Vnode op for VM putpages.
531 * possible bug: all IO done in sync mode
532 * Note that vop_close always invalidate pages before close, so it's
533 * not necessary to open vnode.
535 * smbfs_putpages(struct vnode *a_vp, vm_page_t *a_m, int a_count, int a_sync,
536 * int *a_rtvals, vm_ooffset_t a_offset)
539 smbfs_putpages(struct vop_putpages_args *ap)
542 struct vnode *vp = ap->a_vp;
543 struct thread *td = curthread; /* XXX */
546 #ifdef SMBFS_RWGENERIC
547 KKASSERT(td->td_proc);
548 cred = td->td_proc->p_ucred;
549 VOP_OPEN(vp, FWRITE, cred, NULL, td);
550 error = vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
551 ap->a_sync, ap->a_rtvals);
552 VOP_CLOSE(vp, FWRITE, cred, td);
559 int i, npages, count;
562 struct smbmount *smp;
564 struct smb_cred scred;
567 KKASSERT(td->td_proc);
568 cred = td->td_proc->p_ucred;
569 /* VOP_OPEN(vp, FWRITE, cred, td);*/
571 smp = VFSTOSMBFS(vp->v_mount);
574 rtvals = ap->a_rtvals;
575 npages = btoc(count);
577 for (i = 0; i < npages; i++) {
578 rtvals[i] = VM_PAGER_AGAIN;
581 bp = getpbuf(&smbfs_pbuf_freecnt);
582 kva = (vm_offset_t) bp->b_data;
583 pmap_qenter(kva, pages, npages);
585 iov.iov_base = (caddr_t) kva;
589 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
590 uio.uio_resid = count;
591 uio.uio_segflg = UIO_SYSSPACE;
592 uio.uio_rw = UIO_WRITE;
594 SMBVDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
596 smb_makescred(&scred, td, cred);
599 * This is kinda nasty. Since smbfs is physically closing the
600 * fid on close(), we have to reopen it if necessary. There are
601 * other races here too, such as if another process opens the same
602 * file while we are blocked in read, or the file is open read-only
607 if (np->n_opencount == 0) {
608 error = smbfs_smb_open(np, SMB_AM_OPENRW, &scred);
613 error = smb_write(smp->sm_share, np->n_fid, &uio, &scred);
615 smbfs_smb_close(smp->sm_share, np->n_fid, NULL, &scred);
616 /* VOP_CLOSE(vp, FWRITE, cred, td);*/
617 SMBVDEBUG("paged write done: %d\n", error);
619 pmap_qremove(kva, npages);
620 relpbuf(bp, &smbfs_pbuf_freecnt);
623 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
624 for (i = 0; i < nwritten; i++) {
625 rtvals[i] = VM_PAGER_OK;
630 #endif /* SMBFS_RWGENERIC */
634 * Flush and invalidate all dirty buffers. If another process is already
635 * doing the flush, just wait for completion.
638 smbfs_vinvalbuf(struct vnode *vp, int flags, struct thread *td, int intrflg)
640 struct smbnode *np = VTOSMB(vp);
641 int error = 0, slpflag, slptimeo;
643 if (vp->v_flag & VRECLAIMED)
652 while (np->n_flag & NFLUSHINPROG) {
653 np->n_flag |= NFLUSHWANT;
654 error = tsleep((caddr_t)&np->n_flag, 0, "smfsvinv", slptimeo);
655 error = smb_proc_intr(td);
656 if (error == EINTR && intrflg)
659 np->n_flag |= NFLUSHINPROG;
660 error = vinvalbuf(vp, flags, td, slpflag, 0);
662 if (intrflg && (error == ERESTART || error == EINTR)) {
663 np->n_flag &= ~NFLUSHINPROG;
664 if (np->n_flag & NFLUSHWANT) {
665 np->n_flag &= ~NFLUSHWANT;
666 wakeup((caddr_t)&np->n_flag);
670 error = vinvalbuf(vp, flags, td, slpflag, 0);
672 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
673 if (np->n_flag & NFLUSHWANT) {
674 np->n_flag &= ~NFLUSHWANT;
675 wakeup((caddr_t)&np->n_flag);