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 1.19 2005/08/27 20:23:06 joerg 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>
49 #include <machine/limits.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 <netproto/smb/smb.h>
61 #include <netproto/smb/smb_conn.h>
62 #include <netproto/smb/smb_subr.h>
65 #include "smbfs_node.h"
66 #include "smbfs_subr.h"
70 #include <sys/thread2.h>
72 /*#define SMBFS_RWGENERIC*/
74 extern int smbfs_pbuf_freecnt;
76 static int smbfs_fastlookup = 1;
78 SYSCTL_DECL(_vfs_smbfs);
79 SYSCTL_INT(_vfs_smbfs, OID_AUTO, fastlookup, CTLFLAG_RW, &smbfs_fastlookup, 0, "");
82 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, i, offset, retval/*, *eofflag = ap->a_eofflag*/;
91 SMBVDEBUG("dirname='%s'\n", np->n_name);
92 smb_makescred(&scred, uio->uio_td, cred);
94 if (uio->uio_resid < 0 || uio->uio_offset < 0 ||
95 uio->uio_offset > INT_MAX)
100 offset = uio->uio_offset;
102 if (uio->uio_resid > 0 && offset < 1) {
103 if (vop_write_dirent(&error, uio, np->n_ino, DT_DIR, 1, "."))
110 if (uio->uio_resid > 0 && offset < 2) {
111 if (vop_write_dirent(&error, uio,
112 np->n_parent ? VTOSMB(np->n_parent)->n_ino : 2,
120 if (uio->uio_resid == 0)
123 if (offset != np->n_dirofs || np->n_dirseq == NULL) {
124 SMBVDEBUG("Reopening search %ld:%ld\n", offset, np->n_dirofs);
126 smbfs_findclose(np->n_dirseq, &scred);
130 error = smbfs_findopen(np, "*", 1,
131 SMB_FA_SYSTEM | SMB_FA_HIDDEN | SMB_FA_DIR,
134 SMBVDEBUG("can not open search, error = %d", error);
140 while (np->n_dirofs < offset) {
141 error = smbfs_findnext(ctx, offset - np->n_dirofs++, &scred);
143 smbfs_findclose(np->n_dirseq, &scred);
145 return error == ENOENT ? 0 : error;
149 for (; uio->uio_resid > 0 && !error; i++) {
151 * Overestimate the size of a record a bit, doesn't really
152 * hurt to be wrong here.
154 error = smbfs_findnext(ctx, uio->uio_resid / _DIRENT_RECLEN(255) + 1, &scred);
159 retval = vop_write_dirent(&error, uio, ctx->f_attr.fa_ino,
160 (ctx->f_attr.fa_attr & SMB_FA_DIR) ? DT_DIR : DT_REG,
161 ctx->f_nmlen, ctx->f_name);
164 if (smbfs_fastlookup && !error) {
165 error = smbfs_nget(vp->v_mount, vp, ctx->f_name,
166 ctx->f_nmlen, &ctx->f_attr, &newvp);
180 smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred)
183 struct smbmount *smp = VFSTOSMBFS(vp->v_mount);
184 struct smbnode *np = VTOSMB(vp);
186 struct smb_cred scred;
190 * Protect against method which is not supported for now
192 if (uiop->uio_segflg == UIO_NOCOPY)
195 if (vp->v_type != VREG && vp->v_type != VDIR) {
196 SMBFSERR("vn types other than VREG or VDIR are unsupported !\n");
199 if (uiop->uio_resid == 0)
201 if (uiop->uio_offset < 0)
203 /* if (uiop->uio_offset + uiop->uio_resid > smp->nm_maxfilesize)
206 if (vp->v_type == VDIR) {
207 lks = LK_EXCLUSIVE;/*lockstatus(&vp->v_lock, td);*/
208 if (lks == LK_SHARED)
209 vn_lock(vp, LK_UPGRADE | LK_RETRY, td);
210 error = smbfs_readvdir(vp, uiop, cred);
211 if (lks == LK_SHARED)
212 vn_lock(vp, LK_DOWNGRADE | LK_RETRY, td);
216 /* biosize = SSTOCN(smp->sm_share)->sc_txmax;*/
217 if (np->n_flag & NMODIFIED) {
218 smbfs_attr_cacheremove(vp);
219 error = VOP_GETATTR(vp, &vattr, td);
222 np->n_mtime.tv_sec = vattr.va_mtime.tv_sec;
224 error = VOP_GETATTR(vp, &vattr, td);
227 if (np->n_mtime.tv_sec != vattr.va_mtime.tv_sec) {
228 error = smbfs_vinvalbuf(vp, V_SAVE, td, 1);
231 np->n_mtime.tv_sec = vattr.va_mtime.tv_sec;
234 smb_makescred(&scred, td, cred);
235 return smb_read(smp->sm_share, np->n_fid, uiop, &scred);
239 smbfs_writevnode(struct vnode *vp, struct uio *uiop,
240 struct ucred *cred, int ioflag)
243 struct smbmount *smp = VTOSMBFS(vp);
244 struct smbnode *np = VTOSMB(vp);
245 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)
258 if (ioflag & (IO_APPEND | IO_SYNC)) {
259 if (np->n_flag & NMODIFIED) {
260 smbfs_attr_cacheremove(vp);
261 error = smbfs_vinvalbuf(vp, V_SAVE, td, 1);
265 if (ioflag & IO_APPEND) {
268 * File size can be changed by another client
270 smbfs_attr_cacheremove(vp);
271 error = VOP_GETATTR(vp, &vattr, td);
272 if (error) return (error);
274 uiop->uio_offset = np->n_size;
277 if (uiop->uio_resid == 0)
280 uiop->uio_offset + uiop->uio_resid >
281 td->td_proc->p_rlimit[RLIMIT_FSIZE].rlim_cur) {
282 psignal(td->td_proc, SIGXFSZ);
285 smb_makescred(&scred, td, cred);
286 error = smb_write(smp->sm_share, np->n_fid, uiop, &scred);
287 SMBVDEBUG("after: ofs=%d,resid=%d\n",(int)uiop->uio_offset, uiop->uio_resid);
289 if (uiop->uio_offset > np->n_size) {
290 np->n_size = uiop->uio_offset;
291 vnode_pager_setsize(vp, np->n_size);
298 * Do an I/O operation to/from a cache block.
301 smbfs_doio(struct buf *bp, struct ucred *cr, struct thread *td)
303 struct vnode *vp = bp->b_vp;
304 struct smbmount *smp = VFSTOSMBFS(vp->v_mount);
305 struct smbnode *np = VTOSMB(vp);
306 struct uio uio, *uiop = &uio;
308 struct smb_cred scred;
312 uiop->uio_iovcnt = 1;
313 uiop->uio_segflg = UIO_SYSSPACE;
316 smb_makescred(&scred, td, cr);
318 if (bp->b_flags & B_READ) {
319 io.iov_len = uiop->uio_resid = bp->b_bcount;
320 io.iov_base = bp->b_data;
321 uiop->uio_rw = UIO_READ;
322 switch (vp->v_type) {
324 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
325 error = smb_read(smp->sm_share, np->n_fid, uiop, &scred);
328 if (uiop->uio_resid) {
329 int left = uiop->uio_resid;
330 int nread = bp->b_bcount - left;
332 bzero((char *)bp->b_data + nread, left);
336 printf("smbfs_doio: type %x unexpected\n",vp->v_type);
341 bp->b_flags |= B_ERROR;
344 if (((bp->b_blkno * DEV_BSIZE) + bp->b_dirtyend) > np->n_size)
345 bp->b_dirtyend = np->n_size - (bp->b_blkno * DEV_BSIZE);
347 if (bp->b_dirtyend > bp->b_dirtyoff) {
348 io.iov_len = uiop->uio_resid = bp->b_dirtyend - bp->b_dirtyoff;
349 uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE + bp->b_dirtyoff;
350 io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
351 uiop->uio_rw = UIO_WRITE;
352 error = smb_write(smp->sm_share, np->n_fid, uiop, &scred);
355 * For an interrupted write, the buffer is still valid
356 * and the write hasn't been pushed to the server yet,
357 * so we can't set BIO_ERROR and report the interruption
358 * by setting B_EINTR. For the B_ASYNC case, B_EINTR
359 * is not relevant, so the rpc attempt is essentially
360 * a noop. For the case of a V3 write rpc not being
361 * committed to stable storage, the block is still
362 * dirty and requires either a commit rpc or another
363 * write rpc with iomode == NFSV3WRITE_FILESYNC before
364 * the block is reused. This is indicated by setting
365 * the B_DELWRI and B_NEEDCOMMIT flags.
368 || (!error && (bp->b_flags & B_NEEDCOMMIT))) {
371 bp->b_flags &= ~(B_INVAL|B_NOCACHE);
372 if ((bp->b_flags & B_ASYNC) == 0)
373 bp->b_flags |= B_EINTR;
374 if ((bp->b_flags & B_PAGING) == 0) {
376 bp->b_flags &= ~B_DONE;
378 if ((bp->b_flags & B_ASYNC) == 0)
379 bp->b_flags |= B_EINTR;
383 bp->b_flags |= B_ERROR;
386 bp->b_dirtyoff = bp->b_dirtyend = 0;
394 bp->b_resid = uiop->uio_resid;
400 * Vnode op for VM getpages.
401 * Wish wish .... get rid from multiple IO routines
403 * smbfs_getpages(struct vnode *a_vp, vm_page_t *a_m, int a_count,
404 * int a_reqpage, vm_ooffset_t a_offset)
407 smbfs_getpages(struct vop_getpages_args *ap)
409 #ifdef SMBFS_RWGENERIC
410 return vnode_pager_generic_getpages(ap->a_vp, ap->a_m, ap->a_count,
413 int i, error, nextoff, size, toff, npages, count;
420 struct thread *td = curthread; /* XXX */
422 struct smbmount *smp;
424 struct smb_cred scred;
427 KKASSERT(td->td_proc);
430 cred = td->td_proc->p_ucred;
432 smp = VFSTOSMBFS(vp->v_mount);
436 if (vp->v_object == NULL) {
437 printf("smbfs_getpages: called with non-merged cache vnode??\n");
438 return VM_PAGER_ERROR;
440 smb_makescred(&scred, td, cred);
442 bp = getpbuf(&smbfs_pbuf_freecnt);
443 npages = btoc(count);
444 kva = (vm_offset_t) bp->b_data;
445 pmap_qenter(kva, pages, npages);
447 iov.iov_base = (caddr_t) kva;
451 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
452 uio.uio_resid = count;
453 uio.uio_segflg = UIO_SYSSPACE;
454 uio.uio_rw = UIO_READ;
458 * This is kinda nasty. Since smbfs is physically closing the
459 * fid on close(), we have to reopen it if necessary. There are
460 * other races here too, such as if another process opens the same
461 * file while we are blocked in read. XXX
465 if (np->n_opencount == 0) {
466 error = smbfs_smb_open(np, SMB_AM_OPENREAD, &scred);
471 error = smb_read(smp->sm_share, np->n_fid, &uio, &scred);
473 smbfs_smb_close(smp->sm_share, np->n_fid, NULL, &scred);
474 pmap_qremove(kva, npages);
476 relpbuf(bp, &smbfs_pbuf_freecnt);
478 if (error && (uio.uio_resid == count)) {
479 printf("smbfs_getpages: error %d\n",error);
480 for (i = 0; i < npages; i++) {
481 if (ap->a_reqpage != i)
482 vnode_pager_freepage(pages[i]);
484 return VM_PAGER_ERROR;
487 size = count - uio.uio_resid;
489 for (i = 0, toff = 0; i < npages; i++, toff = nextoff) {
491 nextoff = toff + PAGE_SIZE;
494 m->flags &= ~PG_ZERO;
496 if (nextoff <= size) {
497 m->valid = VM_PAGE_BITS_ALL;
500 int nvalid = ((size + DEV_BSIZE - 1) - toff) & ~(DEV_BSIZE - 1);
501 vm_page_set_validclean(m, 0, nvalid);
504 if (i != ap->a_reqpage) {
506 * Whether or not to leave the page activated is up in
507 * the air, but we should put the page on a page queue
508 * somewhere (it already is in the object). Result:
509 * It appears that emperical results show that
510 * deactivating pages is best.
514 * Just in case someone was asking for this page we
515 * now tell them that it is ok to use.
518 if (m->flags & PG_WANTED)
521 vm_page_deactivate(m);
524 vnode_pager_freepage(m);
529 #endif /* SMBFS_RWGENERIC */
533 * Vnode op for VM putpages.
534 * possible bug: all IO done in sync mode
535 * Note that vop_close always invalidate pages before close, so it's
536 * not necessary to open vnode.
538 * smbfs_putpages(struct vnode *a_vp, vm_page_t *a_m, int a_count, int a_sync,
539 * int *a_rtvals, vm_ooffset_t a_offset)
542 smbfs_putpages(struct vop_putpages_args *ap)
545 struct vnode *vp = ap->a_vp;
546 struct thread *td = curthread; /* XXX */
549 #ifdef SMBFS_RWGENERIC
550 KKASSERT(td->td_proc);
551 cred = td->td_proc->p_ucred;
552 VOP_OPEN(vp, FWRITE, cred, NULL, td);
553 error = vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
554 ap->a_sync, ap->a_rtvals);
555 VOP_CLOSE(vp, FWRITE, cred, td);
562 int i, npages, count;
565 struct smbmount *smp;
567 struct smb_cred scred;
570 KKASSERT(td->td_proc);
571 cred = td->td_proc->p_ucred;
572 /* VOP_OPEN(vp, FWRITE, cred, td);*/
574 smp = VFSTOSMBFS(vp->v_mount);
577 rtvals = ap->a_rtvals;
578 npages = btoc(count);
580 for (i = 0; i < npages; i++) {
581 rtvals[i] = VM_PAGER_AGAIN;
584 bp = getpbuf(&smbfs_pbuf_freecnt);
585 kva = (vm_offset_t) bp->b_data;
586 pmap_qenter(kva, pages, npages);
588 iov.iov_base = (caddr_t) kva;
592 uio.uio_offset = IDX_TO_OFF(pages[0]->pindex);
593 uio.uio_resid = count;
594 uio.uio_segflg = UIO_SYSSPACE;
595 uio.uio_rw = UIO_WRITE;
597 SMBVDEBUG("ofs=%d,resid=%d\n",(int)uio.uio_offset, uio.uio_resid);
599 smb_makescred(&scred, td, cred);
602 * This is kinda nasty. Since smbfs is physically closing the
603 * fid on close(), we have to reopen it if necessary. There are
604 * other races here too, such as if another process opens the same
605 * file while we are blocked in read, or the file is open read-only
610 if (np->n_opencount == 0) {
611 error = smbfs_smb_open(np, SMB_AM_OPENRW, &scred);
616 error = smb_write(smp->sm_share, np->n_fid, &uio, &scred);
618 smbfs_smb_close(smp->sm_share, np->n_fid, NULL, &scred);
619 /* VOP_CLOSE(vp, FWRITE, cred, td);*/
620 SMBVDEBUG("paged write done: %d\n", error);
622 pmap_qremove(kva, npages);
623 relpbuf(bp, &smbfs_pbuf_freecnt);
626 int nwritten = round_page(count - uio.uio_resid) / PAGE_SIZE;
627 for (i = 0; i < nwritten; i++) {
628 rtvals[i] = VM_PAGER_OK;
633 #endif /* SMBFS_RWGENERIC */
637 * Flush and invalidate all dirty buffers. If another process is already
638 * doing the flush, just wait for completion.
641 smbfs_vinvalbuf(struct vnode *vp, int flags, struct thread *td, int intrflg)
643 struct smbnode *np = VTOSMB(vp);
644 int error = 0, slpflag, slptimeo;
646 if (vp->v_flag & VRECLAIMED)
655 while (np->n_flag & NFLUSHINPROG) {
656 np->n_flag |= NFLUSHWANT;
657 error = tsleep((caddr_t)&np->n_flag, 0, "smfsvinv", slptimeo);
658 error = smb_proc_intr(td);
659 if (error == EINTR && intrflg)
662 np->n_flag |= NFLUSHINPROG;
663 error = vinvalbuf(vp, flags, td, slpflag, 0);
665 if (intrflg && (error == ERESTART || error == EINTR)) {
666 np->n_flag &= ~NFLUSHINPROG;
667 if (np->n_flag & NFLUSHWANT) {
668 np->n_flag &= ~NFLUSHWANT;
669 wakeup((caddr_t)&np->n_flag);
673 error = vinvalbuf(vp, flags, td, slpflag, 0);
675 np->n_flag &= ~(NMODIFIED | NFLUSHINPROG);
676 if (np->n_flag & NFLUSHWANT) {
677 np->n_flag &= ~NFLUSHWANT;
678 wakeup((caddr_t)&np->n_flag);