1/*
2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17 */
18#include "xfs.h"
19#include "xfs_fs.h"
20#include "xfs_types.h"
21#include "xfs_bit.h"
22#include "xfs_log.h"
23#include "xfs_inum.h"
24#include "xfs_trans.h"
25#include "xfs_sb.h"
26#include "xfs_ag.h"
27#include "xfs_dir.h"
28#include "xfs_dir2.h"
29#include "xfs_dmapi.h"
30#include "xfs_mount.h"
31#include "xfs_bmap_btree.h"
32#include "xfs_alloc_btree.h"
33#include "xfs_ialloc_btree.h"
34#include "xfs_dir_sf.h"
35#include "xfs_dir2_sf.h"
36#include "xfs_attr_sf.h"
37#include "xfs_dinode.h"
38#include "xfs_inode.h"
39#include "xfs_inode_item.h"
40#include "xfs_itable.h"
41#include "xfs_btree.h"
42#include "xfs_alloc.h"
43#include "xfs_ialloc.h"
44#include "xfs_attr.h"
45#include "xfs_bmap.h"
46#include "xfs_acl.h"
47#include "xfs_mac.h"
48#include "xfs_error.h"
49#include "xfs_buf_item.h"
50#include "xfs_rw.h"
51
52/*
53 * This is a subroutine for xfs_write() and other writers (xfs_ioctl)
54 * which clears the setuid and setgid bits when a file is written.
55 */
56int
57xfs_write_clear_setuid(
58	xfs_inode_t	*ip)
59{
60	xfs_mount_t	*mp;
61	xfs_trans_t	*tp;
62	int		error;
63
64	mp = ip->i_mount;
65	tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID);
66	if ((error = xfs_trans_reserve(tp, 0,
67				      XFS_WRITEID_LOG_RES(mp),
68				      0, 0, 0))) {
69		xfs_trans_cancel(tp, 0);
70		return error;
71	}
72	xfs_ilock(ip, XFS_ILOCK_EXCL);
73	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
74	xfs_trans_ihold(tp, ip);
75	ip->i_d.di_mode &= ~S_ISUID;
76
77	/*
78	 * Note that we don't have to worry about mandatory
79	 * file locking being disabled here because we only
80	 * clear the S_ISGID bit if the Group execute bit is
81	 * on, but if it was on then mandatory locking wouldn't
82	 * have been enabled.
83	 */
84	if (ip->i_d.di_mode & S_IXGRP) {
85		ip->i_d.di_mode &= ~S_ISGID;
86	}
87	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
88	xfs_trans_set_sync(tp);
89	error = xfs_trans_commit(tp, 0, NULL);
90	xfs_iunlock(ip, XFS_ILOCK_EXCL);
91	return 0;
92}
93
94/*
95 * Force a shutdown of the filesystem instantly while keeping
96 * the filesystem consistent. We don't do an unmount here; just shutdown
97 * the shop, make sure that absolutely nothing persistent happens to
98 * this filesystem after this point.
99 */
100
101void
102xfs_do_force_shutdown(
103	bhv_desc_t	*bdp,
104	int		flags,
105	char		*fname,
106	int		lnnum)
107{
108	int		logerror;
109	xfs_mount_t	*mp;
110
111	mp = XFS_BHVTOM(bdp);
112	logerror = flags & XFS_LOG_IO_ERROR;
113
114	if (!(flags & XFS_FORCE_UMOUNT)) {
115		cmn_err(CE_NOTE,
116		"xfs_force_shutdown(%s,0x%x) called from line %d of file %s.  Return address = 0x%p",
117			mp->m_fsname,flags,lnnum,fname,__return_address);
118	}
119	/*
120	 * No need to duplicate efforts.
121	 */
122	if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
123		return;
124
125	/*
126	 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
127	 * queue up anybody new on the log reservations, and wakes up
128	 * everybody who's sleeping on log reservations and tells
129	 * them the bad news.
130	 */
131	if (xfs_log_force_umount(mp, logerror))
132		return;
133
134	if (flags & XFS_CORRUPT_INCORE) {
135		xfs_cmn_err(XFS_PTAG_SHUTDOWN_CORRUPT, CE_ALERT, mp,
136    "Corruption of in-memory data detected.  Shutting down filesystem: %s",
137			mp->m_fsname);
138		if (XFS_ERRLEVEL_HIGH <= xfs_error_level) {
139			xfs_stack_trace();
140		}
141	} else if (!(flags & XFS_FORCE_UMOUNT)) {
142		if (logerror) {
143			xfs_cmn_err(XFS_PTAG_SHUTDOWN_LOGERROR, CE_ALERT, mp,
144			"Log I/O Error Detected.  Shutting down filesystem: %s",
145				mp->m_fsname);
146		} else if (!(flags & XFS_SHUTDOWN_REMOTE_REQ)) {
147			xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp,
148				"I/O Error Detected.  Shutting down filesystem: %s",
149				mp->m_fsname);
150		}
151	}
152	if (!(flags & XFS_FORCE_UMOUNT)) {
153		cmn_err(CE_ALERT,
154		"Please umount the filesystem, and rectify the problem(s)");
155	}
156}
157
158
159/*
160 * Called when we want to stop a buffer from getting written or read.
161 * We attach the EIO error, muck with its flags, and call biodone
162 * so that the proper iodone callbacks get called.
163 */
164int
165xfs_bioerror(
166	xfs_buf_t *bp)
167{
168
169#ifdef XFSERRORDEBUG
170	ASSERT(XFS_BUF_ISREAD(bp) || bp->b_iodone);
171#endif
172
173	/*
174	 * No need to wait until the buffer is unpinned.
175	 * We aren't flushing it.
176	 */
177	xfs_buftrace("XFS IOERROR", bp);
178	XFS_BUF_ERROR(bp, EIO);
179	/*
180	 * We're calling biodone, so delete B_DONE flag. Either way
181	 * we have to call the iodone callback, and calling biodone
182	 * probably is the best way since it takes care of
183	 * GRIO as well.
184	 */
185	XFS_BUF_UNREAD(bp);
186	XFS_BUF_UNDELAYWRITE(bp);
187	XFS_BUF_UNDONE(bp);
188	XFS_BUF_STALE(bp);
189
190	XFS_BUF_CLR_BDSTRAT_FUNC(bp);
191	xfs_biodone(bp);
192
193	return (EIO);
194}
195
196/*
197 * Same as xfs_bioerror, except that we are releasing the buffer
198 * here ourselves, and avoiding the biodone call.
199 * This is meant for userdata errors; metadata bufs come with
200 * iodone functions attached, so that we can track down errors.
201 */
202int
203xfs_bioerror_relse(
204	xfs_buf_t *bp)
205{
206	int64_t fl;
207
208	ASSERT(XFS_BUF_IODONE_FUNC(bp) != xfs_buf_iodone_callbacks);
209	ASSERT(XFS_BUF_IODONE_FUNC(bp) != xlog_iodone);
210
211	xfs_buftrace("XFS IOERRELSE", bp);
212	fl = XFS_BUF_BFLAGS(bp);
213	/*
214	 * No need to wait until the buffer is unpinned.
215	 * We aren't flushing it.
216	 *
217	 * chunkhold expects B_DONE to be set, whether
218	 * we actually finish the I/O or not. We don't want to
219	 * change that interface.
220	 */
221	XFS_BUF_UNREAD(bp);
222	XFS_BUF_UNDELAYWRITE(bp);
223	XFS_BUF_DONE(bp);
224	XFS_BUF_STALE(bp);
225	XFS_BUF_CLR_IODONE_FUNC(bp);
226	XFS_BUF_CLR_BDSTRAT_FUNC(bp);
227	if (!(fl & XFS_B_ASYNC)) {
228		/*
229		 * Mark b_error and B_ERROR _both_.
230		 * Lot's of chunkcache code assumes that.
231		 * There's no reason to mark error for
232		 * ASYNC buffers.
233		 */
234		XFS_BUF_ERROR(bp, EIO);
235		XFS_BUF_V_IODONESEMA(bp);
236	} else {
237		xfs_buf_relse(bp);
238	}
239	return (EIO);
240}
241
242/*
243 * Prints out an ALERT message about I/O error.
244 */
245void
246xfs_ioerror_alert(
247	char			*func,
248	struct xfs_mount	*mp,
249	xfs_buf_t		*bp,
250	xfs_daddr_t		blkno)
251{
252	cmn_err(CE_ALERT,
253 "I/O error in filesystem (\"%s\") meta-data dev %s block 0x%llx"
254 "       (\"%s\") error %d buf count %zd",
255		(!mp || !mp->m_fsname) ? "(fs name not set)" : mp->m_fsname,
256		XFS_BUFTARG_NAME(XFS_BUF_TARGET(bp)),
257		(__uint64_t)blkno, func,
258		XFS_BUF_GETERROR(bp), XFS_BUF_COUNT(bp));
259}
260
261/*
262 * This isn't an absolute requirement, but it is
263 * just a good idea to call xfs_read_buf instead of
264 * directly doing a read_buf call. For one, we shouldn't
265 * be doing this disk read if we are in SHUTDOWN state anyway,
266 * so this stops that from happening. Secondly, this does all
267 * the error checking stuff and the brelse if appropriate for
268 * the caller, so the code can be a little leaner.
269 */
270
271int
272xfs_read_buf(
273	struct xfs_mount *mp,
274	xfs_buftarg_t	 *target,
275	xfs_daddr_t	 blkno,
276	int              len,
277	uint             flags,
278	xfs_buf_t	 **bpp)
279{
280	xfs_buf_t	 *bp;
281	int		 error;
282
283	if (flags)
284		bp = xfs_buf_read_flags(target, blkno, len, flags);
285	else
286		bp = xfs_buf_read(target, blkno, len, flags);
287	if (!bp)
288		return XFS_ERROR(EIO);
289	error = XFS_BUF_GETERROR(bp);
290	if (bp && !error && !XFS_FORCED_SHUTDOWN(mp)) {
291		*bpp = bp;
292	} else {
293		*bpp = NULL;
294		if (error) {
295			xfs_ioerror_alert("xfs_read_buf", mp, bp, XFS_BUF_ADDR(bp));
296		} else {
297			error = XFS_ERROR(EIO);
298		}
299		if (bp) {
300			XFS_BUF_UNDONE(bp);
301			XFS_BUF_UNDELAYWRITE(bp);
302			XFS_BUF_STALE(bp);
303			/*
304			 * brelse clears B_ERROR and b_error
305			 */
306			xfs_buf_relse(bp);
307		}
308	}
309	return (error);
310}
311
312/*
313 * Wrapper around bwrite() so that we can trap
314 * write errors, and act accordingly.
315 */
316int
317xfs_bwrite(
318	struct xfs_mount *mp,
319	struct xfs_buf	 *bp)
320{
321	int	error;
322
323	/*
324	 * XXXsup how does this work for quotas.
325	 */
326	XFS_BUF_SET_BDSTRAT_FUNC(bp, xfs_bdstrat_cb);
327	XFS_BUF_SET_FSPRIVATE3(bp, mp);
328	XFS_BUF_WRITE(bp);
329
330	if ((error = XFS_bwrite(bp))) {
331		ASSERT(mp);
332		/*
333		 * Cannot put a buftrace here since if the buffer is not
334		 * B_HOLD then we will brelse() the buffer before returning
335		 * from bwrite and we could be tracing a buffer that has
336		 * been reused.
337		 */
338		xfs_force_shutdown(mp, XFS_METADATA_IO_ERROR);
339	}
340	return (error);
341}
342