[linux-2.6-block.git] / fs / ocfs2 / buffer_head_io.c

// SPDX-License-Identifier: GPL-2.0-or-later
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * io.c
 *
 * Buffer cache handling
 *
 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 */

#include <linux/fs.h>
#include <linux/types.h>
#include <linux/highmem.h>
#include <linux/bio.h>

#include <cluster/masklog.h>

#include "ocfs2.h"

#include "alloc.h"
#include "inode.h"
#include "journal.h"
#include "uptodate.h"
#include "buffer_head_io.h"
#include "ocfs2_trace.h"

/*
 * Bits on bh->b_state used by ocfs2.
 *
 * These MUST be after the JBD2 bits.  Hence, we use BH_JBDPrivateStart.
 */
enum ocfs2_state_bits {
	BH_NeedsValidate = BH_JBDPrivateStart,
};

/* Expand the magic b_state functions */
BUFFER_FNS(NeedsValidate, needs_validate);

int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh,
		      struct ocfs2_caching_info *ci)
{
	int ret = 0;

	trace_ocfs2_write_block((unsigned long long)bh->b_blocknr, ci);

	BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
	BUG_ON(buffer_jbd(bh));

	/* No need to check for a soft readonly file system here. non
	 * journalled writes are only ever done on system files which
	 * can get modified during recovery even if read-only. */
	if (ocfs2_is_hard_readonly(osb)) {
		ret = -EROFS;
		mlog_errno(ret);
		goto out;
	}

	ocfs2_metadata_cache_io_lock(ci);

	lock_buffer(bh);
	set_buffer_uptodate(bh);

	/* remove from dirty list before I/O. */
	clear_buffer_dirty(bh);

	get_bh(bh); /* for end_buffer_write_sync() */
	bh->b_end_io = end_buffer_write_sync;
	submit_bh(REQ_OP_WRITE, 0, bh);

	wait_on_buffer(bh);

	if (buffer_uptodate(bh)) {
		ocfs2_set_buffer_uptodate(ci, bh);
	} else {
		/* We don't need to remove the clustered uptodate
		 * information for this bh as it's not marked locally
		 * uptodate. */
		ret = -EIO;
		mlog_errno(ret);
	}

	ocfs2_metadata_cache_io_unlock(ci);
out:
	return ret;
}

/* Caller must provide a bhs[] with all NULL or non-NULL entries, so it
 * will be easier to handle read failure.
 */
int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block,
			   unsigned int nr, struct buffer_head *bhs[])
{
	int status = 0;
	unsigned int i;
	struct buffer_head *bh;
	int new_bh = 0;

	trace_ocfs2_read_blocks_sync((unsigned long long)block, nr);

	if (!nr)
		goto bail;

	/* Don't put buffer head and re-assign it to NULL if it is allocated
	 * outside since the caller can't be aware of this alternation!
	 */
	new_bh = (bhs[0] == NULL);

	for (i = 0 ; i < nr ; i++) {
		if (bhs[i] == NULL) {
			bhs[i] = sb_getblk(osb->sb, block++);
			if (bhs[i] == NULL) {
				status = -ENOMEM;
				mlog_errno(status);
				break;
			}
		}
		bh = bhs[i];

		if (buffer_jbd(bh)) {
			trace_ocfs2_read_blocks_sync_jbd(
					(unsigned long long)bh->b_blocknr);
			continue;
		}

		if (buffer_dirty(bh)) {
			/* This should probably be a BUG, or
			 * at least return an error. */
			mlog(ML_ERROR,
			     "trying to sync read a dirty "
			     "buffer! (blocknr = %llu), skipping\n",
			     (unsigned long long)bh->b_blocknr);
			continue;
		}

		lock_buffer(bh);
		if (buffer_jbd(bh)) {
#ifdef CATCH_BH_JBD_RACES
			mlog(ML_ERROR,
			     "block %llu had the JBD bit set "
			     "while I was in lock_buffer!",
			     (unsigned long long)bh->b_blocknr);
			BUG();
#else
			unlock_buffer(bh);
			continue;
#endif
		}

		get_bh(bh); /* for end_buffer_read_sync() */
		bh->b_end_io = end_buffer_read_sync;
		submit_bh(REQ_OP_READ, 0, bh);
	}

read_failure:
	for (i = nr; i > 0; i--) {
		bh = bhs[i - 1];

		if (unlikely(status)) {
			if (new_bh && bh) {
				/* If middle bh fails, let previous bh
				 * finish its read and then put it to
				 * aovoid bh leak
				 */
				if (!buffer_jbd(bh))
					wait_on_buffer(bh);
				put_bh(bh);
				bhs[i - 1] = NULL;
			} else if (bh && buffer_uptodate(bh)) {
				clear_buffer_uptodate(bh);
			}
			continue;
		}

		/* No need to wait on the buffer if it's managed by JBD. */
		if (!buffer_jbd(bh))
			wait_on_buffer(bh);

		if (!buffer_uptodate(bh)) {
			/* Status won't be cleared from here on out,
			 * so we can safely record this and loop back
			 * to cleanup the other buffers. */
			status = -EIO;
			goto read_failure;
		}
	}

bail:
	return status;
}

/* Caller must provide a bhs[] with all NULL or non-NULL entries, so it
 * will be easier to handle read failure.
 */
int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr,
		      struct buffer_head *bhs[], int flags,
		      int (*validate)(struct super_block *sb,
				      struct buffer_head *bh))
{
	int status = 0;
	int i, ignore_cache = 0;
	struct buffer_head *bh;
	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
	int new_bh = 0;

	trace_ocfs2_read_blocks_begin(ci, (unsigned long long)block, nr, flags);

	BUG_ON(!ci);
	BUG_ON((flags & OCFS2_BH_READAHEAD) &&
	       (flags & OCFS2_BH_IGNORE_CACHE));

	if (bhs == NULL) {
		status = -EINVAL;
		mlog_errno(status);
		goto bail;
	}

	if (nr < 0) {
		mlog(ML_ERROR, "asked to read %d blocks!\n", nr);
		status = -EINVAL;
		mlog_errno(status);
		goto bail;
	}

	if (nr == 0) {
		status = 0;
		goto bail;
	}

	/* Don't put buffer head and re-assign it to NULL if it is allocated
	 * outside since the caller can't be aware of this alternation!
	 */
	new_bh = (bhs[0] == NULL);

	ocfs2_metadata_cache_io_lock(ci);
	for (i = 0 ; i < nr ; i++) {
		if (bhs[i] == NULL) {
			bhs[i] = sb_getblk(sb, block++);
			if (bhs[i] == NULL) {
				ocfs2_metadata_cache_io_unlock(ci);
				status = -ENOMEM;
				mlog_errno(status);
				/* Don't forget to put previous bh! */
				break;
			}
		}
		bh = bhs[i];
		ignore_cache = (flags & OCFS2_BH_IGNORE_CACHE);

		/* There are three read-ahead cases here which we need to
		 * be concerned with. All three assume a buffer has
		 * previously been submitted with OCFS2_BH_READAHEAD
		 * and it hasn't yet completed I/O.
		 *
		 * 1) The current request is sync to disk. This rarely
		 *    happens these days, and never when performance
		 *    matters - the code can just wait on the buffer
		 *    lock and re-submit.
		 *
		 * 2) The current request is cached, but not
		 *    readahead. ocfs2_buffer_uptodate() will return
		 *    false anyway, so we'll wind up waiting on the
		 *    buffer lock to do I/O. We re-check the request
		 *    with after getting the lock to avoid a re-submit.
		 *
		 * 3) The current request is readahead (and so must
		 *    also be a caching one). We short circuit if the
		 *    buffer is locked (under I/O) and if it's in the
		 *    uptodate cache. The re-check from #2 catches the
		 *    case that the previous read-ahead completes just
		 *    before our is-it-in-flight check.
		 */

		if (!ignore_cache && !ocfs2_buffer_uptodate(ci, bh)) {
			trace_ocfs2_read_blocks_from_disk(
			     (unsigned long long)bh->b_blocknr,
			     (unsigned long long)ocfs2_metadata_cache_owner(ci));
			/* We're using ignore_cache here to say
			 * "go to disk" */
			ignore_cache = 1;
		}

		trace_ocfs2_read_blocks_bh((unsigned long long)bh->b_blocknr,
			ignore_cache, buffer_jbd(bh), buffer_dirty(bh));

		if (buffer_jbd(bh)) {
			continue;
		}

		if (ignore_cache) {
			if (buffer_dirty(bh)) {
				/* This should probably be a BUG, or
				 * at least return an error. */
				continue;
			}

			/* A read-ahead request was made - if the
			 * buffer is already under read-ahead from a
			 * previously submitted request than we are
			 * done here. */
			if ((flags & OCFS2_BH_READAHEAD)
			    && ocfs2_buffer_read_ahead(ci, bh))
				continue;

			lock_buffer(bh);
			if (buffer_jbd(bh)) {
#ifdef CATCH_BH_JBD_RACES
				mlog(ML_ERROR, "block %llu had the JBD bit set "
					       "while I was in lock_buffer!",
				     (unsigned long long)bh->b_blocknr);
				BUG();
#else
				unlock_buffer(bh);
				continue;
#endif
			}

			/* Re-check ocfs2_buffer_uptodate() as a
			 * previously read-ahead buffer may have
			 * completed I/O while we were waiting for the
			 * buffer lock. */
			if (!(flags & OCFS2_BH_IGNORE_CACHE)
			    && !(flags & OCFS2_BH_READAHEAD)
			    && ocfs2_buffer_uptodate(ci, bh)) {
				unlock_buffer(bh);
				continue;
			}

			get_bh(bh); /* for end_buffer_read_sync() */
			if (validate)
				set_buffer_needs_validate(bh);
			bh->b_end_io = end_buffer_read_sync;
			submit_bh(REQ_OP_READ, 0, bh);
			continue;
		}
	}

read_failure:
	for (i = (nr - 1); i >= 0; i--) {
		bh = bhs[i];

		if (!(flags & OCFS2_BH_READAHEAD)) {
			if (unlikely(status)) {
				/* Clear the buffers on error including those
				 * ever succeeded in reading
				 */
				if (new_bh && bh) {
					/* If middle bh fails, let previous bh
					 * finish its read and then put it to
					 * aovoid bh leak
					 */
					if (!buffer_jbd(bh))
						wait_on_buffer(bh);
					put_bh(bh);
					bhs[i] = NULL;
				} else if (bh && buffer_uptodate(bh)) {
					clear_buffer_uptodate(bh);
				}
				continue;
			}
			/* We know this can't have changed as we hold the
			 * owner sem. Avoid doing any work on the bh if the
			 * journal has it. */
			if (!buffer_jbd(bh))
				wait_on_buffer(bh);

			if (!buffer_uptodate(bh)) {
				/* Status won't be cleared from here on out,
				 * so we can safely record this and loop back
				 * to cleanup the other buffers. Don't need to
				 * remove the clustered uptodate information
				 * for this bh as it's not marked locally
				 * uptodate. */
				status = -EIO;
				clear_buffer_needs_validate(bh);
				goto read_failure;
			}

			if (buffer_needs_validate(bh)) {
				/* We never set NeedsValidate if the
				 * buffer was held by the journal, so
				 * that better not have changed */
				BUG_ON(buffer_jbd(bh));
				clear_buffer_needs_validate(bh);
				status = validate(sb, bh);
				if (status)
					goto read_failure;
			}
		}

		/* Always set the buffer in the cache, even if it was
		 * a forced read, or read-ahead which hasn't yet
		 * completed. */
		ocfs2_set_buffer_uptodate(ci, bh);
	}
	ocfs2_metadata_cache_io_unlock(ci);

	trace_ocfs2_read_blocks_end((unsigned long long)block, nr,
				    flags, ignore_cache);

bail:

	return status;
}

/* Check whether the blkno is the super block or one of the backups. */
static void ocfs2_check_super_or_backup(struct super_block *sb,
					sector_t blkno)
{
	int i;
	u64 backup_blkno;

	if (blkno == OCFS2_SUPER_BLOCK_BLKNO)
		return;

	for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
		backup_blkno = ocfs2_backup_super_blkno(sb, i);
		if (backup_blkno == blkno)
			return;
	}

	BUG();
}

/*
 * Write super block and backups doesn't need to collaborate with journal,
 * so we don't need to lock ip_io_mutex and ci doesn't need to bea passed
 * into this function.
 */
int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
				struct buffer_head *bh)
{
	int ret = 0;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;

	BUG_ON(buffer_jbd(bh));
	ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);

	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) {
		ret = -EROFS;
		mlog_errno(ret);
		goto out;
	}

	lock_buffer(bh);
	set_buffer_uptodate(bh);

	/* remove from dirty list before I/O. */
	clear_buffer_dirty(bh);

	get_bh(bh); /* for end_buffer_write_sync() */
	bh->b_end_io = end_buffer_write_sync;
	ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &di->i_check);
	submit_bh(REQ_OP_WRITE, 0, bh);

	wait_on_buffer(bh);

	if (!buffer_uptodate(bh)) {
		ret = -EIO;
		mlog_errno(ret);
	}

out:
	return ret;
}
Commit	Line	Data
328970de	1	// SPDX-License-Identifier: GPL-2.0-or-later
ccd979bd MF	2	/* -- mode: c; c-basic-offset: 8; --
	3	* vim: noexpandtab sw=8 ts=8 sts=0:
	4	*
	5	* io.c
	6	*
	7	* Buffer cache handling
	8	*
	9	* Copyright (C) 2002, 2004 Oracle. All rights reserved.
ccd979bd MF	10	*/
	11
	12	#include <linux/fs.h>
	13	#include <linux/types.h>
ccd979bd	14	#include <linux/highmem.h>
2f8b5444	15	#include <linux/bio.h>
ccd979bd MF	16
	17	#include <cluster/masklog.h>
	18
	19	#include "ocfs2.h"
	20
	21	#include "alloc.h"
	22	#include "inode.h"
	23	#include "journal.h"
	24	#include "uptodate.h"
ccd979bd	25	#include "buffer_head_io.h"
15057e98	26	#include "ocfs2_trace.h"
ccd979bd	27
970e4936 JB	28	/*
	29	* Bits on bh->b_state used by ocfs2.
	30	*
b86c86fa	31	* These MUST be after the JBD2 bits. Hence, we use BH_JBDPrivateStart.
970e4936 JB	32	*/
970e4936 JB	33	enum ocfs2_state_bits {
b86c86fa	34	BH_NeedsValidate = BH_JBDPrivateStart,
970e4936 JB	35	};
	36
	37	/* Expand the magic b_state functions */
	38	BUFFER_FNS(NeedsValidate, needs_validate);
	39
ccd979bd	40	int ocfs2_write_block(struct ocfs2_super osb, struct buffer_head bh,
8cb471e8	41	struct ocfs2_caching_info *ci)
ccd979bd MF	42	{
	43	int ret = 0;
	44
15057e98	45	trace_ocfs2_write_block((unsigned long long)bh->b_blocknr, ci);
ccd979bd MF	46
	47	BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
	48	BUG_ON(buffer_jbd(bh));
	49
	50	/* No need to check for a soft readonly file system here. non
	51	* journalled writes are only ever done on system files which
	52	* can get modified during recovery even if read-only. */
	53	if (ocfs2_is_hard_readonly(osb)) {
	54	ret = -EROFS;
c1e8d35e	55	mlog_errno(ret);
ccd979bd MF	56	goto out;
	57	}
	58
8cb471e8	59	ocfs2_metadata_cache_io_lock(ci);
ccd979bd MF	60
	61	lock_buffer(bh);
	62	set_buffer_uptodate(bh);
	63
	64	/* remove from dirty list before I/O. */
	65	clear_buffer_dirty(bh);
	66
da1e9098	67	get_bh(bh); /* for end_buffer_write_sync() */
ccd979bd	68	bh->b_end_io = end_buffer_write_sync;
2a222ca9	69	submit_bh(REQ_OP_WRITE, 0, bh);
ccd979bd MF	70
	71	wait_on_buffer(bh);
	72
	73	if (buffer_uptodate(bh)) {
8cb471e8	74	ocfs2_set_buffer_uptodate(ci, bh);
ccd979bd MF	75	} else {
	76	/* We don't need to remove the clustered uptodate
	77	* information for this bh as it's not marked locally
	78	* uptodate. */
	79	ret = -EIO;
c1e8d35e	80	mlog_errno(ret);
ccd979bd MF	81	}
ccd979bd MF	82
8cb471e8	83	ocfs2_metadata_cache_io_unlock(ci);
ccd979bd	84	out:
ccd979bd MF	85	return ret;
	86	}
	87
cf76c785 CG	88	/* Caller must provide a bhs[] with all NULL or non-NULL entries, so it
	89	* will be easier to handle read failure.
	90	*/
da1e9098 JB	91	int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block,
	92	unsigned int nr, struct buffer_head *bhs[])
	93	{
	94	int status = 0;
	95	unsigned int i;
	96	struct buffer_head *bh;
cf76c785	97	int new_bh = 0;
da1e9098	98
15057e98 TM	99	trace_ocfs2_read_blocks_sync((unsigned long long)block, nr);
	100
	101	if (!nr)
da1e9098	102	goto bail;
da1e9098	103
cf76c785 CG	104	/* Don't put buffer head and re-assign it to NULL if it is allocated
	105	* outside since the caller can't be aware of this alternation!
	106	*/
	107	new_bh = (bhs[0] == NULL);
	108
da1e9098 JB	109	for (i = 0 ; i < nr ; i++) {
	110	if (bhs[i] == NULL) {
	111	bhs[i] = sb_getblk(osb->sb, block++);
	112	if (bhs[i] == NULL) {
7391a294	113	status = -ENOMEM;
da1e9098	114	mlog_errno(status);
cf76c785	115	break;
da1e9098 JB	116	}
	117	}
	118	bh = bhs[i];
	119
	120	if (buffer_jbd(bh)) {
15057e98 TM	121	trace_ocfs2_read_blocks_sync_jbd(
15057e98 TM	122	(unsigned long long)bh->b_blocknr);
da1e9098 JB	123	continue;
	124	}
	125
	126	if (buffer_dirty(bh)) {
	127	/* This should probably be a BUG, or
	128	* at least return an error. */
	129	mlog(ML_ERROR,
	130	"trying to sync read a dirty "
	131	"buffer! (blocknr = %llu), skipping\n",
	132	(unsigned long long)bh->b_blocknr);
	133	continue;
	134	}
	135
	136	lock_buffer(bh);
	137	if (buffer_jbd(bh)) {
7186ee06	138	#ifdef CATCH_BH_JBD_RACES
da1e9098 JB	139	mlog(ML_ERROR,
	140	"block %llu had the JBD bit set "
	141	"while I was in lock_buffer!",
	142	(unsigned long long)bh->b_blocknr);
	143	BUG();
7186ee06 GH	144	#else
	145	unlock_buffer(bh);
	146	continue;
	147	#endif
da1e9098 JB	148	}
da1e9098 JB	149
da1e9098 JB	150	get_bh(bh); /* for end_buffer_read_sync() */
da1e9098 JB	151	bh->b_end_io = end_buffer_read_sync;
2a222ca9	152	submit_bh(REQ_OP_READ, 0, bh);
da1e9098 JB	153	}
da1e9098 JB	154
cf76c785	155	read_failure:
da1e9098 JB	156	for (i = nr; i > 0; i--) {
	157	bh = bhs[i - 1];
	158
cf76c785 CG	159	if (unlikely(status)) {
	160	if (new_bh && bh) {
	161	/* If middle bh fails, let previous bh
	162	* finish its read and then put it to
	163	* aovoid bh leak
	164	*/
	165	if (!buffer_jbd(bh))
	166	wait_on_buffer(bh);
	167	put_bh(bh);
	168	bhs[i - 1] = NULL;
	169	} else if (bh && buffer_uptodate(bh)) {
	170	clear_buffer_uptodate(bh);
	171	}
	172	continue;
	173	}
	174
d6b58f89 MF	175	/* No need to wait on the buffer if it's managed by JBD. */
	176	if (!buffer_jbd(bh))
	177	wait_on_buffer(bh);
da1e9098	178
da1e9098 JB	179	if (!buffer_uptodate(bh)) {
	180	/* Status won't be cleared from here on out,
	181	* so we can safely record this and loop back
	182	* to cleanup the other buffers. */
	183	status = -EIO;
cf76c785	184	goto read_failure;
da1e9098 JB	185	}
	186	}
	187
	188	bail:
	189	return status;
	190	}
	191
cf76c785 CG	192	/* Caller must provide a bhs[] with all NULL or non-NULL entries, so it
	193	* will be easier to handle read failure.
	194	*/
8cb471e8	195	int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr,
970e4936 JB	196	struct buffer_head *bhs[], int flags,
	197	int (validate)(struct super_block sb,
	198	struct buffer_head *bh))
ccd979bd MF	199	{
ccd979bd MF	200	int status = 0;
ccd979bd MF	201	int i, ignore_cache = 0;
ccd979bd MF	202	struct buffer_head *bh;
8cb471e8	203	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
cf76c785	204	int new_bh = 0;
ccd979bd	205
15057e98	206	trace_ocfs2_read_blocks_begin(ci, (unsigned long long)block, nr, flags);
ccd979bd	207
8cb471e8	208	BUG_ON(!ci);
d4a8c93c JB	209	BUG_ON((flags & OCFS2_BH_READAHEAD) &&
d4a8c93c JB	210	(flags & OCFS2_BH_IGNORE_CACHE));
aa958874	211
31d33073	212	if (bhs == NULL) {
ccd979bd MF	213	status = -EINVAL;
	214	mlog_errno(status);
	215	goto bail;
	216	}
	217
	218	if (nr < 0) {
	219	mlog(ML_ERROR, "asked to read %d blocks!\n", nr);
	220	status = -EINVAL;
	221	mlog_errno(status);
	222	goto bail;
	223	}
	224
	225	if (nr == 0) {
ccd979bd MF	226	status = 0;
	227	goto bail;
	228	}
	229
cf76c785 CG	230	/* Don't put buffer head and re-assign it to NULL if it is allocated
	231	* outside since the caller can't be aware of this alternation!
	232	*/
	233	new_bh = (bhs[0] == NULL);
	234
8cb471e8	235	ocfs2_metadata_cache_io_lock(ci);
ccd979bd MF	236	for (i = 0 ; i < nr ; i++) {
ccd979bd MF	237	if (bhs[i] == NULL) {
8cb471e8	238	bhs[i] = sb_getblk(sb, block++);
ccd979bd	239	if (bhs[i] == NULL) {
8cb471e8	240	ocfs2_metadata_cache_io_unlock(ci);
7391a294	241	status = -ENOMEM;
ccd979bd	242	mlog_errno(status);
cf76c785 CG	243	/* Don't forget to put previous bh! */
cf76c785 CG	244	break;
ccd979bd MF	245	}
	246	}
	247	bh = bhs[i];
d4a8c93c	248	ignore_cache = (flags & OCFS2_BH_IGNORE_CACHE);
ccd979bd	249
aa958874 MF	250	/* There are three read-ahead cases here which we need to
	251	* be concerned with. All three assume a buffer has
	252	* previously been submitted with OCFS2_BH_READAHEAD
	253	* and it hasn't yet completed I/O.
	254	*
	255	* 1) The current request is sync to disk. This rarely
	256	* happens these days, and never when performance
	257	* matters - the code can just wait on the buffer
	258	* lock and re-submit.
	259	*
	260	* 2) The current request is cached, but not
	261	* readahead. ocfs2_buffer_uptodate() will return
	262	* false anyway, so we'll wind up waiting on the
	263	* buffer lock to do I/O. We re-check the request
	264	* with after getting the lock to avoid a re-submit.
	265	*
	266	* 3) The current request is readahead (and so must
	267	* also be a caching one). We short circuit if the
	268	* buffer is locked (under I/O) and if it's in the
	269	* uptodate cache. The re-check from #2 catches the
	270	* case that the previous read-ahead completes just
	271	* before our is-it-in-flight check.
	272	*/
	273
8cb471e8	274	if (!ignore_cache && !ocfs2_buffer_uptodate(ci, bh)) {
d701485a	275	trace_ocfs2_read_blocks_from_disk(
ccd979bd	276	(unsigned long long)bh->b_blocknr,
8cb471e8	277	(unsigned long long)ocfs2_metadata_cache_owner(ci));
d4a8c93c JB	278	/* We're using ignore_cache here to say
d4a8c93c JB	279	* "go to disk" */
ccd979bd MF	280	ignore_cache = 1;
	281	}
	282
15057e98 TM	283	trace_ocfs2_read_blocks_bh((unsigned long long)bh->b_blocknr,
	284	ignore_cache, buffer_jbd(bh), buffer_dirty(bh));
	285
ccd979bd	286	if (buffer_jbd(bh)) {
ccd979bd MF	287	continue;
	288	}
	289
d4a8c93c	290	if (ignore_cache) {
ccd979bd MF	291	if (buffer_dirty(bh)) {
	292	/* This should probably be a BUG, or
	293	* at least return an error. */
ccd979bd MF	294	continue;
	295	}
	296
aa958874 MF	297	/* A read-ahead request was made - if the
	298	* buffer is already under read-ahead from a
	299	* previously submitted request than we are
	300	* done here. */
	301	if ((flags & OCFS2_BH_READAHEAD)
8cb471e8	302	&& ocfs2_buffer_read_ahead(ci, bh))
aa958874 MF	303	continue;
aa958874 MF	304
ccd979bd MF	305	lock_buffer(bh);
	306	if (buffer_jbd(bh)) {
	307	#ifdef CATCH_BH_JBD_RACES
	308	mlog(ML_ERROR, "block %llu had the JBD bit set "
	309	"while I was in lock_buffer!",
	310	(unsigned long long)bh->b_blocknr);
	311	BUG();
	312	#else
	313	unlock_buffer(bh);
	314	continue;
	315	#endif
	316	}
aa958874 MF	317
	318	/* Re-check ocfs2_buffer_uptodate() as a
	319	* previously read-ahead buffer may have
	320	* completed I/O while we were waiting for the
	321	* buffer lock. */
d4a8c93c	322	if (!(flags & OCFS2_BH_IGNORE_CACHE)
aa958874	323	&& !(flags & OCFS2_BH_READAHEAD)
8cb471e8	324	&& ocfs2_buffer_uptodate(ci, bh)) {
aa958874 MF	325	unlock_buffer(bh);
	326	continue;
	327	}
	328
ccd979bd	329	get_bh(bh); /* for end_buffer_read_sync() */
970e4936 JB	330	if (validate)
970e4936 JB	331	set_buffer_needs_validate(bh);
ccd979bd	332	bh->b_end_io = end_buffer_read_sync;
2a222ca9	333	submit_bh(REQ_OP_READ, 0, bh);
ccd979bd MF	334	continue;
	335	}
	336	}
	337
cf76c785	338	read_failure:
ccd979bd MF	339	for (i = (nr - 1); i >= 0; i--) {
	340	bh = bhs[i];
	341
aa958874	342	if (!(flags & OCFS2_BH_READAHEAD)) {
cf76c785 CG	343	if (unlikely(status)) {
	344	/* Clear the buffers on error including those
	345	* ever succeeded in reading
	346	*/
	347	if (new_bh && bh) {
	348	/* If middle bh fails, let previous bh
	349	* finish its read and then put it to
	350	* aovoid bh leak
	351	*/
	352	if (!buffer_jbd(bh))
	353	wait_on_buffer(bh);
	354	put_bh(bh);
	355	bhs[i] = NULL;
	356	} else if (bh && buffer_uptodate(bh)) {
	357	clear_buffer_uptodate(bh);
	358	}
34237681 GR	359	continue;
34237681 GR	360	}
aa958874	361	/* We know this can't have changed as we hold the
8cb471e8	362	* owner sem. Avoid doing any work on the bh if the
aa958874 MF	363	* journal has it. */
	364	if (!buffer_jbd(bh))
	365	wait_on_buffer(bh);
	366
	367	if (!buffer_uptodate(bh)) {
	368	/* Status won't be cleared from here on out,
	369	* so we can safely record this and loop back
	370	* to cleanup the other buffers. Don't need to
	371	* remove the clustered uptodate information
	372	* for this bh as it's not marked locally
	373	* uptodate. */
	374	status = -EIO;
234b69e3	375	clear_buffer_needs_validate(bh);
cf76c785	376	goto read_failure;
aa958874	377	}
970e4936 JB	378
	379	if (buffer_needs_validate(bh)) {
	380	/* We never set NeedsValidate if the
	381	* buffer was held by the journal, so
	382	* that better not have changed */
	383	BUG_ON(buffer_jbd(bh));
	384	clear_buffer_needs_validate(bh);
8cb471e8	385	status = validate(sb, bh);
cf76c785 CG	386	if (status)
cf76c785 CG	387	goto read_failure;
970e4936	388	}
ccd979bd MF	389	}
ccd979bd MF	390
aa958874 MF	391	/* Always set the buffer in the cache, even if it was
	392	* a forced read, or read-ahead which hasn't yet
	393	* completed. */
8cb471e8	394	ocfs2_set_buffer_uptodate(ci, bh);
ccd979bd	395	}
8cb471e8	396	ocfs2_metadata_cache_io_unlock(ci);
ccd979bd	397
15057e98 TM	398	trace_ocfs2_read_blocks_end((unsigned long long)block, nr,
15057e98 TM	399	flags, ignore_cache);
ccd979bd MF	400
	401	bail:
	402
ccd979bd MF	403	return status;
ccd979bd MF	404	}
d659072f TM	405
	406	/* Check whether the blkno is the super block or one of the backups. */
	407	static void ocfs2_check_super_or_backup(struct super_block *sb,
	408	sector_t blkno)
	409	{
	410	int i;
	411	u64 backup_blkno;
	412
	413	if (blkno == OCFS2_SUPER_BLOCK_BLKNO)
	414	return;
	415
	416	for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
	417	backup_blkno = ocfs2_backup_super_blkno(sb, i);
	418	if (backup_blkno == blkno)
	419	return;
	420	}
	421
	422	BUG();
	423	}
	424
	425	/*
	426	* Write super block and backups doesn't need to collaborate with journal,
8cb471e8	427	* so we don't need to lock ip_io_mutex and ci doesn't need to bea passed
d659072f TM	428	* into this function.
	429	*/
	430	int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
	431	struct buffer_head *bh)
	432	{
	433	int ret = 0;
a42ab8e1	434	struct ocfs2_dinode di = (struct ocfs2_dinode )bh->b_data;
d659072f	435
d659072f TM	436	BUG_ON(buffer_jbd(bh));
	437	ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);
	438
	439	if (ocfs2_is_hard_readonly(osb) \|\| ocfs2_is_soft_readonly(osb)) {
	440	ret = -EROFS;
c1e8d35e	441	mlog_errno(ret);
d659072f TM	442	goto out;
	443	}
	444
	445	lock_buffer(bh);
	446	set_buffer_uptodate(bh);
	447
	448	/* remove from dirty list before I/O. */
	449	clear_buffer_dirty(bh);
	450
	451	get_bh(bh); /* for end_buffer_write_sync() */
	452	bh->b_end_io = end_buffer_write_sync;
a42ab8e1	453	ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &di->i_check);
2a222ca9	454	submit_bh(REQ_OP_WRITE, 0, bh);
d659072f TM	455
	456	wait_on_buffer(bh);
	457
	458	if (!buffer_uptodate(bh)) {
	459	ret = -EIO;
c1e8d35e	460	mlog_errno(ret);
d659072f TM	461	}
	462
	463	out:
d659072f TM	464	return ret;
d659072f TM	465	}