Return to xfs_trans_buf.c CVS log

Up to [Development] / xfs-linux

Fix bug in xfs_trans_brelse() that went in with the
xfs_trans_binval() checkin.  We need to decrement the
refcount on the buf item before we possibly free it.

#include <sys/param.h>
#ifdef SIM
#define _KERNEL
#endif
#include <sys/buf.h>
#include <sys/sysmacros.h>
#ifdef SIM
#undef _KERNEL
#endif
#include <sys/vnode.h>
#include <sys/debug.h>
#include <sys/uuid.h>
#ifndef SIM
#include <sys/sysinfo.h>
#include <sys/kmem.h>
#include <sys/conf.h>
#include <sys/user.h>
#include <sys/systm.h>
#endif
#include "xfs_types.h"
#include "xfs_inum.h"
#include "xfs_log.h"
#include "xfs_trans.h"
#include "xfs_buf_item.h"
#include "xfs_bio.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_mount.h"
#include "xfs_trans_priv.h"

#ifdef SIM
#include "sim.h"
#endif

/*
 * Get and lock the buffer for the caller if it is not already
 * locked within the given transaction.  If it is already locked
 * within the transaction, just increment its lock recursion count
 * and return a pointer to it.
 *
 * Use the buffer cache routine incore_match() to find the buffer
 * if it is already owned by this transaction.
 *
 * If we don't already own the buffer, use xfs_get_buf() to get it.
 * If it doesn't yet have an associated xfs_buf_log_item structure,
 * then allocate one and add the item to this transaction.
 *
 * If the transaction pointer is NULL, make this just a normal
 * get_buf() call.
 */
buf_t *
xfs_trans_get_buf(xfs_trans_t	*tp,
		  dev_t		dev,
		  daddr_t	blkno,
		  int		len,
		  uint		flags)
{
	buf_t			*bp;
	xfs_buf_log_item_t	*bip;
	int			s;

	/*
	 * Default to a normal get_buf() call if the tp is NULL.
	 * Always specify the BUF_BUSY flag so that get_buf() does
	 * not try to push out dirty buffers.  This keeps us from
	 * running out of stack space due to recursive calls into
	 * the buffer cache.
	 */
	if (tp == NULL) {
		return (get_buf(dev, blkno, len, flags | BUF_BUSY));
	}

	/*
	 * If we find the buffer in the cache with this transaction
	 * pointer in its b_fsprivate2 field, then we know we already
	 * have it locked.  In this case we just increment the lock
	 * recursion count and return the buffer to the caller.
	 */
	if ((bp = incore_match(dev, blkno, len, BUF_FSPRIV2, tp)) != NULL) {
		bip = (xfs_buf_log_item_t*)bp->b_fsprivate;
		ASSERT(bip != NULL);
		bip->bli_recur++;
		return (bp);
	}

	/*
	 * We always specify the BUF_BUSY flag within a transaction so
	 * that get_buf does not try to push out a delayed write buffer
	 * which might cause another transaction to take place (if the
	 * buffer was delayed alloc).  Such recursive transactions can
	 * easily deadlock with our current transaction as well as cause
	 * us to run out of stack space.
	 */
	bp = xfs_get_buf(dev, blkno, len, flags | BUF_BUSY);
	if (bp == NULL) {
		return NULL;
	}

	/*
	 * The xfs_buf_log_item pointer is stored in b_fsprivate.  If
	 * it doesn't have one yet, then allocate one and initialize it.
	 * The checks to see if one is there are in xfs_buf_item_init().
	 */
	xfs_buf_item_init(bp, tp->t_mountp);

	/*
	 * Set the recursion count for the buffer within this transaction
	 * to 0.
	 */
	bip = (xfs_buf_log_item_t*)bp->b_fsprivate;
	bip->bli_recur = 0;

	/*
	 * Take a reference for this transaction on the buf item.
	 */
	s = splockspl(xfs_bli_reflock, splhi);
	bip->bli_refcount++;
	spunlockspl(xfs_bli_reflock, s);

	/*
	 * Get a log_item_desc to point at the new item.
	 */
	(void) xfs_trans_add_item(tp, (xfs_log_item_t*)bip);

	/*
	 * Initialize b_fsprivate2 so we can find it with incore_match()
	 * above.
	 */
	bp->b_fsprivate2 = tp;

	return (bp);
}

/*
 * Get and lock the superblock buffer of this file system for the
 * given transaction.
 *
 * We don't need to use incore_match() here, because the superblock
 * buffer is a private buffer which we keep a pointer to in the
 * mount structure.
 */
buf_t *
xfs_trans_getsb(xfs_trans_t *tp)
{
	buf_t			*bp;
	xfs_buf_log_item_t	*bip;
	int			s;

	/*
	 * Default to just trying to lock the superblock buffer
	 * if tp is NULL.
	 */
	if (tp == NULL) {
		return (xfs_getsb(tp->t_mountp));
	}

	/*
	 * If the superblock buffer already has this transaction
	 * pointer in its b_fsprivate2 field, then we know we already
	 * have it locked.  In this case we just increment the lock
	 * recursion count and return the buffer to the caller.
	 */
	bp = tp->t_mountp->m_sb_bp;
	if (bp->b_fsprivate2 == tp) {
		bip = (xfs_buf_log_item_t*)bp->b_fsprivate;
		ASSERT(bip != NULL);
		bip->bli_recur++;
		return (bp);
	}

	bp = xfs_getsb(tp->t_mountp);

	/*
	 * The xfs_buf_log_item pointer is stored in b_fsprivate.  If
	 * it doesn't have one yet, then allocate one and initialize it.
	 * The checks to see if one is there are in xfs_buf_item_init().
	 */
	xfs_buf_item_init(bp, tp->t_mountp);

	/*
	 * Set the recursion count for the buffer within this transaction
	 * to 0.
	 */
	bip = (xfs_buf_log_item_t*)bp->b_fsprivate;
	bip->bli_recur = 0;

	/*
	 * Take a reference for this transaction on the buf item.
	 */
	s = splockspl(xfs_bli_reflock, splhi);
	bip->bli_refcount++;
	spunlockspl(xfs_bli_reflock, s);

	/*
	 * Get a log_item_desc to point at the new item.
	 */
	(void) xfs_trans_add_item(tp, (xfs_log_item_t*)bip);

	/*
	 * Initialize b_fsprivate2 so we can find it with incore_match()
	 * above.
	 */
	bp->b_fsprivate2 = tp;

	return (bp);
}


/*
 * Get and lock the buffer for the caller if it is not already
 * locked within the given transaction.  If it has not yet been
 * read in, read it from disk. If it is already locked
 * within the transaction and already read in, just increment its
 * lock recursion count and return a pointer to it.
 *
 * Use the buffer cache routine incore_match() to find the buffer
 * if it is already owned by this transaction.
 *
 * If we don't already own the buffer, use xfs_read_buf() to get it.
 * If it doesn't yet have an associated xfs_buf_log_item structure,
 * then allocate one and add the item to this transaction.
 *
 * If the transaction pointer is NULL, make this just a normal
 * read_buf() call.
 */
buf_t *
xfs_trans_read_buf(xfs_trans_t	*tp,
		   dev_t	dev,
		   daddr_t	blkno,
		   int		len,
		   uint		flags)
{
	buf_t			*bp;
	xfs_buf_log_item_t	*bip;
	int			s;

	/*
	 * Default to a normal get_buf() call if the tp is NULL.
	 * Always specify the BUF_BUSY flag so that get_buf() does
	 * not try to push out dirty buffers.  This keeps us from
	 * running out of stack space due to recursive calls into
	 * the buffer cache.
	 */
	if (tp == NULL) {
		return (read_buf(dev, blkno, len, flags | BUF_BUSY));
	}

	/*
	 * If we find the buffer in the cache with this transaction
	 * pointer in its b_fsprivate2 field, then we know we already
	 * have it locked.  If it is already read in we just increment
	 * the lock recursion count and return the buffer to the caller.
	 * If the buffer is not yet read in, then we read it in, increment
	 * the lock recursion count, and return it to the caller.
	 */
	if ((bp = incore_match(dev, blkno, len, BUF_FSPRIV2, tp)) != NULL) {
		ASSERT(bp->b_fsprivate != NULL);
		if (!(bp->b_flags & B_DONE)) {
#ifndef SIM
			SYSINFO.lread += len;
#endif

			ASSERT(!(bp->b_flags & B_ASYNC));
			bp->b_flags |= B_READ;
			bdstrat(bmajor(dev), bp);

#ifndef SIM
			u.u_ior++;
			SYSINFO.bread += len;
#endif

			iowait(bp);
		}

		bip = (xfs_buf_log_item_t*)bp->b_fsprivate;
		bip->bli_recur++;

		return (bp);
	}

	/*
	 * We always specify the BUF_BUSY flag within a transaction so
	 * that get_buf does not try to push out a delayed write buffer
	 * which might cause another transaction to take place (if the
	 * buffer was delayed alloc).  Such recursive transactions can
	 * easily deadlock with our current transaction as well as cause
	 * us to run out of stack space.
	 */
	bp = xfs_read_buf(dev, blkno, len, flags | BUF_BUSY);
	if (bp == NULL) {
		return NULL;
	}

	/*
	 * The xfs_buf_log_item pointer is stored in b_fsprivate.  If
	 * it doesn't have one yet, then allocate one and initialize it.
	 * The checks to see if one is there are in xfs_buf_item_init().
	 */
	xfs_buf_item_init(bp, tp->t_mountp);

	/*
	 * Set the recursion count for the buffer within this transaction
	 * to 0.
	 */
	bip = (xfs_buf_log_item_t*)bp->b_fsprivate;
	bip->bli_recur = 0;

	/*
	 * Take a reference for this transaction on the buf item.
	 */
	s = splockspl(xfs_bli_reflock, splhi);
	bip->bli_refcount++;
	spunlockspl(xfs_bli_reflock, s);

	/*
	 * Get a log_item_desc to point at the new item.
	 */
	(void) xfs_trans_add_item(tp, (xfs_log_item_t*)bip);

	/*
	 * Initialize b_fsprivate2 so we can find it with incore_match()
	 * above.
	 */
	bp->b_fsprivate2 = tp;

	return (bp);
}


/*
 * Release the buffer bp which was previously acquired with one of the
 * xfs_trans_... buffer allocation routines if the buffer has not
 * been modified within this transaction.  If the buffer is modified
 * within this transaction, do decrement the recursion count but do
 * not release the buffer even if the count goes to 0.  If the buffer is not
 * modified within the transaction, decrement the recursion count and
 * release the buffer if the recursion count goes to 0.
 *
 * If the buffer is to be released and it was not modified before
 * this transaction began, then free the buf_log_item associated with it.
 *
 * If the transaction pointer is NULL, make this just a normal
 * brelse() call.
 */
void
xfs_trans_brelse(xfs_trans_t	*tp,
		 buf_t		*bp)
{
	xfs_buf_log_item_t	*bip;
	xfs_log_item_desc_t	*lidp;
	int			s;

	/*
	 * Default to a normal brelse() call if the tp is NULL.
	 */
	if (tp == NULL) {
		/*
		 * If there's a buf log item attached to the buffer,
		 * then let the AIL know that the buffer is being
		 * unlocked.
		 */
		if (bp->b_fsprivate != NULL) {
			bip = (xfs_buf_log_item_t*)bp->b_fsprivate;	
			ASSERT(bip->bli_item.li_type == XFS_LI_BUF);
			xfs_trans_unlocked_item(bip->bli_item.li_mountp,
						(xfs_log_item_t*)bip);
		}
		brelse(bp);
		return;
	}

	ASSERT(bp->b_fsprivate2 == tp);
	bip = (xfs_buf_log_item_t*)bp->b_fsprivate;	
	ASSERT(bip->bli_item.li_type == XFS_LI_BUF);
	ASSERT(!(bip->bli_flags & XFS_BLI_STALE));
	/*
	 * Find the item descriptor pointing to this buffer's
	 * log item.  It must be there.
	 */
	lidp = xfs_trans_find_item(tp, (xfs_log_item_t*)bip);
	ASSERT(lidp != NULL);

	/*
	 * If the release is just for a recursive lock,
	 * then decrement the count and return.
	 */
	if (bip->bli_recur > 0) {
		bip->bli_recur--;
		return;
	}

	/*
	 * If the buffer is dirty within this transaction, we can't
	 * release it until we commit.
	 */
	if (lidp->lid_flags & XFS_LID_DIRTY) {
		return;
	}

	/*
	 * Free up the log item descriptor tracking the released item.
	 */
	xfs_trans_free_item(tp, lidp);

	/*
	 * Clear the hold flag in the buf log item if it is set.
	 * We wouldn't want the next user of the buffer to
	 * get confused.
	 */
	if (bip->bli_flags & XFS_BLI_HOLD) {
		bip->bli_flags &= ~XFS_BLI_HOLD;
	}

	/*
	 * Drop our reference to the buf log item.
	 */
	s = splockspl(xfs_bli_reflock, splhi);
	bip->bli_refcount--;
	spunlockspl(xfs_bli_reflock, s);

	/*
	 * If the buf item is not tracking data in the log, then
	 * we must free it before releasing the buffer back to the
	 * free pool.  Before releasing the buffer to the free pool,
	 * clear the transaction pointer in b_fsprivate2 to disolve
	 * its relation to this transaction.
	 */
	if (!xfs_buf_item_dirty(bip)) {
		xfs_buf_item_relse(bp);
	}
	bp->b_fsprivate2 = NULL;

	/*
	 * If we've still got a buf log item on the buffer, then
	 * tell the AIL that the buffer is being unlocked.
	 */
	if (bp->b_fsprivate != NULL) {
		xfs_trans_unlocked_item(bip->bli_item.li_mountp,
					(xfs_log_item_t*)bip);
	}
	brelse(bp);
	return;
}

/*
 * Add the locked buffer to the transaction.
 * The buffer must be locked, and it cannot be associated with any
 * transaction.
 *
 * If the buffer does not yet have a buf log item associated with it,
 * then allocate one for it.  Then add the buf item to the transaction.
 */
void
xfs_trans_bjoin(xfs_trans_t	*tp,
		buf_t		*bp)
{
	int			s;
	xfs_buf_log_item_t	*bip;

	ASSERT(bp->b_flags & B_BUSY);
	ASSERT(bp->b_fsprivate2 == NULL);

	/*
	 * The xfs_buf_log_item pointer is stored in b_fsprivate.  If
	 * it doesn't have one yet, then allocate one and initialize it.
	 * The checks to see if one is there are in xfs_buf_item_init().
	 */
	xfs_buf_item_init(bp, tp->t_mountp);
	bip = bp->b_fsprivate;

	/*
	 * Take a reference for this transaction on the buf item.
	 */
	s = splockspl(xfs_bli_reflock, splhi);
	bip->bli_refcount++;
	spunlockspl(xfs_bli_reflock, s);

	/*
	 * Get a log_item_desc to point at the new item.
	 */
	(void) xfs_trans_add_item(tp, (xfs_log_item_t *)bip);

	/*
	 * Initialize b_fsprivate2 so we can find it with incore_match()
	 * in xfs_trans_get_buf() and friends above.
	 */
	bp->b_fsprivate2 = tp;
}

/*
 * Mark the buffer as not needing to be unlocked when the buf item's
 * IOP_UNLOCK() routine is called.  The buffer must already be locked
 * and associated with the given transaction.
 */
void
xfs_trans_bhold(xfs_trans_t	*tp,
		buf_t		*bp)
{
	xfs_buf_log_item_t	*bip;

	ASSERT(bp->b_flags & B_BUSY);
	ASSERT((xfs_trans_t*)(bp->b_fsprivate2) == tp);
	ASSERT(bp->b_fsprivate != NULL);

	bip = (xfs_buf_log_item_t*)(bp->b_fsprivate);
	bip->bli_flags |= XFS_BLI_HOLD;
}

/*
 * This function is used to indicate that the buffer should not be
 * unlocked until the transaction is committed to disk.
 *
 * It uses the log item descriptor flag XFS_LID_SYNC_UNLOCK to
 * delay the buf items's unlock call until the transaction is
 * committed to disk or aborted.
 */
void
xfs_trans_bhold_until_committed(xfs_trans_t	*tp,
				buf_t		*bp)
{
	xfs_log_item_desc_t	*lidp;
	xfs_buf_log_item_t	*bip;

	ASSERT(bp->b_flags & B_BUSY);
	ASSERT((xfs_trans_t*)(bp->b_fsprivate2) == tp);
	ASSERT(bp->b_fsprivate != NULL);

	bip = (xfs_buf_log_item_t *)(bp->b_fsprivate);
	lidp = xfs_trans_find_item(tp, (xfs_log_item_t*)bip);
	ASSERT(lidp != NULL);

	lidp->lid_flags |= XFS_LID_SYNC_UNLOCK;
}


/*
 * This is called to mark bytes first through last inclusive of the given
 * buffer as needing to be logged when the transaction is committed.
 * The buffer must already be associated with the given transaction.
 *
 * First and last are numbers relative to the beginning of this buffer,
 * so the first byte in the buffer is numbered 0 regardless of the
 * value of b_blkno.
 */
void
xfs_trans_log_buf(xfs_trans_t	*tp,
		  buf_t		*bp,
		  uint		first,
		  uint		last)
{
	xfs_buf_log_item_t	*bip;
	xfs_log_item_desc_t	*lidp;

	ASSERT(bp->b_flags & B_BUSY);
	ASSERT((xfs_trans_t*)bp->b_fsprivate2 == tp);
	ASSERT(bp->b_fsprivate != NULL);
	ASSERT((first <= last) && (last <= bp->b_bcount));
	ASSERT((bp->b_iodone == NULL) ||
	       (bp->b_iodone == xfs_buf_iodone_callbacks));

	/*
	 * Mark the buffer as needing to be written out eventually,
	 * and set its iodone function to remove the buffer's buf log
	 * item from the AIL and free it when the buffer is flushed
	 * to disk.  See xfs_buf_attach_iodone() for more details
	 * on li_cb and xfs_buf_iodone_callbacks().
	 */
	bp->b_flags |= B_DELWRI | B_DONE;
	bip = (xfs_buf_log_item_t*)bp->b_fsprivate;
	if (bp->b_iodone == NULL) {
		bp->b_iodone = xfs_buf_iodone_callbacks;
	}
	bip->bli_item.li_cb = (void(*)(buf_t*,xfs_log_item_t*))xfs_buf_iodone;

	lidp = xfs_trans_find_item(tp, (xfs_log_item_t*)bip);
	ASSERT(lidp != NULL);

	tp->t_flags |= XFS_TRANS_DIRTY;
	lidp->lid_flags |= XFS_LID_DIRTY;
	xfs_buf_item_log(bip, first, last);
}


/*
 * This called to invalidate a buffer that is being used within
 * a transaction.  Typically this is because the blocks in the
 * buffer are being freed, so we need to prevent it from being
 * written out when we're done.  Allowing it to be written again
 * might overwrite data in the free blocks if they are reallocated
 * to a file.
 *
 * We prevent the buffer from being written out by clearing the
 * B_DELWRI flag.  We can't always
 * get rid of the buf log item at this point, though, because
 * the buffer may still be pinned by other transaction.  If that
 * is the case, then we'll wait until the buffer is committed to
 * disk for the last time (we can tell by the ref count) and
 * free it in xfs_buf_item_unpin().  Until it is cleaned up we
 * will keep the buffer locked so that the buffer and buf log item
 * are not reused.
 */
void
xfs_trans_binval(
	xfs_trans_t	*tp,
	buf_t		*bp)
{
	xfs_log_item_desc_t	*lidp;
	xfs_buf_log_item_t	*bip;
	xfs_log_item_chunk_t	*licp;

	ASSERT(bp->b_flags & B_BUSY);
	ASSERT((xfs_trans_t*)(bp->b_fsprivate2) == tp);
	ASSERT(bp->b_fsprivate != NULL);

	bip = (xfs_buf_log_item_t *)(bp->b_fsprivate);
	lidp = xfs_trans_find_item(tp, (xfs_log_item_t*)bip);
	ASSERT(lidp != NULL);

	if (!(bp->b_flags & B_DELWRI)) {
		/*
		 * The buffer isn't dirty so the buf item isn't
		 * in the AIL and the buffer isn't pinned.
		 * In this case we can just free the buf item
		 * and release the buffer from the transaction.
		 */
		ASSERT(bp->b_pincount == 0);
		ASSERT(!(bip->bli_item.li_flags & XFS_LI_IN_AIL));
		ASSERT(bip->bli_refcount == 1);
		xfs_buf_item_relse(bp);
		/*
		 * There had better not be any log items attached to
		 * this buffer waiting for it to be written out, because
		 * it is never going out.
		 */
		ASSERT(bp->b_fsprivate == NULL);
		/*
		 * Free up the descriptor pointing at the buf log item.
		 */
		licp = XFS_LIC_DESC_TO_CHUNK(lidp);
		XFS_LIC_RELSE(licp, XFS_LIC_DESC_TO_SLOT(lidp));
		brelse(bp);
		return;
	}

	/*
	 * The buffer is dirty, so we can't clean it up until this
	 * transaction is permanent on disk.  Mark the buf item descriptor
	 * dirty so that we'll keep it around across the commit.
	 * Clear the B_DELWRI flag in the buffer so that it won't be
	 * written out anymore.  We set the XFS_BLI_STALE flag in the
	 * buf log item to indicate to the xfs_buf_item_unpin() that
	 * it should clean up when the last reference to the buf item
	 * is given up.
	 */
	bp->b_flags &= ~B_DELWRI;
	bip->bli_flags |= XFS_BLI_STALE;
	lidp->lid_flags |= XFS_LID_DIRTY;
	tp->t_flags |= XFS_TRANS_DIRTY;
}