kernel-fxtec-pro1x/fs/xfs/xfs_trans_item.c
Dave Chinner ed3b4d6cdc xfs: Improve scalability of busy extent tracking
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.

Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.

To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.

So, use a per-AG rbtree indexed by block number to track busy
extents.  This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.

However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.

Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.

By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.

The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.

Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.

Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.

Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-24 10:34:00 -05:00

440 lines
11 KiB
C

/*
* Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
* All Rights Reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation.
*
* This program is distributed in the hope that it would be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_types.h"
#include "xfs_log.h"
#include "xfs_inum.h"
#include "xfs_trans.h"
#include "xfs_trans_priv.h"
/* XXX: from here down needed until struct xfs_trans has its own ailp */
#include "xfs_bit.h"
#include "xfs_buf_item.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
STATIC int xfs_trans_unlock_chunk(xfs_log_item_chunk_t *,
int, int, xfs_lsn_t);
/*
* This is called to add the given log item to the transaction's
* list of log items. It must find a free log item descriptor
* or allocate a new one and add the item to that descriptor.
* The function returns a pointer to item descriptor used to point
* to the new item. The log item will now point to its new descriptor
* with its li_desc field.
*/
xfs_log_item_desc_t *
xfs_trans_add_item(xfs_trans_t *tp, xfs_log_item_t *lip)
{
xfs_log_item_desc_t *lidp;
xfs_log_item_chunk_t *licp;
int i=0;
/*
* If there are no free descriptors, allocate a new chunk
* of them and put it at the front of the chunk list.
*/
if (tp->t_items_free == 0) {
licp = (xfs_log_item_chunk_t*)
kmem_alloc(sizeof(xfs_log_item_chunk_t), KM_SLEEP);
ASSERT(licp != NULL);
/*
* Initialize the chunk, and then
* claim the first slot in the newly allocated chunk.
*/
xfs_lic_init(licp);
xfs_lic_claim(licp, 0);
licp->lic_unused = 1;
xfs_lic_init_slot(licp, 0);
lidp = xfs_lic_slot(licp, 0);
/*
* Link in the new chunk and update the free count.
*/
licp->lic_next = tp->t_items.lic_next;
tp->t_items.lic_next = licp;
tp->t_items_free = XFS_LIC_NUM_SLOTS - 1;
/*
* Initialize the descriptor and the generic portion
* of the log item.
*
* Point the new slot at this item and return it.
* Also point the log item at its currently active
* descriptor and set the item's mount pointer.
*/
lidp->lid_item = lip;
lidp->lid_flags = 0;
lidp->lid_size = 0;
lip->li_desc = lidp;
lip->li_mountp = tp->t_mountp;
lip->li_ailp = tp->t_mountp->m_ail;
return lidp;
}
/*
* Find the free descriptor. It is somewhere in the chunklist
* of descriptors.
*/
licp = &tp->t_items;
while (licp != NULL) {
if (xfs_lic_vacancy(licp)) {
if (licp->lic_unused <= XFS_LIC_MAX_SLOT) {
i = licp->lic_unused;
ASSERT(xfs_lic_isfree(licp, i));
break;
}
for (i = 0; i <= XFS_LIC_MAX_SLOT; i++) {
if (xfs_lic_isfree(licp, i))
break;
}
ASSERT(i <= XFS_LIC_MAX_SLOT);
break;
}
licp = licp->lic_next;
}
ASSERT(licp != NULL);
/*
* If we find a free descriptor, claim it,
* initialize it, and return it.
*/
xfs_lic_claim(licp, i);
if (licp->lic_unused <= i) {
licp->lic_unused = i + 1;
xfs_lic_init_slot(licp, i);
}
lidp = xfs_lic_slot(licp, i);
tp->t_items_free--;
lidp->lid_item = lip;
lidp->lid_flags = 0;
lidp->lid_size = 0;
lip->li_desc = lidp;
lip->li_mountp = tp->t_mountp;
lip->li_ailp = tp->t_mountp->m_ail;
return lidp;
}
/*
* Free the given descriptor.
*
* This requires setting the bit in the chunk's free mask corresponding
* to the given slot.
*/
void
xfs_trans_free_item(xfs_trans_t *tp, xfs_log_item_desc_t *lidp)
{
uint slot;
xfs_log_item_chunk_t *licp;
xfs_log_item_chunk_t **licpp;
slot = xfs_lic_desc_to_slot(lidp);
licp = xfs_lic_desc_to_chunk(lidp);
xfs_lic_relse(licp, slot);
lidp->lid_item->li_desc = NULL;
tp->t_items_free++;
/*
* If there are no more used items in the chunk and this is not
* the chunk embedded in the transaction structure, then free
* the chunk. First pull it from the chunk list and then
* free it back to the heap. We didn't bother with a doubly
* linked list here because the lists should be very short
* and this is not a performance path. It's better to save
* the memory of the extra pointer.
*
* Also decrement the transaction structure's count of free items
* by the number in a chunk since we are freeing an empty chunk.
*/
if (xfs_lic_are_all_free(licp) && (licp != &(tp->t_items))) {
licpp = &(tp->t_items.lic_next);
while (*licpp != licp) {
ASSERT(*licpp != NULL);
licpp = &((*licpp)->lic_next);
}
*licpp = licp->lic_next;
kmem_free(licp);
tp->t_items_free -= XFS_LIC_NUM_SLOTS;
}
}
/*
* This is called to find the descriptor corresponding to the given
* log item. It returns a pointer to the descriptor.
* The log item MUST have a corresponding descriptor in the given
* transaction. This routine does not return NULL, it panics.
*
* The descriptor pointer is kept in the log item's li_desc field.
* Just return it.
*/
/*ARGSUSED*/
xfs_log_item_desc_t *
xfs_trans_find_item(xfs_trans_t *tp, xfs_log_item_t *lip)
{
ASSERT(lip->li_desc != NULL);
return lip->li_desc;
}
/*
* Return a pointer to the first descriptor in the chunk list.
* This does not return NULL if there are none, it panics.
*
* The first descriptor must be in either the first or second chunk.
* This is because the only chunk allowed to be empty is the first.
* All others are freed when they become empty.
*
* At some point this and xfs_trans_next_item() should be optimized
* to quickly look at the mask to determine if there is anything to
* look at.
*/
xfs_log_item_desc_t *
xfs_trans_first_item(xfs_trans_t *tp)
{
xfs_log_item_chunk_t *licp;
int i;
licp = &tp->t_items;
/*
* If it's not in the first chunk, skip to the second.
*/
if (xfs_lic_are_all_free(licp)) {
licp = licp->lic_next;
}
/*
* Return the first non-free descriptor in the chunk.
*/
ASSERT(!xfs_lic_are_all_free(licp));
for (i = 0; i < licp->lic_unused; i++) {
if (xfs_lic_isfree(licp, i)) {
continue;
}
return xfs_lic_slot(licp, i);
}
cmn_err(CE_WARN, "xfs_trans_first_item() -- no first item");
return NULL;
}
/*
* Given a descriptor, return the next descriptor in the chunk list.
* This returns NULL if there are no more used descriptors in the list.
*
* We do this by first locating the chunk in which the descriptor resides,
* and then scanning forward in the chunk and the list for the next
* used descriptor.
*/
/*ARGSUSED*/
xfs_log_item_desc_t *
xfs_trans_next_item(xfs_trans_t *tp, xfs_log_item_desc_t *lidp)
{
xfs_log_item_chunk_t *licp;
int i;
licp = xfs_lic_desc_to_chunk(lidp);
/*
* First search the rest of the chunk. The for loop keeps us
* from referencing things beyond the end of the chunk.
*/
for (i = (int)xfs_lic_desc_to_slot(lidp) + 1; i < licp->lic_unused; i++) {
if (xfs_lic_isfree(licp, i)) {
continue;
}
return xfs_lic_slot(licp, i);
}
/*
* Now search the next chunk. It must be there, because the
* next chunk would have been freed if it were empty.
* If there is no next chunk, return NULL.
*/
if (licp->lic_next == NULL) {
return NULL;
}
licp = licp->lic_next;
ASSERT(!xfs_lic_are_all_free(licp));
for (i = 0; i < licp->lic_unused; i++) {
if (xfs_lic_isfree(licp, i)) {
continue;
}
return xfs_lic_slot(licp, i);
}
ASSERT(0);
/* NOTREACHED */
return NULL; /* keep gcc quite */
}
/*
* This is called to unlock all of the items of a transaction and to free
* all the descriptors of that transaction.
*
* It walks the list of descriptors and unlocks each item. It frees
* each chunk except that embedded in the transaction as it goes along.
*/
void
xfs_trans_free_items(
xfs_trans_t *tp,
int flags)
{
xfs_log_item_chunk_t *licp;
xfs_log_item_chunk_t *next_licp;
int abort;
abort = flags & XFS_TRANS_ABORT;
licp = &tp->t_items;
/*
* Special case the embedded chunk so we don't free it below.
*/
if (!xfs_lic_are_all_free(licp)) {
(void) xfs_trans_unlock_chunk(licp, 1, abort, NULLCOMMITLSN);
xfs_lic_all_free(licp);
licp->lic_unused = 0;
}
licp = licp->lic_next;
/*
* Unlock each item in each chunk and free the chunks.
*/
while (licp != NULL) {
ASSERT(!xfs_lic_are_all_free(licp));
(void) xfs_trans_unlock_chunk(licp, 1, abort, NULLCOMMITLSN);
next_licp = licp->lic_next;
kmem_free(licp);
licp = next_licp;
}
/*
* Reset the transaction structure's free item count.
*/
tp->t_items_free = XFS_LIC_NUM_SLOTS;
tp->t_items.lic_next = NULL;
}
/*
* This is called to unlock the items associated with a transaction.
* Items which were not logged should be freed.
* Those which were logged must still be tracked so they can be unpinned
* when the transaction commits.
*/
void
xfs_trans_unlock_items(xfs_trans_t *tp, xfs_lsn_t commit_lsn)
{
xfs_log_item_chunk_t *licp;
xfs_log_item_chunk_t *next_licp;
xfs_log_item_chunk_t **licpp;
int freed;
freed = 0;
licp = &tp->t_items;
/*
* Special case the embedded chunk so we don't free.
*/
if (!xfs_lic_are_all_free(licp)) {
freed = xfs_trans_unlock_chunk(licp, 0, 0, commit_lsn);
}
licpp = &(tp->t_items.lic_next);
licp = licp->lic_next;
/*
* Unlock each item in each chunk, free non-dirty descriptors,
* and free empty chunks.
*/
while (licp != NULL) {
ASSERT(!xfs_lic_are_all_free(licp));
freed += xfs_trans_unlock_chunk(licp, 0, 0, commit_lsn);
next_licp = licp->lic_next;
if (xfs_lic_are_all_free(licp)) {
*licpp = next_licp;
kmem_free(licp);
freed -= XFS_LIC_NUM_SLOTS;
} else {
licpp = &(licp->lic_next);
}
ASSERT(*licpp == next_licp);
licp = next_licp;
}
/*
* Fix the free descriptor count in the transaction.
*/
tp->t_items_free += freed;
}
/*
* Unlock each item pointed to by a descriptor in the given chunk.
* Stamp the commit lsn into each item if necessary.
* Free descriptors pointing to items which are not dirty if freeing_chunk
* is zero. If freeing_chunk is non-zero, then we need to unlock all
* items in the chunk.
*
* Return the number of descriptors freed.
*/
STATIC int
xfs_trans_unlock_chunk(
xfs_log_item_chunk_t *licp,
int freeing_chunk,
int abort,
xfs_lsn_t commit_lsn)
{
xfs_log_item_desc_t *lidp;
xfs_log_item_t *lip;
int i;
int freed;
freed = 0;
lidp = licp->lic_descs;
for (i = 0; i < licp->lic_unused; i++, lidp++) {
if (xfs_lic_isfree(licp, i)) {
continue;
}
lip = lidp->lid_item;
lip->li_desc = NULL;
if (commit_lsn != NULLCOMMITLSN)
IOP_COMMITTING(lip, commit_lsn);
if (abort)
lip->li_flags |= XFS_LI_ABORTED;
IOP_UNLOCK(lip);
/*
* Free the descriptor if the item is not dirty
* within this transaction and the caller is not
* going to just free the entire thing regardless.
*/
if (!(freeing_chunk) &&
(!(lidp->lid_flags & XFS_LID_DIRTY) || abort)) {
xfs_lic_relse(licp, i);
freed++;
}
}
return freed;
}