2018-06-06 05:42:14 +03:00
// SPDX-License-Identifier: GPL-2.0
2005-04-17 02:20:36 +04:00
/*
2005-11-02 06:58:39 +03:00
* Copyright ( c ) 2000 - 2005 Silicon Graphics , Inc .
* All Rights Reserved .
2005-04-17 02:20:36 +04:00
*/
# include "xfs.h"
2005-11-02 06:38:42 +03:00
# include "xfs_fs.h"
2013-10-23 03:36:05 +04:00
# include "xfs_shared.h"
2013-10-23 03:50:10 +04:00
# include "xfs_format.h"
# include "xfs_log_format.h"
# include "xfs_trans_resv.h"
2005-11-02 06:38:42 +03:00
# include "xfs_bit.h"
2005-04-17 02:20:36 +04:00
# include "xfs_mount.h"
# include "xfs_inode.h"
# include "xfs_bmap.h"
2013-10-23 03:51:50 +04:00
# include "xfs_bmap_btree.h"
2013-10-23 03:50:10 +04:00
# include "xfs_trans.h"
2005-04-17 02:20:36 +04:00
# include "xfs_trans_space.h"
2012-10-08 14:56:11 +04:00
# include "xfs_icache.h"
2013-10-15 02:17:56 +04:00
# include "xfs_rtalloc.h"
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Read and return the summary information for a given extent size ,
* bitmap block combination .
* Keeps track of a current summary block , so we don ' t keep reading
* it from the buffer cache .
2005-04-17 02:20:36 +04:00
*/
2014-09-23 10:11:43 +04:00
static int
2013-10-15 02:17:56 +04:00
xfs_rtget_summary (
xfs_mount_t * mp , /* file system mount structure */
xfs_trans_t * tp , /* transaction pointer */
int log , /* log2 of extent size */
xfs_rtblock_t bbno , /* bitmap block number */
xfs_buf_t * * rbpp , /* in/out: summary block buffer */
xfs_fsblock_t * rsb , /* in/out: summary block number */
xfs_suminfo_t * sum ) /* out: summary info for this block */
2005-04-17 02:20:36 +04:00
{
2014-09-09 05:58:42 +04:00
return xfs_rtmodify_summary_int ( mp , tp , log , bbno , 0 , rbpp , rsb , sum ) ;
2005-04-17 02:20:36 +04:00
}
/*
2013-10-15 02:17:56 +04:00
* Return whether there are any free extents in the size range given
* by low and high , for the bitmap block bbno .
2005-04-17 02:20:36 +04:00
*/
STATIC int /* error */
2013-10-15 02:17:56 +04:00
xfs_rtany_summary (
xfs_mount_t * mp , /* file system mount structure */
2005-04-17 02:20:36 +04:00
xfs_trans_t * tp , /* transaction pointer */
2013-10-15 02:17:56 +04:00
int low , /* low log2 extent size */
int high , /* high log2 extent size */
2005-04-17 02:20:36 +04:00
xfs_rtblock_t bbno , /* bitmap block number */
xfs_buf_t * * rbpp , /* in/out: summary block buffer */
xfs_fsblock_t * rsb , /* in/out: summary block number */
2013-10-15 02:17:56 +04:00
int * stat ) /* out: any good extents here? */
2005-04-17 02:20:36 +04:00
{
int error ; /* error value */
2013-10-15 02:17:56 +04:00
int log ; /* loop counter, log2 of ext. size */
xfs_suminfo_t sum ; /* summary data */
2005-04-17 02:20:36 +04:00
xfs: cache minimum realtime summary level
The realtime summary is a two-dimensional array on disk, effectively:
u32 rsum[log2(number of realtime extents) + 1][number of blocks in the bitmap]
rsum[log][bbno] is the number of extents of size 2**log which start in
bitmap block bbno.
xfs_rtallocate_extent_near() uses xfs_rtany_summary() to check whether
rsum[log][bbno] != 0 for any log level. However, the summary array is
stored in row-major order (i.e., like an array in C), so all of these
entries are not adjacent, but rather spread across the entire summary
file. In the worst case (a full bitmap block), xfs_rtany_summary() has
to check every level.
This means that on a moderately-used realtime device, an allocation will
waste a lot of time finding, reading, and releasing buffers for the
realtime summary. In particular, one of our storage services (which runs
on servers with 8 very slow CPUs and 15 8 TB XFS realtime filesystems)
spends almost 5% of its CPU cycles in xfs_rtbuf_get() and
xfs_trans_brelse() called from xfs_rtany_summary().
One solution would be to also store the summary with the dimensions
swapped. However, this would require a disk format change to a very old
component of XFS.
Instead, we can cache the minimum size which contains any extents. We do
so lazily; rather than guaranteeing that the cache contains the precise
minimum, it always contains a loose lower bound which we tighten when we
read or update a summary block. This only uses a few kilobytes of memory
and is already serialized via the realtime bitmap and summary inode
locks, so the cost is minimal. With this change, the same workload only
spends 0.2% of its CPU cycles in the realtime allocator.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2018-12-12 19:46:32 +03:00
/* There are no extents at levels < m_rsum_cache[bbno]. */
if ( mp - > m_rsum_cache & & low < mp - > m_rsum_cache [ bbno ] )
low = mp - > m_rsum_cache [ bbno ] ;
2005-04-17 02:20:36 +04:00
/*
xfs: cache minimum realtime summary level
The realtime summary is a two-dimensional array on disk, effectively:
u32 rsum[log2(number of realtime extents) + 1][number of blocks in the bitmap]
rsum[log][bbno] is the number of extents of size 2**log which start in
bitmap block bbno.
xfs_rtallocate_extent_near() uses xfs_rtany_summary() to check whether
rsum[log][bbno] != 0 for any log level. However, the summary array is
stored in row-major order (i.e., like an array in C), so all of these
entries are not adjacent, but rather spread across the entire summary
file. In the worst case (a full bitmap block), xfs_rtany_summary() has
to check every level.
This means that on a moderately-used realtime device, an allocation will
waste a lot of time finding, reading, and releasing buffers for the
realtime summary. In particular, one of our storage services (which runs
on servers with 8 very slow CPUs and 15 8 TB XFS realtime filesystems)
spends almost 5% of its CPU cycles in xfs_rtbuf_get() and
xfs_trans_brelse() called from xfs_rtany_summary().
One solution would be to also store the summary with the dimensions
swapped. However, this would require a disk format change to a very old
component of XFS.
Instead, we can cache the minimum size which contains any extents. We do
so lazily; rather than guaranteeing that the cache contains the precise
minimum, it always contains a loose lower bound which we tighten when we
read or update a summary block. This only uses a few kilobytes of memory
and is already serialized via the realtime bitmap and summary inode
locks, so the cost is minimal. With this change, the same workload only
spends 0.2% of its CPU cycles in the realtime allocator.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2018-12-12 19:46:32 +03:00
* Loop over logs of extent sizes .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
for ( log = low ; log < = high ; log + + ) {
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Get one summary datum .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
error = xfs_rtget_summary ( mp , tp , log , bbno , rbpp , rsb , & sum ) ;
2005-04-17 02:20:36 +04:00
if ( error ) {
return error ;
}
/*
2013-10-15 02:17:56 +04:00
* If there are any , return success .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
if ( sum ) {
* stat = 1 ;
xfs: cache minimum realtime summary level
The realtime summary is a two-dimensional array on disk, effectively:
u32 rsum[log2(number of realtime extents) + 1][number of blocks in the bitmap]
rsum[log][bbno] is the number of extents of size 2**log which start in
bitmap block bbno.
xfs_rtallocate_extent_near() uses xfs_rtany_summary() to check whether
rsum[log][bbno] != 0 for any log level. However, the summary array is
stored in row-major order (i.e., like an array in C), so all of these
entries are not adjacent, but rather spread across the entire summary
file. In the worst case (a full bitmap block), xfs_rtany_summary() has
to check every level.
This means that on a moderately-used realtime device, an allocation will
waste a lot of time finding, reading, and releasing buffers for the
realtime summary. In particular, one of our storage services (which runs
on servers with 8 very slow CPUs and 15 8 TB XFS realtime filesystems)
spends almost 5% of its CPU cycles in xfs_rtbuf_get() and
xfs_trans_brelse() called from xfs_rtany_summary().
One solution would be to also store the summary with the dimensions
swapped. However, this would require a disk format change to a very old
component of XFS.
Instead, we can cache the minimum size which contains any extents. We do
so lazily; rather than guaranteeing that the cache contains the precise
minimum, it always contains a loose lower bound which we tighten when we
read or update a summary block. This only uses a few kilobytes of memory
and is already serialized via the realtime bitmap and summary inode
locks, so the cost is minimal. With this change, the same workload only
spends 0.2% of its CPU cycles in the realtime allocator.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2018-12-12 19:46:32 +03:00
goto out ;
2005-04-17 02:20:36 +04:00
}
}
/*
2013-10-15 02:17:56 +04:00
* Found nothing , return failure .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
* stat = 0 ;
xfs: cache minimum realtime summary level
The realtime summary is a two-dimensional array on disk, effectively:
u32 rsum[log2(number of realtime extents) + 1][number of blocks in the bitmap]
rsum[log][bbno] is the number of extents of size 2**log which start in
bitmap block bbno.
xfs_rtallocate_extent_near() uses xfs_rtany_summary() to check whether
rsum[log][bbno] != 0 for any log level. However, the summary array is
stored in row-major order (i.e., like an array in C), so all of these
entries are not adjacent, but rather spread across the entire summary
file. In the worst case (a full bitmap block), xfs_rtany_summary() has
to check every level.
This means that on a moderately-used realtime device, an allocation will
waste a lot of time finding, reading, and releasing buffers for the
realtime summary. In particular, one of our storage services (which runs
on servers with 8 very slow CPUs and 15 8 TB XFS realtime filesystems)
spends almost 5% of its CPU cycles in xfs_rtbuf_get() and
xfs_trans_brelse() called from xfs_rtany_summary().
One solution would be to also store the summary with the dimensions
swapped. However, this would require a disk format change to a very old
component of XFS.
Instead, we can cache the minimum size which contains any extents. We do
so lazily; rather than guaranteeing that the cache contains the precise
minimum, it always contains a loose lower bound which we tighten when we
read or update a summary block. This only uses a few kilobytes of memory
and is already serialized via the realtime bitmap and summary inode
locks, so the cost is minimal. With this change, the same workload only
spends 0.2% of its CPU cycles in the realtime allocator.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2018-12-12 19:46:32 +03:00
out :
/* There were no extents at levels < log. */
if ( mp - > m_rsum_cache & & log > mp - > m_rsum_cache [ bbno ] )
mp - > m_rsum_cache [ bbno ] = log ;
2013-10-15 02:17:56 +04:00
return 0 ;
}
2005-04-17 02:20:36 +04:00
2013-10-15 02:17:56 +04:00
/*
* Copy and transform the summary file , given the old and new
* parameters in the mount structures .
*/
STATIC int /* error */
xfs_rtcopy_summary (
xfs_mount_t * omp , /* old file system mount point */
xfs_mount_t * nmp , /* new file system mount point */
xfs_trans_t * tp ) /* transaction pointer */
{
xfs_rtblock_t bbno ; /* bitmap block number */
xfs_buf_t * bp ; /* summary buffer */
int error ; /* error return value */
int log ; /* summary level number (log length) */
xfs_suminfo_t sum ; /* summary data */
xfs_fsblock_t sumbno ; /* summary block number */
bp = NULL ;
for ( log = omp - > m_rsumlevels - 1 ; log > = 0 ; log - - ) {
for ( bbno = omp - > m_sb . sb_rbmblocks - 1 ;
( xfs_srtblock_t ) bbno > = 0 ;
bbno - - ) {
error = xfs_rtget_summary ( omp , tp , log , bbno , & bp ,
& sumbno , & sum ) ;
if ( error )
return error ;
if ( sum = = 0 )
continue ;
error = xfs_rtmodify_summary ( omp , tp , log , bbno , - sum ,
& bp , & sumbno ) ;
if ( error )
return error ;
error = xfs_rtmodify_summary ( nmp , tp , log , bbno , sum ,
& bp , & sumbno ) ;
if ( error )
return error ;
ASSERT ( sum > 0 ) ;
}
}
return 0 ;
}
/*
* Mark an extent specified by start and len allocated .
* Updates all the summary information as well as the bitmap .
*/
STATIC int /* error */
xfs_rtallocate_range (
xfs_mount_t * mp , /* file system mount point */
xfs_trans_t * tp , /* transaction pointer */
xfs_rtblock_t start , /* start block to allocate */
xfs_extlen_t len , /* length to allocate */
xfs_buf_t * * rbpp , /* in/out: summary block buffer */
xfs_fsblock_t * rsb ) /* in/out: summary block number */
{
xfs_rtblock_t end ; /* end of the allocated extent */
int error ; /* error value */
xfs_rtblock_t postblock = 0 ; /* first block allocated > end */
xfs_rtblock_t preblock = 0 ; /* first block allocated < start */
end = start + len - 1 ;
/*
* Assume we ' re allocating out of the middle of a free extent .
* We need to find the beginning and end of the extent so we can
* properly update the summary .
*/
error = xfs_rtfind_back ( mp , tp , start , 0 , & preblock ) ;
if ( error ) {
return error ;
}
/*
* Find the next allocated block ( end of free extent ) .
*/
error = xfs_rtfind_forw ( mp , tp , end , mp - > m_sb . sb_rextents - 1 ,
& postblock ) ;
if ( error ) {
return error ;
}
/*
* Decrement the summary information corresponding to the entire
* ( old ) free extent .
*/
error = xfs_rtmodify_summary ( mp , tp ,
XFS_RTBLOCKLOG ( postblock + 1 - preblock ) ,
XFS_BITTOBLOCK ( mp , preblock ) , - 1 , rbpp , rsb ) ;
if ( error ) {
return error ;
}
/*
* If there are blocks not being allocated at the front of the
* old extent , add summary data for them to be free .
*/
if ( preblock < start ) {
error = xfs_rtmodify_summary ( mp , tp ,
XFS_RTBLOCKLOG ( start - preblock ) ,
XFS_BITTOBLOCK ( mp , preblock ) , 1 , rbpp , rsb ) ;
if ( error ) {
return error ;
}
}
/*
* If there are blocks not being allocated at the end of the
* old extent , add summary data for them to be free .
*/
if ( postblock > end ) {
error = xfs_rtmodify_summary ( mp , tp ,
XFS_RTBLOCKLOG ( postblock - end ) ,
XFS_BITTOBLOCK ( mp , end + 1 ) , 1 , rbpp , rsb ) ;
if ( error ) {
return error ;
}
}
/*
* Modify the bitmap to mark this extent allocated .
*/
error = xfs_rtmodify_range ( mp , tp , start , len , 0 ) ;
return error ;
}
/*
* Attempt to allocate an extent minlen < = len < = maxlen starting from
* bitmap block bbno . If we don ' t get maxlen then use prod to trim
* the length , if given . Returns error ; returns starting block in * rtblock .
* The lengths are all in rtextents .
*/
STATIC int /* error */
xfs_rtallocate_extent_block (
xfs_mount_t * mp , /* file system mount point */
xfs_trans_t * tp , /* transaction pointer */
xfs_rtblock_t bbno , /* bitmap block number */
xfs_extlen_t minlen , /* minimum length to allocate */
xfs_extlen_t maxlen , /* maximum length to allocate */
xfs_extlen_t * len , /* out: actual length allocated */
xfs_rtblock_t * nextp , /* out: next block to try */
xfs_buf_t * * rbpp , /* in/out: summary block buffer */
xfs_fsblock_t * rsb , /* in/out: summary block number */
xfs_extlen_t prod , /* extent product factor */
xfs_rtblock_t * rtblock ) /* out: start block allocated */
{
xfs_rtblock_t besti ; /* best rtblock found so far */
xfs_rtblock_t bestlen ; /* best length found so far */
xfs_rtblock_t end ; /* last rtblock in chunk */
int error ; /* error value */
xfs_rtblock_t i ; /* current rtblock trying */
xfs_rtblock_t next ; /* next rtblock to try */
int stat ; /* status from internal calls */
/*
* Loop over all the extents starting in this bitmap block ,
* looking for one that ' s long enough .
*/
for ( i = XFS_BLOCKTOBIT ( mp , bbno ) , besti = - 1 , bestlen = 0 ,
end = XFS_BLOCKTOBIT ( mp , bbno + 1 ) - 1 ;
i < = end ;
i + + ) {
2020-09-10 00:21:06 +03:00
/* Make sure we don't scan off the end of the rt volume. */
maxlen = min ( mp - > m_sb . sb_rextents , i + maxlen ) - i ;
2013-10-15 02:17:56 +04:00
/*
* See if there ' s a free extent of maxlen starting at i .
* If it ' s not so then next will contain the first non - free .
*/
error = xfs_rtcheck_range ( mp , tp , i , maxlen , 1 , & next , & stat ) ;
if ( error ) {
return error ;
}
if ( stat ) {
/*
* i for maxlen is all free , allocate and return that .
*/
error = xfs_rtallocate_range ( mp , tp , i , maxlen , rbpp ,
rsb ) ;
if ( error ) {
return error ;
}
* len = maxlen ;
* rtblock = i ;
return 0 ;
}
/*
* In the case where we have a variable - sized allocation
* request , figure out how big this free piece is ,
* and if it ' s big enough for the minimum , and the best
* so far , remember it .
*/
if ( minlen < maxlen ) {
xfs_rtblock_t thislen ; /* this extent size */
thislen = next - i ;
if ( thislen > = minlen & & thislen > bestlen ) {
besti = i ;
bestlen = thislen ;
}
}
/*
* If not done yet , find the start of the next free space .
*/
if ( next < end ) {
error = xfs_rtfind_forw ( mp , tp , next , end , & i ) ;
if ( error ) {
return error ;
}
} else
break ;
}
/*
* Searched the whole thing & didn ' t find a maxlen free extent .
*/
if ( minlen < maxlen & & besti ! = - 1 ) {
xfs_extlen_t p ; /* amount to trim length by */
/*
* If size should be a multiple of prod , make that so .
*/
2018-06-08 19:54:22 +03:00
if ( prod > 1 ) {
div_u64_rem ( bestlen , prod , & p ) ;
if ( p )
bestlen - = p ;
}
2005-04-17 02:20:36 +04:00
/*
* Allocate besti for bestlen & return that .
*/
error = xfs_rtallocate_range ( mp , tp , besti , bestlen , rbpp , rsb ) ;
if ( error ) {
return error ;
}
* len = bestlen ;
* rtblock = besti ;
return 0 ;
}
/*
* Allocation failed . Set * nextp to the next block to try .
*/
* nextp = next ;
* rtblock = NULLRTBLOCK ;
return 0 ;
}
/*
* Allocate an extent of length minlen < = len < = maxlen , starting at block
* bno . If we don ' t get maxlen then use prod to trim the length , if given .
* Returns error ; returns starting block in * rtblock .
* The lengths are all in rtextents .
*/
STATIC int /* error */
xfs_rtallocate_extent_exact (
xfs_mount_t * mp , /* file system mount point */
xfs_trans_t * tp , /* transaction pointer */
xfs_rtblock_t bno , /* starting block number to allocate */
xfs_extlen_t minlen , /* minimum length to allocate */
xfs_extlen_t maxlen , /* maximum length to allocate */
xfs_extlen_t * len , /* out: actual length allocated */
xfs_buf_t * * rbpp , /* in/out: summary block buffer */
xfs_fsblock_t * rsb , /* in/out: summary block number */
xfs_extlen_t prod , /* extent product factor */
xfs_rtblock_t * rtblock ) /* out: start block allocated */
{
int error ; /* error value */
xfs_extlen_t i ; /* extent length trimmed due to prod */
int isfree ; /* extent is free */
xfs_rtblock_t next ; /* next block to try (dummy) */
ASSERT ( minlen % prod = = 0 & & maxlen % prod = = 0 ) ;
/*
* Check if the range in question ( for maxlen ) is free .
*/
error = xfs_rtcheck_range ( mp , tp , bno , maxlen , 1 , & next , & isfree ) ;
if ( error ) {
return error ;
}
if ( isfree ) {
/*
* If it is , allocate it and return success .
*/
error = xfs_rtallocate_range ( mp , tp , bno , maxlen , rbpp , rsb ) ;
if ( error ) {
return error ;
}
* len = maxlen ;
* rtblock = bno ;
return 0 ;
}
/*
* If not , allocate what there is , if it ' s at least minlen .
*/
maxlen = next - bno ;
if ( maxlen < minlen ) {
/*
* Failed , return failure status .
*/
* rtblock = NULLRTBLOCK ;
return 0 ;
}
/*
* Trim off tail of extent , if prod is specified .
*/
if ( prod > 1 & & ( i = maxlen % prod ) ) {
maxlen - = i ;
if ( maxlen < minlen ) {
/*
* Now we can ' t do it , return failure status .
*/
* rtblock = NULLRTBLOCK ;
return 0 ;
}
}
/*
* Allocate what we can and return it .
*/
error = xfs_rtallocate_range ( mp , tp , bno , maxlen , rbpp , rsb ) ;
if ( error ) {
return error ;
}
* len = maxlen ;
* rtblock = bno ;
return 0 ;
}
/*
* Allocate an extent of length minlen < = len < = maxlen , starting as near
* to bno as possible . If we don ' t get maxlen then use prod to trim
* the length , if given . The lengths are all in rtextents .
*/
STATIC int /* error */
xfs_rtallocate_extent_near (
xfs_mount_t * mp , /* file system mount point */
xfs_trans_t * tp , /* transaction pointer */
xfs_rtblock_t bno , /* starting block number to allocate */
xfs_extlen_t minlen , /* minimum length to allocate */
xfs_extlen_t maxlen , /* maximum length to allocate */
xfs_extlen_t * len , /* out: actual length allocated */
xfs_buf_t * * rbpp , /* in/out: summary block buffer */
xfs_fsblock_t * rsb , /* in/out: summary block number */
xfs_extlen_t prod , /* extent product factor */
xfs_rtblock_t * rtblock ) /* out: start block allocated */
{
int any ; /* any useful extents from summary */
xfs_rtblock_t bbno ; /* bitmap block number */
int error ; /* error value */
int i ; /* bitmap block offset (loop control) */
int j ; /* secondary loop control */
int log2len ; /* log2 of minlen */
xfs_rtblock_t n ; /* next block to try */
xfs_rtblock_t r ; /* result block */
ASSERT ( minlen % prod = = 0 & & maxlen % prod = = 0 ) ;
/*
* If the block number given is off the end , silently set it to
* the last block .
*/
if ( bno > = mp - > m_sb . sb_rextents )
bno = mp - > m_sb . sb_rextents - 1 ;
2020-09-10 00:21:06 +03:00
/* Make sure we don't run off the end of the rt volume. */
maxlen = min ( mp - > m_sb . sb_rextents , bno + maxlen ) - bno ;
if ( maxlen < minlen ) {
* rtblock = NULLRTBLOCK ;
return 0 ;
}
2005-04-17 02:20:36 +04:00
/*
* Try the exact allocation first .
*/
error = xfs_rtallocate_extent_exact ( mp , tp , bno , minlen , maxlen , len ,
rbpp , rsb , prod , & r ) ;
if ( error ) {
return error ;
}
/*
* If the exact allocation worked , return that .
*/
if ( r ! = NULLRTBLOCK ) {
* rtblock = r ;
return 0 ;
}
bbno = XFS_BITTOBLOCK ( mp , bno ) ;
i = 0 ;
2008-08-13 09:41:12 +04:00
ASSERT ( minlen ! = 0 ) ;
2005-04-17 02:20:36 +04:00
log2len = xfs_highbit32 ( minlen ) ;
/*
* Loop over all bitmap blocks ( bbno + i is current block ) .
*/
for ( ; ; ) {
/*
* Get summary information of extents of all useful levels
* starting in this bitmap block .
*/
error = xfs_rtany_summary ( mp , tp , log2len , mp - > m_rsumlevels - 1 ,
bbno + i , rbpp , rsb , & any ) ;
if ( error ) {
return error ;
}
/*
* If there are any useful extents starting here , try
* allocating one .
*/
if ( any ) {
/*
* On the positive side of the starting location .
*/
if ( i > = 0 ) {
/*
* Try to allocate an extent starting in
* this block .
*/
error = xfs_rtallocate_extent_block ( mp , tp ,
bbno + i , minlen , maxlen , len , & n , rbpp ,
rsb , prod , & r ) ;
if ( error ) {
return error ;
}
/*
* If it worked , return it .
*/
if ( r ! = NULLRTBLOCK ) {
* rtblock = r ;
return 0 ;
}
}
/*
* On the negative side of the starting location .
*/
else { /* i < 0 */
/*
* Loop backwards through the bitmap blocks from
* the starting point - 1 up to where we are now .
* There should be an extent which ends in this
* bitmap block and is long enough .
*/
for ( j = - 1 ; j > i ; j - - ) {
/*
* Grab the summary information for
* this bitmap block .
*/
error = xfs_rtany_summary ( mp , tp ,
log2len , mp - > m_rsumlevels - 1 ,
bbno + j , rbpp , rsb , & any ) ;
if ( error ) {
return error ;
}
/*
* If there ' s no extent given in the
* summary that means the extent we
* found must carry over from an
* earlier block . If there is an
* extent given , we ' ve already tried
* that allocation , don ' t do it again .
*/
if ( any )
continue ;
error = xfs_rtallocate_extent_block ( mp ,
tp , bbno + j , minlen , maxlen ,
len , & n , rbpp , rsb , prod , & r ) ;
if ( error ) {
return error ;
}
/*
* If it works , return the extent .
*/
if ( r ! = NULLRTBLOCK ) {
* rtblock = r ;
return 0 ;
}
}
/*
* There weren ' t intervening bitmap blocks
* with a long enough extent , or the
* allocation didn ' t work for some reason
* ( i . e . it ' s a little * too short ) .
* Try to allocate from the summary block
* that we found .
*/
error = xfs_rtallocate_extent_block ( mp , tp ,
bbno + i , minlen , maxlen , len , & n , rbpp ,
rsb , prod , & r ) ;
if ( error ) {
return error ;
}
/*
* If it works , return the extent .
*/
if ( r ! = NULLRTBLOCK ) {
* rtblock = r ;
return 0 ;
}
}
}
/*
* Loop control . If we were on the positive side , and there ' s
* still more blocks on the negative side , go there .
*/
if ( i > 0 & & ( int ) bbno - i > = 0 )
i = - i ;
/*
* If positive , and no more negative , but there are more
* positive , go there .
*/
else if ( i > 0 & & ( int ) bbno + i < mp - > m_sb . sb_rbmblocks - 1 )
i + + ;
/*
* If negative or 0 ( just started ) , and there are positive
* blocks to go , go there . The 0 case moves to block 1.
*/
else if ( i < = 0 & & ( int ) bbno - i < mp - > m_sb . sb_rbmblocks - 1 )
i = 1 - i ;
/*
* If negative or 0 and there are more negative blocks ,
* go there .
*/
else if ( i < = 0 & & ( int ) bbno + i > 0 )
i - - ;
/*
* Must be done . Return failure .
*/
else
break ;
}
* rtblock = NULLRTBLOCK ;
return 0 ;
}
/*
* Allocate an extent of length minlen < = len < = maxlen , with no position
* specified . If we don ' t get maxlen then use prod to trim
* the length , if given . The lengths are all in rtextents .
*/
STATIC int /* error */
xfs_rtallocate_extent_size (
xfs_mount_t * mp , /* file system mount point */
xfs_trans_t * tp , /* transaction pointer */
xfs_extlen_t minlen , /* minimum length to allocate */
xfs_extlen_t maxlen , /* maximum length to allocate */
xfs_extlen_t * len , /* out: actual length allocated */
xfs_buf_t * * rbpp , /* in/out: summary block buffer */
xfs_fsblock_t * rsb , /* in/out: summary block number */
xfs_extlen_t prod , /* extent product factor */
xfs_rtblock_t * rtblock ) /* out: start block allocated */
{
int error ; /* error value */
int i ; /* bitmap block number */
int l ; /* level number (loop control) */
xfs_rtblock_t n ; /* next block to be tried */
xfs_rtblock_t r ; /* result block number */
xfs_suminfo_t sum ; /* summary information for extents */
ASSERT ( minlen % prod = = 0 & & maxlen % prod = = 0 ) ;
2008-08-13 09:41:12 +04:00
ASSERT ( maxlen ! = 0 ) ;
2005-04-17 02:20:36 +04:00
/*
* Loop over all the levels starting with maxlen .
* At each level , look at all the bitmap blocks , to see if there
* are extents starting there that are long enough ( > = maxlen ) .
* Note , only on the initial level can the allocation fail if
* the summary says there ' s an extent .
*/
for ( l = xfs_highbit32 ( maxlen ) ; l < mp - > m_rsumlevels ; l + + ) {
/*
* Loop over all the bitmap blocks .
*/
for ( i = 0 ; i < mp - > m_sb . sb_rbmblocks ; i + + ) {
/*
* Get the summary for this level / block .
*/
error = xfs_rtget_summary ( mp , tp , l , i , rbpp , rsb ,
& sum ) ;
if ( error ) {
return error ;
}
/*
* Nothing there , on to the next block .
*/
if ( ! sum )
continue ;
/*
* Try allocating the extent .
*/
error = xfs_rtallocate_extent_block ( mp , tp , i , maxlen ,
maxlen , len , & n , rbpp , rsb , prod , & r ) ;
if ( error ) {
return error ;
}
/*
* If it worked , return that .
*/
if ( r ! = NULLRTBLOCK ) {
* rtblock = r ;
2013-10-15 02:17:56 +04:00
return 0 ;
}
/*
* If the " next block to try " returned from the
* allocator is beyond the next bitmap block ,
* skip to that bitmap block .
*/
if ( XFS_BITTOBLOCK ( mp , n ) > i + 1 )
i = XFS_BITTOBLOCK ( mp , n ) - 1 ;
}
2005-04-17 02:20:36 +04:00
}
/*
2013-10-15 02:17:56 +04:00
* Didn ' t find any maxlen blocks . Try smaller ones , unless
* we ' re asking for a fixed size extent .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
if ( minlen > - - maxlen ) {
* rtblock = NULLRTBLOCK ;
return 0 ;
}
ASSERT ( minlen ! = 0 ) ;
ASSERT ( maxlen ! = 0 ) ;
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Loop over sizes , from maxlen down to minlen .
* This time , when we do the allocations , allow smaller ones
* to succeed .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
for ( l = xfs_highbit32 ( maxlen ) ; l > = xfs_highbit32 ( minlen ) ; l - - ) {
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Loop over all the bitmap blocks , try an allocation
* starting in that block .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
for ( i = 0 ; i < mp - > m_sb . sb_rbmblocks ; i + + ) {
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Get the summary information for this level / block .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
error = xfs_rtget_summary ( mp , tp , l , i , rbpp , rsb ,
& sum ) ;
2005-04-17 02:20:36 +04:00
if ( error ) {
return error ;
}
/*
2013-10-15 02:17:56 +04:00
* If nothing there , go on to next .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
if ( ! sum )
continue ;
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Try the allocation . Make sure the specified
* minlen / maxlen are in the possible range for
* this summary level .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
error = xfs_rtallocate_extent_block ( mp , tp , i ,
XFS_RTMAX ( minlen , 1 < < l ) ,
XFS_RTMIN ( maxlen , ( 1 < < ( l + 1 ) ) - 1 ) ,
len , & n , rbpp , rsb , prod , & r ) ;
2005-04-17 02:20:36 +04:00
if ( error ) {
return error ;
}
/*
2013-10-15 02:17:56 +04:00
* If it worked , return that extent .
*/
if ( r ! = NULLRTBLOCK ) {
* rtblock = r ;
return 0 ;
}
/*
* If the " next block to try " returned from the
* allocator is beyond the next bitmap block ,
* skip to that bitmap block .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
if ( XFS_BITTOBLOCK ( mp , n ) > i + 1 )
i = XFS_BITTOBLOCK ( mp , n ) - 1 ;
2005-04-17 02:20:36 +04:00
}
}
/*
2013-10-15 02:17:56 +04:00
* Got nothing , return failure .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
* rtblock = NULLRTBLOCK ;
2005-04-17 02:20:36 +04:00
return 0 ;
}
/*
2013-10-15 02:17:56 +04:00
* Allocate space to the bitmap or summary file , and zero it , for growfs .
2005-04-17 02:20:36 +04:00
*/
2015-08-19 03:01:40 +03:00
STATIC int
2013-10-15 02:17:56 +04:00
xfs_growfs_rt_alloc (
2015-08-19 03:01:40 +03:00
struct xfs_mount * mp , /* file system mount point */
xfs_extlen_t oblocks , /* old count of blocks */
xfs_extlen_t nblocks , /* new count of blocks */
struct xfs_inode * ip ) /* inode (bitmap/summary) */
2005-04-17 02:20:36 +04:00
{
2015-08-19 03:01:40 +03:00
xfs_fileoff_t bno ; /* block number in file */
struct xfs_buf * bp ; /* temporary buffer for zeroing */
xfs_daddr_t d ; /* disk block address */
int error ; /* error return value */
xfs_fsblock_t fsbno ; /* filesystem block for bno */
struct xfs_bmbt_irec map ; /* block map output */
int nmap ; /* number of block maps */
int resblks ; /* space reservation */
2020-09-16 06:50:42 +03:00
enum xfs_blft buf_type ;
2015-08-19 03:01:40 +03:00
struct xfs_trans * tp ;
2005-04-17 02:20:36 +04:00
2020-09-16 06:50:42 +03:00
if ( ip = = mp - > m_rsumip )
buf_type = XFS_BLFT_RTSUMMARY_BUF ;
else
buf_type = XFS_BLFT_RTBITMAP_BUF ;
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Allocate space to the file , as necessary .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
while ( oblocks < nblocks ) {
resblks = XFS_GROWFSRT_SPACE_RES ( mp , nblocks - oblocks ) ;
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Reserve space & log for one extent added to the file .
2005-04-17 02:20:36 +04:00
*/
2016-04-06 02:19:55 +03:00
error = xfs_trans_alloc ( mp , & M_RES ( mp ) - > tr_growrtalloc , resblks ,
0 , 0 , & tp ) ;
2013-10-15 02:17:56 +04:00
if ( error )
2016-04-06 02:19:55 +03:00
return error ;
2005-04-17 02:20:36 +04:00
/*
2013-10-15 02:17:56 +04:00
* Lock the inode .
2005-04-17 02:20:36 +04:00
*/
2013-10-15 02:17:56 +04:00
xfs_ilock ( ip , XFS_ILOCK_EXCL ) ;
xfs_trans_ijoin ( tp , ip , XFS_ILOCK_EXCL ) ;
/*
* Allocate blocks to the bitmap file .
*/
nmap = 1 ;
error = xfs_bmapi_write ( tp , ip , oblocks , nblocks - oblocks ,
xfs: don't set bmapi total block req where minleft is
xfs_bmapi_write() takes a total block requirement parameter that is
passed down to the block allocation code and is used to specify the
total block requirement of the associated transaction. This is used
to try and select an AG that can not only satisfy the requested
extent allocation, but can also accommodate subsequent allocations
that might be required to complete the transaction. For example,
additional bmbt block allocations may be required on insertion of
the resulting extent to an inode data fork.
While it's important for callers to calculate and reserve such extra
blocks in the transaction, it is not necessary to pass the total
value to xfs_bmapi_write() in all cases. The latter automatically
sets minleft to ensure that sufficient free blocks remain after the
allocation attempt to expand the format of the associated inode
(i.e., such as extent to btree conversion, btree splits, etc).
Therefore, any callers that pass a total block requirement of the
bmap mapping length plus worst case bmbt expansion essentially
specify the additional reservation requirement twice. These callers
can pass a total of zero to rely on the bmapi minleft policy.
Beyond being superfluous, the primary motivation for this change is
that the total reservation logic in the bmbt code is dubious in
scenarios where minlen < maxlen and a maxlen extent cannot be
allocated (which is more common for data extent allocations where
contiguity is not required). The total value is based on maxlen in
the xfs_bmapi_write() caller. If the bmbt code falls back to an
allocation between minlen and maxlen, that allocation will not
succeed until total is reset to minlen, which essentially throws
away any additional reservation included in total by the caller. In
addition, the total value is not reset until after alignment is
dropped, which means that such callers drop alignment far too
aggressively than necessary.
Update all callers of xfs_bmapi_write() that pass a total block
value of the mapping length plus bmbt reservation to instead pass
zero and rely on xfs_bmapi_minleft() to enforce the bmbt reservation
requirement. This trades off slightly less conservative AG selection
for the ability to preserve alignment in more scenarios.
xfs_bmapi_write() callers that incorporate unrelated or additional
reservations in total beyond what is already included in minleft
must continue to use the former.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-10-21 19:26:48 +03:00
XFS_BMAPI_METADATA , 0 , & map , & nmap ) ;
2013-10-15 02:17:56 +04:00
if ( ! error & & nmap < 1 )
2014-06-25 08:58:08 +04:00
error = - ENOSPC ;
2013-10-15 02:17:56 +04:00
if ( error )
2018-07-24 23:43:13 +03:00
goto out_trans_cancel ;
2013-10-15 02:17:56 +04:00
/*
* Free any blocks freed up in the transaction , then commit .
*/
2015-06-04 06:48:08 +03:00
error = xfs_trans_commit ( tp ) ;
2013-10-15 02:17:56 +04:00
if ( error )
2015-08-19 03:01:40 +03:00
return error ;
2013-10-15 02:17:56 +04:00
/*
* Now we need to clear the allocated blocks .
* Do this one block per transaction , to keep it simple .
*/
for ( bno = map . br_startoff , fsbno = map . br_startblock ;
bno < map . br_startoff + map . br_blockcount ;
bno + + , fsbno + + ) {
/*
* Reserve log for one block zeroing .
*/
2016-04-06 02:19:55 +03:00
error = xfs_trans_alloc ( mp , & M_RES ( mp ) - > tr_growrtzero ,
0 , 0 , 0 , & tp ) ;
2013-10-15 02:17:56 +04:00
if ( error )
2016-04-06 02:19:55 +03:00
return error ;
2013-10-15 02:17:56 +04:00
/*
* Lock the bitmap inode .
*/
xfs_ilock ( ip , XFS_ILOCK_EXCL ) ;
xfs_trans_ijoin ( tp , ip , XFS_ILOCK_EXCL ) ;
/*
* Get a buffer for the block .
*/
d = XFS_FSB_TO_DADDR ( mp , fsbno ) ;
2020-01-24 04:01:18 +03:00
error = xfs_trans_get_buf ( tp , mp - > m_ddev_targp , d ,
mp - > m_bsize , 0 , & bp ) ;
if ( error )
2015-08-19 03:01:40 +03:00
goto out_trans_cancel ;
2020-09-16 06:50:42 +03:00
xfs_trans_buf_set_type ( tp , bp , buf_type ) ;
2013-10-15 02:17:56 +04:00
memset ( bp - > b_addr , 0 , mp - > m_sb . sb_blocksize ) ;
xfs_trans_log_buf ( tp , bp , 0 , mp - > m_sb . sb_blocksize - 1 ) ;
/*
* Commit the transaction .
*/
2015-06-04 06:48:08 +03:00
error = xfs_trans_commit ( tp ) ;
2013-10-15 02:17:56 +04:00
if ( error )
2015-08-19 03:01:40 +03:00
return error ;
2005-04-17 02:20:36 +04:00
}
2013-10-15 02:17:56 +04:00
/*
* Go on to the next extent , if any .
*/
oblocks = map . br_startoff + map . br_blockcount ;
2005-04-17 02:20:36 +04:00
}
2015-08-19 03:01:40 +03:00
2005-04-17 02:20:36 +04:00
return 0 ;
2013-10-15 02:17:56 +04:00
2015-08-19 03:01:40 +03:00
out_trans_cancel :
xfs_trans_cancel ( tp ) ;
2013-10-15 02:17:56 +04:00
return error ;
2005-04-17 02:20:36 +04:00
}
2018-12-22 05:45:18 +03:00
static void
xfs_alloc_rsum_cache (
xfs_mount_t * mp , /* file system mount structure */
xfs_extlen_t rbmblocks ) /* number of rt bitmap blocks */
{
/*
* The rsum cache is initialized to all zeroes , which is trivially a
* lower bound on the minimum level with any free extents . We can
* continue without the cache if it couldn ' t be allocated .
*/
2020-09-01 21:47:12 +03:00
mp - > m_rsum_cache = kvzalloc ( rbmblocks , GFP_KERNEL ) ;
2018-12-22 05:45:18 +03:00
if ( ! mp - > m_rsum_cache )
xfs_warn ( mp , " could not allocate realtime summary cache " ) ;
}
2005-04-17 02:20:36 +04:00
/*
* Visible ( exported ) functions .
*/
/*
* Grow the realtime area of the filesystem .
*/
int
xfs_growfs_rt (
xfs_mount_t * mp , /* mount point for filesystem */
xfs_growfs_rt_t * in ) /* growfs rt input struct */
{
xfs_rtblock_t bmbno ; /* bitmap block number */
xfs_buf_t * bp ; /* temporary buffer */
int error ; /* error return value */
xfs_mount_t * nmp ; /* new (fake) mount structure */
2014-07-30 03:12:05 +04:00
xfs_rfsblock_t nrblocks ; /* new number of realtime blocks */
2005-04-17 02:20:36 +04:00
xfs_extlen_t nrbmblocks ; /* new number of rt bitmap blocks */
2014-07-30 03:12:05 +04:00
xfs_rtblock_t nrextents ; /* new number of realtime extents */
2005-04-17 02:20:36 +04:00
uint8_t nrextslog ; /* new log2 of sb_rextents */
xfs_extlen_t nrsumblocks ; /* new number of summary blocks */
uint nrsumlevels ; /* new rt summary levels */
uint nrsumsize ; /* new size of rt summary, bytes */
xfs_sb_t * nsbp ; /* new superblock */
xfs_extlen_t rbmblocks ; /* current number of rt bitmap blocks */
xfs_extlen_t rsumblocks ; /* current number of rt summary blks */
xfs_sb_t * sbp ; /* old superblock */
xfs_fsblock_t sumbno ; /* summary block number */
2018-12-22 05:45:18 +03:00
uint8_t * rsum_cache ; /* old summary cache */
2005-04-17 02:20:36 +04:00
sbp = & mp - > m_sb ;
/*
* Initial error checking .
*/
2008-11-26 06:20:06 +03:00
if ( ! capable ( CAP_SYS_ADMIN ) )
2014-06-25 08:58:08 +04:00
return - EPERM ;
2006-06-28 02:42:26 +04:00
if ( mp - > m_rtdev_targp = = NULL | | mp - > m_rbmip = = NULL | |
2005-04-17 02:20:36 +04:00
( nrblocks = in - > newblocks ) < = sbp - > sb_rblocks | |
( sbp - > sb_rblocks & & ( in - > extsize ! = sbp - > sb_rextsize ) ) )
2014-06-25 08:58:08 +04:00
return - EINVAL ;
2007-05-14 12:24:02 +04:00
if ( ( error = xfs_sb_validate_fsb_count ( sbp , nrblocks ) ) )
return error ;
2005-04-17 02:20:36 +04:00
/*
* Read in the last block of the device , make sure it exists .
*/
2014-10-02 03:05:32 +04:00
error = xfs_buf_read_uncached ( mp - > m_rtdev_targp ,
2010-09-22 04:47:20 +04:00
XFS_FSB_TO_BB ( mp , nrblocks - 1 ) ,
2014-10-02 03:05:32 +04:00
XFS_FSB_TO_BB ( mp , 1 ) , 0 , & bp , NULL ) ;
if ( error )
2012-11-12 15:54:02 +04:00
return error ;
2005-04-17 02:20:36 +04:00
xfs_buf_relse ( bp ) ;
2010-09-22 04:47:20 +04:00
2005-04-17 02:20:36 +04:00
/*
* Calculate new parameters . These are the final values to be reached .
*/
nrextents = nrblocks ;
do_div ( nrextents , in - > extsize ) ;
2006-09-28 05:03:53 +04:00
nrbmblocks = howmany_64 ( nrextents , NBBY * sbp - > sb_blocksize ) ;
2005-04-17 02:20:36 +04:00
nrextslog = xfs_highbit32 ( nrextents ) ;
nrsumlevels = nrextslog + 1 ;
nrsumsize = ( uint ) sizeof ( xfs_suminfo_t ) * nrsumlevels * nrbmblocks ;
nrsumblocks = XFS_B_TO_FSB ( mp , nrsumsize ) ;
nrsumsize = XFS_FSB_TO_B ( mp , nrsumblocks ) ;
/*
* New summary size can ' t be more than half the size of
* the log . This prevents us from getting a log overflow ,
* since we ' ll log basically the whole summary file at once .
*/
if ( nrsumblocks > ( mp - > m_sb . sb_logblocks > > 1 ) )
2014-06-25 08:58:08 +04:00
return - EINVAL ;
2005-04-17 02:20:36 +04:00
/*
* Get the old block counts for bitmap and summary inodes .
* These can ' t change since other growfs callers are locked out .
*/
rbmblocks = XFS_B_TO_FSB ( mp , mp - > m_rbmip - > i_d . di_size ) ;
rsumblocks = XFS_B_TO_FSB ( mp , mp - > m_rsumip - > i_d . di_size ) ;
/*
* Allocate space to the bitmap and summary files , as necessary .
*/
2011-02-13 16:25:31 +03:00
error = xfs_growfs_rt_alloc ( mp , rbmblocks , nrbmblocks , mp - > m_rbmip ) ;
if ( error )
2005-04-17 02:20:36 +04:00
return error ;
2011-02-13 16:25:31 +03:00
error = xfs_growfs_rt_alloc ( mp , rsumblocks , nrsumblocks , mp - > m_rsumip ) ;
if ( error )
2005-04-17 02:20:36 +04:00
return error ;
2018-12-22 05:45:18 +03:00
rsum_cache = mp - > m_rsum_cache ;
if ( nrbmblocks ! = sbp - > sb_rbmblocks )
xfs_alloc_rsum_cache ( mp , nrbmblocks ) ;
2006-09-28 05:03:44 +04:00
/*
* Allocate a new ( fake ) mount / sb .
*/
2019-08-26 22:06:22 +03:00
nmp = kmem_alloc ( sizeof ( * nmp ) , 0 ) ;
2005-04-17 02:20:36 +04:00
/*
* Loop over the bitmap blocks .
* We will do everything one bitmap block at a time .
* Skip the current block if it is exactly full .
* This also deals with the case where there were no rtextents before .
*/
for ( bmbno = sbp - > sb_rbmblocks -
( ( sbp - > sb_rextents & ( ( 1 < < mp - > m_blkbit_log ) - 1 ) ) ! = 0 ) ;
bmbno < nrbmblocks ;
bmbno + + ) {
2008-11-28 06:23:34 +03:00
xfs_trans_t * tp ;
2005-04-17 02:20:36 +04:00
* nmp = * mp ;
nsbp = & nmp - > m_sb ;
/*
* Calculate new sb and mount fields for this round .
*/
nsbp - > sb_rextsize = in - > extsize ;
nsbp - > sb_rbmblocks = bmbno + 1 ;
nsbp - > sb_rblocks =
XFS_RTMIN ( nrblocks ,
nsbp - > sb_rbmblocks * NBBY *
nsbp - > sb_blocksize * nsbp - > sb_rextsize ) ;
nsbp - > sb_rextents = nsbp - > sb_rblocks ;
do_div ( nsbp - > sb_rextents , nsbp - > sb_rextsize ) ;
2008-08-13 09:41:12 +04:00
ASSERT ( nsbp - > sb_rextents ! = 0 ) ;
2005-04-17 02:20:36 +04:00
nsbp - > sb_rextslog = xfs_highbit32 ( nsbp - > sb_rextents ) ;
nrsumlevels = nmp - > m_rsumlevels = nsbp - > sb_rextslog + 1 ;
nrsumsize =
( uint ) sizeof ( xfs_suminfo_t ) * nrsumlevels *
nsbp - > sb_rbmblocks ;
nrsumblocks = XFS_B_TO_FSB ( mp , nrsumsize ) ;
nmp - > m_rsumsize = nrsumsize = XFS_FSB_TO_B ( mp , nrsumblocks ) ;
/*
* Start a transaction , get the log reservation .
*/
2016-04-06 02:19:55 +03:00
error = xfs_trans_alloc ( mp , & M_RES ( mp ) - > tr_growrtfree , 0 , 0 , 0 ,
& tp ) ;
2013-08-12 14:49:59 +04:00
if ( error )
2016-04-06 02:19:55 +03:00
break ;
2005-04-17 02:20:36 +04:00
/*
* Lock out other callers by grabbing the bitmap inode lock .
*/
2011-02-13 16:25:31 +03:00
xfs_ilock ( mp - > m_rbmip , XFS_ILOCK_EXCL ) ;
2011-09-19 19:00:54 +04:00
xfs_trans_ijoin ( tp , mp - > m_rbmip , XFS_ILOCK_EXCL ) ;
2005-04-17 02:20:36 +04:00
/*
* Update the bitmap inode ' s size .
*/
mp - > m_rbmip - > i_d . di_size =
nsbp - > sb_rbmblocks * nsbp - > sb_blocksize ;
xfs_trans_log_inode ( tp , mp - > m_rbmip , XFS_ILOG_CORE ) ;
/*
* Get the summary inode into the transaction .
*/
2011-02-13 16:25:31 +03:00
xfs_ilock ( mp - > m_rsumip , XFS_ILOCK_EXCL ) ;
2011-09-19 19:00:54 +04:00
xfs_trans_ijoin ( tp , mp - > m_rsumip , XFS_ILOCK_EXCL ) ;
2005-04-17 02:20:36 +04:00
/*
* Update the summary inode ' s size .
*/
mp - > m_rsumip - > i_d . di_size = nmp - > m_rsumsize ;
xfs_trans_log_inode ( tp , mp - > m_rsumip , XFS_ILOG_CORE ) ;
/*
* Copy summary data from old to new sizes .
* Do this when the real size ( not block - aligned ) changes .
*/
if ( sbp - > sb_rbmblocks ! = nsbp - > sb_rbmblocks | |
mp - > m_rsumlevels ! = nmp - > m_rsumlevels ) {
error = xfs_rtcopy_summary ( mp , nmp , tp ) ;
if ( error )
2008-11-28 06:23:34 +03:00
goto error_cancel ;
2005-04-17 02:20:36 +04:00
}
/*
* Update superblock fields .
*/
if ( nsbp - > sb_rextsize ! = sbp - > sb_rextsize )
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_REXTSIZE ,
nsbp - > sb_rextsize - sbp - > sb_rextsize ) ;
if ( nsbp - > sb_rbmblocks ! = sbp - > sb_rbmblocks )
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_RBMBLOCKS ,
nsbp - > sb_rbmblocks - sbp - > sb_rbmblocks ) ;
if ( nsbp - > sb_rblocks ! = sbp - > sb_rblocks )
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_RBLOCKS ,
nsbp - > sb_rblocks - sbp - > sb_rblocks ) ;
if ( nsbp - > sb_rextents ! = sbp - > sb_rextents )
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_REXTENTS ,
nsbp - > sb_rextents - sbp - > sb_rextents ) ;
if ( nsbp - > sb_rextslog ! = sbp - > sb_rextslog )
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_REXTSLOG ,
nsbp - > sb_rextslog - sbp - > sb_rextslog ) ;
/*
* Free new extent .
*/
bp = NULL ;
error = xfs_rtfree_range ( nmp , tp , sbp - > sb_rextents ,
nsbp - > sb_rextents - sbp - > sb_rextents , & bp , & sumbno ) ;
2008-11-28 06:23:34 +03:00
if ( error ) {
error_cancel :
2015-06-04 06:47:56 +03:00
xfs_trans_cancel ( tp ) ;
2006-09-28 05:03:44 +04:00
break ;
2008-11-28 06:23:34 +03:00
}
2005-04-17 02:20:36 +04:00
/*
* Mark more blocks free in the superblock .
*/
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_FREXTENTS ,
nsbp - > sb_rextents - sbp - > sb_rextents ) ;
/*
* Update mp values into the real mp structure .
*/
mp - > m_rsumlevels = nrsumlevels ;
mp - > m_rsumsize = nrsumsize ;
2008-04-10 06:21:18 +04:00
2015-06-04 06:48:08 +03:00
error = xfs_trans_commit ( tp ) ;
2008-11-28 06:23:34 +03:00
if ( error )
2008-04-10 06:21:18 +04:00
break ;
2005-04-17 02:20:36 +04:00
}
2006-09-28 05:03:44 +04:00
2005-04-17 02:20:36 +04:00
/*
2006-09-28 05:03:44 +04:00
* Free the fake mp structure .
2005-04-17 02:20:36 +04:00
*/
2008-05-19 10:31:57 +04:00
kmem_free ( nmp ) ;
2006-09-28 05:03:44 +04:00
2018-12-22 05:45:18 +03:00
/*
* If we had to allocate a new rsum_cache , we either need to free the
* old one ( if we succeeded ) or free the new one and restore the old one
* ( if there was an error ) .
*/
if ( rsum_cache ! = mp - > m_rsum_cache ) {
if ( error ) {
kmem_free ( mp - > m_rsum_cache ) ;
mp - > m_rsum_cache = rsum_cache ;
} else {
kmem_free ( rsum_cache ) ;
}
}
2005-04-17 02:20:36 +04:00
return error ;
}
/*
* Allocate an extent in the realtime subvolume , with the usual allocation
* parameters . The length units are all in realtime extents , as is the
* result block number .
*/
int /* error */
xfs_rtallocate_extent (
xfs_trans_t * tp , /* transaction pointer */
xfs_rtblock_t bno , /* starting block number to allocate */
xfs_extlen_t minlen , /* minimum length to allocate */
xfs_extlen_t maxlen , /* maximum length to allocate */
xfs_extlen_t * len , /* out: actual length allocated */
int wasdel , /* was a delayed allocation extent */
xfs_extlen_t prod , /* extent product factor */
xfs_rtblock_t * rtblock ) /* out: start block allocated */
{
2011-01-25 12:06:19 +03:00
xfs_mount_t * mp = tp - > t_mountp ;
2005-04-17 02:20:36 +04:00
int error ; /* error value */
xfs_rtblock_t r ; /* result allocated block */
xfs_fsblock_t sb ; /* summary file block number */
xfs_buf_t * sumbp ; /* summary file block buffer */
2011-01-25 12:06:19 +03:00
ASSERT ( xfs_isilocked ( mp - > m_rbmip , XFS_ILOCK_EXCL ) ) ;
2005-04-17 02:20:36 +04:00
ASSERT ( minlen > 0 & & minlen < = maxlen ) ;
2011-01-25 12:06:19 +03:00
2005-04-17 02:20:36 +04:00
/*
* If prod is set then figure out what to do to minlen and maxlen .
*/
if ( prod > 1 ) {
xfs_extlen_t i ;
if ( ( i = maxlen % prod ) )
maxlen - = i ;
if ( ( i = minlen % prod ) )
minlen + = prod - i ;
if ( maxlen < minlen ) {
* rtblock = NULLRTBLOCK ;
return 0 ;
}
}
2011-01-25 12:06:19 +03:00
2017-02-17 19:21:06 +03:00
retry :
2005-04-17 02:20:36 +04:00
sumbp = NULL ;
2017-02-17 19:21:06 +03:00
if ( bno = = 0 ) {
2005-04-17 02:20:36 +04:00
error = xfs_rtallocate_extent_size ( mp , tp , minlen , maxlen , len ,
& sumbp , & sb , prod , & r ) ;
2017-02-17 19:21:06 +03:00
} else {
2005-04-17 02:20:36 +04:00
error = xfs_rtallocate_extent_near ( mp , tp , bno , minlen , maxlen ,
len , & sumbp , & sb , prod , & r ) ;
}
2017-02-17 19:21:06 +03:00
2011-01-25 12:06:19 +03:00
if ( error )
2005-04-17 02:20:36 +04:00
return error ;
2011-01-25 12:06:19 +03:00
2005-04-17 02:20:36 +04:00
/*
* If it worked , update the superblock .
*/
if ( r ! = NULLRTBLOCK ) {
long slen = ( long ) * len ;
ASSERT ( * len > = minlen & & * len < = maxlen ) ;
if ( wasdel )
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_RES_FREXTENTS , - slen ) ;
else
xfs_trans_mod_sb ( tp , XFS_TRANS_SB_FREXTENTS , - slen ) ;
2017-02-17 19:21:06 +03:00
} else if ( prod > 1 ) {
prod = 1 ;
goto retry ;
2005-04-17 02:20:36 +04:00
}
2017-02-17 19:21:06 +03:00
2005-04-17 02:20:36 +04:00
* rtblock = r ;
return 0 ;
}
/*
* Initialize realtime fields in the mount structure .
*/
int /* error */
xfs_rtmount_init (
2014-10-02 03:05:32 +04:00
struct xfs_mount * mp ) /* file system mount structure */
2005-04-17 02:20:36 +04:00
{
2014-10-02 03:05:32 +04:00
struct xfs_buf * bp ; /* buffer for last block of subvolume */
struct xfs_sb * sbp ; /* filesystem superblock copy in mount */
xfs_daddr_t d ; /* address of last block of subvolume */
int error ;
2005-04-17 02:20:36 +04:00
sbp = & mp - > m_sb ;
if ( sbp - > sb_rblocks = = 0 )
return 0 ;
if ( mp - > m_rtdev_targp = = NULL ) {
2011-03-07 02:08:35 +03:00
xfs_warn ( mp ,
" Filesystem has a realtime volume, use rtdev=device option " ) ;
2014-06-25 08:58:08 +04:00
return - ENODEV ;
2005-04-17 02:20:36 +04:00
}
mp - > m_rsumlevels = sbp - > sb_rextslog + 1 ;
mp - > m_rsumsize =
( uint ) sizeof ( xfs_suminfo_t ) * mp - > m_rsumlevels *
sbp - > sb_rbmblocks ;
mp - > m_rsumsize = roundup ( mp - > m_rsumsize , sbp - > sb_blocksize ) ;
mp - > m_rbmip = mp - > m_rsumip = NULL ;
/*
* Check that the realtime section is an ok size .
*/
d = ( xfs_daddr_t ) XFS_FSB_TO_BB ( mp , mp - > m_sb . sb_rblocks ) ;
if ( XFS_BB_TO_FSB ( mp , d ) ! = mp - > m_sb . sb_rblocks ) {
2011-03-07 02:08:35 +03:00
xfs_warn ( mp , " realtime mount -- %llu != %llu " ,
2005-04-17 02:20:36 +04:00
( unsigned long long ) XFS_BB_TO_FSB ( mp , d ) ,
( unsigned long long ) mp - > m_sb . sb_rblocks ) ;
2014-06-25 08:58:08 +04:00
return - EFBIG ;
2005-04-17 02:20:36 +04:00
}
2014-10-02 03:05:32 +04:00
error = xfs_buf_read_uncached ( mp - > m_rtdev_targp ,
2010-09-22 04:47:20 +04:00
d - XFS_FSB_TO_BB ( mp , 1 ) ,
2014-10-02 03:05:32 +04:00
XFS_FSB_TO_BB ( mp , 1 ) , 0 , & bp , NULL ) ;
if ( error ) {
2011-03-07 02:08:35 +03:00
xfs_warn ( mp , " realtime device size check failed " ) ;
2014-10-02 03:05:32 +04:00
return error ;
2005-04-17 02:20:36 +04:00
}
xfs_buf_relse ( bp ) ;
return 0 ;
}
/*
xfs: cache minimum realtime summary level
The realtime summary is a two-dimensional array on disk, effectively:
u32 rsum[log2(number of realtime extents) + 1][number of blocks in the bitmap]
rsum[log][bbno] is the number of extents of size 2**log which start in
bitmap block bbno.
xfs_rtallocate_extent_near() uses xfs_rtany_summary() to check whether
rsum[log][bbno] != 0 for any log level. However, the summary array is
stored in row-major order (i.e., like an array in C), so all of these
entries are not adjacent, but rather spread across the entire summary
file. In the worst case (a full bitmap block), xfs_rtany_summary() has
to check every level.
This means that on a moderately-used realtime device, an allocation will
waste a lot of time finding, reading, and releasing buffers for the
realtime summary. In particular, one of our storage services (which runs
on servers with 8 very slow CPUs and 15 8 TB XFS realtime filesystems)
spends almost 5% of its CPU cycles in xfs_rtbuf_get() and
xfs_trans_brelse() called from xfs_rtany_summary().
One solution would be to also store the summary with the dimensions
swapped. However, this would require a disk format change to a very old
component of XFS.
Instead, we can cache the minimum size which contains any extents. We do
so lazily; rather than guaranteeing that the cache contains the precise
minimum, it always contains a loose lower bound which we tighten when we
read or update a summary block. This only uses a few kilobytes of memory
and is already serialized via the realtime bitmap and summary inode
locks, so the cost is minimal. With this change, the same workload only
spends 0.2% of its CPU cycles in the realtime allocator.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2018-12-12 19:46:32 +03:00
* Get the bitmap and summary inodes and the summary cache into the mount
* structure at mount time .
2005-04-17 02:20:36 +04:00
*/
int /* error */
xfs_rtmount_inodes (
xfs_mount_t * mp ) /* file system mount structure */
{
int error ; /* error return value */
xfs_sb_t * sbp ;
sbp = & mp - > m_sb ;
2010-06-24 05:35:17 +04:00
error = xfs_iget ( mp , NULL , sbp - > sb_rbmino , 0 , 0 , & mp - > m_rbmip ) ;
2005-04-17 02:20:36 +04:00
if ( error )
return error ;
ASSERT ( mp - > m_rbmip ! = NULL ) ;
2018-12-13 02:18:52 +03:00
2010-06-24 05:35:17 +04:00
error = xfs_iget ( mp , NULL , sbp - > sb_rsumino , 0 , 0 , & mp - > m_rsumip ) ;
2005-04-17 02:20:36 +04:00
if ( error ) {
2018-07-25 22:52:32 +03:00
xfs_irele ( mp - > m_rbmip ) ;
2005-04-17 02:20:36 +04:00
return error ;
}
ASSERT ( mp - > m_rsumip ! = NULL ) ;
2018-12-22 05:45:18 +03:00
xfs_alloc_rsum_cache ( mp , sbp - > sb_rbmblocks ) ;
2005-04-17 02:20:36 +04:00
return 0 ;
}
2009-02-04 11:33:58 +03:00
void
xfs_rtunmount_inodes (
struct xfs_mount * mp )
{
xfs: cache minimum realtime summary level
The realtime summary is a two-dimensional array on disk, effectively:
u32 rsum[log2(number of realtime extents) + 1][number of blocks in the bitmap]
rsum[log][bbno] is the number of extents of size 2**log which start in
bitmap block bbno.
xfs_rtallocate_extent_near() uses xfs_rtany_summary() to check whether
rsum[log][bbno] != 0 for any log level. However, the summary array is
stored in row-major order (i.e., like an array in C), so all of these
entries are not adjacent, but rather spread across the entire summary
file. In the worst case (a full bitmap block), xfs_rtany_summary() has
to check every level.
This means that on a moderately-used realtime device, an allocation will
waste a lot of time finding, reading, and releasing buffers for the
realtime summary. In particular, one of our storage services (which runs
on servers with 8 very slow CPUs and 15 8 TB XFS realtime filesystems)
spends almost 5% of its CPU cycles in xfs_rtbuf_get() and
xfs_trans_brelse() called from xfs_rtany_summary().
One solution would be to also store the summary with the dimensions
swapped. However, this would require a disk format change to a very old
component of XFS.
Instead, we can cache the minimum size which contains any extents. We do
so lazily; rather than guaranteeing that the cache contains the precise
minimum, it always contains a loose lower bound which we tighten when we
read or update a summary block. This only uses a few kilobytes of memory
and is already serialized via the realtime bitmap and summary inode
locks, so the cost is minimal. With this change, the same workload only
spends 0.2% of its CPU cycles in the realtime allocator.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2018-12-12 19:46:32 +03:00
kmem_free ( mp - > m_rsum_cache ) ;
2009-02-04 11:33:58 +03:00
if ( mp - > m_rbmip )
2018-07-25 22:52:32 +03:00
xfs_irele ( mp - > m_rbmip ) ;
2009-02-04 11:33:58 +03:00
if ( mp - > m_rsumip )
2018-07-25 22:52:32 +03:00
xfs_irele ( mp - > m_rsumip ) ;
2009-02-04 11:33:58 +03:00
}
2005-04-17 02:20:36 +04:00
/*
* Pick an extent for allocation at the start of a new realtime file .
* Use the sequence number stored in the atime field of the bitmap inode .
* Translate this to a fraction of the rtextents , and return the product
* of rtextents and the fraction .
* The fraction sequence is 0 , 1 / 2 , 1 / 4 , 3 / 4 , 1 / 8 , . . . , 7 / 8 , 1 / 16 , . . .
*/
int /* error */
xfs_rtpick_extent (
xfs_mount_t * mp , /* file system mount point */
xfs_trans_t * tp , /* transaction pointer */
xfs_extlen_t len , /* allocation length (rtextents) */
xfs_rtblock_t * pick ) /* result rt extent */
{
xfs_rtblock_t b ; /* result block */
int log2 ; /* log of sequence number */
2017-06-16 21:00:05 +03:00
uint64_t resid ; /* residual after log removed */
uint64_t seq ; /* sequence number of file creation */
uint64_t * seqp ; /* pointer to seqno in inode */
2005-04-17 02:20:36 +04:00
2011-01-25 12:06:19 +03:00
ASSERT ( xfs_isilocked ( mp - > m_rbmip , XFS_ILOCK_EXCL ) ) ;
2017-06-16 21:00:05 +03:00
seqp = ( uint64_t * ) & VFS_I ( mp - > m_rbmip ) - > i_atime ;
2011-01-25 12:06:19 +03:00
if ( ! ( mp - > m_rbmip - > i_d . di_flags & XFS_DIFLAG_NEWRTBM ) ) {
mp - > m_rbmip - > i_d . di_flags | = XFS_DIFLAG_NEWRTBM ;
2005-04-17 02:20:36 +04:00
* seqp = 0 ;
}
seq = * seqp ;
if ( ( log2 = xfs_highbit64 ( seq ) ) = = - 1 )
b = 0 ;
else {
resid = seq - ( 1ULL < < log2 ) ;
b = ( mp - > m_sb . sb_rextents * ( ( resid < < 1 ) + 1ULL ) ) > >
( log2 + 1 ) ;
if ( b > = mp - > m_sb . sb_rextents )
2018-06-08 19:54:22 +03:00
div64_u64_rem ( b , mp - > m_sb . sb_rextents , & b ) ;
2005-04-17 02:20:36 +04:00
if ( b + len > mp - > m_sb . sb_rextents )
b = mp - > m_sb . sb_rextents - len ;
}
* seqp = seq + 1 ;
2011-01-25 12:06:19 +03:00
xfs_trans_log_inode ( tp , mp - > m_rbmip , XFS_ILOG_CORE ) ;
2005-04-17 02:20:36 +04:00
* pick = b ;
return 0 ;
}