xfs: fixes for 3.10 merge window.
cover xfs: fixes for 3.10 merge window.
Hi Ben,
The following 4 patches are for the 3.10 merge window. The first is a repost of the CONFIG_XFS_QUOTA build fix, the second is the CONFIG_XFS_WARN functionality, and the remaining two are updates to the CRC patchset in response to your review comments.
The first patch really needs to go in ASAP to fix the linux-next build problems, the others aren't quite so critical.
Cheers,
Dave.
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
PATCH 1/4 xfs: Teach dquot recovery about CONFIG_XFS_QUOTA
From: Dave Chinner <[email protected]>
Fix a build error when CONFIG_XFS_QUOTA=n:
fs/built-in.o: In function `xlog_recovery_validate_buf_type':/home/dave/src/build/x86-64/xfsdev/fs/xfs/xfs_log_recover.c:1948: undefined reference to `xfs_dquot_buf_ops'
--- fs/xfs/xfs_log_recover.c | 6 ++++++ 1 file changed, 6 insertions(+)
fs/xfs/xfs_log_recover.c +6 -0
@@ -1940,12 +1940,18 @@ xlog_recovery_validate_buf_type(case XFS_BLFT_UDQUOT_BUF :case XFS_BLFT_PDQUOT_BUF :case XFS_BLFT_GDQUOT_BUF : +#ifdef CONFIG_XFS_QUOTA if (magic16 !=XFS_DQUOT_MAGIC ) {xfs_warn (mp,"Bad DQUOT block magic!" );ASSERT (0 );break ; } bp->b_ops = &xfs_dquot_buf_ops; +#else +xfs_alert (mp, +"Trying to recover dquots without QUOTA support built in!" ); +ASSERT (0 ); +#endif break ;case XFS_BLFT_DINO_BUF :/*
Signature · 6 lines
1.7.10.4 _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Re: [PATCH 1/4] xfs: Teach dquot recovery about CONFIG_XFS_QUOTA
Quoted 10 lines · Dave Chinner
From: Dave Chinner <[email protected]>
Fix a build error when
CONFIG_XFS_QUOTA=n:
fs/built-in.o: In function `xlog_recovery_validate_buf_type':/home/dave/src/build/x86-64/xfsdev/fs/xfs/xfs_log_recover.c:1948: undefined reference to `xfs_dquot_buf_ops'Reported-by Michael L. Semon <[email protected]>Signed-off-by Dave Chinner <[email protected]>
Applied fine this time. Must have been a PEBKAC.
Quoted 13 lines
--- fs/xfs/xfs_log_recover.c | 6 ++++++ 1 file changed, 6 insertions(+)fs/xfs/xfs_log_recover.c +1
-0@@ -1940,12 +1940,18 @@ xlog_recovery_validate_buf_type(case XFS_BLFT_UDQUOT_BUF :case XFS_BLFT_PDQUOT_BUF :case XFS_BLFT_GDQUOT_BUF : +#ifdef CONFIG_XFS_QUOTA
fs/xfs/xfs_log_recover.c xlog_recovery_validate_buf_type( · line 1953
Quoted 10 lines
fs/xfs/xfs_log_recover.c +4
-0if (magic16 !=XFS_DQUOT_MAGIC ) {xfs_warn (mp,"Bad DQUOT block magic!" );ASSERT (0 );break ; } bp->b_ops = &xfs_dquot_buf_ops; +#else +xfs_alert (mp, +"Trying to recover dquots without QUOTA support built in!" ); +ASSERT (0 );
Maybe it makes sense for recovery to fail in that case.
-Ben
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
PATCH 2/4 xfs: introduce CONFIG_XFS_WARN
In reply to Dave ChinnerFrom: Dave Chinner <[email protected]>
Running a CONFIG_XFS_DEBUG kernel in production environments is not the best idea as it introduces significant overhead, can change the behaviour of algorithms (such as allocation) to improve test coverage, and (most importantly) panic the machine on non-fatal errors.
There are many cases where all we want to do is run a kernel with more bounds checking enabled, such as is provided by the ASSERT() statements throughout the code, but without all the potential overhead and drawbacks.
This patch converts all the ASSERT statements to evaluate as WARN_ON(1) statements and hence if they fail dump a warning and a stack trace to the log. This has minimal overhead and does not change any algorithms, and will allow us to find strange "out of bounds" problems more easily on production machines.
There are a few places where assert statements contain debug only code. These are converted to be debug-or-warn only code so that we still get all the assert checks in the code.
--- fs/xfs/Kconfig | 13 +++++++++++++ fs/xfs/mrlock.h | 12 ++++++------ fs/xfs/xfs.h | 5 +++++ fs/xfs/xfs_alloc_btree.c | 4 ++-- fs/xfs/xfs_bmap_btree.c | 4 ++-- fs/xfs/xfs_btree.h | 2 +- fs/xfs/xfs_dir2_node.c | 4 ++-- fs/xfs/xfs_ialloc_btree.c | 4 ++-- fs/xfs/xfs_inode.c | 2 +- fs/xfs/xfs_linux.h | 24 ++++++++++++++++++------ fs/xfs/xfs_message.c | 8 ++++++++ fs/xfs/xfs_message.h | 1 + fs/xfs/xfs_trans.h | 4 ++-- 13 files changed, 63 insertions(+), 24 deletions(-)
fs/xfs/Kconfig +13 -0
@@ -69,6 +69,19 @@ config XFS_RT If unsure, say N. +config XFS_WARN + bool "XFS Verbose Warnings" + depends on XFS_FS && !XFS_DEBUG + help + Say Y here to get an XFS build with many additional warnings. + It converts ASSERT checks to WARN, so will log any out-of-bounds + conditions that occur that would otherwise be missed. It is much + lighter weight than XFS_DEBUG and does not modify algorithms and will + not cause the kernel to panic on non-fatal errors. + + However, similar to XFS_DEBUG, it is only advisable to use this if you + are debugging a particular problem. + config XFS_DEBUG bool "XFS Debugging support" depends on XFS_FS
fs/xfs/mrlock.h +6 -6
@@ -22,12 +22,12 @@typedef struct {struct rw_semaphore mr_lock; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN )int mr_writer;#endif }mrlock_t ; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN )#define mrinit (mrp, name) \ do { (mrp)->mr_writer = 0; init_rwsem(&(mrp)->mr_lock); } while (0)#else
@@ -46,7 +46,7 @@ static inline void mraccess_nested(mrlock_t *mrp, int subclass)static inline void mrupdate_nested (mrlock_t *mrp,int subclass) {down_write_nested (&mrp->mr_lock, subclass); -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) mrp->mr_writer =1 ;#endif }
@@ -60,7 +60,7 @@ static inline int mrtryupdate(mrlock_t *mrp) {if (!down_write_trylock (&mrp->mr_lock))return 0 ; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) mrp->mr_writer =1 ;#endif return 1 ;
@@ -68,7 +68,7 @@ static inline int mrtryupdate(mrlock_t *mrp)static inline void mrunlock_excl (mrlock_t *mrp) { -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) mrp->mr_writer =0 ;#endif up_write (&mrp->mr_lock);
@@ -81,7 +81,7 @@ static inline void mrunlock_shared(mrlock_t *mrp)static inline void mrdemote (mrlock_t *mrp) { -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) mrp->mr_writer =0 ;#endif downgrade_write (&mrp->mr_lock);
fs/xfs/xfs.h +5 -0
@@ -24,6 +24,11 @@#define XFS_BUF_LOCK_TRACKING 1#endif +#ifdef CONFIG_XFS_WARN +#define XFS_WARN 1 +#endif + +#include "xfs_linux.h" #endif /* __XFS_H__ */
fs/xfs/xfs_alloc_btree.c +2 -2
@@ -386,7 +386,7 @@ const struct xfs_buf_ops xfs_allocbt_buf_ops = { }; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) STATIC int xfs_allocbt_keys_inorder( struct xfs_btree_cur *cur,
@@ -442,7 +442,7 @@ static const struct xfs_btree_ops xfs_allocbt_ops = { .init_ptr_from_cur = xfs_allocbt_init_ptr_from_cur, .key_diff = xfs_allocbt_key_diff, .buf_ops = &xfs_allocbt_buf_ops, -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) .keys_inorder = xfs_allocbt_keys_inorder, .recs_inorder = xfs_allocbt_recs_inorder,#endif
fs/xfs/xfs_bmap_btree.c +2 -2
@@ -813,7 +813,7 @@ const struct xfs_buf_ops xfs_bmbt_buf_ops = { }; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) STATIC int xfs_bmbt_keys_inorder( struct xfs_btree_cur *cur,
@@ -853,7 +853,7 @@ static const struct xfs_btree_ops xfs_bmbt_ops = { .init_ptr_from_cur = xfs_bmbt_init_ptr_from_cur, .key_diff = xfs_bmbt_key_diff, .buf_ops = &xfs_bmbt_buf_ops, -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) .keys_inorder = xfs_bmbt_keys_inorder, .recs_inorder = xfs_bmbt_recs_inorder,#endif
fs/xfs/xfs_btree.h +1 -1
@@ -215,7 +215,7 @@ struct xfs_btree_ops {const struct xfs_buf_ops *buf_ops; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN )/* check that k1 is lower than k2 */ int (*keys_inorder)(struct xfs_btree_cur *cur, union xfs_btree_key *k1,
fs/xfs/xfs_dir2_node.c +2 -2
@@ -993,7 +993,7 @@ xfs_dir2_leafn_rebalance(xfs_dir2_leaf_t *leaf1;/* first leaf structure */ xfs_dir2_leaf_t *leaf2;/* second leaf structure */ int mid;/* midpoint leaf index */ -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN )int oldstale;/* old count of stale leaves */ #endif int oldsum;/* old total leaf count */
@@ -1022,7 +1022,7 @@ xfs_dir2_leafn_rebalance( ents2 =xfs_dir3_leaf_ents_p (leaf2); oldsum = hdr1.count + hdr2.count; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) oldstale = hdr1.stale + hdr2.stale;#endif mid = oldsum >>1 ;
fs/xfs/xfs_ialloc_btree.c +2 -2
@@ -272,7 +272,7 @@ const struct xfs_buf_ops xfs_inobt_buf_ops = { .verify_write = xfs_inobt_write_verify, }; -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) STATIC int xfs_inobt_keys_inorder( struct xfs_btree_cur *cur,
@@ -310,7 +310,7 @@ static const struct xfs_btree_ops xfs_inobt_ops = { .init_ptr_from_cur = xfs_inobt_init_ptr_from_cur, .key_diff = xfs_inobt_key_diff, .buf_ops = &xfs_inobt_buf_ops, -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) .keys_inorder = xfs_inobt_keys_inorder, .recs_inorder = xfs_inobt_recs_inorder,#endif
fs/xfs/xfs_inode.c +1 -1
@@ -287,7 +287,7 @@ xfs_ilock_demote(trace_xfs_ilock_demote (ip, lock_flags, _RET_IP_); } -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN ) int xfs_isilocked( xfs_inode_t *ip,
fs/xfs/xfs_linux.h +18 -6
@@ -293,22 +293,34 @@ static inline __uint64_t howmany_64(__uint64_t x, __uint32_t y)#define ASSERT_ALWAYS (expr) \ (unlikely(expr) ? (void)0 : assfail(#expr, __FILE__, __LINE__)) -#ifndef DEBUG -#define ASSERT (expr) ((void)0) +#ifdef DEBUG +#define ASSERT (expr) \ + (unlikely(expr) ? (void)0 : assfail(#expr, __FILE__, __LINE__))#ifndef STATIC -# define STATIC static noinline +# define STATIC noinline#endif -#else /* DEBUG */ +#else /* !DEBUG */ + +#ifdef XFS_WARN #define ASSERT (expr) \ - (unlikely(expr) ? (void)0 : assfail(#expr, __FILE__, __LINE__)) + (unlikely(expr) ? (void)0 : asswarn(#expr, __FILE__, __LINE__))#ifndef STATIC -# define STATIC noinline +# define STATIC static noinline +#endif + +#else /* !DEBUG && !XFS_WARN */ + +#define ASSERT (expr) ((void)0) + +#ifndef STATIC +# define STATIC static noinline#endif +#endif /* XFS_WARN */ #endif /* DEBUG */ #endif /* __XFS_LINUX__ */
fs/xfs/xfs_message.c +8 -0
@@ -93,6 +93,14 @@ xfs_alert_tag( }void +asswarn (char *expr,char *file,int line) +{ +xfs_warn (NULL ,"Assertion failed: %s, file: %s, line: %d" , + expr, file, line); +WARN_ON (1 ); +} + +void assfail (char *expr,char *file,int line) { xfs_emerg(NULL, "Assertion failed: %s, file: %s, line: %d",
fs/xfs/xfs_message.h +1 -0
@@ -57,6 +57,7 @@ do { \ xfs_printk_ratelimited(xfs_debug, dev, fmt, ##__VA_ARGS__)extern void assfail (char *expr,char *f,int l); +extern void asswarn (char *expr,char *f,int l);extern void xfs_hex_dump (void *p,int length);
fs/xfs/xfs_trans.h +2 -2
@@ -405,7 +405,7 @@ typedef struct xfs_trans {int64_t t_res_fdblocks_delta;/* on-disk only chg */ int64_t t_frextents_delta;/* superblock freextents chg*/ int64_t t_res_frextents_delta;/* on-disk only chg */ -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN )int64_t t_ag_freeblks_delta;/* debugging counter */ int64_t t_ag_flist_delta;/* debugging counter */ int64_t t_ag_btree_delta;/* debugging counter */
@@ -433,7 +433,7 @@ typedef struct xfs_trans {#define xfs_trans_get_block_res (tp) ((tp)->t_blk_res)#define xfs_trans_set_sync (tp) ((tp)->t_flags |= XFS_TRANS_SYNC) -#ifdef DEBUG +#if defined(DEBUG ) || defined(XFS_WARN )#define xfs_trans_agblocks_delta (tp, d) ((tp)->t_ag_freeblks_delta += (int64_t)d)#define xfs_trans_agflist_delta (tp, d) ((tp)->t_ag_flist_delta += (int64_t)d)#define xfs_trans_agbtree_delta (tp, d) ((tp)->t_ag_btree_delta += (int64_t)d)
Signature · 6 lines
1.7.10.4 _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Ben Myers
AppliedRe: [PATCH 2/4] xfs: introduce CONFIG_XFS_WARN
Quoted 19 lines · Dave Chinner
From: Dave Chinner <[email protected]>
Running a
CONFIG_XFS_DEBUGkernel in production environments is not the best idea as it introduces significant overhead, can change the behaviour of algorithms (such as allocation) to improve test coverage, and (most importantly) panic the machine on non-fatal errors.There are many cases where all we want to do is run a kernel with more bounds checking enabled, such as is provided by the
ASSERT()statements throughout the code, but without all the potential overhead and drawbacks.This patch converts all the ASSERT statements to evaluate as
WARN_ON(1)statements and hence if they fail dump a warning and a stack trace to the log. This has minimal overhead and does not change any algorithms, and will allow us to find strange "out of bounds" problems more easily on production machines.
Quoted 5 lines
There are a few places where assert statements contain debug only code. These are converted to be debug-or-warn only code so that we still get all the assert checks in the code.
Signed-off-by Dave Chinner <[email protected]>
Applied.
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
PATCH 3/4 xfs: Remote attr validation fixes and optimisations
In reply to Dave ChinnerFrom: Dave Chinner <[email protected]>
- optimise the calcuation for the number of blocks in a remote xattr.
- check attribute length against MAX_XATTR_SIZE, not MAXPATHLEN
- whitespace fixes
--- fs/xfs/xfs_attr_remote.c | 19 +++++-------------- 1 file changed, 5 insertions(+), 14 deletions(-)
fs/xfs/xfs_attr_remote.c +5 -14
@@ -52,15 +52,9 @@ xfs_attr3_rmt_blocks(struct xfs_mount *mp, int attrlen) { - int fsblocks = 0; -int len = attrlen; - -do { - fsblocks++; - len -=XFS_ATTR3_RMT_BUF_SPACE (mp, mp->m_sb.sb_blocksize); - }while (len >0 ); - -return fsblocks; + int buflen = XFS_ATTR3_RMT_BUF_SPACE(mp, + mp->m_sb.sb_blocksize); +return (attrlen + buflen -1 ) / buflen; } static bool
@@ -79,7 +73,7 @@ xfs_attr3_rmt_verify(if (bp->b_bn !=be64_to_cpu (rmt->rm_blkno))return false;if (be32_to_cpu (rmt->rm_offset) + -be32_to_cpu (rmt->rm_bytes) >=MAXPATHLEN ) +be32_to_cpu (rmt->rm_bytes) >=XATTR_SIZE_MAX )return false;if (rmt->rm_owner ==0 )return false;
@@ -183,7 +177,6 @@ xfs_attr3_rmt_hdr_ok(/* ok */ return true; - }/*
@@ -367,7 +360,6 @@ xfs_attr_rmtval_set(* spill for another block every 9 headers we require in this * loop. */ -if (crcs && blkcnt ==0 ) {int total_len;
@@ -422,9 +414,8 @@ xfs_attr_rmtval_set( byte_cnt =BBTOB (bp->b_length); byte_cnt =XFS_ATTR3_RMT_BUF_SPACE (mp, byte_cnt); -if (valuelen < byte_cnt) { +if (valuelen < byte_cnt) byte_cnt = valuelen; - } buf = bp->b_addr; buf += xfs_attr3_rmt_hdr_set(mp, dp->i_ino, offset,
Signature · 6 lines
1.7.10.4 _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Re: [PATCH 3/4] xfs: Remote attr validation fixes and optimisations
Hi Dave,
Quoted 8 lines · Dave Chinner
From: Dave Chinner <[email protected]>
- optimise the calcuation for the number of blocks in a remote xattr.
- check attribute length againstMAX_XATTR_SIZE, not MAXPATHLEN
- whitespace fixesReviewed-by Dave Chinner <[email protected]>
Signed-off-by:
I guess this is what you intended?
Else, looks good to me.
Thanks,
Ben
Quoted 70 lines
--- fs/xfs/xfs_attr_remote.c | 19 +++++-------------- 1 file changed, 5 insertions(+), 14 deletions(-)fs/xfs/xfs_attr_remote.c +5
-14@@ -52,15 +52,9 @@ xfs_attr3_rmt_blocks(struct xfs_mount *mp, int attrlen) { - int fsblocks = 0; -int len = attrlen; - -do { - fsblocks++; - len -=XFS_ATTR3_RMT_BUF_SPACE (mp, mp->m_sb.sb_blocksize); - }while (len >0 ); - -return fsblocks; + int buflen = XFS_ATTR3_RMT_BUF_SPACE(mp, + mp->m_sb.sb_blocksize); +return (attrlen + buflen -1 ) / buflen; } static bool@@ -79,7 +73,7 @@ xfs_attr3_rmt_verify(if (bp->b_bn !=be64_to_cpu (rmt->rm_blkno))return false;if (be32_to_cpu (rmt->rm_offset) + -be32_to_cpu (rmt->rm_bytes) >=MAXPATHLEN ) +be32_to_cpu (rmt->rm_bytes) >=XATTR_SIZE_MAX )return false;if (rmt->rm_owner ==0 )return false;@@ -183,7 +177,6 @@ xfs_attr3_rmt_hdr_ok(/* ok */ return true; - }/* @@ -367,7 +360,6 @@ xfs_attr_rmtval_set(* spill for another block every 9 headers we require in this * loop. */ -if (crcs && blkcnt ==0 ) {int total_len;@@ -422,9 +414,8 @@ xfs_attr_rmtval_set( byte_cnt =BBTOB (bp->b_length); byte_cnt =XFS_ATTR3_RMT_BUF_SPACE (mp, byte_cnt); -if (valuelen < byte_cnt) { +if (valuelen < byte_cnt) byte_cnt = valuelen; - } buf = bp->b_addr; buf += xfs_attr3_rmt_hdr_set(mp, dp->i_ino, offset,Signature · 6 lines
1.7.10.4 _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
Re: [PATCH 3/4] xfs: Remote attr validation fixes and optimisations
Quoted 10 lines · Ben Myers
Hi Dave,
Quoted 7 lines · Dave Chinner
From: Dave Chinner <[email protected]>
- optimise the calcuation for the number of blocks in a remote xattr.
- check attribute length againstMAX_XATTR_SIZE, not MAXPATHLEN
- whitespace fixes
Quoted 4 lines
Quoted 1 line
Reviewed-by Dave Chinner <[email protected]>Signed-off-by:
I guess this is what you intended?
Er, yeah. Wrong vim paste macro....
Cheers,
Dave.
Signature · 7 lines
Dave Chinner [email protected] _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
PATCH 4/4 xfs: fix da node magic number mismatches
In reply to Dave ChinnerFrom: Dave Chinner <[email protected]>
--- fs/xfs/xfs_da_btree.c | 2 +- fs/xfs/xfs_log_recover.c | 3 +-- 2 files changed, 2 insertions(+), 3 deletions(-)
fs/xfs/xfs_da_btree.c +1 -1
@@ -310,8 +310,8 @@ xfs_da3_node_read(int type;switch (be16_to_cpu (info->magic)) { -case XFS_DA3_NODE_MAGIC :case XFS_DA_NODE_MAGIC : +case XFS_DA3_NODE_MAGIC : type =XFS_BLFT_DA_NODE_BUF ;break ;case XFS_ATTR_LEAF_MAGIC :
fs/xfs/xfs_log_recover.c +1 -2
@@ -2039,7 +2039,7 @@ xlog_recovery_validate_buf_type(case XFS_BLFT_ATTR_RMT_BUF :if (!xfs_sb_version_hascrc (&mp->m_sb))break ; -if (magicda !=XFS_ATTR3_RMT_MAGIC ) { +if (magic32 !=XFS_ATTR3_RMT_MAGIC ) {xfs_warn (mp,"Bad attr remote magic!" );ASSERT (0 );break ;
@@ -2135,7 +2135,6 @@ xlog_recover_do_reg_buffer(ASSERT (i == item->ri_total);xlog_recovery_validate_buf_type (mp, bp, buf_f); - }/*
Signature · 6 lines
1.7.10.4 _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Re: [PATCH 4/4] xfs: fix da node magic number mismatches
Quoted 3 lines · Dave Chinner
From: Dave Chinner <[email protected]>
Signed-off-by Dave Chinner <[email protected]>
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs