metadump: don't verify metadata being dumped
PATCH 1/1 metadump: don't verify metadata being dumped
From: Dave Chinner <[email protected]>
The discontiguous buffer support series added a verifier check on the metadata buffers before they go written to the metadump image. If this failed, it returned an error, and the restul woul dbe that we stopped processing the metadata and exited, resulting in a truncated dump.
xfs_metadump is supposed to dump the metadata in the filesystem for forensic analysis purposes, which means we actually want it to retain any corruptions it finds in the filesystem. Hence running the verifier - even to recalculate CRCs - is the wrong thing to be doing. And stopping the dum pwhen we come across an error is even worse.
Therefore remove the code tha truns the verifier and causes all these problems and replace it with a comment explaining why we don't want to run verifiers in the metadump process.
--- db/metadump.c | 23 +++++++---------------- 1 file changed, 7 insertions(+), 16 deletions(-)
db/metadump.c +7 -16
@@ -190,6 +190,13 @@ write_buf_segment(return 0 ; } +/* +* we want to preserve the state of the metadata in the dump - whether it is +* intact or corrupt, so even if the buffer has a verifier attached to it we +* don't want to run it prior to writing the buffer to the metadump image. +* Even just recalculating the CRCs is the wrong thing to do here as it can +* hide errors that only the CRCs were detecting. +*/ static int write_buf (iocur_t *buf)
@@ -197,22 +204,6 @@ write_buf(int i;int ret; -/* -* Run the write verifier to recalculate the buffer CRCs and check -* we are writing something valid to disk -*/ -if (buf->bp && buf->bp->b_ops) { - buf->bp->b_error =0 ; - buf->bp->b_ops->verify_write (buf->bp); -if (buf->bp->b_error) { -fprintf (stderr, -_ ("%s: write verifer failed on bno 0x%llx/0x%x\n" ), - __func__, (long long )buf->bp->b_bn, - buf->bp->b_bcount); -return -buf->bp->b_error; - } - } -/* handle discontiguous buffers */ if (!buf->bbmap) { ret =write_buf_segment (buf->data, buf->bb, buf->blen);
Signature · 6 lines
1.9.0 _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Quoted 15 lines · Dave Chinner
From: Dave Chinner <[email protected]>
The discontiguous buffer support series added a verifier check on the metadata buffers before they go written to the metadump image. If this failed, it returned an error, and the restul woul dbe that we stopped processing the metadata and exited, resulting in a truncated dump.
xfs_metadumpis supposed to dump the metadata in the filesystem for forensic analysis purposes, which means we actually want it to retain any corruptions it finds in the filesystem. Hence running the verifier - even to recalculate CRCs - is the wrong thing to be doing. And stopping the dum pwhen we come across an error is even worse.
Quoted 3 lines
Therefore remove the code tha truns the verifier and causes all these problems and replace it with a comment explaining why we don't want to run verifiers in the metadump process.
This leaves the net functional change from
8ab75c db: enable metadump on CRC filesystems
as:
@@ -1727,6 +1743,9 @@ copy_inode_chunk(
if (!process_inode(agno, agino + i, dip))
goto pop_out;
+
+ /* calculate the new CRC for the inode */
+ xfs_dinode_calc_crc(mp, dip);
}
skip_processing:
if (!write_buf(iocur_top))which seems a) minimal, but also b) like we shouldn't be recalculating CRCs if the point is to copy out existing fs state...?
OTOH if we're obfuscating, we would HAVE to recalculate CRCs, but then would lose the info that the CRC was bad before.
So probably should skip CRC recalculating if the original CRC was bad, in the obfuscating case?
Maybe this patch stands ok on its own but it seems like there's more work to do. :)
-Eric
Quoted 37 lines
Signed-off-by Dave Chinner <[email protected]>--- db/metadump.c | 23 +++++++---------------- 1 file changed, 7 insertions(+), 16 deletions(-)db/metadump.c +7
-9@@ -190,6 +190,13 @@ write_buf_segment(return 0 ; } +/* +* we want to preserve the state of the metadata in the dump - whether it is +* intact or corrupt, so even if the buffer has a verifier attached to it we +* don't want to run it prior to writing the buffer to the metadump image. +* Even just recalculating the CRCs is the wrong thing to do here as it can +* hide errors that only the CRCs were detecting. +*/ static int write_buf (iocur_t *buf)@@ -197,22 +204,6 @@ write_buf(int i;int ret; -/* -* Run the write verifier to recalculate the buffer CRCs and check -* we are writing something valid to disk -*/ -if (buf->bp && buf->bp->b_ops) { - buf->bp->b_error =0 ; - buf->bp->b_ops->verify_write (buf->bp); -if (buf->bp->b_error) { - fprintf(stderr,
db/metadump.c write_buf( · line 209
Quoted 10 lines
db/metadump.c +0
-7-_ ("%s: write verifer failed on bno 0x%llx/0x%x\n" ), - __func__, (long long )buf->bp->b_bn, - buf->bp->b_bcount); -return -buf->bp->b_error; - } - } -/* handle discontiguous buffers */ if (!buf->bbmap) { ret =write_buf_segment (buf->data, buf->bb, buf->blen);
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
Quoted 35 lines · Eric Sandeen
Quoted 15 lines · Dave Chinner
From: Dave Chinner <[email protected]>
The discontiguous buffer support series added a verifier check on the metadata buffers before they go written to the metadump image. If this failed, it returned an error, and the restul woul dbe that we stopped processing the metadata and exited, resulting in a truncated dump.
xfs_metadumpis supposed to dump the metadata in the filesystem for forensic analysis purposes, which means we actually want it to retain any corruptions it finds in the filesystem. Hence running the verifier - even to recalculate CRCs - is the wrong thing to be doing. And stopping the dum pwhen we come across an error is even worse.Quoted 3 lines
Therefore remove the code tha truns the verifier and causes all these problems and replace it with a comment explaining why we don't want to run verifiers in the metadump process.
This leaves the net functional change from
8ab75c db: enable metadump on CRC filesystems
as:(quoted diff) +3
-0@@ -1727,6 +1743,9 @@ copy_inode_chunk( if (!process_inode(agno, agino + i, dip)) goto pop_out; + + /* calculate the new CRC for the inode */ + xfs_dinode_calc_crc(mp, dip); } skip_processing: if (!write_buf(iocur_top))
Quoted 8 lines
which seems a) minimal, but also b) like we shouldn't be recalculating CRCs if the point is to copy out existing fs state...?
OTOH if we're obfuscating, we would HAVE to recalculate CRCs, but then would lose the info that the CRC was bad before.
So probably should skip CRC recalculating if the original CRC was bad, in the obfuscating case?
Yeah, you are right - I wasn't thinking of the obfuscation case changing the metadata....
Quoted 2 lines
Maybe this patch stands ok on its own but it seems like there's more work to do. :)
I'll rework it.
Cheers,
Dave.
Signature · 7 lines
Dave Chinner [email protected] _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
PATCH v2 1/1 metadump: don't verify metadata being dumped
In reply to Dave Chinnermetadump: don't verify metadata being dumped
From: Dave Chinner <[email protected]>
The discontiguous buffer support series added a verifier check on the metadata buffers before they go written to the metadump image. If this failed, it returned an error, and the restul woul dbe that we stopped processing the metadata and exited, resulting in a truncated dump.
xfs_metadump is supposed to dump the metadata in the filesystem for forensic analysis purposes, which means we actually want it to retain any corruptions it finds in the filesystem. Hence running the verifier - even to recalculate CRCs - when the metadata is unmodified is the wrong thing to be doing. And stopping the dump when we come across an error is even worse.
We still need to do CRC recalculation when obfuscating names and attributes. Hence we need to make running the verifier conditional on the buffer or inode:
a) being uncorrupted when read, and
b) modified by the obfuscation code.
If either of these conditions is not true, then we don't run the verifier or recalculate the CRCs.
---
V2: run verifiers on buffers and inodes modified by obfuscation, but only if they are not corrupt before obfuscation. Thanks, Eric!
db/io.h | 1 + db/metadump.c | 57 ++++++++++++++++++++++++++++++++++++++++++++------------- 2 files changed, 45 insertions(+), 13 deletions(-)
db/io.h +1 -0
@@ -41,6 +41,7 @@ typedef struct iocur { int ino_crc_ok:1; int ino_buf:1; int dquot_buf:1; + int need_crc:1; } iocur_t;#define DB_RING_ADD 1/* add to ring on set_cur */
db/metadump.c +44 -13
126 lines not on this page · raw message
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
Quoted 1 line · Dave Chinner
metadump: don't verify metadata being dumped
Not a complete summary anymore...
Quoted 5 lines
From: Dave Chinner <[email protected]>
The discontiguous buffer support series added a verifier check on the metadata buffers before they go written to the metadump image. If this failed, it returned an error, and the restul woul dbe that
wheny ou type too fasty ou end up with thes ame sort of typosa sIdo. ;)
Quoted 10 lines
we stopped processing the metadata and exited, resulting in a truncated dump.
xfs_metadumpis supposed to dump the metadata in the filesystem for forensic analysis purposes, which means we actually want it to retain any corruptions it finds in the filesystem. Hence running the verifier - even to recalculate CRCs - when the metadata is unmodified is the wrong thing to be doing. And stopping the dump when we come across an error is even worse.
Quoted 8 lines
We still need to do CRC recalculation when obfuscating names and attributes. Hence we need to make running the verifier conditional on the buffer or inode:
a) being uncorrupted when read, and
b) modified by the obfuscation code.If either of these conditions is not true, then we don't run the verifier or recalculate the CRCs.
I think this looks mostly ok; small questions & nitpicks below.
Quoted 73 lines
Signed-off-by Dave Chinner <[email protected]>---
V2: run verifiers on buffers and inodes modified by obfuscation, but only if they are not corrupt before obfuscation. Thanks, Eric!db/io.h | 1 + db/metadump.c | 57 ++++++++++++++++++++++++++++++++++++++++++++------------- 2 files changed, 45 insertions(+), 13 deletions(-)db/io.h +1
-0@@ -41,6 +41,7 @@ typedef struct iocur { int ino_crc_ok:1; int ino_buf:1; int dquot_buf:1; + int need_crc:1; } iocur_t;#define DB_RING_ADD 1/* add to ring on set_cur */ db/metadump.c +20
-10@@ -190,26 +190,36 @@ write_buf_segment(return 0 ; } +/* +* we want to preserve the state of the metadata in the dump - whether it is +* intact or corrupt, so even if the buffer has a verifier attached to it we +* don't want to run it prior to writing the buffer to the metadump image. +* +* The only reason for running the verifier is to recalculate the CRCs on a +* buffer that has been obfuscated. i.e. a buffer than metadump modified itself. +* In this case, we only run the verifier if the buffer was not corrupt to begin +* with so that we don't accidentally correct buffers with CRC or errors in them +* when we are obfuscating them. +*/ static int write_buf (iocur_t *buf) { +struct xfs_buf *bp = buf->bp;int i;int ret;/* * Run the write verifier to recalculate the buffer CRCs and check -* we are writing something valid to disk +* metadump didn't introduce a new corruption. Warn if the verifier +* failed, but still continue to dump it into the output file. */ -if (buf->bp && buf->bp->b_ops) { - buf->bp->b_error =0 ; - buf->bp->b_ops->verify_write (buf->bp); -if (buf->bp->b_error) { -fprintf (stderr, -_ ("%s: write verifer failed on bno 0x%llx/0x%x\n" ), - __func__, (long long )buf->bp->b_bn, - buf->bp->b_bcount); -return -buf->bp->b_error; +if (buf->need_crc && bp && bp->b_ops && !bp->b_error) { + bp->b_ops->verify_write (bp); +if (bp->b_error) { +print_warning ( +"obfuscation corrupted block at bno 0x%llx/0x%x" , + (long long )bp->b_bn, bp->b_bcount); } }
db/metadump.c process_single_fsb_objects( · line 1387
Quoted 4 lines
db/metadump.c +1
-0@@ -1374,6 +1384,7 @@ process_single_fsb_objects( o++; dp += mp->m_sb.sb_blocksize; } + iocur_top->need_crc =1 ;
in the
default:
case we don't obfuscate. Should it still get need_crc?
Quoted 26 lines
db/metadump.c +8
-0ret =write_buf (iocur_top); out_pop:@@ -1442,6 +1453,7 @@ process_multi_fsb_objects(obfuscate_dir_data_block (iocur_top->data, o, last == mp->m_dirblkfsbs); + iocur_top->need_crc =1 ; ret =write_buf (iocur_top); out_pop:pop_cur ();@@ -1722,6 +1734,13 @@ process_inode_data(return 1 ; } +/* +* when we process the inode, we may change the data in the data and/or +* attribute fork if they are in short form and we are obfuscating names. +* In this case we need to recalculate the CRC of the inode, but we should +* only do that if the CRC in the inode is good to begin with. If the crc +* is not ok, we just leave it alone. +*/ static int process_inode( xfs_agnumber_t agno,
db/metadump.c process_inode( · line 1752
Quoted 5 lines
db/metadump.c +2
-0@@ -1729,17 +1748,28 @@ process_inode( xfs_dinode_t *dip) {int success; +bool crc_ok = false;/* don't recalc by default */ +bool need_crc = false;
I might do
+ bool crc_was_ok = false; /* don't recalc by default */
+ bool need_new_crc = false;
for clarity...?
Quoted 4 lines
db/metadump.c +1
-0success =1 ; cur_ino =XFS_AGINO_TO_INO (mp, agno, agino); +/* we only care about crc recalculation if we are obfuscating names. */
db/metadump.c process_inode( · line 1767
Quoted 10 lines
db/metadump.c +6
-0+if (!dont_obfuscate) + crc_ok =xfs_verify_cksum ((char *)dip, mp->m_sb.sb_inodesize, + offsetof(struct xfs_dinode , di_crc)); +/* copy appropriate data fork metadata */ switch (be16_to_cpu (dip->di_mode) &S_IFMT ) {case S_IFDIR : success =process_inode_data (dip,TYP_DIR2 ); +if (dip->di_format ==XFS_DINODE_FMT_LOCAL ) + need_crc =1 ;
I wish this were closer to the point of obfuscation, but oh well.
Quoted 37 lines
db/metadump.c +6
-3break ;case S_IFLNK : success =process_inode_data (dip,TYP_SYMLINK ); +if (dip->di_format ==XFS_DINODE_FMT_LOCAL ) + need_crc =1 ;break ;case S_IFREG : success =process_inode_data (dip,TYP_DATA );@@ -1754,6 +1784,7 @@ process_inode( attr_data.remote_val_count =0 ;switch (dip->di_aformat) {case XFS_DINODE_FMT_LOCAL : + need_crc =1 ;if (!dont_obfuscate)obfuscate_sf_attr (dip);break ;@@ -1768,6 +1799,9 @@ process_inode( }nametable_clear (); } + +if (crc_ok && need_crc) +xfs_dinode_calc_crc (mp, dip);return success; }@@ -1838,9 +1872,6 @@ copy_inode_chunk(if (!process_inode (agno, agino + i, dip)) goto pop_out; - -/* calculate the new CRC for the inode */ -xfs_dinode_calc_crc (mp, dip); } skip_processing:if (write_buf (iocur_top))
Quoted 5 lines
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
Quoted 9 lines · Eric Sandeen
db/metadump.c process_single_fsb_objects( · line 1387
Quoted 5 lines · Dave Chinner
db/metadump.c +1
-0@@ -1374,6 +1384,7 @@ process_single_fsb_objects( o++; dp += mp->m_sb.sb_blocksize; } + iocur_top->need_crc =1 ;in the
Quoted 3 lines
default:
case we don't obfuscate. Should it still get
need_crc?
Ah, right. I looked at the "dont_obfuscate" jump, not the switch statement. Will fix.
Quoted 22 lines
Quoted 11 lines
db/metadump.c +7
-0+/* +* when we process the inode, we may change the data in the data and/or +* attribute fork if they are in short form and we are obfuscating names. +* In this case we need to recalculate the CRC of the inode, but we should +* only do that if the CRC in the inode is good to begin with. If the crc +* is not ok, we just leave it alone. +*/ static int process_inode( xfs_agnumber_t agno,db/metadump.c process_inode( · line 1752
Quoted 5 lines
db/metadump.c +2
-0@@ -1729,17 +1748,28 @@ process_inode( xfs_dinode_t *dip) {int success; +bool crc_ok = false;/* don't recalc by default */ +bool need_crc = false;I might do
(quoted diff) +2
-0+ bool crc_was_ok = false; /* don't recalc by default */ + bool need_new_crc = false;
Quoted 1 line
for clarity...?
*nod*
Quoted 16 lines
Quoted 4 lines
db/metadump.c +1
-0success =1 ; cur_ino =XFS_AGINO_TO_INO (mp, agno, agino); +/* we only care about crc recalculation if we are obfuscating names. */ db/metadump.c process_inode( · line 1767
Quoted 10 lines
db/metadump.c +6
-0+if (!dont_obfuscate) + crc_ok =xfs_verify_cksum ((char *)dip, mp->m_sb.sb_inodesize, + offsetof(struct xfs_dinode , di_crc)); +/* copy appropriate data fork metadata */ switch (be16_to_cpu (dip->di_mode) &S_IFMT ) {case S_IFDIR : success =process_inode_data (dip,TYP_DIR2 ); +if (dip->di_format ==XFS_DINODE_FMT_LOCAL ) + need_crc =1 ;I wish this were closer to the point of obfuscation, but oh well.
It means carrying a tristate or extra varaible through the stack, which doesn't make it any simpler to understand because we have to do the CRC here once we know we've finished all the modifications to the inode....
Quoted 16 lines
Quoted 10 lines
db/metadump.c +2
-0break ;case S_IFLNK : success =process_inode_data (dip,TYP_SYMLINK ); +if (dip->di_format ==XFS_DINODE_FMT_LOCAL ) + need_crc =1 ;break ;case S_IFREG : success =process_inode_data (dip,TYP_DATA );db/metadump.c process_inode( · line 1789
Quoted 6 lines
db/metadump.c +1
-0@@ -1754,6 +1784,7 @@ process_inode( attr_data.remote_val_count =0 ;switch (dip->di_aformat) {case XFS_DINODE_FMT_LOCAL : + need_crc =1 ;if (!dont_obfuscate)obfuscate_sf_attr (dip);
And this one kind of forces us to do it here if we want to only have one place where we recalc CRCs...
Cheers,
Dave.
Signature · 7 lines
Dave Chinner [email protected] _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs