linux-stable.git/drivers/md/dm-snap-persistent.c, branch linux-3.4.y

dm snapshot: fix data corruption

2013-11-04T12:23:42+00:00

commit e9c6a182649f4259db704ae15a91ac820e63b0ca upstream.

This patch fixes a particular type of data corruption that has been
encountered when loading a snapshot's metadata from disk.

When we allocate a new chunk in persistent_prepare, we increment
ps->next_free and we make sure that it doesn't point to a metadata area
by further incrementing it if necessary.

When we load metadata from disk on device activation, ps->next_free is
positioned after the last used data chunk. However, if this last used
data chunk is followed by a metadata area, ps->next_free is positioned
erroneously to the metadata area. A newly-allocated chunk is placed at
the same location as the metadata area, resulting in data or metadata
corruption.

This patch changes the code so that ps->next_free skips the metadata
area when metadata are loaded in function read_exceptions.

The patch also moves a piece of code from persistent_prepare_exception
to a separate function skip_metadata to avoid code duplication.

CVE-2013-4299

Signed-off-by: Mikulas Patocka 
Cc: Mike Snitzer 
Signed-off-by: Alasdair G Kergon 
Signed-off-by: Greg Kroah-Hartman

dm snapshot: workaround for a false positive lockdep warning

2013-10-05T14:06:54+00:00

commit 5ea330a75bd86b2b2a01d7b85c516983238306fb upstream.

The kernel reports a lockdep warning if a snapshot is invalidated because
it runs out of space.

The lockdep warning was triggered by commit 0976dfc1d0cd80a4e9dfaf87bd87
("workqueue: Catch more locking problems with flush_work()") in v3.5.

The warning is false positive.  The real cause for the warning is that
the lockdep engine treats different instances of md->lock as a single
lock.

This patch is a workaround - we use flush_workqueue instead of flush_work.
This code path is not performance sensitive (it is called only on
initialization or invalidation), thus it doesn't matter that we flush the
whole workqueue.

The real fix for the problem would be to teach the lockdep engine to treat
different instances of md->lock as separate locks.

Signed-off-by: Mikulas Patocka 
Acked-by: Alasdair G Kergon 
Signed-off-by: Mike Snitzer 
Signed-off-by: Greg Kroah-Hartman

md: Add in export.h for files using EXPORT_SYMBOL

2011-10-31T23:31:19+00:00

These files were getting the defines for EXPORT_SYMBOL because
device.h was including module.h.  But we are going to put an
end to that.  So add the proper export.h include now.

Signed-off-by: Paul Gortmaker

dm snapshot: style cleanups

2011-08-02T11:32:03+00:00

Coding style cleanups.

Signed-off-by: Alasdair G Kergon 
Signed-off-by: Jonathan Brassow

dm: use vzalloc

2011-08-02T11:32:02+00:00

Use vzalloc() instead of vmalloc()+memset().

Signed-off-by: Joe Perches 
Signed-off-by: Alasdair G Kergon

dm: suppress endian warnings

2011-08-02T11:32:01+00:00

Suppress sparse warnings about cpu_to_le32() by using __le32 types for
on-disk data etc.

Signed-off-by: Alasdair G Kergon

dm snapshot: flush disk cache when merging

2011-08-02T11:32:00+00:00

This patch makes dm-snapshot flush disk cache when writing metadata for
merging snapshot.

Without cache flushing the disk may reorder metadata write and other
data writes and there is a possibility of data corruption in case of
power fault.

Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka 
Signed-off-by: Alasdair G Kergon

dm io: use fixed initial mempool size

2011-05-29T12:03:09+00:00

Replace the arbitrary calculation of an initial io struct mempool size
with a constant.

The code calculated the number of reserved structures based on the request
size and used a "magic" multiplication constant of 4.  This patch changes
it to reserve a fixed number - itself still chosen quite arbitrarily.
Further testing might show if there is a better number to choose.

Note that if there is no memory pressure, we can still allocate an
arbitrary number of "struct io" structures.  One structure is enough to
process the whole request.

Signed-off-by: Mikulas Patocka 
Signed-off-by: Alasdair G Kergon

dm snapshot: persistent make metadata_wq multithreaded

2011-01-13T19:59:59+00:00

metadata_wq serves on-stack work items from chunk_io().  Even if
multiple chunk_io() are simultaneously in progress, each is
independent and queued only once, so multithreaded workqueue can be
safely used.

Switch metadata_wq to multithread and flush the work item instead of
the workqueue in chunk_io().

Signed-off-by: Tejun Heo 
Signed-off-by: Mike Snitzer 
Signed-off-by: Alasdair G Kergon

dm: convert workqueues to alloc_ordered

2011-01-13T19:59:57+00:00

Convert all create[_singlethread]_work() users to the new
alloc[_ordered]_workqueue().  This conversion is mechanical and
doesn't introduce any behavior change.

Signed-off-by: Tejun Heo 
Signed-off-by: Mike Snitzer 
Signed-off-by: Alasdair G Kergon