This ensures that the checks catch all invalid cluster indexes
instead of returning the refcount of a wrong cluster.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
free_cluster_index is only correct if update_refcount() was called from
an allocation function, and even there it's brittle because it's used to
protect unfinished allocations which still have a refcount of 0 - if it
moves in the wrong place, the unfinished allocation can be corrupted.
So not using it any more seems to be a good idea. Instead, use the
first requested cluster to do the calculations. Return -EAGAIN if
unfinished allocations could become invalid and let the caller restart
its search for some free clusters.
The context of creating a snapsnot is one situation where
update_refcount() is called outside of a cluster allocation. For this
case, the change fixes a buffer overflow if a cluster is referenced in
an L2 table that cannot be represented by an existing refcount block.
(new_table[refcount_table_index] was out of bounds)
[Bump the qemu-iotests 026 refblock_alloc.write leak count from 10 to
11.
--Stefan]
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Limit the in-memory reference count table size to 8 MB, it's enough in
practice. This fixes an unbounded allocation as well as a buffer
overflow in qcow2_refcount_init().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
If qcow2_alloc_clusters() fails, new_offset and ret will both be
negative after the fail label, thus passing the first if condition and
subsequently resulting in a call of qcow2_free_clusters() with an
invalid (negative) offset parameter. Fix this by introducing a new label
"fail_free_cluster" which is only invoked if new_offset is indeed
pointing to a newly allocated cluster that should be cleaned up by
freeing it.
While we're at it, clean up the whole fail path. qcow2_cache_put()
should (and actually can) never fail, hence the return value can safely
be ignored (aside from asserting that it indeed did not fail).
Furthermore, there is no reason to give QCOW2_DISCARD_ALWAYS to
qcow2_free_clusters(), a mere QCOW2_DISCARD_OTHER will suffice.
Ultimately, rename the "fail" label to "done", as it is invoked both on
failure and success.
Suggested-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Contrary to the comment describing this function's behavior, it does not
return 0 on success, but rather the offset of the newly allocated
cluster. This patch adjusts the comment accordingly to reflect the
actual behavior.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When reading the refcount table entry in get_refcount(), only bits which
are actually significant for the refcount block offset should be taken
into account.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
When cluster size is big enough it can lead to an offset overflow
in qcow2_alloc_clusters_at(). This patch fixes it.
The allocation is stopped each time at L2 table boundary
(see handle_alloc()), so the possible maximum bytes could be
2^(cluster_bits - 3 + cluster_bits)
cluster_bits - 3 is used to compute the number of entry by L2
and the additional cluster_bits is to take into account each
clusters referenced by the L2 entries.
so int is safe for cluster_bits<=17, unsafe otherwise.
Signed-off-by: Hu Tao <hutao@cn.fujitsu.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Replace the QCOW2_OL_DEFAULT macro by a variable overlap_check in
BDRVQcowState.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In qcow2_check_metadata_overlap and qcow2_pre_write_overlap_check,
change the parameter signifying the checks to perform from its current
positive form to a negative one, i.e., it will no longer explicitly
specify every check to perform but rather a mask of checks not to
perform.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In qcow2_free_any_clusters, preallocated zero clusters should be freed
just as normal clusters are.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Currently, qcow2_check_metadata_overlap uses bdrv_read to read inactive
L1 tables from disk. The number of sectors to read is calculated through
a truncating integer division, therefore, if the L1 table size is not a
multiple of the sector size, the final entries will not be read and
their entries in memory remain undefined (from the g_malloc).
Using bdrv_pread fixes this.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
CHECK_OFLAG_COPIED as a parameter to check_refcounts_l1 and
check_refcounts_l2 is obselete now, since the OFLAG_COPIED consistency
check is actually no longer performed by these functions (but by
check_oflag_copied).
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
If an inactive L1 table is loaded from disk, its entries are in big
endian and have to be converted to host byte order before using them.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
When trying to update the refcounts for a snapshot, the return value of
update_refcount on a compressed cluster was pretty much ignored,
cancelling the update on error but returning 0. This is caused by an
inner "ret" variable shadowing the outer one (the latter is used in the
return statement).
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add functionality for expanding zero clusters. This is necessary for
downgrading the image version to one without zero cluster support.
For non-backed images, this function may also just discard zero clusters
instead of truly expanding them.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
If the refcount of a refcount block is greater than one, we can at least
try to repair that problem by duplicating the affected block.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Since the OFLAG_COPIED checks are now executed after the refcounts have
been repaired (if repairing), it is safe to assume that they are correct
but the OFLAG_COPIED flag may be not. Therefore, if its value differs
from what it should be (considering the according refcount), that
discrepancy can be repaired by correctly setting (or clearing that flag.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Move the OFLAG_COPIED checks out of check_refcounts_l1 and
check_refcounts_l2 and after the actual refcount checks/fixes (since the
refcounts might actually change there).
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Two new functions are added; the first one checks a given range in the
image file for overlaps with metadata (main header, L1 tables, L2
tables, refcount table and blocks).
The second one should be used immediately before writing to the image
file as it calls the first function and, upon collision, marks the
image as corrupt and makes the BDS unusable, thereby preventing
further access.
Both functions take a bitmask argument specifying the structures which
should be checked for overlaps, making it possible to also check
metadata writes against colliding with other structures.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Account for all cluster types in qcow2_update_snapshot_refcounts;
this prevents this function from updating the refcount of unallocated
zero clusters which effectively led to wrong adjustments of the refcount
of cluster 0 (the main qcow2 header). This in turn resulted in images
with (unallocated) zero clusters having a cluster 0 refcount greater
than one after creating a snapshot.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This optimises the discard operation for freed clusters by batching
discard requests (both snapshot deletion and bdrv_discard end up
updating the refcounts cluster by cluster).
Note that we don't discard asynchronously, but keep s->lock held. This
is to avoid that a freed cluster is reallocated and written to while the
discard is still in flight.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Deleted snapshots are discarded in the image file by default, discard
requests take their default from the -drive discard=... option and other
places that free clusters must always be enabled explicitly.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This adds a refcount update reason to all callers of update_refcounts(),
so that a follow-up patch can use this information to decide whether
clusters that reach a refcount of 0 should be discarded in the image
file.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
It ignored the error code, and at least the 'goto fail' is obvious
nonsense as it creates an endless loop (if the next attempt doesn't
magically succeed) and leaves the in-memory L1 table in big-endian
instead of converting it back.
In error cases, there's no point in writing an updated L1 table, so
skip this part for them.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This should be based on the virtual disk size, not on the size of the
image.
Interesting observation: With some VM state stored in the image file,
percentages higher than 100% are possible, even though snapshots
themselves are ignored. This is a qcow2 bug to be fixed another day: The
VM state should be discarded in the active L2 tables after completing
the snapshot creation.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
We already flush when the function completes. There is no need to flush
after every compressed cluster.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The update_cluster_refcount() function increments/decrements a cluster's
refcount and then returns the new refcount value.
There is no need to flush since both update_cluster_refcount() callers
already take care of this:
1. qcow2_alloc_bytes() calls update_cluster_refcount() when compressed
sectors will be appended to an existing cluster with enough free
space. qcow2_alloc_bytes() already flushes so there is no need to do
so in update_cluster_refcount().
2. qcow2_update_snapshot_refcount() sets a cache dependency on refcounts
if it needs to update L2 entries. It also flushes before completing.
Removing this flush significantly speeds up qcow2 snapshot creation:
$ qemu-img create -f qcow2 test.qcow2 -o size=50G,preallocation=metadata
$ time qemu-img snapshot -c new test.qcow2
Time drops from more than 3 minutes to under 1 second.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Users of qcow2_update_snapshot_refcount() do not flush consistently.
qcow2_snapshot_create() flushes but qcow2_snapshot_goto() and
qcow2_snapshot_delete() do not.
Solve this by moving the bdrv_flush() into
qcow2_update_snapshot_refcount().
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Compressed writes use qcow2_alloc_bytes() to allocate space with byte
granularity. The affected clusters' refcounts will be incremented but
we do not need to flush yet.
Set a L2 cache dependency on the refcount block cache, so that the
refcounts get written out before the L2 updates.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
update_refcount() affects the refcount cache, it does not write to disk.
Therefore bdrv_flush(bs->file) does nothing. We need to flush the
refcount cache in order to write out the refcount updates!
While we're here also add error returns when qcow2_cache_flush() fails.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The qemu-img check command can display fragmentation statistics:
* Total number of clusters in virtual disk
* Number of allocated clusters
* Number of fragmented clusters
This patch adds fragmentation statistics support to qcow2.
Compressed and normal clusters count as allocated. Zero clusters are
not counted as allocated unless their L2 entry has a non-zero offset
(e.g. preallocation).
Only the current L1 table counts towards the statistics - snapshots are
ignored.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The check_refcounts_l1/l2() functions have a check_copied argument to
check that the QCOW_O_COPIED flag is consistent with refcount == 1.
This should be a bool, not an int.
However, the next patch introduces qcow2 fragmentation statistics and
also needs to pass an option to check_refcounts_l1/l2(). This is a good
opportunity to use an int flags field.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This patch adds the support for reporting the image end offset (in
bytes). This is particularly useful after a conversion (or a rebase)
where the destination is a block device in order to find the first
unused byte at the end of the image.
Signed-off-by: Federico Simoncelli <fsimonce@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Once upon a time, it was decided that qemu_malloc(0) should abort.
Switching to glib retired that bright idea. Some code that was added
to cope with it (e.g. in commits 702ef63, b76b6e9) is still around.
Bury it.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
A missing factor for the refcount table entry size in the calculation
could mean that too little memory was allocated for the in-memory
representation of the table, resulting in a buffer overflow.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
Tested-by: Michael Tokarev <mjt@tls.msk.ru>
When qcow2_alloc_clusters() error handling code was introduced in commit
5d757b563d, the value of free_byte_offset
was clobbered in the error case. This patch keeps free_byte_offset at 0
so we will try to allocate clusters again next time this function is
called.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Writethrough does not need special-casing anymore in the qcow2 caches.
The block layer adds flushes after every guest-initiated data write,
and these will also flush the qcow2 caches to the OS.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Also don't infer the cluster type directly from the L2 entries, but use
qcow2_get_cluster_type() to keep everything in a single place.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Zero clusters will add another cluster type. Refactor the open-coded
cluster type detection into a switch of QCOW2_CLUSTER_* options so that
the detection is in a single place. This makes it easier to add new
cluster types.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This changes the still existing places that assume that the only flags
are QCOW_OFLAG_COPIED and QCOW_OFLAG_COMPRESSED to properly mask out
reserved bits.
It does not convert bdrv_check yet.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Refcount block allocation and refcount table growth rely on
s->free_cluster_index pointing to somewhere after the current
allocation. Change qcow2_alloc_cluster_at() to fulfill this
assumption.
Without this change it could happen that a newly allocated refcount
block and the allocated data block point to the same area in the image
file, causing data corruption in the long run.
This fixes a bug that became first visible after commit 250196f1.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Someone forgot something in commit 29c1a730... Documenting the right
return value is not enough, you also need to actually return it in the
code.
This bug sometimes causes error return values even when everything has
succeeded: The new offset of the refcount block is truncated to 32 bits
and interpreted as signed. At least with small cluster sizes it's easy
to get a negative return value this way.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Kevin Wolf
Max Reitz
Hu Tao
Stefan Hajnoczi
Federico Simoncelli
Markus Armbruster
Paolo Bonzini
Stefan Hajnoczi
Zhi Yong Wu