Move the various memalign-related functions out of osdep.h and into
their own header, which we include only where they are used.
While we're doing this, add some brief documentation comments.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220226180723.1706285-10-peter.maydell@linaro.org
With -m32, size_t is generally only a uint32_t. That makes clang
complain that in the assertion
assert(qiov->size <= INT64_MAX);
the range of the type of qiov->size (size_t) is too small for any of its
values to ever exceed INT64_MAX.
Cast qiov->size to uint64_t to silence clang.
Fixes: f7ef38dd13
("block: use int64_t instead of uint64_t in driver read
handlers")
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20211011155031.149158-1-hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
We are generally moving to int64_t for both offset and bytes parameters
on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for
fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and
with possibility for signed return type (where negative value means
error).
So, convert driver read handlers parameters which are already 64bit to
signed type.
While being here, convert also flags parameter to be BdrvRequestFlags.
Now let's consider all callers. Simple
git grep '\->bdrv_\(aio\|co\)_preadv\(_part\)\?'
shows that's there three callers of driver function:
bdrv_driver_preadv() in block/io.c, passes int64_t, checked by
bdrv_check_qiov_request() to be non-negative.
qcow2_load_vmstate() does bdrv_check_qiov_request().
do_perform_cow_read() has uint64_t argument. And a lot of things in
qcow2 driver are uint64_t, so converting it is big job. But we must
not work with requests that don't satisfy bdrv_check_qiov_request(),
so let's just assert it here.
Still, the functions may be called directly, not only by drv->...
Let's check:
git grep '\.bdrv_\(aio\|co\)_preadv\(_part\)\?\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
The only one such caller:
QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, &data, 1);
...
ret = bdrv_replace_test_co_preadv(bs, 0, 1, &qiov, 0);
in tests/unit/test-bdrv-drain.c, and it's OK obviously.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210903102807.27127-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: fix typos]
Signed-off-by: Eric Blake <eblake@redhat.com>
Add helper to parse compressed l2_entry and use it everywhere instead
of open-coding.
Note, that in most places we move to precise coffset/csize instead of
sector-aligned. Still it should work good enough for updating
refcounts.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210914122454.141075-4-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Let's pass the whole L2 entry and not bother with
L2E_COMPRESSED_OFFSET_SIZE_MASK.
It also helps further refactoring that adds generic
qcow2_parse_compressed_l2_entry() helper.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210914122454.141075-3-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
There is no conflict and no dependency if we have parallel writes to
different subclusters of one cluster when the cluster itself is already
allocated. So, relax extra dependency.
Measure performance:
First, prepare build/qemu-img-old and build/qemu-img-new images.
cd scripts/simplebench
./img_bench_templater.py
Paste the following to stdin of running script:
qemu_img=../../build/qemu-img-{old|new}
$qemu_img create -f qcow2 -o extended_l2=on /ssd/x.qcow2 1G
$qemu_img bench -c 100000 -d 8 [-s 2K|-s 2K -o 512|-s $((1024*2+512))] \
-w -t none -n /ssd/x.qcow2
The result:
All results are in seconds
------------------ --------- ---------
old new
-s 2K 6.7 ± 15% 6.2 ± 12%
-7%
-s 2K -o 512 13 ± 3% 11 ± 5%
-16%
-s $((1024*2+512)) 9.5 ± 4% 8.4
-12%
------------------ --------- ---------
So small writes are more independent now and that helps to keep deeper
io queue which improves performance.
271 iotest output becomes racy for three allocation in one cluster.
Second and third writes may finish in different order. Second and
third requests don't depend on each other any more. Still they both
depend on first request anyway. Filter out second and third write
offsets to cover both possible outputs.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210824101517.59802-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
[hreitz: s/ an / and /]
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
No logic change, just prepare for the following commit. While being
here do also small grammar fix in a comment.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210824101517.59802-3-vsementsov@virtuozzo.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Commit 205fa50750 ("qcow2: Add subcluster support to zero_in_l2_slice()")
introduced a subtle change to code in zero_in_l2_slice:
It swapped the order of
1. qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_slice);
2. set_l2_entry(s, l2_slice, l2_index + i, QCOW_OFLAG_ZERO);
3. qcow2_free_any_clusters(bs, old_offset, 1, QCOW2_DISCARD_REQUEST);
To
1. qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_slice);
2. qcow2_free_any_clusters(bs, old_offset, 1, QCOW2_DISCARD_REQUEST);
3. set_l2_entry(s, l2_slice, l2_index + i, QCOW_OFLAG_ZERO);
It seems harmless, however the call to qcow2_free_any_clusters can
trigger a cache flush which can mark the L2 table as clean, and
assuming that this was the last write to it, a stale version of it
will remain on the disk.
Now we have a valid L2 entry pointing to a freed cluster. Oops.
Fixes: 205fa50750 ("qcow2: Add subcluster support to zero_in_l2_slice()")
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
[ kwolf: Fixed to restore the correct original order from before
205fa50750; added comments like in discard_in_l2_slice(). ]
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201124092815.39056-1-kwolf@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The QCowL2Meta structure is used to store information about a part of
a write request that touches clusters that need changes in their L2
entries. This happens with newly-allocated clusters or subclusters.
This structure has changed a bit since it was first created and its
current documentation is not quite up-to-date.
A write request can span a region consisting of a combination of
clusters of different types, and qcow2_alloc_host_offset() can
repeatedly call handle_copied() and handle_alloc() to add more
clusters to the mix as long as they all are contiguous on the image
file.
Because of this a write request has a list of QCowL2Meta structures,
one for each part of the request that needs changes in the L2
metadata.
Each one of them spans nb_clusters and has two copy-on-write regions
located immediately before and after the middle region touched by that
part of the write request. Even when those regions themselves are
empty their offsets must be correct because they are used to know the
location of the middle region.
This was not always the case but it is not a problem anymore
because the only two places where QCowL2Meta structures are created
(calculate_l2_meta() and qcow2_co_truncate()) ensure that the
copy-on-write regions are correctly defined, and so do assertions like
the ones in perform_cow().
The conditional initialization of the 'written_to' variable is
therefore unnecessary and is removed by this patch.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201007161323.4667-1-berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
We overlooked these in 02b1ecfa10
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <20200928162333.14998-1-berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
qcow2_alloc_cluster_offset() takes an (unaligned) guest offset and
returns the (aligned) offset of the corresponding cluster in the qcow2
image.
In practice none of the callers need to know where the cluster starts
so this patch makes the function calculate and return the final host
offset directly. The function is also renamed accordingly.
See 388e581615 for a similar change to qcow2_get_cluster_offset().
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <9bfef50ec9200d752413be4fc2aeb22a28378817.1599833007.git.berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
This function takes an L2 entry and a number of clusters to free.
Although in principle it can free any type of cluster (using the L2
entry to determine its type) in practice the API is broken because
compressed clusters have a variable size and there is no way to free
more than one without having the L2 entry of each one of them.
The good news all callers are passing nb_clusters=1 so we can simply
get rid of that parameter.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <77cea0f4616f921d37e971b3c5b18a2faa24b173.1599573989.git.berto@igalia.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
The current text corresponds to an earlier, simpler version of this
function and it does not explain how it works now.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <bb5bd06f07c5a05b0818611de0d06ec5b66c8df3.1599150873.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
When a write request needs to allocate new clusters (or change the L2
bitmap of existing ones) a QCowL2Meta structure is created so the L2
metadata can be later updated and any copy-on-write can be performed
if necessary.
A write request can span a region consisting of an arbitrary
combination of previously unallocated and allocated clusters, and if
the unallocated ones can be put contiguous to the existing ones then
QEMU will do so in order to minimize the number of write operations.
In practice this means that a write request has not just one but a
number of QCowL2Meta structures. All of them are added to the
cluster_allocs list that is stored in BDRVQcow2State and is used to
detect overlapping requests. After the write request finishes all its
associated QCowL2Meta are removed from that list. calculate_l2_meta()
takes care of creating and putting those structures in the list, and
qcow2_handle_l2meta() takes care of removing them.
The problem is that the error path in handle_alloc() also tries to
remove an item in that list, a remnant from the time when this was
handled there (that code would not even be correct anymore because
it only removes one struct and not all the ones from the same write
request).
This can trigger a double removal of the same item from the list,
causing a crash. This is not easy to reproduce in practice because
it requires that do_alloc_cluster_offset() fails after a successful
previous allocation during the same write request, but it can be
reproduced with the included test case.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <3440a1c4d53c4fe48312b478c96accb338cbef7c.1599150873.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
This patch replaces instances of sizeof(uint64_t) in the qcow2 driver
with macros that indicate what those sizes are actually referring to.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <20200828110828.13833-1-berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
When compiling with -Werror=implicit-fallthrough, the compiler currently
complains:
../../devel/qemu/block/qcow2-cluster.c: In function ‘cluster_needs_new_alloc’:
../../devel/qemu/block/qcow2-cluster.c:1320:12: error: this statement may fall
through [-Werror=implicit-fallthrough=]
if (l2_entry & QCOW_OFLAG_COPIED) {
^
../../devel/qemu/block/qcow2-cluster.c:1323:5: note: here
case QCOW2_CLUSTER_UNALLOCATED:
^~~~
It's quite obvious that the fallthrough is intended here, so let's add
a comment to silence the compiler warning.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20200908070028.193298-1-thuth@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This function is only used by qcow2_expand_zero_clusters() to
downgrade a qcow2 image to a previous version. This would require
transforming all extended L2 entries into normal L2 entries but this
is not a simple task and there are no plans to implement this at the
moment.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <15e65112b4144381b4d8c0bdf8fb76b0d813e3d1.1594396418.git.berto@igalia.com>
[mreitz: Fixed comment style]
Signed-off-by: Max Reitz <mreitz@redhat.com>
This field allows us to indicate that the L2 metadata update does not
come from a write request with actual data but from a preallocation
request.
For traditional images this does not make any difference, but for
images with extended L2 entries this means that the clusters are
allocated normally in the L2 table but individual subclusters are
marked as unallocated.
This will allow preallocating images that have a backing file.
There is one special case: when we resize an existing image we can
also request that the new clusters are preallocated. If the image
already had a backing file then we have to hide any possible stale
data and zero out the new clusters (see commit 955c7d6687 for more
details).
In this case the subclusters cannot be left as unallocated so the L2
bitmap must be updated.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <960d4c444a4f5a870e2b47e5da322a73cd9a2f5a.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
This works now at the subcluster level and pwrite_zeroes_alignment is
updated accordingly.
qcow2_cluster_zeroize() is turned into qcow2_subcluster_zeroize() with
the following changes:
- The request can now be subcluster-aligned.
- The cluster-aligned body of the request is still zeroized using
zero_in_l2_slice() as before.
- The subcluster-aligned head and tail of the request are zeroized
with the new zero_l2_subclusters() function.
There is just one thing to take into account for a possible future
improvement: compressed clusters cannot be partially zeroized so
zero_l2_subclusters() on the head or the tail can return -ENOTSUP.
This makes the caller repeat the *complete* request and write actual
zeroes to disk. This is sub-optimal because
1) if the head area was compressed we would still be able to use
the fast path for the body and possibly the tail.
2) if the tail area was compressed we are writing zeroes to the
head and the body areas, which are already zeroized.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <17e05e2ee7e12f10dcf012da81e83ebe27eb3bef.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Compressed clusters always have the bitmap part of the extended L2
entry set to 0.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <04455b3de5dfeb9d1cfe1fc7b02d7060a6e09710.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
The L2 bitmap needs to be updated after each write to indicate what
new subclusters are now allocated. This needs to happen even if the
cluster was already allocated and the L2 entry was otherwise valid.
In some cases however a write operation doesn't need change the L2
bitmap (because all affected subclusters were already allocated). This
is detected in calculate_l2_meta(), and qcow2_alloc_cluster_link_l2()
is never called in those cases.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <0875620d49f44320334b6a91c73b3f301f975f38.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Two things need to be taken into account here:
1) With full_discard == true the L2 entry must be cleared completely.
This also includes the L2 bitmap if the image has extended L2
entries.
2) With full_discard == false we have to make the discarded cluster
read back as zeroes. With normal L2 entries this is done with the
QCOW_OFLAG_ZERO bit, whereas with extended L2 entries this is done
with the individual 'all zeroes' bits for each subcluster.
Note however that QCOW_OFLAG_ZERO is not supported in v2 qcow2
images so, if there is a backing file, discard cannot guarantee
that the image will read back as zeroes. If this is important for
the caller it should forbid it as qcow2_co_pdiscard() does (see
80f5c01183 for more details).
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <5ef8274e628aa3ab559bfac467abf488534f2b76.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
The QCOW_OFLAG_ZERO bit that indicates that a cluster reads as
zeroes is only used in standard L2 entries. Extended L2 entries use
individual 'all zeroes' bits for each subcluster.
This must be taken into account when updating the L2 entry and also
when deciding that an existing entry does not need to be updated.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <b61d61606d8c9b367bd641ab37351ddb9172799a.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
The logic of this function remains pretty much the same, except that
it uses count_contiguous_subclusters(), which combines the logic of
count_contiguous_clusters() / count_contiguous_clusters_unallocated()
and checks individual subclusters.
qcow2_cluster_to_subcluster_type() is not necessary as a separate
function anymore so it's inlined into its caller.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <d2193fd48653a350d80f0eca1c67b1d9053fb2f3.1594396418.git.berto@igalia.com>
[mreitz: Initialize expected_type to anything]
Signed-off-by: Max Reitz <mreitz@redhat.com>
If an image has subclusters then there are more copy-on-write
scenarios that we need to consider. Let's say we have a write request
from the middle of subcluster #3 until the end of the cluster:
1) If we are writing to a newly allocated cluster then we need
copy-on-write. The previous contents of subclusters #0 to #3 must
be copied to the new cluster. We can optimize this process by
skipping all leading unallocated or zero subclusters (the status of
those skipped subclusters will be reflected in the new L2 bitmap).
2) If we are overwriting an existing cluster:
2.1) If subcluster #3 is unallocated or has the all-zeroes bit set
then we need copy-on-write (on subcluster #3 only).
2.2) If subcluster #3 was already allocated then there is no need
for any copy-on-write. However we still need to update the L2
bitmap to reflect possible changes in the allocation status of
subclusters #4 to #31. Because of this, this function checks
if all the overwritten subclusters are already allocated and
in this case it returns without creating a new QCowL2Meta
structure.
After all these changes l2meta_cow_start() and l2meta_cow_end()
are not necessarily cluster-aligned anymore. We need to update the
calculation of old_start and old_end in handle_dependencies() to
guarantee that no two requests try to write on the same cluster.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <4292dd56e4446d386a2fe307311737a711c00708.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
In order to support extended L2 entries some functions of the qcow2
driver need to start dealing with subclusters instead of clusters.
qcow2_get_host_offset() is modified to return the subcluster type
instead of the cluster type, and all callers are updated to replace
all values of QCow2ClusterType with their QCow2SubclusterType
equivalents.
This patch only changes the data types, there are no semantic changes.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <f6c29737c295f32cbee74c903c30b01820363b34.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
This function returns an integer that can be either an error code or a
cluster type (a value from the QCow2ClusterType enum).
We are going to start using subcluster types instead of cluster types
in some functions so it's better to use the exact data types instead
of integers for clarity and in order to detect errors more easily.
This patch makes qcow2_get_host_offset() return 0 on success and
puts the returned cluster type in a separate parameter. There are no
semantic changes.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <396b6eab1859a271551dcd7dcba77f8934aa3c3f.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
There are situations in which we want to know how many contiguous
subclusters of the same type there are in a given cluster. This can be
done by simply iterating over the subclusters and repeatedly calling
qcow2_get_subcluster_type() for each one of them.
However once we determined the type of a subcluster we can check the
rest efficiently by counting the number of adjacent ones (or zeroes)
in the bitmap. This is what this function does.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <db917263d568ec6ffb4a41cac3c9100f96bf6c18.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
qcow2 images with subclusters have 128-bit L2 entries. The first 64
bits contain the same information as traditional images and the last
64 bits form a bitmap with the status of each individual subcluster.
Because of that we cannot assume that L2 entries are sizeof(uint64_t)
anymore. This function returns the proper value for the image.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <d34d578bd0380e739e2dde3e8dd6187d3d249fa9.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
The size of an L2 entry is 64 bits, but if we want to have subclusters
we need extended L2 entries. This means that we have to access L2
tables and slices differently depending on whether an image has
extended L2 entries or not.
This patch replaces all l2_slice[] accesses with calls to
get_l2_entry() and set_l2_entry().
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <9586363531fec125ba1386e561762d3e4224e9fc.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
When writing to a qcow2 file there are two functions that take a
virtual offset and return a host offset, possibly allocating new
clusters if necessary:
- handle_copied() looks for normal data clusters that are already
allocated and have a reference count of 1. In those clusters we
can simply write the data and there is no need to perform any
copy-on-write.
- handle_alloc() looks for clusters that do need copy-on-write,
either because they haven't been allocated yet, because their
reference count is != 1 or because they are ZERO_ALLOC clusters.
The ZERO_ALLOC case is a bit special because those are clusters that
are already allocated and they could perfectly be dealt with in
handle_copied() (as long as copy-on-write is performed when required).
In fact, there is extra code specifically for them in handle_alloc()
that tries to reuse the existing allocation if possible and frees them
otherwise.
This patch changes the handling of ZERO_ALLOC clusters so the
semantics of these two functions are now like this:
- handle_copied() looks for clusters that are already allocated and
which we can overwrite (NORMAL and ZERO_ALLOC clusters with a
reference count of 1).
- handle_alloc() looks for clusters for which we need a new
allocation (all other cases).
One important difference after this change is that clusters found
in handle_copied() may now require copy-on-write, but this will be
necessary anyway once we add support for subclusters.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <eb17fc938f6be7be2e8d8ff42763d2c19241f866.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
We are going to need it in other places.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <65e5d9627ca2ebe7e62deaeddf60949c33067d9d.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
handle_alloc() creates a QCowL2Meta structure in order to update the
image metadata and perform the necessary copy-on-write operations.
This patch moves that code to a separate function so it can be used
from other places.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <e5bc4a648dac31972bfa7a0e554be8064be78799.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
qcow2_get_cluster_offset() takes an (unaligned) guest offset and
returns the (aligned) offset of the corresponding cluster in the qcow2
image.
In practice none of the callers need to know where the cluster starts
so this patch makes the function calculate and return the final host
offset directly. The function is also renamed accordingly.
There is a pre-existing exception with compressed clusters: in this
case the function returns the complete cluster descriptor (containing
the offset and size of the compressed data). This does not change with
this patch but it is now documented.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <ffae6cdc5ca8950e8280ac0f696dcc376cb07095.1594396418.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
When calculating the offset, the result of left shift operation will be promoted
to type int64 automatically because the left operand of + operator is uint64_t.
but the result after integer promotion may be produce an error value for us and
trigger the following asserting error.
For example, consider i=0x2000, cluster_bits=18, the result of left shift
operation will be 0x80000000. Cause argument i is of signed integer type,
the result is automatically promoted to 0xffffffff80000000 which is not
we expected
The way to trigger the assertion error:
qemu-img create -f qcow2 -o preallocation=full,cluster_size=256k tmpdisk 10G
This patch fix it by casting @i to uint64_t before doing left shift operation
Signed-off-by: Guoyi Tu <tu.guoyi@h3c.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 81ba90fe0c014f269621c283269b42ad@h3c.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
qcow2 version 2 images don't support the zero flag for clusters, so for
write_zeroes requests, we return -ENOTSUP and get explicit zero buffer
writes. If the image doesn't have a backing file, we can do better: Just
discard the respective clusters.
This is relevant for 'qemu-img convert -O qcow2 -n', where qemu-img has
to assume that the existing target image may contain any data, so it has
to write zeroes. Without this patch, this results in a fully allocated
target image, even if the source image was empty.
Reported-by: Nir Soffer <nsoffer@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200721135520.72355-2-kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
If BDRV_REQ_ZERO_WRITE is set and we're extending the image, calling
qcow2_cluster_zeroize() with flags=0 does the right thing: It doesn't
undo any previous preallocation, but just adds the zero flag to all
relevant L2 entries. If an external data file is in use, a write_zeroes
request to the data file is made instead.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200424125448.63318-5-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
handle_alloc() reuses preallocated zero clusters. If anything goes
wrong during the data write, we do not change their L2 entry, so we
must not let qcow2_alloc_cluster_abort() free them.
Fixes: 8b24cd1415
Cc: qemu-stable@nongnu.org
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200225143130.111267-2-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For external data file, cluster allocations return an offset in the data
file and are not refcounted. In this case, there is nothing to do for
qcow2_alloc_cluster_abort(). Freeing the same offset in the qcow2 file
is wrong and causes crashes in the better case or image corruption in
the worse case.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200211094900.17315-3-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When updating an L1 entry the qcow2 driver writes a (512-byte) sector
worth of data to avoid a read-modify-write cycle. Instead of always
writing 512 bytes we should follow the alignment requirements of the
storage backend.
(the only exception is when the alignment is larger than the cluster
size because then we could be overwriting data after the L1 table)
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 71f34d4ae4b367b32fb36134acbf4f4f7ee681f4.1579374329.git.berto@igalia.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
The L1 table is read from disk using the byte-based bdrv_pread() and
is never accessed beyond its last element, so there's no need to
allocate more memory than that.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: b2e27214ec7b03a585931bcf383ee1ac3a641a10.1579374329.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
The standard cluster descriptor in L2 table entries has a field to
store the host cluster offset. When we need to get that offset from an
entry we use L2E_OFFSET_MASK to ensure that we only use the bits that
belong to that field.
But while that mask is used every time we read from an L2 entry, it
is never used when we write to it. Due to the QCOW_MAX_CLUSTER_OFFSET
limit set in the cluster allocation code QEMU can never produce
offsets that don't fit in that field so any such offset would indicate
a bug in QEMU.
Compressed cluster descriptors contain two fields (host cluster offset
and size of the compressed data) and the situation with them is
similar. In this case the masks are not constant but are stored in the
csize_mask and cluster_offset_mask fields of BDRVQcow2State.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20200113161146.20099-1-berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
When the COW areas are included, the size of an allocation can exceed
INT_MAX. This is kind of limited by handle_alloc() in that it already
caps avail_bytes at INT_MAX, but the number of clusters still reflects
the original length.
This can have all sorts of effects, ranging from the storage layer write
call failing to image corruption. (If there were no image corruption,
then I suppose there would be data loss because the .cow_end area is
forced to be empty, even though there might be something we need to
COW.)
Fix all of it by limiting nb_clusters so the equivalent number of bytes
will not exceed INT_MAX.
Cc: qemu-stable@nongnu.org
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
* Change the qcow2_co_{encrypt|decrypt} to just receive full host and
guest offsets and use this function directly instead of calling
do_perform_cow_encrypt (which is removed by that patch).
* Adjust qcow2_co_encdec to take full host and guest offsets as well.
* Document the qcow2_co_{encrypt|decrypt} arguments
to prevent the bug fixed in former commit from hopefully
happening again.
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Message-id: 20190915203655.21638-3-mlevitsk@redhat.com
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
[mreitz: Let perform_cow() return the error value returned by
qcow2_co_encrypt(), as proposed by Vladimir]
Signed-off-by: Max Reitz <mreitz@redhat.com>
This fixes subtle corruption introduced by luks threaded encryption
in commit 8ac0f15f33
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1745922
The corruption happens when we do a write that
* writes to two or more unallocated clusters at once
* doesn't fully cover the first sector
* doesn't fully cover the last sector
* uses luks encryption
In this case, when allocating the new clusters we COW both areas
prior to the write and after the write, and we encrypt them.
The above mentioned commit accidentally made it so we encrypt the
second COW area using the physical cluster offset of the first area.
The problem is that offset_in_cluster in do_perform_cow_encrypt
can be larger that the cluster size, thus cluster_offset
will no longer point to the start of the cluster at which encrypted
area starts.
Next patch in this series will refactor the code to avoid all these
assumptions.
In the bugreport that was triggered by rebasing a luks image to new,
zero filled base, which lot of such writes, and causes some files
with zero areas to contain garbage there instead.
But as described above it can happen elsewhere as well
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-id: 20190915203655.21638-2-mlevitsk@redhat.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Replace instances of:
(n & (BDRV_SECTOR_SIZE - 1)) == 0
And:
(n & ~BDRV_SECTOR_MASK) == 0
With:
QEMU_IS_ALIGNED(n, BDRV_SECTOR_SIZE)
Which reveals the intent of the code better, and makes it easier to
locate the code checking alignment.
Signed-off-by: Nir Soffer <nsoffer@redhat.com>
Message-id: 20190827185913.27427-2-nsoffer@redhat.com
Reviewed-by: John Snow <jsnow@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
handle_alloc() tries to find as many contiguous clusters that need
copy-on-write as possible in order to allocate all of them at the same
time.
However, compressed clusters are only overwritten one by one, so let's
say that we have an image with 1024 consecutive compressed clusters:
qemu-img create -f qcow2 hd.qcow2 64M
for f in `seq 0 64 65472`; do
qemu-io -c "write -c ${f}k 64k" hd.qcow2
done
In this case trying to overwrite the whole image with one large write
request results in 1024 separate allocations:
qemu-io -c "write 0 64M" hd.qcow2
This restriction comes from commit 095a9c58ce from 2008.
Nowadays QEMU can overwrite multiple compressed clusters just fine,
and in fact it already does: as long as the first cluster that
handle_alloc() finds is not compressed, all other compressed clusters
in the same batch will be overwritten in one go:
qemu-img create -f qcow2 hd.qcow2 64M
qemu-io -c "write -z 0 64k" hd.qcow2
for f in `seq 64 64 65472`; do
qemu-io -c "write -c ${f}k 64k" hd.qcow2
done
Compared to the previous one, overwriting this image on my computer
goes from 8.35s down to 230ms.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: John Snow <jsnow@redhat.com
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Implement and use new interface to get rid of hd_qiov.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20190604161514.262241-13-vsementsov@virtuozzo.com
Message-Id: <20190604161514.262241-13-vsementsov@virtuozzo.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
