similar to: [PATCH] v2v: ova: Don't rely on qemu-img version, test "offset" and

v1 -> v2: - Use 'qemu-img info' which avoids one of the temporary files (thanks Tomáš). - Add a unit test. Unfortunately when I run 'make -C v2v check' I don't think this code is being run at all, so there may be something else going on here which I don't understand. Rich.

See: https://www.redhat.com/archives/libguestfs/2017-February/msg00064.html --- v2v/input_ova.ml | 5 ++--- v2v/utils.ml | 51 +++++++++++++++++++++++++++++++-------------------- v2v/utils.mli | 7 +++---- 3 files changed, 36 insertions(+), 27 deletions(-) diff --git a/v2v/input_ova.ml b/v2v/input_ova.ml index 3c11cd0..3beaffc 100644 --- a/v2v/input_ova.ml +++ b/v2v/input_ova.ml @@

This series is related to the problem of inefficient import of OVA files. The needed enhancements of QEMU were merged into the codebase and should be available in QEMU 2.8. From there we can use 'size' and 'offset' options in raw driver to tell QEMU to use only subset of a file as an image. The patch set is more or less complete. The only outstanding issue is the missing detection

v7: - rebased because patch 1/3 has been pushed - changes to nsplit have been dropped (2/3) - addressed Richard's comments, notably the subfolder function was moved to mllib and renamed to subdirectory v6: - just rebase v5: - rebase, patches 1,3,5 were merged - 1/3: we still need to discuss whether to detect compressed discs - 2/3: - renamed argument noempty to keep_empty - tests were

v6: - just rebase v5: - rebase, patches 1,3,5 were merged - 1/3: we still need to discuss whether to detect compressed discs - 2/3: - renamed argument noempty to keep_empty - tests were not run - 3/3: - using JSON module to generate JSON (as suggested by Pino) - all the other comments raised by Pino v4: - rebase to more recent master - 1/6: using just "quote" instead of

v4: - rebase to more recent master - 1/6: using just "quote" instead of "Filename.quote" - 2/6: reformated block of code according to Richards suggestion - 4/6: added tests for nsplit v3: Addressed Pino's comments, namely: - input_ova.ml - untar takes list of paths - renamed untar_partial to untar_metadata - replaced uggly regex with nsplit - tests - test changes

Since this new temporary directory will contain UNIX sockets for communicating with nbdkit, then its path must not be too long. Use the existing directory that libguestfs exposes for this, i.e. sockdir. --- v2v/nbdkit.ml | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/v2v/nbdkit.ml b/v2v/nbdkit.ml index 65317f9b..46b20c9d 100644 --- a/v2v/nbdkit.ml +++ b/v2v/nbdkit.ml

v5: - rebase, patches 1,3,5 were merged - 1/3: we still need to discuss whether to detect compressed discs - 2/3: - renamed argument noempty to keep_empty - tests were not run - 3/3: - using JSON module to generate JSON (as suggested by Pino) - all the other comments raised by Pino v4: - rebase to more recent master - 1/6: using just "quote" instead of

v2: - 1/3: same as in previous version - 2/3: add check for libvirt version - 3/3: restore the disabled test The libvirt patch [2] that should fix our problem with 'raw' driver that is mentioned in [1] was merged. To realy fix things, there is at least one thing we have to change on our side. We have to be explicit about the driver in the JSON we generate. [1]

v3: Addressed Pino's comments, namely: - input_ova.ml - untar takes list of paths - renamed untar_partial to untar_metadata - replaced uggly regex with nsplit - tests - test changes are part of the main commit - renamed test-data/guestfs-hashsums.sh to test-data/test-utils.sh - renamed qemu_version to qemu_is_version and moved it to test-data/test-utils.sh - normalize paths

v8: - split the big patch into several commits v7: - rebased because patch 1/3 has been pushed - changes to nsplit have been dropped (2/3) - addressed Richard's comments, notably the subfolder function was moved to mllib and renamed to subdirectory v6: - just rebase v5: - rebase, patches 1,3,5 were merged - 1/3: we still need to discuss whether to detect compressed discs - 2/3: -

On Saturday, 12 November 2016 16:37:52 CET Tomáš Golembiovský wrote: > We don't have to always extract all files from the OVA archive. The OVA, > as defined in the standard, is plain tar. We can work directly over the > tar archive if we use correct 'offset' and 'size' options when defining > the backing file for QEMU. > > This leads to improvements in speed

v1 was here: https://www.redhat.com/archives/libguestfs/2020-April/msg00007.html There's a BZ for this now which I forgot to add to the commit message: https://bugzilla.redhat.com/show_bug.cgi?id=1814611 For v2: - Fix incorrect reference to $TMPDIR in existing manual. - Separate handling for small temporary files and large temporary files. Small temporary files go into $TMPDIR

We don't have to always extract all files from the OVA archive. The OVA, as defined in the standard, is plain tar. We can work directly over the tar archive if we use correct 'offset' and 'size' options when defining the backing file for QEMU. This leads to improvements in speed and puts much lower requirement on available disk space. Signed-off-by: Tomáš Golembiovský

This is still a draft, not ready for commit yet. But feedback is welcomed. This series is related to the problem of inefficient import of OVA files. The needed enhancements of QEMU was merged into the codebase and should be available in QEMU 2.8. From there we can use 'size' and 'offset' options in raw driver to tell QEMU to use only subset of a file as an image. The first three

https://bugzilla.redhat.com/show_bug.cgi?id=1570407 This turned into quite an in-depth refactoring of how we handle OVAs. It also fixes a potential security issue. Rich.

Originally an attempt to fix: https://bugzilla.redhat.com/show_bug.cgi?id=1570407 However this isn't a complete fix. The OVA supplied doesn't even conform to VMware's own "specification" (I use the word loosely). The OVF inside the OVA references the disk.vmdk file, but the OVA doesn't contain that disk.vmdk file, only a snapshot called disk.vmdk.000000000. Therefore

We don't have to always extract all files from the OVA archive. The OVA, as defined in the standard, is plain tar. We can work directly over the tar archive if we use correct 'offset' and 'size' options when defining the backing file for QEMU. This puts much lower requirement on available disk space. Since the virt-v2v behaviour for OVA input now depends on QEMU version

We don't have to always extract all files from the OVA archive. The OVA, as defined in the standard, is plain tar. We can work directly over the tar archive if we use correct 'offset' and 'size' options when defining the backing file for QEMU. This puts much lower requirement on available disk space. Since the virt-v2v behaviour for OVA input now depends on QEMU version

We don't have to always extract all files from the OVA archive. The OVA, as defined in the standard, is plain tar. We can work directly over the tar archive if we use correct 'offset' and 'size' options when defining the backing file for QEMU. This puts much lower requirement on available disk space. Since the virt-v2v behaviour for OVA input now depends on QEMU version