[Libguestfs] [v2v PATCH v2] input_vmx: cleanly reject guests with snapshots when using "-it ssh"

[Laszlo Ersek]

For traversing a backing chain of VMDK descriptor files over ssh, two
things are necessary:

- qemu-nbd with the ssh block driver, rather than nbdkit-ssh-plugin,

- a remote SSH URL (for qemu-nbd) without a query string appended, as
  qemu-nbd cannot update the last pathname component (for tracking the
  relative pathnames of VMDK descriptor files) if a query string is
  appended.

Before commit 7a6f6113a25f ("v2v: -i vmx -it ssh: Replace qemu block ssh
driver with nbdkit-ssh-plugin.", 2019-10-08), we passed the
"?host_key_check=no" query string in the URL to qemu-nbd, so we can't just
return to that, for accessing snapshotted guests with vmx+ssh.

But, we shouldn't return to qemu-nbd for vmx+ssh even without a query
string, as that would undo the other benefit(s) of commit 7a6f6113a25f.

Instead, clearly document that snapshotted guests are not supported over
vmx+ssh, and cleanly reject this situation in the code as well. Recommend
the two alternative transports that allow the user to convert such guests.

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1774386
Signed-off-by: Laszlo Ersek <lersek at redhat.com>
---

Notes:
    v2:
    
    - in the manual, centralize the description in a new NOTES section, and
      reference that section from "-i vmx -it ssh" and "INPUT FROM VMWARE
      VMX" [Rich]
    
    - in the code, bail out with error (f_"...") rather than failwith, for
      enabling translations [Rich]
    
    - in the error message printed from the code, recommend actions to the
      user (reference the new NOTES section of the manual) [Rich]

 input/input_vmx.ml             | 27 ++++++++------------
 docs/virt-v2v-input-vmware.pod | 18 ++++++++++++-
 2 files changed, 27 insertions(+), 18 deletions(-)

diff --git a/input/input_vmx.ml b/input/input_vmx.ml
index 9921419b5a85..eed8a4335622 100644
--- a/input/input_vmx.ml
+++ b/input/input_vmx.ml
@@ -81,23 +81,16 @@ module VMX = struct
 
             let vmx_path = path_of_uri uri in
             let abs_path = absolute_path_from_other_file vmx_path filename in
-            let format = "vmdk" in
+            let flat_vmdk = PCRE.replace (PCRE.compile "\\.vmdk$")
+                              "-flat.vmdk" abs_path in
 
-            (* XXX This is a hack to work around qemu / VMDK limitation
-             *   "Cannot use relative extent paths with VMDK descriptor file"
-             * We can remove this if the above is fixed.
-             *)
-            let abs_path, format =
-              let flat_vmdk =
-                PCRE.replace (PCRE.compile "\\.vmdk$") "-flat.vmdk" abs_path in
-              if remote_file_exists uri flat_vmdk then (flat_vmdk, "raw")
-              else (abs_path, format) in
-
-            (* XXX In virt-v2v 1.42+ importing from VMX over SSH
-             * was broken if the -flat.vmdk file did not exist.
-             * It is still broken here.
-             *)
-            ignore format;
+            (* RHBZ#1774386 *)
+            if not (remote_file_exists uri flat_vmdk) then
+              error (f_"This transport does not support guests with snapshots. \
+                        Either collapse the snapshots for this guest and try \
+                        the conversion again, or use one of the alternate \
+                        conversion methods described in \
+                        virt-v2v-input-vmware(1) section \"NOTES\".");
 
             let server = server_of_uri uri in
             let port = Option.map string_of_int (port_of_uri uri) in
@@ -110,7 +103,7 @@ module VMX = struct
             let cor = dir // "convert" in
             let bandwidth = options.bandwidth in
             let nbdkit = Nbdkit_ssh.create_ssh ?bandwidth ~cor ~password
-                           ~server ?port ?user abs_path in
+                           ~server ?port ?user flat_vmdk in
             let _, pid = Nbdkit.run_unix socket nbdkit in
             On_exit.kill pid
         ) filenames
diff --git a/docs/virt-v2v-input-vmware.pod b/docs/virt-v2v-input-vmware.pod
index 2c90e0afc7cd..66ca5f00d28a 100644
--- a/docs/virt-v2v-input-vmware.pod
+++ b/docs/virt-v2v-input-vmware.pod
@@ -51,6 +51,9 @@ to ESXi to read the F<GUEST.vmx> file and associated disks.  This
 requires that you have enabled SSH access to the VMware ESXi
 hypervisor - in the default ESXi configuration this is turned off.
 
+This transport is incompatible with guests that have snapshots; refer
+to L</NOTES>.
+
 =item B<-ic vpx://...> B<-it vddk>
 
 =item B<-ic esx://...> B<-it vddk>
@@ -84,6 +87,18 @@ import a guest from VMware vCenter.  This is the slowest method.
 
 =back
 
+=head1 NOTES
+
+When accessing the F<I<guest>.vmx> file on ESXi over an SSH connection
+(that is, when using the B<-i vmx> B<-it ssh> options), the conversion
+will not work if the guest has snapshots (files called
+F<I<guest>-000001.vmdk> and similar).  Either collapse the snapshots
+for the guest and retry the conversion with the same
+B<-i vmx> B<-it ssh> options, or leave the snapshots intact and use a
+transport different from SSH: just B<-i vmx>, or
+B<-ic vpx://...> B<-it vddk> or B<-ic esx://...> B<-it vddk>.  Refer
+to L<https://bugzilla.redhat.com/1774386>.
+
 =head1 INPUT FROM VMWARE VMX
 
 Virt-v2v is able to import guests from VMware’s vmx files.
@@ -106,7 +121,8 @@ a C</vmfs/volumes> folder containing the virtual machines.
 
 If you find a folder of files called F<I<guest>.vmx>,
 F<I<guest>.vmxf>, F<I<guest>.nvram> and one or more F<.vmdk> disk
-images, then you can use this method.
+images, then you can use this method.  The SSH transport is not usable
+if the guest has snapshots; refer to L</NOTES>.
 
 =head2 VMX: Guest must be shut down
 
-- 
2.19.1.3.g30247aa5d201