fix(OVA import): support big files (#3504)

Fixes #3468

- fix bug in VMDK header parsing
- parse extended tar headers

CHANGELOG.md

@@ -11,9 +11,11 @@
 - [New SR] No redirection if the SR creation failed or canceled [#3843](https://github.com/vatesfr/xen-orchestra/issues/3843) (PR [#3853](https://github.com/vatesfr/xen-orchestra/pull/3853))
 - [Home] Fix two tabs opened by middle click in Firefox [#3450](https://github.com/vatesfr/xen-orchestra/issues/3450) (PR [#3825](https://github.com/vatesfr/xen-orchestra/pull/3825))
 - [XOA] Enable downgrade for ending trial (PR [#3867](https://github.com/vatesfr/xen-orchestra/pull/3867))
+- [OVA import] allow import of big files [#3468](https://github.com/vatesfr/xen-orchestra/issues/3468) (PR [#3504](https://github.com/vatesfr/xen-orchestra/pull/3504))
 
 ### Released packages
 
+- xo-vmdk-to-vhd v0.1.6
 - xo-server v5.34.0
 - xo-web v5.34.0
 

packages/xo-vmdk-to-vhd/src/vmdk-read-table.js

@@ -54,10 +54,9 @@ export default async function readVmdkGrainTable(fileAccessor) {
   const grainSize =
     getLongLong(headerBuffer, GRAIN_SIZE_OFFSET, 'grain size') * SECTOR_SIZE
   const grainCount = Math.ceil(capacity / grainSize)
-  const numGTEsPerGT = getLongLong(
+  const numGTEsPerGT = new DataView(headerBuffer).getUint32(
-    headerBuffer,
     NUM_GTE_PER_GT_OFFSET,
-    'num GTE per GT'
+    true
   )
   const grainTablePhysicalSize = numGTEsPerGT * 4
   const grainDirectoryEntries = Math.ceil(grainCount / numGTEsPerGT)

packages/xo-web/src/xo-app/vm-import/ova/index.js

@@ -102,7 +102,10 @@ function parseTarHeader(header) {
     Buffer.from(header.slice(124, 124 + 11)).toString('ascii'),
     8
   )
-  return { fileName, fileSize }
+  // normal files are either the char '0' (charcode 48) or the char null (charcode zero)
+  const typeSlice = new Uint8Array(header.slice(156, 156 + 1))[0]
+  const fileType = typeSlice === 0 ? '0' : String.fromCharCode(typeSlice)
+  return { fileName, fileSize, fileType }
 }
 
 async function parseOVF(fileFragment) {
@@ -172,27 +175,56 @@ async function parseOVF(fileFragment) {
   )
 }
 
+// tar spec: https://www.gnu.org/software/tar/manual/html_node/Standard.html
 async function parseTarFile(file) {
   let offset = 0
   const HEADER_SIZE = 512
   let data = { tables: {} }
   while (offset + HEADER_SIZE <= file.size) {
-    const header = parseTarHeader(
+    let header = parseTarHeader(
       await readFileFragment(file, offset, offset + HEADER_SIZE)
     )
     offset += HEADER_SIZE
     if (header === null) {
       break
     }
-    if (header.fileName.toLowerCase().endsWith('.ovf')) {
+    // extended header: it's a text file named 'PaxHeader/<filename.ext>' appearing before the file.
-      const res = await parseOVF(file.slice(offset, offset + header.fileSize))
+    // the attribute are ascii lines of form: "charcount key=value\n". Charcount is the number of chars in the line.
-      data = { ...data, ...res }
+    // Parsing it is the only way to get the size of big files because  normal headers store the size in a
+    // 12 char octal string, they can't store big sizes.
+    if (header.fileType === 'x') {
+      const paxFile = Buffer.from(
+        await readFileFragment(file, offset, offset + header.fileSize)
+      ).toString()
+      console.log('pax header', paxFile)
+      const lines = paxFile.split('\n')
+      // "<charcount> key=value\n"
+      const foundSize = lines.find(l => l.match(/^[0-9]+ size=/))
+      // go to next header
+      offset += Math.ceil(header.fileSize / 512) * 512
+      header = parseTarHeader(
+        await readFileFragment(file, offset, offset + HEADER_SIZE)
+      )
+      offset += HEADER_SIZE
+      if (foundSize) {
+        header.fileSize = parseInt(foundSize.split('=')[1])
+      }
     }
-    if (header.fileName.toLowerCase().endsWith('.vmdk')) {
+    // remove mac os X forks https://stackoverflow.com/questions/8766730/tar-command-in-mac-os-x-adding-hidden-files-why
-      const fileSlice = file.slice(offset, offset + header.fileSize)
+    if (
-      const readFile = async (start, end) =>
+      header.fileType === '0' &&
-        readFileFragment(fileSlice, start, end)
+      !header.fileName.toLowerCase().startsWith('./._')
-      data.tables[header.fileName] = await readVmdkGrainTable(readFile)
+    ) {
+      if (header.fileName.toLowerCase().endsWith('.ovf')) {
+        const res = await parseOVF(file.slice(offset, offset + header.fileSize))
+        data = { ...data, ...res }
+      }
+      if (header.fileName.toLowerCase().endsWith('.vmdk')) {
+        const fileSlice = file.slice(offset, offset + header.fileSize)
+        const readFile = async (start, end) =>
+          readFileFragment(fileSlice, start, end)
+        data.tables[header.fileName] = await readVmdkGrainTable(readFile)
+      }
     }
     offset += Math.ceil(header.fileSize / 512) * 512
   }