fix(OVA import): fix tar file size parsing (#3941)

Avoids relying on PAX header, uses a weird encoding in the normal filesize header.

Fixes #3900

CHANGELOG.unreleased.md

@@ -9,6 +9,7 @@
 
 ### Bug fixes
 
+- [Import] Fix import of big OVA files
 - [Host] Show the host's memory usage instead of the sum of the VMs' memory usage (PR [#3924](https://github.com/vatesfr/xen-orchestra/pull/3924))
 - [SAML] Make `AssertionConsumerServiceURL` matches the callback URL
 - [Backup NG] Correctly delete broken VHD chains [#3875](https://github.com/vatesfr/xen-orchestra/issues/3875) (PR [#3939](https://github.com/vatesfr/xen-orchestra/pull/3939))

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

@@ -231,6 +231,7 @@ const parseFile = async (file, type, func) => {
       type,
     }
   } catch (error) {
+    console.error(error)
     return { error, file, type }
   }
 }

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

@@ -98,9 +98,23 @@ function parseTarHeader(header) {
   if (fileName.length === 0) {
     return null
   }
-  const fileSize = parseInt(
-    Buffer.from(header.slice(124, 124 + 11)).toString('ascii'),
-    8
+  // https://stackoverflow.com/a/2511526/72637
+  const sizeBuffer = Buffer.from(header.slice(124, 124 + 12))
+  let fileSize = 0
+  if (sizeBuffer[0] === 0x80) {
+    // https://github.com/chrisdickinson/tar-parse/blob/master/header.js#L271
+    // remove the head byte and go in decreasing power order.
+    for (let i = 1; i < sizeBuffer.length; i++) {
+      fileSize = fileSize * 256 + sizeBuffer[i]
+    }
+  } else fileSize = parseInt(sizeBuffer.slice(0, -1).toString('ascii'), 8)
+  console.log(
+    'fileSize',
+    fileName,
+    fileSize,
+    'B',
+    fileSize / Math.pow(1024, 3),
+    'GB'
   )
   // 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]
@@ -177,58 +191,43 @@ 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) {
-    let header = parseTarHeader(
-      await readFileFragment(file, offset, offset + HEADER_SIZE)
-    )
-    offset += HEADER_SIZE
-    if (header === null) {
-      break
-    }
-    // extended header: it's a text file named 'PaxHeader/<filename.ext>' appearing before the file.
-    // the attribute are ascii lines of form: "charcount key=value\n". Charcount is the number of chars in the line.
-    // 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(
+  document.body.style.cursor = 'wait'
+  try {
+    let offset = 0
+    const HEADER_SIZE = 512
+    let data = { tables: {} }
+    while (offset + HEADER_SIZE <= file.size) {
+      const header = parseTarHeader(
         await readFileFragment(file, offset, offset + HEADER_SIZE)
       )
       offset += HEADER_SIZE
-      if (foundSize) {
-        header.fileSize = parseInt(foundSize.split('=')[1])
+      if (header === null) {
+        break
       }
+      // remove mac os X forks https://stackoverflow.com/questions/8766730/tar-command-in-mac-os-x-adding-hidden-files-why
+      if (
+        header.fileType === '0' &&
+        !header.fileName.toLowerCase().startsWith('./._')
+      ) {
+        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
     }
-    // remove mac os X forks https://stackoverflow.com/questions/8766730/tar-command-in-mac-os-x-adding-hidden-files-why
-    if (
-      header.fileType === '0' &&
-      !header.fileName.toLowerCase().startsWith('./._')
-    ) {
-      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
+    return data
+  } finally {
+    document.body.style.cursor = null
   }
-  return data
 }
 
 const parseOvaFile = async file => parseTarFile(file)