debian-forge/sources/org.osbuild.curl

#!/usr/bin/python3
"""
Source for downloading files from URLs.

The files are indexed by their content hash. Can download files
that require secrets. The only secret provider currently supported
is `org.osbuild.rhsm` for downloading Red Hat content that requires
a subscriptions.

Internally use curl to download the files; the files are cached in
an internal cache. Multiple parallel connections are used to speed
up the download.
"""


import concurrent.futures
import itertools
import os
import subprocess
import sys
import tempfile

from osbuild import sources

from osbuild.util.checksum import verify_file
from osbuild.util.rhsm import Subscriptions


SCHEMA = """
"additionalProperties": false,
"definitions": {
  "item": {
    "description": "The files to fetch indexed their content checksum",
    "type": "object",
    "additionalProperties": false,
    "patternProperties": {
      "(md5|sha1|sha256|sha384|sha512):[0-9a-f]{32,128}": {
        "oneOf": [
          {
            "type": "string",
            "description": "URL to download the file from."
          },
          {
            "type": "object",
            "additionalProperties": false,
            "required": [
              "url"
            ],
            "properties": {
              "url": {
                "type": "string",
                "description": "URL to download the file from."
              },
              "secrets": {
                "type": "object",
                "additionalProperties": false,
                "required": [
                  "name"
                ],
                "properties": {
                  "name": {
                    "type": "string",
                    "description": "Name of the secrets provider."
                  }
                }
              }
            }
          }
        ]
      }
    }
  }
},
"properties": {
  "items": {"$ref": "#/definitions/item"},
  "urls": {"$ref": "#/definitions/item"}
},
"oneOf": [{
  "required": ["items"]
}, {
  "required": ["urls"]
}]
"""


def fetch(url, checksum, directory):
    secrets = url.get("secrets")
    url_path = url.get("url")
    # Download to a temporary directory until we have verified the checksum. Use a
    # subdirectory, so we avoid copying across block devices.
    with tempfile.TemporaryDirectory(prefix="osbuild-unverified-file-", dir=directory) as tmpdir:
        # some mirrors are sometimes broken. retry manually, because we could be
        # redirected to a different, working, one on retry.
        return_code = 0
        for _ in range(10):
            curl_command = [
                "curl",
                "--silent",
                "--speed-limit", "1000",
                "--connect-timeout", "30",
                "--fail",
                "--location",
                "--output", checksum,
            ]
            if secrets:
                if secrets.get('ssl_ca_cert'):
                    curl_command.extend(["--cacert", secrets.get('ssl_ca_cert')])
                if secrets.get('ssl_client_cert'):
                    curl_command.extend(["--cert", secrets.get('ssl_client_cert')])
                if secrets.get('ssl_client_key'):
                    curl_command.extend(["--key", secrets.get('ssl_client_key')])
            # url must follow options
            curl_command.append(url_path)

            curl = subprocess.run(curl_command, encoding="utf-8", cwd=tmpdir, check=False)
            return_code = curl.returncode
            if return_code == 0:
                break
        else:
            raise RuntimeError(f"curl: error downloading {url}: error code {return_code}")

        if not verify_file(f"{tmpdir}/{checksum}", checksum):
            raise RuntimeError(f"checksum mismatch: {checksum} {url}")

        # The checksum has been verified, move the file into place. in case we race
        # another download of the same file, we simply ignore the error as their
        # contents are guaranteed to be  the same.
        try:
            os.rename(f"{tmpdir}/{checksum}", f"{directory}/{checksum}")
        except FileExistsError:
            pass


def download(items, cache):
    with concurrent.futures.ProcessPoolExecutor(max_workers=4) as executor:
        requested_urls = []
        requested_checksums = []
        subscriptions = None

        for (checksum, url) in items.items():

            # Invariant: all files in @directory must be named after their (verified) checksum.
            # Check this before secrets so that if everything is pre-downloaded we don't need secrets
            if os.path.isfile(f"{cache}/{checksum}"):
                continue

            if not isinstance(url, dict):
                url = {"url": url}

            # check if url needs rhsm secrets
            if url.get("secrets", {}).get("name") == "org.osbuild.rhsm":
                # rhsm secrets only need to be retrieved once and can then be reused
                if subscriptions is None:
                    subscriptions = Subscriptions.from_host_system()
                url["secrets"] = subscriptions.get_secrets(url.get("url"))

            requested_urls.append(url)
            requested_checksums.append(checksum)

        results = executor.map(fetch, requested_urls, requested_checksums, itertools.repeat(cache))

        for _ in results:
            pass


class CurlSource(sources.SourceService):

    def download(self, items, cache, _options):
        cache = os.path.join(cache, "org.osbuild.files")
        os.makedirs(cache, exist_ok=True)

        download(items, cache)


def main():
    service = CurlSource.from_args(sys.argv[1:])
    service.main()


if __name__ == '__main__':
    main()