fix(egress-proxy): mint CA via openssl req so leaf AKI matches CA SKI
test / unit (pull_request) Successful in 18s
test / integration (pull_request) Successful in 1m1s

Root cause of the persistent SSL handshake failure: pipelock's
`tls init` stamps a non-standard `Subject Key Identifier` on the
CAs it generates (random rather than SHA-1 of the pubkey).
mitmproxy computes the `Authority Key Identifier` on each leaf
cert it mints as SHA-1(issuer's pubkey). openssl's chain validator
uses the leaf's AKI to find the issuer cert by SKI; with pipelock's
SKI off by definition, the lookup fails and openssl returns
"unable to get local issuer certificate" — even though the CA is
right there in the trust store with the matching SHA-256
fingerprint. (Also, pipelock generates EC CAs; the cert+key concat
fit in 834 bytes vs ~2.3KB for RSA, which was the first red
flag.)

Diagnostic from a live bottle confirmed:

  leaf cert AKI:   A8:F0:D5:E3:B5:B9:C2:38:2B:9F:DD:4A:DF:26:8C:72:19:A2:5E:94
  CA cert SKI:     81:CA:6D:4C:ED:5C:C2:B1:48:0C:3E:E8:8D:73:86:97:B9:89:B4:3D
  CA cert + leaf cert: same Pipelock-named subject, same public key bytes
  openssl verify -CAfile <our CA> <leaf>: error 20

Fix: switch `egress_proxy_tls_init` from `pipelock tls init` to
host `openssl req` with an explicit `subjectKeyIdentifier=hash`
extension. SHA-1(pubkey) for the SKI matches what mitmproxy puts
in the AKI, so chain validation works. The generated CA is also
RSA-2048 / sha256WithRSAEncryption — mitmproxy's most-tested
configuration.

The new generator is independent of pipelock entirely (no docker
run on the pipelock image to mint the CA), so the egress-proxy
CA generation now requires only `openssl` on the host. macOS +
Linux dev images both have it.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
2026-05-25 16:29:27 -04:00
parent b9c70f7daa
commit 5dc33f3acc
+63 -37
View File
@@ -75,9 +75,7 @@ def build_egress_proxy_image() -> None:
def egress_proxy_tls_init(stage_dir: Path) -> tuple[Path, Path]:
"""Mint the per-bottle egress-proxy MITM CA. Reuses the pipelock
binary's `tls init` subcommand — a known-good RSA CA minter we
already pin and run on this host.
"""Mint the per-bottle egress-proxy MITM CA via host `openssl req`.
Returns `(mitmproxy_pem, cert_only_pem)`:
- `mitmproxy_pem` is the single-PEM concat (cert + key)
@@ -86,47 +84,75 @@ def egress_proxy_tls_init(stage_dir: Path) -> tuple[Path, Path]:
trust store by `provision_ca` so the agent trusts the bumped
CONNECT cert egress-proxy presents.
Both files live under `<stage_dir>/egress-proxy-ca/` (mode 600).
Private keys never leave the host stage dir until
`DockerEgressProxy.start` docker-cps the concat file into the
sidecar; start.py's outer finally `shutil.rmtree`s the stage dir
after teardown.
Why openssl req (not the pipelock binary's `tls init`):
pipelock's CA generator stamps a non-standard `Subject Key
Identifier` on the CA (random rather than SHA-1 of the pubkey).
mitmproxy computes the `Authority Key Identifier` on each leaf
it mints as SHA-1(issuer's pubkey). openssl's chain validator
uses the leaf's AKI to find the issuer cert by SKI; pipelock's
SKI doesn't match → openssl reports "unable to get local issuer
certificate" even though the CA is right there in the trust
store. openssl req's `subjectKeyIdentifier=hash` extension uses
SHA-1(pubkey), matching mitmproxy's computation.
Imported lazily inside the function so test patchers in
pipelock-land don't need to know about us."""
# Local import keeps the module-import graph free of a hard
# pipelock-image dependency at top of file (we don't actually
# need pipelock's *runtime* here, just its tls-init subcommand).
from .pipelock import PIPELOCK_IMAGE
Both files live under `<stage_dir>/egress-proxy-ca/` (mode 644 —
`docker cp` preserves the mode into the container, where the
mitmproxy user (uid 1000) reads them; the host stage_dir is
mode 700 so the private key isn't world-exposed)."""
work = stage_dir / "egress-proxy-ca"
work.mkdir(exist_ok=True)
result = subprocess.run(
["docker", "run", "--rm",
"-v", f"{work}:/h",
"-e", "PIPELOCK_HOME=/h",
PIPELOCK_IMAGE, "tls", "init"],
capture_output=True,
text=True,
check=False,
key_path = work / "ca-key.pem"
cert_path = work / "ca.pem"
cnf_path = work / "ca.cnf"
# RSA-2048 — broad mitmproxy compatibility (its default leaf-cert
# config matches RSA CAs without surprise), and openssl req's
# default behavior here is exactly what we want.
keygen = subprocess.run(
["openssl", "genrsa", "-out", str(key_path), "2048"],
capture_output=True, text=True, check=False,
)
if result.returncode != 0:
die(f"egress-proxy tls init failed: {result.stderr.strip()}")
cert = work / "ca.pem"
key = work / "ca-key.pem"
if not cert.is_file() or not key.is_file():
die(f"egress-proxy tls init did not produce ca files in {work}")
# Mode 644 (not 600) so `docker cp` preserves world-readability
# inside the container — the mitmproxy user (uid 1000) needs to
# read the file, and the host uid `docker cp` propagates from the
# source doesn't match. The host stage_dir is mode 700 so other
# host users still can't traverse in; the private key isn't
# exposed despite the file mode.
cert.chmod(0o644)
if keygen.returncode != 0:
die(f"egress-proxy ca keygen failed: {keygen.stderr.strip()}")
# `subjectKeyIdentifier=hash` makes openssl compute the SKI as
# SHA-1(pubkey), matching how mitmproxy computes the AKI on the
# leaves it later mints. Without this, chain validation breaks
# despite the CA being present in the trust store.
cnf_path.write_text(
"[req]\n"
"distinguished_name = req_dn\n"
"prompt = no\n"
"x509_extensions = v3_ca\n"
"\n"
"[req_dn]\n"
"O = claude-bottle\n"
"CN = claude-bottle egress-proxy CA\n"
"\n"
"[v3_ca]\n"
"basicConstraints = critical, CA:TRUE\n"
"keyUsage = critical, keyCertSign, cRLSign\n"
"subjectKeyIdentifier = hash\n"
)
cnf_path.chmod(0o644)
req = subprocess.run(
["openssl", "req", "-x509", "-new", "-nodes",
"-key", str(key_path),
"-sha256", "-days", "365",
"-config", str(cnf_path),
"-out", str(cert_path)],
capture_output=True, text=True, check=False,
)
if req.returncode != 0:
die(f"egress-proxy ca cert generation failed: {req.stderr.strip()}")
cert_path.chmod(0o644)
# mitmproxy reads cert + key from a single concatenated PEM file.
mitm = work / "mitmproxy-ca.pem"
mitm.write_bytes(cert.read_bytes() + key.read_bytes())
mitm.write_bytes(cert_path.read_bytes() + key_path.read_bytes())
mitm.chmod(0o644)
return (mitm, cert)
return (mitm, cert_path)
class DockerEgressProxy(EgressProxy):