wip account creation

millipds_dev.dockerfile

@@ -20,6 +20,9 @@ RUN python3 -m pip install -v src/
 # init the db with dev presets
 RUN python3 -m millipds init millipds.test --dev
 
+# create a test user
+RUN python3 -m millipds account create bob.test did:web:bob.test --unsafe_password=hunter2
+
 # do the thing
 CMD python3 -m millipds run --listen_host=0.0.0.0 --listen_port=8123
 

src/millipds/__main__.py

@@ -3,6 +3,7 @@
 Usage:
   millipds init <hostname> [--dev|--sandbox]
   millipds config [--pds_pfx=URL] [--pds_did=DID] [--bsky_appview_pfx=URL] [--bsky_appview_did=DID]
+  millipds account create <did> <handle> [--unsafe_password=PW]
   millipds run [--sock_path=PATH] [--listen_host=HOST] [--listen_port=PORT]
   millipds (-h | --help)
   millipds --version
@@ -19,11 +20,22 @@
   Any options not specified will be left at their previous values. Once changes
   have been made (or even if they haven't), the new config will be printed.
 
+  Do not change the config while the PDS is running (TODO: enforce this in code (or make sure it's harmless?))
+
   --pds_pfx=URL           The HTTP URL prefix that this PDS is publicly accessible at (e.g. mypds.example)
   --pds_did=DID           This PDS's DID (e.g. did:web:mypds.example)
   --bsky_appview_pfx=URL  AppView URL prefix e.g. "https://api.bsky-sandbox.dev"
   --bsky_appview_did=DID  AppView DID e.g. did:web:api.bsky-sandbox.dev
 
+Account create:
+  Create a new user account on the PDS. Bring your own DID and corresponding
+  handle - millipds will not (yet?) attempt to validate either.
+  You'll be prompted for a password interactively.
+
+  TODO: consider bring-your-own signing key?
+
+  --unsafe_password=PW  Specify password non-iteractively, for use in test scripts etc.
+
 Run:
   Launch the service (in the foreground)
 
@@ -36,12 +48,21 @@
   --version           Show version.
 """
 
-from docopt import docopt
 import importlib.metadata
 import asyncio
+import sys
+import logging
+from getpass import getpass
+
+from docopt import docopt
 
 from . import service
 from . import database
+from . import crypto
+
+
+logging.basicConfig(level=logging.DEBUG) # TODO: make this configurable?
+
 
 """
 This is the entrypoint for the `millipds` command (declared in project.scripts)
@@ -75,7 +96,7 @@ def main():
 				bsky_appview_pfx="https://api.bsky-sandbox.dev",
 				bsky_appview_did="did:web:api.bsky-sandbox.dev",
 			)
-		else:
+		else: # "prod" presets
 			db.update_config(
 				pds_pfx=f'https://{args["<hostname>"]}',
 				pds_did=f'did:web:{args["<hostname>"]}',
@@ -98,6 +119,24 @@ def main():
 			bsky_appview_did=args["--bsky_appview_did"],
 		)
 		db.print_config()
+	elif args["account"]:
+		if args["create"]:
+			pw = args["--unsafe_password"]
+			if pw:
+				# rationale: only allow non-iteractive password input from scripts etc.
+				if sys.stdin.buffer.isatty():
+					print("error: --unsafe_password can't be used from an interactive shell")
+					return
+			else:
+				pw = getpass(f"Password for new account: ")
+			db.account_create(
+				did=args["<did>"],
+				handle=args["<handle>"],
+				password=pw,
+				privkey=crypto.keygen_p256() # TODO: supply from arg
+			)
+		else:
+			print("CLI arg parse error?!")
 	elif args["run"]:
 		asyncio.run(service.run(
 			sock_path=args["--sock_path"],

src/millipds/crypto.py

@@ -1,5 +1,6 @@
 from cryptography.hazmat.primitives.asymmetric import ec
 from cryptography.hazmat.primitives import hashes
+from cryptography.hazmat.primitives import serialization
 from cryptography.hazmat.primitives.asymmetric.utils import decode_dss_signature, encode_dss_signature
 from cryptography.exceptions import InvalidSignature
 
@@ -47,3 +48,24 @@ def raw_sign(privkey: ec.EllipticCurvePrivateKey, data: bytes) -> bytes:
 	)
 	signature = r.to_bytes(32, "big") + s.to_bytes(32, "big")
 	return signature
+
+
+def keygen_p256() -> ec.EllipticCurvePrivateKey:
+	return ec.generate_private_key(ec.SECP256R1())
+
+
+def privkey_to_pem(privkey: ec.EllipticCurvePrivateKey) -> str:
+	return privkey.private_bytes(
+		encoding=serialization.Encoding.PEM,
+		format=serialization.PrivateFormat.PKCS8,
+		encryption_algorithm=serialization.NoEncryption()
+	).decode()
+
+
+def privkey_from_pem(pem: str) -> ec.EllipticCurvePrivateKey:
+	privkey = serialization.load_pem_private_key(pem.encode(), password=None)
+	if not isinstance(privkey, ec.EllipticCurvePrivateKey):
+		raise TypeError("unsupported key type")
+	if not isinstance(privkey.curve, (ec.SECP256R1, ec.SECP256K1)):
+		raise TypeError("unsupported key type")
+	return privkey

src/millipds/database.py

@@ -1,20 +1,36 @@
 """
-Ideally, all SQL statements are contained within this file
+Ideally, all SQL statements are contained within this file.
+
+Password hashing also happens in here, because it doesn't make much sense to do
+it anywhere else.
 """
 
 from typing import Optional, Dict
 from functools import cached_property
 import secrets
+import os
+import logging
 
+from argon2 import PasswordHasher # maybe this should come from .crypto?
 import apsw
+import apsw.bestpractice
+
+from atmst.blockstore import BlockStore
 
 from . import static_config
 from . import util
+from . import crypto
+
+logger = logging.getLogger(__name__)
+
+# https://rogerbinns.github.io/apsw/bestpractice.html
+apsw.bestpractice.apply(apsw.bestpractice.recommended)
 
 class Database:
 	def __init__(self, path: str=static_config.MAIN_DB_PATH) -> None:
 		util.mkdirs_for_file(path)
 		self.con = apsw.Connection(path)
+		self.pw_hasher = PasswordHasher()
 
 		try:
 			if self.config["db_version"] != static_config.MILLIPDS_DB_VERSION:
@@ -27,6 +43,7 @@ def __init__(self, path: str=static_config.MAIN_DB_PATH) -> None:
 				self._init_central_tables()
 
 	def _init_central_tables(self):
+		logger.info("initing central tables")
 		self.con.execute(
 			"""
 			CREATE TABLE config(
@@ -42,7 +59,7 @@ def _init_central_tables(self):
 
 		self.con.execute(
 			"""
-			INSERT INTO config (
+			INSERT INTO config(
 				db_version,
 				jwt_access_secret
 			) VALUES (?, ?)
@@ -54,6 +71,7 @@ def _init_central_tables(self):
 			"""
 			CREATE TABLE user(
 				did TEXT PRIMARY KEY NOT NULL,
+				handle TEXT NOT NULL,
 				prefs BLOB NOT NULL,
 				pw_hash TEXT NOT NULL,
 				repo_path TEXT NOT NULL,
@@ -62,6 +80,8 @@ def _init_central_tables(self):
 			"""
 		)
 
+		self.con.execute("CREATE UNIQUE INDEX user_by_handle ON user(handle)")
+
 		self.con.execute(
 			"""
 			CREATE TABLE firehose(
@@ -84,11 +104,20 @@ def update_config(self,
 			if pds_did is not None:
 				self.con.execute("UPDATE config SET pds_did=?", (pds_did,))
 			if bsky_appview_pfx is not None:
-				self.con.execute("UPDATE config SET bsky_appview_pfx=?", (bsky_appview_pfx,))
+				self.con.execute(
+					"UPDATE config SET bsky_appview_pfx=?",
+					(bsky_appview_pfx,)
+				)
 			if bsky_appview_did is not None:
-				self.con.execute("UPDATE config SET bsky_appview_did=?", (bsky_appview_did,))
+				self.con.execute(
+					"UPDATE config SET bsky_appview_did=?",
+					(bsky_appview_did,)
+				)
 
-		del self.config # invalidate the cached value
+		try:
+			del self.config # invalidate the cached value
+		except AttributeError:
+			pass
 
 	@cached_property
 	def config(self) -> Dict[str, object]:
@@ -117,3 +146,68 @@ def print_config(self, redact_secrets: bool=True) -> None:
 			if redact_secrets and "secret" in k:
 				v = "[REDACTED]"
 			print(f"{k:<{maxlen}} : {v!r}")
+
+	def account_create(self,
+		did: str,
+		handle: str,
+		password: str,
+		privkey: crypto.ec.EllipticCurvePrivateKey
+	) -> None:
+		pw_hash = self.pw_hasher.hash(password)
+		privkey_pem = crypto.privkey_to_pem(privkey)
+		repo_path = f"{static_config.REPOS_DIR}/{util.did_to_safe_filename(did)}.sqlite3"
+		logger.info(
+			f"creating account for did={did}, handle={handle} at {repo_path}"
+		)
+		with self.con:
+			self.con.execute(
+				"""
+				INSERT INTO user(
+					did,
+					handle,
+					prefs,
+					pw_hash,
+					repo_path,
+					signing_key
+				) VALUES (?, ?, ?, ?, ?, ?)
+				""",
+				(did, handle, b"{}", pw_hash, repo_path, privkey_pem)
+			)
+			UserDatabase.init_tables(self.con, did, repo_path)
+		self.con.execute("DETACH spoke")
+
+
+class UserDBBlockStore(BlockStore):
+	pass # TODO
+
+
+class UserDatabase:
+	def __init__(self, wcon: apsw.Connection, did: str, path: str) -> None:
+		self.wcon = wcon # writes go via the hub database connection, using ATTACH
+		self.rcon = apsw.Connection(path, flags=apsw.SQLITE_OPEN_READONLY)
+
+		# TODO: check db version and did match
+
+	@staticmethod
+	def init_tables(wcon: apsw.Connection, did: str, path: str) -> None:
+		util.mkdirs_for_file(path)
+		wcon.execute("ATTACH ? AS spoke", (path,))
+
+		wcon.execute(
+			"""
+			CREATE TABLE spoke.repo(
+				db_version INTEGER NOT NULL,
+				did TEXT NOT NULL
+			)
+			"""
+		)
+
+		wcon.execute(
+			"INSERT INTO spoke.repo(db_version, did) VALUES (?, ?)",
+			(static_config.MILLIPDS_DB_VERSION, did)
+		)
+
+		# TODO: the other tables
+
+		# nb: caller is responsible for running "DETACH spoke", after the end
+		# of the transaction

src/millipds/service.py

@@ -9,8 +9,6 @@
 
 from . import static_config
 
-logging.basicConfig(level=logging.DEBUG) # TODO: make this configurable?
-
 
 async def hello(request: web.Request):
 	version = importlib.metadata.version("millipds")
@@ -46,7 +44,7 @@ async def server_describe_server(request: web.Request):
 This gets invoked via millipds.__main__.py
 """
 async def run(sock_path: Optional[str], host: str, port: int):
-	runner = web.AppRunner(app, access_log_format=static_config.LOG_FMT)
+	runner = web.AppRunner(app, access_log_format=static_config.HTTP_LOG_FMT)
 	await runner.setup()
 
 	if sock_path is None:

src/millipds/static_config.py

@@ -4,12 +4,12 @@
 (some of this stuff might want to be broken out into a proper config file, eventually)
 """
 
-LOG_FMT = '%{X-Forwarded-For}i %t (%Tf) "%r" %s %b "%{Referer}i" "%{User-Agent}i"'
+HTTP_LOG_FMT = '%{X-Forwarded-For}i %t (%Tf) "%r" %s %b "%{Referer}i" "%{User-Agent}i"'
 
 GROUPNAME = "millipds-sock"
 
 MILLIPDS_DB_VERSION = 1 # this gets bumped if we make breaking changes to the db schema
 
-DATA_DIR = "./data/"
-MAIN_DB_PATH = DATA_DIR + "millipds.sqlite3"
-REPOS_DIR = DATA_DIR + "repos/"
+DATA_DIR = "./data"
+MAIN_DB_PATH = DATA_DIR + "/millipds.sqlite3"
+REPOS_DIR = DATA_DIR + "/repos"

src/millipds/util.py

@@ -1,4 +1,20 @@
 import os
+import hashlib
 
 def mkdirs_for_file(path: str) -> None:
 	os.makedirs(os.path.dirname(path), exist_ok=True)
+
+FILANEME_SAFE_CHARS = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_"
+
+def did_to_safe_filename(did: str) -> str:
+	"""
+	The format is <sha256(did)>_<filtered_did>
+	The former guarantees uniqueness, and the latter makes it human-recognizeable (ish)
+	"""
+
+	hexdigest = hashlib.sha256(did.encode()).hexdigest()
+	filtered = "".join(char for char in did if char in FILANEME_SAFE_CHARS)
+
+	# Truncate to make sure we're staying within PATH_MAX
+	# (with room to spare, in case the caller appends a file extension)
+	return f"{hexdigest}_{filtered}"[:200]