Coverage for bzfs_main/detect.py: 96%
209 statements
« prev ^ index » next coverage.py v7.11.0, created at 2025-11-07 04:44 +0000
1# Copyright 2024 Wolfgang Hoschek AT mac DOT com
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7# http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14#
15"""Detection of ZFS features and system capabilities on local and remote hosts."""
17from __future__ import (
18 annotations,
19)
20import re
21import subprocess
22import threading
23import time
24from dataclasses import (
25 dataclass,
26 field,
27)
28from subprocess import (
29 DEVNULL,
30 PIPE,
31)
32from typing import (
33 TYPE_CHECKING,
34 Final,
35)
37from bzfs_main.connection import (
38 DEDICATED,
39 SHARED,
40 ConnectionPools,
41 run_ssh_command,
42 try_ssh_command,
43)
44from bzfs_main.utils import (
45 LOG_TRACE,
46 PROG_NAME,
47 SynchronousExecutor,
48 die,
49 drain,
50 list_formatter,
51 stderr_to_str,
52 xprint,
53)
55if TYPE_CHECKING: # pragma: no cover - for type hints only
56 from bzfs_main.bzfs import (
57 Job,
58 )
59 from bzfs_main.configuration import (
60 Params,
61 Remote,
62 )
64# constants:
65DISABLE_PRG: Final[str] = "-"
66DUMMY_DATASET: Final[str] = "dummy"
67ZFS_VERSION_IS_AT_LEAST_2_1_0: Final[str] = "zfs>=2.1.0"
68ZFS_VERSION_IS_AT_LEAST_2_2_0: Final[str] = "zfs>=2.2.0"
71#############################################################################
72@dataclass(frozen=True)
73class RemoteConfCacheItem:
74 """Caches detected programs, zpool features and connection pools, per remote."""
76 connection_pools: ConnectionPools
77 available_programs: dict[str, str]
78 zpool_features: dict[str, dict[str, str]]
79 timestamp_nanos: int = field(default_factory=time.monotonic_ns)
82def detect_available_programs(job: Job) -> None:
83 """Detects programs, zpool features and connection pools for local and remote hosts."""
84 p = params = job.params
85 log = p.log
86 available_programs: dict[str, dict[str, str]] = params.available_programs
87 if "local" not in available_programs:
88 cmd: list[str] = [p.shell_program_local, "-c", _find_available_programs(p)]
89 sp = job.subprocesses
90 proc = sp.subprocess_run(cmd, stdin=DEVNULL, stdout=PIPE, stderr=PIPE, text=True)
91 xprint(log=log, value=stderr_to_str(proc.stderr), end="")
92 stdout: str = proc.stdout
93 available_programs["local"] = dict.fromkeys(stdout.splitlines(), "")
94 cmd = [p.shell_program_local, "-c", "exit"]
95 proc = sp.subprocess_run(cmd, stdin=DEVNULL, stdout=PIPE, stderr=PIPE, text=True)
96 xprint(log=log, value=stderr_to_str(proc.stderr), end="")
97 if proc.returncode != 0:
98 _disable_program(p, "sh", ["local"])
100 todo: list[Remote] = []
101 for r in [p.dst, p.src]:
102 loc: str = r.location
103 remote_conf_cache_key: tuple = r.cache_key()
104 cache_item: RemoteConfCacheItem | None = job.remote_conf_cache.get(remote_conf_cache_key)
105 if cache_item is not None:
106 # startup perf: cache avoids ssh connect setup and feature detection roundtrips on revisits to same site
107 p.connection_pools[loc] = cache_item.connection_pools
108 p.available_programs[loc] = cache_item.available_programs
109 p.zpool_features[loc] = cache_item.zpool_features
110 if time.monotonic_ns() - cache_item.timestamp_nanos < p.remote_conf_cache_ttl_nanos:
111 if r.pool in cache_item.zpool_features: 111 ↛ 119line 111 didn't jump to line 119 because the condition on line 111 was always true
112 continue # cache hit, skip remote detection
113 else:
114 p.zpool_features[loc] = {} # cache miss, invalidate features of zpools before refetching from remote
115 else:
116 p.connection_pools[loc] = ConnectionPools(
117 r, {SHARED: r.max_concurrent_ssh_sessions_per_tcp_connection, DEDICATED: 1}
118 )
119 todo.append(r)
121 lock: threading.Lock = threading.Lock()
123 def run_detect(r: Remote) -> None: # thread-safe
124 loc: str = r.location
125 remote_conf_cache_key: tuple = r.cache_key()
126 available_programs: dict[str, str] = _detect_available_programs_remote(job, r, r.ssh_user_host)
127 zpool_features: dict[str, str] = _detect_zpool_features(job, r, available_programs)
128 with lock:
129 r.params.available_programs[loc] = available_programs
130 r.params.zpool_features[loc][r.pool] = zpool_features
131 job.remote_conf_cache[remote_conf_cache_key] = RemoteConfCacheItem(
132 p.connection_pools[loc], available_programs, r.params.zpool_features[loc]
133 )
134 if r.use_zfs_delegation and zpool_features.get("delegation") == "off":
135 die(
136 f"Permission denied as ZFS delegation is disabled for {r.location} "
137 f"dataset: {r.basis_root_dataset}. Manually enable it via 'sudo zpool set delegation=on {r.pool}'"
138 )
140 with SynchronousExecutor.executor_for(max_workers=max(1, len(todo))) as executor:
141 drain(executor.map(run_detect, todo)) # detect ZFS features + system capabilities on src+dst in parallel
143 locations = ["src", "dst", "local"]
144 if params.compression_program == DISABLE_PRG:
145 _disable_program(p, "zstd", locations)
146 if params.mbuffer_program == DISABLE_PRG:
147 _disable_program(p, "mbuffer", locations)
148 if params.ps_program == DISABLE_PRG:
149 _disable_program(p, "ps", locations)
150 if params.pv_program == DISABLE_PRG:
151 _disable_program(p, "pv", locations)
152 if params.shell_program == DISABLE_PRG:
153 _disable_program(p, "sh", locations)
154 if params.sudo_program == DISABLE_PRG:
155 _disable_program(p, "sudo", locations)
156 if params.zpool_program == DISABLE_PRG:
157 _disable_program(p, "zpool", locations)
159 for key, programs in available_programs.items():
160 for program in list(programs.keys()):
161 if program.startswith("uname-"):
162 # uname-Linux foo 5.15.0-69-generic #76-Ubuntu SMP Fri Mar 17 17:19:29 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux
163 # uname-FreeBSD freebsd 14.1-RELEASE FreeBSD 14.1-RELEASE releng/14.1-n267679-10e31f0946d8 GENERIC amd64
164 # uname-Darwin foo 23.6.0 Darwin Kernel Version 23.6.0: Mon Jul 29 21:13:04 PDT 2024; root:xnu-10063.141.2~1/RELEASE_ARM64_T6020 arm64
165 programs.pop(program)
166 uname: str = program[len("uname-") :]
167 programs["uname"] = uname
168 log.log(LOG_TRACE, f"available_programs[{key}][uname]: %s", uname)
169 programs["os"] = uname.split(" ", maxsplit=1)[0] # Linux|FreeBSD|Darwin
170 log.log(LOG_TRACE, f"available_programs[{key}][os]: %s", programs["os"])
171 elif program.startswith("default_shell-"):
172 programs.pop(program)
173 default_shell: str = program[len("default_shell-") :]
174 programs["default_shell"] = default_shell
175 log.log(LOG_TRACE, f"available_programs[{key}][default_shell]: %s", default_shell)
176 ssh_user_host = p.src.ssh_user_host if key == "src" else p.dst.ssh_user_host if key == "dst" else ""
177 _validate_default_shell(default_shell, key, ssh_user_host)
178 elif program.startswith("getconf_cpu_count-"):
179 programs.pop(program)
180 getconf_cpu_count: str = program[len("getconf_cpu_count-") :]
181 programs["getconf_cpu_count"] = getconf_cpu_count
182 log.log(LOG_TRACE, f"available_programs[{key}][getconf_cpu_count]: %s", getconf_cpu_count)
184 for key, programs in available_programs.items():
185 log.debug(f"available_programs[{key}]: %s", list_formatter(programs, separator=", "))
187 for r in [p.dst, p.src]:
188 if is_dummy(r):
189 continue
190 if r.sudo and not p.is_program_available("sudo", r.location):
191 die(f"{p.sudo_program} CLI is not available on {r.location} host: {r.ssh_user_host or 'localhost'}")
193 if (
194 len(p.args.preserve_properties) > 0
195 and any(prop in p.zfs_send_program_opts for prop in ["--props", "-p"])
196 and not p.is_program_available(ZFS_VERSION_IS_AT_LEAST_2_2_0, p.dst.location)
197 ):
198 die(
199 "Cowardly refusing to proceed as --preserve-properties is unreliable on destination ZFS < 2.2.0 when using "
200 "'zfs send --props'. Either upgrade destination ZFS, or remove '--props' from --zfs-send-program-opt(s)."
201 )
204def _disable_program(p: Params, program: str, locations: list[str]) -> None:
205 """Removes the given program from the available_programs mapping."""
206 for location in locations:
207 p.available_programs[location].pop(program, None)
210def _find_available_programs(p: Params) -> str:
211 """POSIX shell script that checks for the existence of various programs; It uses `if` statements instead of `&&` plus
212 `printf` instead of `echo` to ensure maximum compatibility across shells."""
213 cmds: list[str] = []
214 cmds.append("printf 'default_shell-%s\n' \"$SHELL\"")
215 cmds.append("if command -v echo > /dev/null; then printf 'echo\n'; fi")
216 cmds.append(f"if command -v {p.zpool_program} > /dev/null; then printf 'zpool\n'; fi")
217 cmds.append(f"if command -v {p.ssh_program} > /dev/null; then printf 'ssh\n'; fi")
218 cmds.append(f"if command -v {p.shell_program} > /dev/null; then printf 'sh\n'; fi")
219 cmds.append(f"if command -v {p.sudo_program} > /dev/null; then printf 'sudo\n'; fi")
220 cmds.append(f"if command -v {p.compression_program} > /dev/null; then printf 'zstd\n'; fi")
221 cmds.append(f"if command -v {p.mbuffer_program} > /dev/null; then printf 'mbuffer\n'; fi")
222 cmds.append(f"if command -v {p.pv_program} > /dev/null; then printf 'pv\n'; fi")
223 cmds.append(f"if command -v {p.ps_program} > /dev/null; then printf 'ps\n'; fi")
224 cmds.append(
225 f"if command -v {p.getconf_program} > /dev/null; then "
226 f"printf 'getconf_cpu_count-'; {p.getconf_program} _NPROCESSORS_ONLN; "
227 "fi"
228 )
229 cmds.append(f"if command -v {p.uname_program} > /dev/null; then printf 'uname-'; {p.uname_program} -a || true; fi")
230 return "; ".join(cmds)
233def _detect_available_programs_remote(job: Job, remote: Remote, ssh_user_host: str) -> dict[str, str]:
234 """Detects CLI tools available on ``remote`` and updates mapping correspondingly."""
235 p, log = job.params, job.params.log
236 location = remote.location
237 available_programs_minimum = {"sudo": ""}
238 available_programs: dict[str, str] = {}
239 if is_dummy(remote):
240 return available_programs
241 lines: str | None = None
242 try:
243 # on Linux, 'zfs --version' returns with zero status and prints the correct info
244 # on FreeBSD, 'zfs --version' always prints the same (correct) info as Linux, but nonetheless sometimes
245 # returns with non-zero status (sometimes = if the zfs kernel module is not loaded)
246 lines = run_ssh_command(job, remote, LOG_TRACE, print_stderr=False, cmd=[p.zfs_program, "--version"])
247 assert lines
248 except (FileNotFoundError, PermissionError): # location is local and program file was not found
249 die(f"{p.zfs_program} CLI is not available on {location} host: {ssh_user_host or 'localhost'}")
250 except subprocess.CalledProcessError as e:
251 if "unrecognized command '--version'" in e.stderr and "run: zfs help" in e.stderr:
252 die(f"Unsupported ZFS platform: {e.stderr}") # solaris is unsupported
253 elif not e.stdout.startswith("zfs"): 253 ↛ 256line 253 didn't jump to line 256 because the condition on line 253 was always true
254 die(f"{p.zfs_program} CLI is not available on {location} host: {ssh_user_host or 'localhost'}")
255 else:
256 lines = e.stdout # FreeBSD if the zfs kernel module is not loaded
257 assert lines
258 if lines: 258 ↛ 270line 258 didn't jump to line 270 because the condition on line 258 was always true
259 # Examples that should parse: "zfs-2.1.5~rc5-ubuntu3", "zfswin-2.2.3rc5"
260 first_line: str = lines.splitlines()[0] if lines.splitlines() else ""
261 match = re.search(r"(\d+)\.(\d+)\.(\d+)", first_line)
262 if not match:
263 die("Unparsable zfs version string: '" + first_line + "'")
264 version = ".".join(match.groups())
265 available_programs["zfs"] = version
266 if is_version_at_least(version, "2.1.0"):
267 available_programs[ZFS_VERSION_IS_AT_LEAST_2_1_0] = ""
268 if is_version_at_least(version, "2.2.0"):
269 available_programs[ZFS_VERSION_IS_AT_LEAST_2_2_0] = ""
270 log.log(LOG_TRACE, f"available_programs[{location}][zfs]: %s", available_programs["zfs"])
272 if p.shell_program != DISABLE_PRG:
273 try:
274 cmd: list[str] = [p.shell_program, "-c", _find_available_programs(p)]
275 stdout: str = run_ssh_command(job, remote, LOG_TRACE, cmd=cmd)
276 available_programs.update(dict.fromkeys(stdout.splitlines(), ""))
277 return available_programs
278 except (FileNotFoundError, PermissionError) as e: # location is local and shell program file was not found
279 if e.filename != p.shell_program: 279 ↛ 280line 279 didn't jump to line 280 because the condition on line 279 was never true
280 raise
281 except subprocess.CalledProcessError:
282 pass
283 log.warning("%s", f"Failed to find {p.shell_program} on {location}. Continuing with minimal assumptions...")
284 available_programs.update(available_programs_minimum)
285 return available_programs
288def is_dummy(r: Remote) -> bool:
289 """Returns True if ``remote`` refers to the synthetic dummy dataset."""
290 return r.root_dataset == DUMMY_DATASET
293def _detect_zpool_features(job: Job, remote: Remote, available_programs: dict) -> dict[str, str]:
294 """Fills ``job.params.zpool_features`` with detected zpool capabilities."""
295 p = params = job.params
296 r, loc, log = remote, remote.location, p.log
297 lines: list[str] = []
298 features: dict[str, str] = {}
299 if is_dummy(r):
300 return {}
301 if params.zpool_program != DISABLE_PRG and (params.shell_program == DISABLE_PRG or "zpool" in available_programs):
302 cmd: list[str] = params.split_args(f"{params.zpool_program} get -Hp -o property,value all", r.pool)
303 try:
304 lines = run_ssh_command(job, remote, LOG_TRACE, check=False, cmd=cmd).splitlines()
305 except (FileNotFoundError, PermissionError) as e:
306 if e.filename != params.zpool_program:
307 raise
308 log.warning("%s", f"Failed to detect zpool features on {loc}: {r.pool}. Continuing with minimal assumptions ...")
309 else:
310 props: dict[str, str] = dict(line.split("\t", 1) for line in lines)
311 features = {k: v for k, v in props.items() if k.startswith("feature@") or k == "delegation"}
312 if len(lines) == 0:
313 cmd = p.split_args(f"{p.zfs_program} list -t filesystem -Hp -o name -s name", r.pool)
314 if try_ssh_command(job, remote, LOG_TRACE, cmd=cmd) is None:
315 die(f"Pool does not exist for {loc} dataset: {r.basis_root_dataset}. Manually create the pool first!")
316 return features
319def is_zpool_feature_enabled_or_active(p: Params, remote: Remote, feature: str) -> bool:
320 """Returns True if the given zpool feature is active or enabled on ``remote``."""
321 return p.zpool_features[remote.location][remote.pool].get(feature) in ("active", "enabled")
324def are_bookmarks_enabled(p: Params, remote: Remote) -> bool:
325 """Checks if bookmark related features are enabled on ``remote``."""
326 return is_zpool_feature_enabled_or_active(p, remote, "feature@bookmark_v2") and is_zpool_feature_enabled_or_active(
327 p, remote, "feature@bookmark_written"
328 )
331def is_caching_snapshots(p: Params, remote: Remote) -> bool:
332 """Returns True if snapshot caching is supported and enabled on ``remote``."""
333 return p.is_caching_snapshots and p.is_program_available(ZFS_VERSION_IS_AT_LEAST_2_2_0, remote.location)
336def is_version_at_least(version_str: str, min_version_str: str) -> bool:
337 """Checks if the version string is at least the minimum version string."""
338 return tuple(map(int, version_str.split("."))) >= tuple(map(int, min_version_str.split(".")))
341def _validate_default_shell(path_to_default_shell: str, location: str, ssh_user_host: str) -> None:
342 """Fails if the remote user uses csh or tcsh as the default shell."""
343 if path_to_default_shell in ("csh", "tcsh") or path_to_default_shell.endswith(("/csh", "/tcsh")):
344 # On some old FreeBSD systems the default shell is still csh. Also see https://www.grymoire.com/unix/CshTop10.txt
345 die(
346 f"Cowardly refusing to proceed because {PROG_NAME} is not compatible with csh-style quoting of special "
347 f"characters. The safe workaround is to first manually set 'sh' instead of '{path_to_default_shell}' as "
348 f"the default shell of the Unix user on {location} host: {ssh_user_host or 'localhost'}, like so: "
349 "chsh -s /bin/sh YOURUSERNAME"
350 )