[supervisord] implement sd_notify

.gitmodules

@@ -142,3 +142,6 @@
 [submodule "platform/marvell-teralynx/mrvl-teralynx"]
 	path = platform/marvell-teralynx/mrvl-teralynx
 	url = https://github.com/Marvell-switching/mrvl-teralynx.git
+[submodule "src/supervisor"]
+	path = src/supervisor
+	url = https://github.com/Supervisor/supervisor

dockers/docker-base-bookworm/Dockerfile.j2

@@ -60,12 +60,17 @@ COPY ["pip.conf", "/etc/pip.conf"]
 RUN pip3 install --upgrade pip
 RUN apt purge -y python3-pip
 
+{% if docker_base_bookworm_whls.strip() -%}
+# Copy locally-built Python wheel dependencies
+{{ copy_files("python-wheels/", docker_base_bookworm_whls.split(' '), "/python-wheels/") }}
+
+# Install locally-built Python wheel dependencies
+{{ install_python_wheels(docker_base_bookworm_whls.split(' ')) }}
+{% endif %}
+
 # For templating
 RUN pip3 install j2cli
 
-# Install supervisor
-RUN pip3 install supervisor==4.2.5
-
 # Add support for supervisord to handle startup dependencies
 RUN pip3 install supervisord-dependent-startup==1.4.0
 

dockers/docker-base-bookworm/etc/supervisor/supervisord.conf

@@ -7,6 +7,7 @@ chmod=0700                     ; socket file mode (default 0700)
 [supervisord]
 logfile=/var/log/supervisor/supervisord.log  ; (main log file;default $CWD/supervisord.log)
 pidfile=/var/run/supervisord.pid             ; (supervisord pidfile;default supervisord.pid)
+notifysock=/var/run/supervisord_notify.sock  ; (path to the notify socket file; default None)
 childlogdir=/var/log/supervisor              ; ('AUTO' child log dir, default $TEMP)
 user=root
 

rules/docker-base-bookworm.mk

@@ -16,6 +16,8 @@ ifeq ($(INCLUDE_FIPS), y)
 $(DOCKER_BASE_BOOKWORM)_DEPENDS += $(FIPS_OPENSSL_LIBSSL) $(FIPS_OPENSSL_LIBSSL_DEV) $(FIPS_OPENSSL) $(SYMCRYPT_OPENSSL) $(FIPS_KRB5)
 endif
 
+$(DOCKER_BASE_BOOKWORM)_PYTHON_WHEELS += $(SUPERVISOR)
+
 $(DOCKER_BASE_BOOKWORM)_DBG_IMAGE_PACKAGES += $(GDB) $(GDBSERVER) $(VIM) $(OPENSSH) $(SSHPASS) $(STRACE)
 
 SONIC_DOCKER_IMAGES += $(DOCKER_BASE_BOOKWORM)

rules/supervisor.dep

@@ -0,0 +1,10 @@
+SPATH       := $($(SUPERVISOR)_SRC_PATH)
+DEP_FILES   := $(SONIC_COMMON_FILES_LIST) rules/supervisor.mk rules/supervisor.dep
+DEP_FILES   += $(SONIC_COMMON_BASE_FILES_LIST)
+SMDEP_FILES := $(addprefix $(SPATH)/,$(shell cd $(SPATH) && git ls-files -- ':!:doc/*'))
+
+$(SUPERVISOR)_CACHE_MODE  := GIT_CONTENT_SHA
+$(SUPERVISOR)_DEP_FLAGS   := $(SONIC_COMMON_FLAGS_LIST)
+$(SUPERVISOR)_DEP_FILES   := $(DEP_FILES)
+$(SUPERVISOR)_SMDEP_FILES := $(SMDEP_FILES)
+$(SUPERVISOR)_SMDEP_PATHS := $(SPATH)

rules/supervisor.mk

@@ -0,0 +1,6 @@
+# supervisor python3 wheel
+
+SUPERVISOR = supervisor-4.2.5-py2.py3-none-any.whl
+$(SUPERVISOR)_SRC_PATH = $(SRC_PATH)/supervisor
+$(SUPERVISOR)_PYTHON_VERSION = 3
+SONIC_PYTHON_WHEELS += $(SUPERVISOR)

src/supervisor.patch/0002-Implement-SD_NOTIFY.patch

@@ -0,0 +1,207 @@
+From 3ab7d6f2007188ddce80fa3c1d6196cf888b3051 Mon Sep 17 00:00:00 2001
+From: Stepan Blyschak <[email protected]>
+Date: Mon, 29 Sep 2025 14:09:22 +0000
+Subject: [PATCH] Implement SD_NOTIFY
+
+Signed-off-by: Stepan Blyschak <[email protected]>
+---
+ supervisor/options.py     | 20 ++++++++++++++++++++
+ supervisor/process.py     | 23 +++++++++++++++++++++++
+ supervisor/supervisord.py | 17 +++++++++++++++++
+ supervisor/tests/base.py  |  5 +++++
+ 4 files changed, 65 insertions(+)
+
+diff --git a/supervisor/options.py b/supervisor/options.py
+index f5cb0c9..49415fb 100644
+--- a/supervisor/options.py
++++ b/supervisor/options.py
+@@ -410,6 +410,8 @@ class ServerOptions(Options):
+     unlink_pidfile = False
+     unlink_socketfiles = False
+     mood = states.SupervisorStates.RUNNING
++    notify_sock_path = None
++    notify_sock = None
+
+     def __init__(self):
+         Options.__init__(self)
+@@ -521,6 +523,8 @@ class ServerOptions(Options):
+
+         self.server_configs = sconfigs = section.server_configs
+
++        self.notify_sock_path = section.notify_sock_path
++
+         # we need to set a fallback serverurl that process.spawn can use
+
+         # prefer a unix domain socket
+@@ -645,6 +649,9 @@ class ServerOptions(Options):
+         section.identifier = get('identifier', 'supervisor')
+         section.nodaemon = boolean(get('nodaemon', 'false'))
+         section.silent = boolean(get('silent', 'false'))
++        section.notify_sock_path = get('notifysock', None)
++        if section.notify_sock_path is not None:
++            section.notify_sock_path = existing_dirpath(section.notify_sock_path)
+
+         tempdir = tempfile.gettempdir()
+         section.childlogdir = existing_directory(get('childlogdir', tempdir))
+@@ -1246,6 +1253,11 @@ class ServerOptions(Options):
+             # also https://web.archive.org/web/20160729222427/http://www.plope.com/software/collector/253
+             server.close()
+
++    def close_notify_socket(self):
++        if self.notify_sock is None:
++            return
++        self.notify_sock.close()
++
+     def close_logger(self):
+         self.logger.close()
+
+@@ -1282,6 +1294,14 @@ class ServerOptions(Options):
+         except ValueError as why:
+             self.usage(why.args[0])
+
++    def open_notify_socket(self):
++        if self.notify_sock_path is None:
++            return
++        self._try_unlink(self.notify_sock_path)
++        self.notify_sock = socket.socket(socket.AF_UNIX, socket.SOCK_DGRAM)
++        self.notify_sock.bind(self.notify_sock_path)
++        self.notify_sock.setsockopt(socket.SOL_SOCKET, socket.SO_PASSCRED, 1)
++
+     def get_autochildlog_name(self, name, identifier, channel):
+         prefix='%s-%s---%s-' % (name, channel, identifier)
+         logfile = self.mktempfile(
+diff --git a/supervisor/process.py b/supervisor/process.py
+index b394be8..bb6a69c 100644
+--- a/supervisor/process.py
++++ b/supervisor/process.py
+@@ -310,6 +310,9 @@ class Subprocess(object):
+             # set environment
+             env = os.environ.copy()
+             env['SUPERVISOR_ENABLED'] = '1'
++            notify_sock_path = self.config.options.notify_sock_path
++            if notify_sock_path is not None:
++                env['NOTIFY_SOCKET'] = notify_sock_path
+             serverurl = self.config.serverurl
+             if serverurl is None: # unset
+                 serverurl = self.config.options.serverurl # might still be None
+@@ -717,6 +720,17 @@ class Subprocess(object):
+                                                           self.pid))
+                 self.kill(signal.SIGKILL)
+
++    def handle_sd_notify(self, msg):
++        for kv in msg.splitlines():
++            if kv == "READY=1":
++                if self.state == ProcessStates.STARTING:
++                    self.delay = 0
++                    self.backoff = 0
++                    self.change_state(ProcessStates.RUNNING)
++                    msg = 'entered RUNNING state, process sent READY=1'
++                    self.config.options.logger.info('success: %s %s' % (self.config.name, msg))
++
++
+ class FastCGISubprocess(Subprocess):
+     """Extends Subprocess class to handle FastCGI subprocesses"""
+
+@@ -841,11 +855,20 @@ class ProcessGroupBase(object):
+     def before_remove(self):
+         pass
+
++    def handle_sd_notify(self, _msg, _pid):
++        pass
++
+ class ProcessGroup(ProcessGroupBase):
+     def transition(self):
+         for proc in self.processes.values():
+             proc.transition()
+
++    def handle_sd_notify(self, msg, pid):
++        for proc in self.processes.values():
++            if proc.pid == pid:
++                proc.handle_sd_notify(msg)
++                break
++
+ class FastCGIProcessGroup(ProcessGroup):
+
+     def __init__(self, config, **kwargs):
+diff --git a/supervisor/supervisord.py b/supervisor/supervisord.py
+index 0a4f3e6..968e4d2 100755
+--- a/supervisor/supervisord.py
++++ b/supervisor/supervisord.py
+@@ -85,6 +85,7 @@ class Supervisor:
+             for config in self.options.process_group_configs:
+                 self.add_process_group(config)
+             self.options.openhttpservers(self)
++            self.options.open_notify_socket()
+             self.options.setsignals()
+             if (not self.options.nodaemon) and self.options.first:
+                 self.options.daemonize()
+@@ -176,6 +177,7 @@ class Supervisor:
+         timeout = 1 # this cannot be fewer than the smallest TickEvent (5)
+
+         socket_map = self.options.get_socket_map()
++        notify_sock = self.options.notify_sock
+
+         while 1:
+             combined_map = {}
+@@ -200,6 +202,9 @@ class Supervisor:
+                     # killing everything), it's OK to shutdown or reload
+                     raise asyncore.ExitNow
+
++            if self.options.notify_sock:
++                self.options.poller.register_readable(notify_sock.fileno())
++
+             for fd, dispatcher in combined_map.items():
+                 if dispatcher.readable():
+                     self.options.poller.register_readable(fd)
+@@ -238,6 +243,17 @@ class Supervisor:
+                     except:
+                         combined_map[fd].handle_error()
+
++            if notify_sock and notify_sock.fileno() in r:
++                import socket
++                import struct
++                data, ancdata, _, _ = notify_sock.recvmsg(4096, socket.CMSG_SPACE(struct.calcsize("3i")))
++                msg = data.decode("utf-8")
++                for cmsg_level, cmsg_type, cmsg_data in ancdata:
++                    if cmsg_level == socket.SOL_SOCKET and cmsg_type == socket.SCM_CREDENTIALS:
++                        pid, _, _ = struct.unpack("3i", cmsg_data)
++                for group in pgroups:
++                    group.handle_sd_notify(msg, pid)
++
+             for group in pgroups:
+                 group.transition()
+
+@@ -358,6 +374,7 @@ def main(args=None, test=False):
+         else:
+             go(options)
+         options.close_httpservers()
++        options.close_notify_socket()
+         options.close_logger()
+         first = False
+         if test or (options.mood < SupervisorStates.RESTARTING):
+diff --git a/supervisor/tests/base.py b/supervisor/tests/base.py
+index f608b2b..ed795db 100644
+--- a/supervisor/tests/base.py
++++ b/supervisor/tests/base.py
+@@ -31,6 +31,8 @@ class DummyOptions:
+     make_pipes_exception = None
+     remove_exception = None
+     write_exception = None
++    notify_sock_path = None
++    notify_sock = None
+
+     def __init__(self):
+         self.identifier = 'supervisor'
+@@ -117,6 +119,9 @@ class DummyOptions:
+     def openhttpservers(self, supervisord):
+         self.httpservers_opened = True
+
++    def open_notify_socket(self):
++        pass
++
+     def daemonize(self):
+         self.daemonized = True
+
+-- 
+2.39.5
+

src/supervisor.patch/series

@@ -0,0 +1 @@
+0002-Implement-SD_NOTIFY.patch