June 2009 - launchd-dev - lists.macosforge.org

SampleD failure with dynamic SockServiceName
by David Anderson 13 Aug '09

13 Aug '09

My intention is have launchd start a daemon on a dynamic port, publish the port number via bonjour, and then connect to this port after determining the port number using CFNetService (Bonjour). Starting from SampleD (and using the included client.pl), I can advertise the service on a hard-coded port via bonjour, lookup the port number using Bonjour Browser, modify client.pl to use this port, and connect using client.pl. The SampleD binary starts correctly in an OnDemand fashion correctly. If I change <key>SockServiceName</key> <string>12345</string> to <key>SockServiceName</key> <string>0</string> then lookup the dynamic port number using Bonjour Browser and configure client.pl to use this port, client.pl will NOT connect to this port and outputs this error: connect: Connection refused at ./client.pl line 77. With a dynamic port, I am unable to connect to this port using a modified client.pl. I can find no evidence that the SampleD binary is starting in an OnDemand fashion. Adding <key>OnDemand</key> <false/> has no effect. Is there anything different about the default configuration of a dynamic port vs a hard-coded port that would impeded the ability to connect to the dynamically assigned port? Thanks, Dave

3 2

About the ServiceIPC key
by Thomas Clement 30 Jun '09

30 Jun '09

Hi, I'm trying to figure out when the ServiceIPC key should be used. This key is not documented but it is used in the various sample codes. My understanding is that this key is required on 10.4 to check-in with launchd but its use is deprecated on 10.5 and later. Am I right? Thanks, Thomas

3 5

launchd: how is it supposed to be done?
by Iceberg-Dev 30 Jun '09

30 Jun '09

What I can do: -------------- Create a plist file for launchd to launch a tool every time at startup (even though the required plist contents are apparently changing with each OS release because it's probably more funny like that). What I am trying to do: ----------------------- I'm trying to launch a helper daemon (not an agent) on demand. Ideally, it would work this way: - Process (P) needs to talk with Daemon (D) - Daemon (D) is launched on demand by launchd based on some socket or mach port magic. - (D) notifies (P) that it's ready to listen - (P) talks to (D). (P) does what it is supposed to do then. - After that (P) waits for a while and if nothing happens, it quits. What I have tried so far: ------------------------- - Read the launchd.plist and launchdman pages. It just gives me headaches as like most man pages it assumes you know how to do something and is not didactic. - Tried the sample code from the ADC Website. It does not event built and the bug report I filed to suggest to get it fixed was probably sent to /dev/null as I haven't heard about it since. - Looked for sample code using google. Was not able to find a single good example. There are plenty of examples showing how to write equivalents of crontab, StartupItems but none for triggering a daemon with sockets or a mach port. Some of them apparently try to deal with file dependencies it is not clear if they are even sure with this. By the way: ----------- I attended the session on launchd at WWDC and was still like Alicia Silverstone in that 1995 movie after that. I must be an idiot. Questions: ---------- - Is there an example somewhere that _works_ and covers launching a daemon with launchd using sockets or mach port as dependencies? - Is there a tutorial somewhere that covers this subject? - Are these mach port and socket things really working? I dare ask because socket dependency was supposed to work on 10.4.0 and AFAIK, it was only supported way later.

3 7

early boot hook in launchd/launchctl?
by Greg Shenaut 24 Jun '09

24 Jun '09

Until recently, it was possible to put arbitrary code into /etc/ rc.server to perform scripted maintenance activities in the early boot context (i.e., system disk still write-only, limited driver access, etc). In the version distributed with 10.5.7, rc.server shows up later in the sequence (i.e., system disk mounted read-write). In any case, adding code to rc.server always was a kludge. Now, the cleanest way to do it seems to be to put code into /.env and use "nvram boot-args=-s ; reboot" to run it. This is actually a much cooler kludge, because /.env (the value of ENV in single-user mode) is normally run only in single-user mode, because otherwise the HOME variable will be set, and for root that will cause ENV normally to be defined as /var/root/.env. Because /.env only gets run in single-user mode, you can put in fairly elaborate checks to make sure you're in in the right place without negatively impacting startup time for normal root shell scripts. But I'm concerned that this method, like rc.server, will someday be seen by the launchd developers as a bug (that is, HOME not being defined in single-user mode); in this case, it too will suddenly disappear and some other kludge will have to be found. So what I'm wondering is whether it might not be a good idea to provide a conventional hook for early-boot maintenance. It could simply be a shell script in a standard location such as /etc/ rc.earlyboot which gets executed by launchd if present, owned by root, and not writable except by root. This would keep people who insist on doing this kind of thing from having to keep finding new and better kludges. Or, of course, the /.env thing and the HOME-less single-user shell could become a feature of launchd that is reserved for this purpose. Thank you, Greg Shenaut

2 2

Re: [launchd-dev] launch_msg(): Socket is not connected
by Arjun SM 24 Jun '09

24 Jun '09

>> Keep in mind that the multiple GUI logins via fast user switching is an edge case that 90% of your users will never encounter. I will take your word for it . I was of the notion that i have to support multiple users for my application. Also i had one more scenario and this is while Uninstalling . The scenario is If i install my application on One User and i log in to other User via Fast User Switching and Other user being the System Administrator (root) and Uninstall for all Users being in root ... i get these log messages (com.mycmpny.appname) Throttling respawn: Will restart in 10 seconds. (com.mycmpny.appname) Throttling respawn: Will restart in 10 seconds. (com.mycmpny.appname) Throttling respawn: Will restart in 10 seconds. i have used the below code for removing the job from launchd su -l $userName -c "launchctl unload -S Aqua /Library/LaunchAgents/com.mycmpny.appname.plist This code works fine for normal and Admin Users though. First of all I want to ask you if it is a valid Usecase. If so how should i be removing the job form launchd. Thanks, Arjun On Thu, Jun 18, 2009 at 9:58 PM, Quinn <eskimo1(a)apple.com> wrote: > At 18:44 +0530 8/6/09, Arjun SM wrote: > >> I am getting this message when i am trying to launch my application which >> is a LaunchAgent. This message comes only when multiple Users are logged in. >> > > Yeah, your script, as its written, is never going to work. The problem is > that, except when run as root, launchctl finds the right launchd to talk to > by looking it up in the Mach bootstrap namespace. Your script is always > running in one user's namespace, so will always end up talk to that user's > launchd (which ends badly because of permissions). > > The launchd story is a quite weak when it comes to installing and upgrading > agents where there are users logged in. This is a well known problem > <rdar://problem/5476420> but, alas, I don't have any good news to share with > you on this front. > > This problem breaks down as follows: > > o upgrade -- There /is/ a reasonable way to handle the upgrade scenario. > Most agents are running on behalf of a daemon. In that case you can > overwrite the agent on disk and then tell the daemon to signal all of the > agents to quit. They will be relaunched by their respective launchd's, this > time running the new code. You have to be a little careful, but this > approach works reasonably well. > > o first install -- There isn't a good way to handle the first install case. > You can start the agent in the GUI context that you're running in, but > starting agents in other existing GUI contexts is tricky. In general I'd > recommend you /not/ try to solve this. Non-foreground GUI users will have > to log out and log back in to pick up your agent. If that's a problem (for > example, you're installing security software), force a restart in that case. > > Keep in mind that the multiple GUI logins via fast user switching is an > edge case that 90% of your users will never encounter. > > S+E > -- > Quinn "The Eskimo!" <http://www.apple.com/developer/> > Apple Developer Relations, Developer Technical Support, Core OS/Hardware > > _______________________________________________ > launchd-dev mailing list > launchd-dev(a)lists.macosforge.org > http://lists.macosforge.org/mailman/listinfo.cgi/launchd-dev >

2 2

launch_msg(): Socket is not connected
by Arjun SM 18 Jun '09

18 Jun '09

Hi all, I am getting this message when i am trying to launch my application which is a LaunchAgent. This message comes only when multiple Users are logged in. This is a small script that i use in my postflight to launch for all the logged in Users for userName in `/usr/bin/users` do echo "launch My_app" su -l $userName -c "launchctl load -S Aqua /Library/LaunchAgents/com.myname.myappname.plist done should i check for anything more than the Users who have logged in so that i can avoid "launch_msg(): Socket is not connected" message.??? thanks, Arjun

3 2

Re: [launchd-dev] launchd-dev Digest, Vol 24, Issue 5
by Nehemiah Dacres 07 Jun '09

07 Jun '09

Not to dissagree with the engeneer, but this falls out of LaunchD domain and in the domain of IOKit. Whats to stop the driver programmer from implementing this level of granularity? What does the origenal poster mean by "Interaction"? On Fri, Jun 5, 2009 at 9:00 AM, <launchd-dev-request(a)lists.macosforge.org>wrote: > Send launchd-dev mailing list submissions to > launchd-dev(a)lists.macosforge.org > > To subscribe or unsubscribe via the World Wide Web, visit > http://lists.macosforge.org/mailman/listinfo.cgi/launchd-dev > or, via email, send a message with subject or body 'help' to > launchd-dev-request(a)lists.macosforge.org > > You can reach the person managing the list at > launchd-dev-owner(a)lists.macosforge.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of launchd-dev digest..." > > > Today's Topics: > > 1. Re: Launch daemon 0n-Demand on USB Connection/Write > (Damien Sorresso) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Thu, 04 Jun 2009 17:00:11 -0700 > From: Damien Sorresso <dsorresso(a)apple.com> > To: Kiran B <Kiran.b(a)robosoftin.com> > Cc: launchd-dev(a)lists.macosforge.org > Subject: Re: [launchd-dev] Launch daemon 0n-Demand on USB > Connection/Write > Message-ID: <BF826895-D2AB-4AFD-A763-898CDB189165(a)apple.com> > Content-Type: text/plain; charset="us-ascii"; Format="flowed"; > DelSp="yes" > > On Jun 3, 2009, at 12:45 AM, Kiran B wrote: > > Hello, > > > > From the recent posts in "launchd-dev" lists, it is evident that > > launchd does not yet support launching of processes based on USB > > device getting connected. > > > > I have a requirement to launch a daemon whenever a device with > > specific VendorID is connected. > > > > I am OK even if the daemon gets launched for any device getting > > connected via USB. In that case, the daemon shall ensure that it is > > the right device on which it is interested in. In other case it may > > terminate. > > For now, you should have a monitor daemon that keeps track of what's > happening with IOKit. When the device you want appears, it asserts > demand for your other daemon by sending a message to a Mach port or > connecting to a socket. Then your other daemon gets launched on demand. > > > I am also curious to know whether a daemon can be launched whenever > > the connected device writes any data on the USB port. (Eg: User > > interaction with the device would make the device write on USB). > > No. If we do implement hardware event-based notifications, it wouldn't > be quite this granular. > -- > Damien Sorresso > BSD Engineering > Apple Inc. > >

2 1

Launch daemon 0n-Demand on USB Connection/Write
by Kiran B 05 Jun '09

05 Jun '09

Hello, From the recent posts in "launchd-dev" lists, it is evident that launchd does not yet support launching of processes based on USB device getting connected. I have a requirement to launch a daemon whenever a device with specific VendorID is connected. I am OK even if the daemon gets launched for any device getting connected via USB. In that case, the daemon shall ensure that it is the right device on which it is interested in. In other case it may terminate. I am also curious to know whether a daemon can be launched whenever the connected device writes any data on the USB port. (Eg: User interaction with the device would make the device write on USB). However I am satisfied that there is no direct means to achieve any of these. Correct me if I am wrong. In that case, I would like to know what would be the best approach to achieve this. I would like to know any work around if it exists. The requirement is for 10.4 and above. However, I may be able to persuade the concerned to make that 10.5 in case this can only be achieved in 10.5. I did look for the same in Net. I could find that it was mentioned that it is not possible, I could not get any suggestion on the best approach to be followed with the available infrastructure. Thanks in advance. Regards, Kiran ----------------------------------------------- Robosoft Technologies - Come home to Technology Disclaimer: This email may contain confidential material. If you were not an intended recipient, please notify the sender and delete all copies. Emails to and from our network may be logged and monitored. This email and its attachments are scanned for virus by our scanners and are believed to be safe. However, no warranty is given that this email is free of malicious content or virus.

2 1

Is there anything that can prevent launchd from working
by Scott Haneda 01 Jun '09

01 Jun '09

In a case where launchd works just fine all the time, for example: <dict> <key>Label</key> <string>com.host.ntpdate</string> <key>ProgramArguments</key> <array> <string>/bin/bash</string> <string>/Users/me/bin/set_date</string> </array> <key>RunAtLoad</key> <true/> <key>StartInterval</key> <integer>3600</integer> </dict> script... (slightly shortened) #!/bin/bash # Create a place to store this log file log_path=/Users/me/Library/Logs/ntpdate.log echo `date` >> $log_path # Sleep for a duration so the system can get the interfaces up # Running in chunks just so the log and get updated # and I can see that it is working echo "Sleeping...90 Seconds `date`" >> $log_path /bin/sleep 90; # Do the work echo "About to run ntpdate -u" >> $log_path /usr/sbin/ntpdate -u echo "Finished running ntpdate -u" >> $log_path log... (Full output, non shortened) ----- BEGIN ----- Started to set time and date Mon Jun 1 11:18:35 PDT 2009 Sleeping...30 Seconds Mon Jun 1 11:18:36 PDT 2009 Sleeping...30 Seconds Mon Jun 1 11:19:06 PDT 2009 Sleeping...15 Seconds Mon Jun 1 11:19:36 PDT 2009 About to run ntpdate -u Finished running ntpdate -u Finished setting time and date ----- END ----- Syslog tells me it works: Jun 1 10:57:41 host com.host.ntpdate[19696]: 1 Jun 10:57:41 ntpdate[19776]: adjust time server 17.151.16.21 offset -0.070842 sec It works all the time, sans one case, which is when I need it most. I hope there are people here who know a but about the kernel of OS X. When the machine panics, which I am still working on determining why, I suspect ram, though on the 3rd batch. I have 100% ruled out a bad pram battery. After a panic, the "automatically restart after power failure" feature does not work. It does work any other time, only after panics does not it. I have to physically reboot. This also adds more credence to the battery being fine. When the machine comes up, there is a message on screen about the date and time being at 1969. Doing a google search for "panic 1969" and you see there are a good deal of others with panic logs at 1969. I conclude, panics mess with the date and time for some reason, or nuke nvram, not sure. No big deal really, though this is an email server, and if I forget, or do not have access, the date and time will continue to be wrong, and all emails will be wrongly dated. My plist is set to run at load, and also to run every 60 minutes, which I see it does do. I can find a syslog line that shows it ran at load, and ones that show it runs every hour as well. launchd totally fails to run this item after a kernel panic and the date is set into the 1969 range. I have the sys log for where the dates are all Dec 31, which is just after the panic, and after I called in a reboot, which then foiled my script from setting the date correctly. I see named started, which uses launchd, so I know it is working. I see a lot of other things start. But then I see this: Dec 31 16:00:54 host-domain-com com.apple.launchd[1] (com.domain.ntpdate[50]): Exited with exit code: 1 Right after that, I have a launchd item that checks S.M.A.R.T. which runs. Since this is after bind came up, I am pretty sure the interfaces are up enough that I could talk to a time server. Is there something about ntpdate that will error out on insanely large time offsets? All I end up doing to solve it is `sudo /usr/sbin/ ntpdate -u`, the same thing launchd should do. I will solve the panics, in the meantime, I would love to know how to protect myself from living in the past for too long. Thank you all. -- Scott * If you contact me off list replace talklists@ with scott@ *

4 7

launchd and system sleep
by Thomas Clement 01 Jun '09

01 Jun '09

Hello, The manual page for the "StartInterval" key mentions that "if the system is asleep, the job will be started the next time the computer wakes up". However I'm receiving bug reports from users which seem to indicate this is not always the case (the job is not launched for a long period of time). The problem does not seem to occur with the "StartCalendarInterval" key. I've also noticed the "com.apple.backupd-wake.plist" Time Machine job which is apparently triggered at wake time although it advertises no way to be launched. Why the need for this additional job since "com.apple.backupd-auto.plist" uses the "StartInterval" key? Is it a workaround for a launchd issue? Is this a known issue or am I missing something? Can something be done about this? Regards, Thomas

2 2