May 2006 - Darwin-dev - lists.apple.com

Re: launchd and fork/exec?
by Brian Bergstrand 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On May 24, 2006, at 6:04 PM, Dave Zarzycki wrote: On May 24, 2006, at 12:24 PM, Brian Bergstrand wrote: Reading the launchd Getting Started doc I see: Thanks. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFEdO5EedHYW7bHtqIRAkDUAJ9tbPlQ/q/sXgXel50PeWY2swoHdQCguxiR eXwuTLOdYqHs6LJt3afFysw= =TjgB -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I have a tool that loads a non-IOKit kext via executing kextload. For various reasons, using kextload directly from the launchd plist won't work (the tool does some pre and post processing). "The only things that are strictly prohibited, however, are fork()/ exit() combinations (including indirect methods, such as the daemon (3) library call)." I'm not forking a child of the tool, but of course my tool fork/ exec's kextload (and kextunload and possibly kextstat). When this happens, launchd starts spewing errors in system.log and re- spawning my tool. Is there anything I can do to work around this? Or do I just have to live with the spurious errors/respawns? That's not really what launchd was meant for, Really? The mantra coming from Apple is use launchd for all startup needs. As StartupItems are going away and I have a need to load a non- IOKit kext at boot, launchd seems to be the only game in town. but if you must, RunAtLoad is the flag you want to set to true and you want OnDemand also set to true (or not set at all). OK, I'll try that. I have RunAtLoad set to true now, but OnDemand is set (explicitly) to false. davez This email sent to site_archiver(a)lists.apple.com

1 0

Re: launchd and fork/exec?
by Dave Zarzycki 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 24, 2006, at 12:24 PM, Brian Bergstrand wrote: Reading the launchd Getting Started doc I see: davez _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I have a tool that loads a non-IOKit kext via executing kextload. For various reasons, using kextload directly from the launchd plist won't work (the tool does some pre and post processing). "The only things that are strictly prohibited, however, are fork()/ exit() combinations (including indirect methods, such as the daemon (3) library call)." I'm not forking a child of the tool, but of course my tool fork/ exec's kextload (and kextunload and possibly kextstat). When this happens, launchd starts spewing errors in system.log and re-spawning my tool. Is there anything I can do to work around this? Or do I just have to live with the spurious errors/respawns? That's not really what launchd was meant for, but if you must, RunAtLoad is the flag you want to set to true and you want OnDemand also set to true (or not set at all). This email sent to site_archiver(a)lists.apple.com

1 0

launchd and fork/exec?
by Brian Bergstrand 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Reading the launchd Getting Started doc I see: Thanks. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFEdLMGedHYW7bHtqIRAsa+AJ9oqqUvyH65NNpWezZTxwylvaADlwCgm4oy z6iKf4FxCjuRvraH2cXH91E= =02b9 -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I have a tool that loads a non-IOKit kext via executing kextload. For various reasons, using kextload directly from the launchd plist won't work (the tool does some pre and post processing). "The only things that are strictly prohibited, however, are fork()/ exit() combinations (including indirect methods, such as the daemon (3) library call)." I'm not forking a child of the tool, but of course my tool fork/ exec's kextload (and kextunload and possibly kextstat). When this happens, launchd starts spewing errors in system.log and re-spawning my tool. Is there anything I can do to work around this? Or do I just have to live with the spurious errors/respawns? This email sent to site_archiver(a)lists.apple.com

1 0

Re: StartupItems
by Chris Ridd 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: AcZ/SPwROqayLOs8EdqK7gARJHlZbA== Thread-topic: StartupItems User-agent: Microsoft-Entourage/11.2.3.060209 On 24/5/06 12:42, Andrew Gallatin <gallatin(a)cs.duke.edu> wrote: > > Garance A Drosihn writes: >> machines this summer. What does SMF do? How does that work? Are > > It replaces (most) startup scripts.. > > There is a good presentation at > http://www.filibeto.org/~aduritz/truetrue/solaris10/sol-smf.pdf Great link, thanks. I've found this next one of some use: <http://home.arcor.de/bnsmb/public/htdocs/My_Little_SMF_FAQ.html> >> all the /etc/rc scripts gone? > > No, /etc/rc is preserved so that 3rd parties can use it. Strictly for "legacy" scripts though, as I understand it. Cheers, Chris _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: StartupItems
by Daniel J. Luke 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com On May 23, 2006, at 9:42 PM, Peter Seebach wrote: In message <4473B7CE.6000204(a)cse.unsw.edu.au>, Charles Gray writes: The over-engineering possibilities are endless, but the base app seems trivial. I'm sure there are a few sysadmins and other people with similar problems who's like to see such an app available, too. Yes. And if it were done well it could go into the system, so everyone could use it, and it could be available for startup tasks. The darwinports project has daemondo (by James Berry) which does some of this sort of thing already: "daemondo is a wrapper program that runs daemons. It starts the specified daemon on launch, stops it when given SIGTERM, and restarts it on SIGHUP. It can also watch for transitions in system state, such as a change in network availability or system power state, and restart the daemon on such an event. daemondo works well as an adapter between darwin 8's launchd, and daemons that are normally started via traditional rc.d style scripts or parameters." http://cvs.opendarwin.org//cgi-bin/cvsweb.cgi/proj/darwinports/base/ src/programs/daemondo/ -- Daniel J. Luke +========================================================+ | *---------------- dluke(a)geeklair.net ----------------* | | *-------------- http://www.geeklair.net -------------* | +========================================================+ | Opinions expressed are mine and do not necessarily | | reflect the opinions of my employer. | +========================================================+ PGP.sig

1 0

Re: StartupItems
by Andrew Gallatin 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Garance A Drosihn writes: > machines this summer. What does SMF do? How does that work? Are It replaces (most) startup scripts.. There is a good presentation at http://www.filibeto.org/~aduritz/truetrue/solaris10/sol-smf.pdf > all the /etc/rc scripts gone? No, /etc/rc is preserved so that 3rd parties can use it. Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: StartupItems
by Lawrence Sica 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com That is the hardest part. --Larry _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… On May 23, 2006, at 2:54 PM, Kevin Van Vechten wrote: Sure, but the issue you're overlooking is that there's no universal definition for when startup is done. Disk and network availability are events, like keystrokes. When do you know that keyboard input is done? You could possibly wait for the keyboard to be quiescent. You could alternatively look for a marker like newline. But you'll never get an EOF. Not OS X, but with Solaris here. But we have issues with when a VCS (Veritas) Cluster FS is fully mounted on some systems. It got to the point we were doingall sorts of rc gymnastics to make sure it was mounted before certain other things started. I understand fully what you mean here. This email sent to site_archiver(a)lists.apple.com

1 0

Re: StartupItems
by Lawrence Sica 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 23, 2006, at 2:12 PM, Chris Ridd wrote: On 23/5/06 7:03, Dan Shoop <shoop(a)iwiring.net> wrote: And if you think that the rc.d mechanism used so commonly today is an improvement or so great try answering the following common question: "what letter/number/name should I make my rc file to start up X? Which rc directory should it go under?" Such mechanisms are a huge step backwards and unbelievably fragile. It's light years better than the old way. --Larry _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… It is interesting that Sun also came to the same conclusion, and replaced it with their Service Management Framework (SMF) in Solaris 10. Has anyone compared launchd with SMF? Is there anything SMF does better than launchd? I've not gotten too deeply into SMF yet. We are starting to roll out Solaris 10 at work though. It strikes me as a very similar idea though, and I do prefer it to rcX.d. It was always frustrating how you cannot truly guarantee startup order of items starting at the same SX level. Bascially you register with SMF what you want started and when and it goes. There is an admin interface for most things though you can also edit some files if you wish. This email sent to site_archiver(a)lists.apple.com

1 0

Re: gdb and mach exceptions
by Jonas Maebe 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 23 mei 2006, at 22:33, Brendan Creane wrote: Hello All, I have a simple but still unsolved problem involving the interaction of the EXC_BAD_ACCESS exception and gdb: I can't figure out how to get gdb to pass the exception to my signal handler. The way I debugged my signal handling code was by Not ideal, but it works. Jonas _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… In theory, I think it should work by playing with the "inferior- handle-exceptions" setting (see "help obscure" and "show inferior- handle-exceptions" for slightly more info). It's not clear to me which kind of exceptions are deemed inferior though (mach or unix- style), since "show inferior-handle-exceptions" prints: "If GDB should handle exceptions or pass them to the UNIX handler is on." That's just as informative as "Are you running Mac OS X, Linux or Windows" - "Yes" In practice, all I ever managed to do by changing that setting was to get gdb in an internal error state when a SIGSEGV occurs. Never got around to file a bug against it though. a) adding a print to my signal handlers (so I could see when a signal handler was entered) followed by a sleep(5) b) keep a separate terminal window ready with "gdp -p <process id>" typed on the command line (but not yet executed) c) when the signal handler was entered, press enter so gdb attaches This email sent to site_archiver(a)lists.apple.com

1 0

Re: StartupItems
by Garance A Drosihn 24 May '06

24 May '06

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 23/5/06 7:03, Dan Shoop <shoop(a)iwiring.net> wrote: And if you think that the rc.d mechanism used so commonly today is an improvement or so great try answering the following common question: "what letter/number/name should I make my rc file to start up X? Which rc directory should it go under?" Such mechanisms are a huge step backwards and unbelievably fragile. It is interesting that Sun also came to the same conclusion, and replaced it with their Service Management Framework (SMF) in Solaris 10. Has anyone compared launchd with SMF? Is there anything SMF does better than launchd? -- Garance Alistair Drosehn = gad(a)gilead.netel.rpi.edu Senior Systems Programmer or gad(a)freebsd.org Rensselaer Polytechnic Institute or drosih(a)rpi.edu _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… At 7:12 PM +0100 5/23/06, Chris Ridd wrote: Having watched a few hundred messages go by on this topic, and having read maybe 1/3 of them, I was surprised to find an interesting question in the babbling stream. I have no idea of SMF, but we (at RPI, where I work) do intend to be moving to solaris 10 on some machines this summer. What does SMF do? How does that work? Are all the /etc/rc scripts gone? This email sent to site_archiver(a)lists.apple.com

1 0