On Nov 6, 2009, at 2:50 AM, Anil Kumar K wrote: > quoted text
Why do you want it always alive? > quoted text
OnDemand services are triggered when their clients attempt to connect. launchd proxies your listening activities (socket listens, mach ports, etc. [...]]]>Anil, Your requirement depends on other keys which are used along with OnDemand/KeepAlive keys set to false. There are some keys which support on demand launching of job like 'watchpaths', 'startonmount', 'startinterval', etc. There is no on-demand trigger that can start a job based on CPU load. [...]]]>
Hi Ethan Thanks for the response. My queries on your suggestion of launch-on-demand launch daemon are below: 1. The daemon I was referring to is an always-load launch daemon (i.e., OnDemand is set to false). We want the daemon to be always alive. So, is there a way that I make OnDemand setting to true but want the daemon to be always alive? 2. My understanding of OnDemand setting is that, the daemon will be triggered on demand if OnDemand is set to true. In such case, I guess I need to write the logic to trigger the daemon. Is there any simple logic that I can add so that daemon will be triggered without impacting boot times? Thanks Anil Kumar [...] Yahoo! India has a new look. Take a sneak peek http://in.yahoo.com/trynew]]>
Hi Anil, Shantonu- The SystemLoadAdvisory mechanism is an inappropriate choice here. It will report when it's appropriate to run speculative work (like re-building an index, or refreshing a web page, or XCode doing a speculative build...). The goal is for processes to take the "Good", "OK", "Bad" responses as guidelines, and change the frequency of their speculative actions. A lot of these things still need to happen when the system is "Bad", but should be done less frequently. Given that, and given that the system might never reach an "OK" level (it could always be Great, or Bad), you should find another solution to this problem. SystemLoadAdvisory has nothing to do with when the system is initially available. So - you want to delay the CPU-intensive initialization time of your service until after boot, so as not to affect boot times. Have you considered becoming a launch-on-demand launch daemon? That way you could defer your initialization costs until you're first accessed. This is what I'd recommend you look into, without knowing more about your problem. -Ethan On Nov 5, 2009, at 3:35 AM, Anil Kumar K wrote: [...]]]>
another workaround would be to simply start immediately but give your backgroundtask a very low priority (nice) so it never gets to run until system load drops to normal. On 05.11.2009, at 12:35, Anil Kumar K wrote: [...]]]>
Hi Shantonu Sorry for the late response. Can you please provide sample code on this. I had seen the header file you suggested but I am confused as to what APIs to call. Thanks Anil Kumar [...] Connect more, do more and share more with Yahoo! India Mail. Learn more. http://in.overview.mail. [...]]]>
I am porting a file system driver to 64-bit 10.6. Â When I try to load the kernel extension, it fails with the following error:
#sudo kextutil -s fsd/symdir fsd/fsd.kext Notice: Using running kernel architecture x86_64 to generate symbols. (kernel) Kext com.company.kext. [...]]]>I have boot-args debugging turned on but I haven't noticed any additional messages. The system does not panic. I am catch the situation and return an error. But this is still a situation we would like to avoid. I think I understand what you're saying about the physical addresses. [...]]]>
Godfrey, Please note the SIOCPROTODETACH ioctl mentioned by Josh is marked PRIVATE which means that it behavior may change with any release of OS X... Vincent On Nov 3, 2009, at 9:37 PM, Josh Graessley wrote: [...]]]>
Vinay, As already discussed at email@hidden ever, receiving router adv behavior is little bit different from other BSDs. Even if accept_rtadv=0, the darwin could receive RA because it is able to be configured whether each interface receives or not. [...]]]>
Hi Godfrey,
The following is based on code from the bootp project:
static int
siocprotodetach(const char * name)
{
int sockfd;
struct ifreq ifr;
sockfd = socket(PF_INET, SOCK_DGRAM, IPPROTO_UDP); // check that
this didn't return -1
bzero(&ifr, sizeof(ifr)); [...]]]>G'day, Josh. Thanks for the clues. I'd really, really like to be able to develop from a ssh into the host machine; this allows me to do everything from the primary development system. I have found the ioctl you mentioned SIOCPROTODETACH but I'm frankly lost as to how to use it? [...]]]>
Hi All,          Please let me know on How to set/change the Router solicitation interval on MAC OS darwin kernel ver 9.6.1? . My intention is to increase the router solicitaion interval .  Regards, Vinay    On 10/30/09, Vincent Lubet <email@hidden> wrote: [. [...]]]>
This may be caused by the IP protocol still being attached to the ethernet interface. When you run ifconfig en2 inet... and an IP address is attached to the interface, that implicitly attaches IP to the ethernet interface. If you instead set up a network service for [...]]]>
While trying to debug a USB network kext I find that I can't unload as the dlil isn't being destroyed. Following the steps of the documentation. Steps to reproduce (using a pegasus USB enet, but the same steps are reproducable with the Air's ENET dongle) [...] [...]]]>