September 2002 - Darwin-kernel

Re: Can Darwin Live Without Disks?
by Bernie Zenis 08 Sep '02

08 Sep '02

On Saturday, September 7, 2002, at 07:34 PM, Shawn Erickson wrote: On Saturday, September 7, 2002, at 01:22 PM, Bernie Zenis wrote: Operating systems seem to have a strong reliance on having dIsk(s) available. Can Darwin boot on a diskless system? What do you mean by disk? The OS has to be read from something and it doesn't have to be a "disk" in the traditional sense of the word. Hard disk, CD, DVD, ZIP, etc. Networking booting of Mac OS X is supported see... http://a160.g.akamai.net/7/160/51/48f6318de5a381/www.apple.com/server/ pdfs/L21951A_NetBoot_TB.pdf I've heard of netbooting but I don't know any details. Is nothing loaded from a local disk? That is, does firmware make a network connection and download the needed files? Does a netbooted machine still need a local writable disk? Can VM work over a network? Does Darwin need to have swap file(s)? If RAM is large enough, can everything happen over the network? Can Darwin continue to operate on a system where the disk(s) have become unavailable (gone offline, crashed, etc.)? If the "disk" that contains the OS image and more importantly the VM swap space goes away the system will most likely fail soon afterwards. That is the answer I was expecting. I just want to know how robust or fail-safe Darwin is. What do you want to do and how does it not require some type of "disk" to function. I'm mostly seeking knowledge. If Darwin can startup and run without a disk, then the whole Mac OS is closer to being able to do the same, then it might be useful to write Mac software that is "diskless friendly". I guess a better, more general question might be, "Can Darwin startup and run without a writable disk, local or remote?" Does any of this make sense to you or just to me? Thanks, Bernie _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Can Darwin Live Without Disks?
by Bernie Zenis 07 Sep '02

07 Sep '02

Hi, Operating systems seem to have a strong reliance on having dIsk(s) available. Can Darwin boot on a diskless system? Can Darwin continue to operate on a system where the disk(s) have become unavailable (gone offline, crashed, etc.)? I doubt the whole Mac OS can live without disk(s); but, I was hoping that atleast the kernel could. Thanks, Bernie _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Figuring out unresolved kernel trap
by Chris Edgington 06 Sep '02

06 Sep '02

I've got a USB network driver that is working well, for the most part. Running OS X 10.1.5. When resuming from a system sleep, if I generate a lot of disk activity immediately after wakeup (by starting IE for example), our driver crashes. So far, we've been able to debug everything with IOLog, not having to figure out kernel debugging and symbols. However, this one is elusive, so its time to do kernel debugging. I've got the source for my kernel and I'm successfully building and running a version of the kernel that I've built. The bug exhibits itself exactly the same with my kernel as with the stock kernel from my machine. I also can successfully connect gdb remotely once the trap has occurred. I've got the symbols / source for my driver loaded into gdb. I can see exactly what line in the driver is crashing - I just don't know why. I don't know how to understand what the kernel is telling me - ie: what the real problem is. Here's the trap info - can someone enlighten me on the 0x300 Data access trap? Thanks, -Chris --------------------- Unresolved kernel trap(cpu 0): 0x300 - Data access DAR=0x00000108 PC=0x21a70584 Latest crash info for cpu 0: Exception state (sv=0x1d146800) PC=0x21A70584; MSR=0x00009030; DAR=0x00000108; DSISR=0x40000000; LR=0x21a704b4; R1=0x0d8db5b0; XCP=0x0000000c (0x300 - Data access) _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: socketcall()
by Vincent Lubet 05 Sep '02

05 Sep '02

On Wednesday, September 4, 2002, at 08:47 PM, Brian Wotring wrote: I need to send a small UDP message to a user level daemon from an NKE. Is a dlil_inject function the easiest way to do this, or is there an easier way? Does Darwin have a socketcall(), or equivalent? Does this really need to be UDP? It would be easier to use a local socket (AF_UNIX) or an IP divert socket (man 4 divert). Vincent _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: socketcall()
by Dave Zarzycki 05 Sep '02

05 Sep '02

On Wed, 4 Sep 2002, Justin C. Walker wrote: > > Does Darwin have a socketcall(), or equivalent? > > I don't recognize the call. What is it (and which system has it)? socketcall is a common kernel entry point for the socket system calls on Linux. davez -- Dave Zarzycki Darwin & Mac OS X Apple Computer, Inc. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: socketcall()
by Justin C. Walker 05 Sep '02

05 Sep '02

On Wednesday, September 4, 2002, at 08:47 PM, Brian Wotring wrote: I need to send a small UDP message to a user level daemon from an NKE. Is a dlil_inject function the easiest way to do this, or is there an easier way? Ew...that's gross. Is the daemon listening for UDP datagrams, or is this a non-sequitur? Do you know which process holds the daemon? All in all, it sounds powerfully kinky. Does Darwin have a socketcall(), or equivalent? I don't recognize the call. What is it (and which system has it)? regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | It's not whether you win or lose... | It's whether *I* win or lose. *--------------------------------------*-------------------------------* _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

socketcall()
by Brian Wotring 05 Sep '02

05 Sep '02

I need to send a small UDP message to a user level daemon from an NKE. Is a dlil_inject function the easiest way to do this, or is there an easier way? Does Darwin have a socketcall(), or equivalent? -- Brian Wotring ( brian(a)shmoo.com ) PGP KeyID: 0x9674763D _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

addNotification in 10.2
by Daniel-Olivier Simard 03 Sep '02

03 Sep '02

I have written code that was working on Mac OS 10.1 that was using the addNotification function. When I tried it on 10.2 the same code produced some weird results. My code is an IOService (classA) that is waiting on another (classB) by defining a property into the IORegistry. When classB starts, it writes the regkey and register him-self. The classA addNotification on gIOFirstMatchNotification with a dictionary that contains the regkey. On 10.1 my classA was deblocking only when the classB was registering, but on 10.2 the classA deblock several times event if the classB is not already registered. Here is my code: classA: OSDictionary* pDict = OSDictionary::withCapacity(1); pDict->setObject("regkey", "string"); IONotifier* pNotifier = addNotification(gIOFirstMatchNotification, pDict, classA::NotificationHandler, this); classB: setProperty("regkey", "string"); registerService(); I have also tried with resourceMatching("regkey") instead of using a custom dictionary, and my classA::Notificationhandler is not event deblocking when I start the classB. Can you tell me why this code is not working on 10.2 like on 10.1 ? Is there another way to wait on an IOService based on a custom dictionary or a resource ? Daniel-Olivier Simard, ing. Kernel developer Matrox Electronic Systems Ltd Tel: (514) 822-6000 ext. 7899 Fax: (514) 822-6024 Email: Daniel-Olivier.Simard(a)matrox.com _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: darwin-kernel digest, Vol 2 #116 - 6 msgs
by Christopher Taylor 03 Sep '02

03 Sep '02

>Cool!!! >I'd like to try kdb, but I can't seem to make it work. If its as good >as the FreeBSD/NetBSD ddb, I'd rather just use it than gdb. However, >I can't seem to make it work. >% nvram boot-args >boot-args -v debug=0x11e >However, when I send a break, it still ends up waiting for kdp. AFAIK from playing with 10.1, you need to have a kernel that has been completely in DEBUG. If you are debugging Apple's shipping kernel, some debug stuff is on (like IOKit I believe), but ddb support is not enabled. -Chris _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0