February 2003 - Darwin-kernel

Re: Finding one parent for a vnode born under X
by Stéphane Sudre 17 Feb '03

17 Feb '03

On Friday, February 14, 2003, at 08:35 PM, Umesh Vaishampayan wrote: On Friday, February 14, 2003, at 10:41 AM, Quinn wrote: At 12:15 +0100 14/2/03, Stiphane Sudre wrote: Would there be a way to find 1 vnode parent of a vnode (on any FS)? Not reliably. [In presence of hard links on files, *which* parent would you like to know about? One. I didn't really care about which one. All of the following is based on my somewhat limited understanding of VFS. Caveat emptor. If the vnode references a directory, you can just VOP_LOOKUP "..". This is, however, subject to vnode locking constraints. Conceptually, yes. In practice, it's not that simple. Just doing a VOP_LOOKUP() in a KEXT will give you a directory vnode, however you will fail to detect escaping out of a chroot(2) [*security hole*]. And you may fail to cross the mount points [which may or may not be what you want]. For files, I don't think there's a file system independent way. For Macintosh file systems (HFS, HFS Plus, AFP), you can probably doing this by: 1. using VOP_GETATTRLIST with ATTR_CMN_PAROBJID to get the parent's CNID and then It has the same problem described above... Well, I think I will give up the vnode parent solution and work directly with the vnode and the VOP_GETATTRLIST command to get the info I need then. _______________________________________________ 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: Pointer Testing
by Quinn 17 Feb '03

17 Feb '03

At 14:11 -0800 14/2/03, Brent Marykuca wrote: Given a pointer value, is there some relatively trivial way to determine whether or not it points to a valid readable or writeable address? It'd be nice to have idea of the big picture of what you're trying to do. S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Technical Support * Networking, Communications, Hardware _______________________________________________ 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: SEGV handler
by Paul Ripke 16 Feb '03

16 Feb '03

On Sunday, Feb 16, 2003, at 07:22 Australia/Sydney, Brian Tabone wrote: Brent, One trick you could try is to install a SEGV signal handler. If I remember correctly, you can actually trap this signal in your application and handle it without core dumping. I'm not sure how risky it would be, nor how efficient, but you may be able to use this method to check for validity of pointers. Now, the question here is would this work in kernel space, and that I am doubtful, but it should work for you in user space. I've never tried this myself, so your milage may vary. I remember doing this an age back under Linux, NetBSD 0.9 and Solaris, while developing a reasonably sized application. Our handler merely attempted a graceful shutdown of active network connections and closed our listening socket. After all that, it dumped core. The main reason for all this? There was an annoying delay before Solaris would clean up our listening socket, preventing the app from restarting. Cheers, -- Paul Ripke Unix/OpenVMS/TSM/DBA 101 reasons why you can't find your Sysadmin: 68: It's 9AM. He/She is not working that late. -- Koos van den Hout _______________________________________________ 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: Pointer Testing
by Jeff Jenkins 15 Feb '03

15 Feb '03

On Saturday, February 15, 2003, at 09:07 AM, Brent Marykuca wrote: On Friday, February 14, 2003, at 10:00 PM, Jeff Jenkins wrote: I think you are asking a question that everyone who has ever used the C language would like to have a simple solution to. The answer is no, there is nothing inherent in the language that does this. Yes, I know that C itself has nothing to help, but surely the information is available to the kernel. The virtual memory system must maintain tables of valid/invalid address ranges and read/write permissions for every task, so it's not inconceivable that such an API could be made available. Win32, for example, has routines called IsBadReadPtr(), IsBadWritePtr(), IsBadCodePtr() and IsBadStringPtr() that do this. Install a SIGSEGV handler if you don't want your process to quit. Also, if these calls result in a dive into the kernel from user space, you would have some serious performance issues to deal with. This might be OK for an app, but doesn't sound very desirable in general. -- jrj _______________________________________________ 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

SEGV handler
by Brian Tabone 15 Feb '03

15 Feb '03

Brent, One trick you could try is to install a SEGV signal handler. If I remember correctly, you can actually trap this signal in your application and handle it without core dumping. I'm not sure how risky it would be, nor how efficient, but you may be able to use this method to check for validity of pointers. Now, the question here is would this work in kernel space, and that I am doubtful, but it should work for you in user space. I've never tried this myself, so your milage may vary. -Brian Tabone On Friday, February 14, 2003, at 04:11 PM, Brent Marykuca wrote: Hello, Given a pointer value, is there some relatively trivial way to determine whether or not it points to a valid readable or writeable address? i.e. is there a way to find out whether accessing the address will cause a panic? I'm interested in how to do this from a KEXT, but it also seems like it would be a useful ability to have in user-space too. Thanks, Brent Marykuca _______________________________________________ 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. _______________________________________________ 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: Pointer Testing
by Brian Bergstrand 15 Feb '03

15 Feb '03

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Saturday, February 15, 2003, at 11:07 AM, Brent Marykuca wrote: > On Friday, February 14, 2003, at 10:00 PM, Jeff Jenkins wrote: > >> I think you are asking a question that everyone who has ever used the >> C >> language would like to have a simple solution to. The answer is no, >> there is nothing inherent in the language that does this. > > Yes, I know that C itself has nothing to help, but surely the > information is available to the kernel. The virtual memory system must > maintain tables of valid/invalid address ranges and read/write > permissions for every task, so it's not inconceivable that such an API > could be made available. Win32, for example, has routines called > IsBadReadPtr(), IsBadWritePtr(), IsBadCodePtr() and IsBadStringPtr() > that do this. There are some functions in the Carbon framework on X that are similar to these W32 functions. I think they are declared in Debugging.h. But, I also think they only work with Carbon Ptr and Handle types. IOW, they won't work with malloc'd memory. AFAIK, there is nothing in kernel that you give you this info.. HTH. Brian Bergstrand <http://www.classicalguitar.net/brian/> PGP Key ID: 0xB6C7B6A2 For every human problem, there is a neat, simple solution; and it is always wrong - H. L. Mencken -----BEGIN PGP SIGNATURE----- Version: PGP 8.0 iQA/AwUBPk5yknnR2Fu2x7aiEQKbvQCgrQiS+ZIIJZjknAPxHRjMW0U9co4AmgND qGSqi7jLYMCwOmBJEggOvzTC =gn0k -----END PGP SIGNATURE----- _______________________________________________ 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: Pointer Testing
by Jonas Maebe 15 Feb '03

15 Feb '03

On Sat, 15 Feb 2003, Brent Marykuca wrote: > Yes, I know that C itself has nothing to help, but surely the > information is available to the kernel. The virtual memory system must > maintain tables of valid/invalid address ranges and read/write > permissions for every task, so it's not inconceivable that such an API > could be made available. I don't think that the information that the kernel has is very useful and would only give you a false sense of security. The reason is that most memory allocators request big blocks from the kernel and then give pointers inside these blocks to the application when it requests memory. So according to the kernel, a pointer may be perfectly OK, while it could point to a non-allocated or freed block. Jonas _______________________________________________ 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: Pointer Testing
by Brent Marykuca 15 Feb '03

15 Feb '03

On Friday, February 14, 2003, at 10:00 PM, Jeff Jenkins wrote: I think you are asking a question that everyone who has ever used the C language would like to have a simple solution to. The answer is no, there is nothing inherent in the language that does this. Yes, I know that C itself has nothing to help, but surely the information is available to the kernel. The virtual memory system must maintain tables of valid/invalid address ranges and read/write permissions for every task, so it's not inconceivable that such an API could be made available. Win32, for example, has routines called IsBadReadPtr(), IsBadWritePtr(), IsBadCodePtr() and IsBadStringPtr() that do this. Brent _______________________________________________ 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: Mach Reference Materials/Tutorials?
by David Blanton 15 Feb '03

15 Feb '03

Topically I would look at Mach IPC as implemented via Mach RPC with MIG support. _______________________________________________ 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: Mach Reference Materials/Tutorials?
by Emmett James 15 Feb '03

15 Feb '03

You might want to start at the place where it was developed, CMU. There's quite a lot of reference material there, and pointers to other material and books. http://www.cs.cmu.edu/afs/cs/project/mach/public/ On Fri, Feb 14, 2003 at 02:12:13PM -0800, Brent Marykuca wrote: > Hi, > > I'm interested in learning the nitty-gritty of how the mach kernel > works, both the theory of it, and the practical side of how to program > using mach interfaces. I've found some of the mach kernel interface > documentation in the Darwin repository, so that's a start, but I'm > looking for more details. (I'll eventually get around to reading the > source in detail, but I'm looking for some higher-bandwidth materials > first.) > > Can you point me toward useful reference material for answers to such > questions as: > > a) Where are the headers for mach kernel apis? > b) How does a user process communicate with the kernel? > c) What is MIG? Where do I find it and how do I use it? > > Thanks, > > Brent Marykuca _______________________________________________ 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