December 2003 - Darwin-kernel

Re: Kernel questions: Syscalls, ABIs, Signals, and Threads
by Jim Magee 12 Dec '03

12 Dec '03

On Dec 11, 2003, at 8:02 PM, Andrew Begel wrote: 4. Are NSThreads implemented using pthreads or directly on top of Mach threads? i.e. if I call pthread_self() from an NSThread, will that give me a real answer? They sit on top of pthreads. So, an NSThread can call pthread functions without worry. --Jim PS: The ABI questions are best answered by others. However, there is Mach-O ABI documentation on the http://developer.apple.com website that you should definitely take a look at. _______________________________________________ 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

opening a vnode at mount time?
by Chris Bergmann 12 Dec '03

12 Dec '03

I'm trying to add additional system-level files to HFS+. I've added code (see below) to the end of hfs_MountHFSPlusVolume() to obtain a vnode for the file. Currently i'm not using the vnode, and i'm releasing it in hfsUnmount() using ReleaseMetaFileVNode(). In console mode, this appears to work except that hfs_unmount() claims a file is remaining open. This is not a primary concern right now, although it suggests perhaps that my vnode may not be considered VSYSTEM. The real problem i'm having is that if i try this out in the GUI, i score a kernel panic as soon as i attempt to open the volume in finder. (note that the mount operation itself is successful.) I ran some basic file operations (ls, cp) in console mode without reproducing the panic. I have yet to attempt to dual-machine debug the panic; thought it saner to ask here first in case i'm doing something visibly stupid in my attempt to open the file.. chris cnid = GetFileInfo(vcb, kRootDirID, ".hfse_backing_file", &cnattr, &cfork); if (cnid) { bzero(&cndesc, sizeof(cndesc)); cndesc.cd_parentcnid = kRootParID; cndesc.cd_flags |= CD_ISMETA; cndesc.cd_nameptr = ".hfse_backing_file"; cndesc.cd_namelen = strlen(".hfse_backing_file"); cndesc.cd_cnid = cnid; retval = hfs_getnewvnode(hfsmp, NULL, &cndesc, 0, &cnattr, &cfork, &vnode); if (retval) { ... } else { VOP_UNLOCK(vnode, 0, p); } } _______________________________________________ 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

Kernel questions: Syscalls, ABIs, Signals, and Threads
by Andrew Begel 12 Dec '03

12 Dec '03

I hope this list proves to be more productive for kernel questions than darwin-development. 1. What is the syscall # for thread_create() and thread_create_running() on Darwin 7 PPC? Even better would be a pointer on where to find this number in the source code. In /usr/include/mach/task.h is a table of function names to numbers, indicating that thread_create() was 4311, but when I tried using that syscall number to set an emulation point (using task_set_emulation()) and redirect thread_create() to my own function, my own function was never called. 2. How many bytes is the red zone in the Darwin i386 ABI? Actually, a better question would be where can I find documentation (or source code) as to the ABI and calling conventions used by Darwin i386? 3. When a signal is delivered to a Mach-O application's thread, does the call to the signal handler obey the runtime calling conventions outlined in the Mach-O Runtime Architecture manual? Specifically, is the linkage area with the LR and CR registers stored at 8(SP) and 4(SP) (respectively?). Does the stack pointer stored in the signal handler's stack frame also point back to the stack pointer of the stack frame that was executing when the signal was triggered? Are there any stack frames in between them/is the original thread's stack still accessible from within the signal handler? 4. Are NSThreads implemented using pthreads or directly on top of Mach threads? i.e. if I call pthread_self() from an NSThread, will that give me a real answer? _______________________________________________ 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: Getting the PID
by Justin Walker 11 Dec '03

11 Dec '03

On Wednesday, December 10, 2003, at 07:09 PM, Matt Jaffa wrote: Well I have patched the Systemcall table, the sendto method, and so I needed a process id to go with the socket decsriptor to get back the socket, but I believe patching the syscall table is not a smart idea. Good call. So now I am not concerned with this approach anymore. If I use a nke to intercept network stuff, is it possible that if I dont like where they are trying to access i can just drop the data? Well, there will likely be differing opinions on this one. My feeling is that just dropping packets is A Bad Thing. There are legitimate reasons why network operations appear to hang, but they are difficult to track down. Adding another one makes it that much more difficult. You should probably return an error when you see a bad (i.e., administratively prohibited) address. The error EHOSTUNREACH is a possibility. So is EADDRNOTAVAIL. Depending on your specific goals, you might want to avoid blanket rejection of datagrams with offending addresses (again, it makes debugging problems more difficult). For example, consider returning an error for TCP connection requests (from either end), but not for other datagrams. That way, things like ping and traceroute will continue to work, but web browsing will not work. Regards, Justin -- /~\ The ASCII Justin C. Walker, Curmudgeon-at-Large \ / Ribbon Campaign X Help cure HTML Email / \ _______________________________________________ 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: Getting the PID
by Matt Jaffa 11 Dec '03

11 Dec '03

Well I have patched the Systemcall table, the sendto method, and so I needed a process id to go with the socket decsriptor to get back the socket, but I believe patching the syscall table is not a smart idea. So now I am not concerned with this approach anymore. If I use a nke to intercept network stuff, is it possible that if I dont like where they are trying to access i can just drop the data? Matt On Dec 10, 2003, at 5:56 PM, Justin Walker wrote: On Wednesday, December 10, 2003, at 03:45 PM, Matt Jaffa wrote: I am trying to get the PID of something. All I basically have to work with is a socket file descriptor. say when you do this int so = socket(....,....., ...., ... etc.); you get back the socket file descriptor, now how can i take that number and see which process created it? You really must enjoy poking yourself in the eye with sharp sticks. This is one of many reasons why doing "user-mode things" in "kernel mode" is not A Good Idea. In the first place, there is almost no reasonable way to determine, from a file descriptor, what process "created" it. Even in user-mode, this is indeterminate. For example: scenario 1: parent creates (open/socket/...) a file descriptor parent forks child execs parent exits Now, the creator of the descriptor is long gone, and the code in the child has no idea where the file descriptor came from, or, in general, what it describes. scenario 2: process A gets a file descriptor somehow process A passes this descriptor to process B process A exits process B execs Again, the code left running (the new code in process B) is left "holding the bag", without knowing much about the bag. What are you trying to do? Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | Some people have a mental | horizon of radius zero, and | call it their point of view. | -- David Hilbert *-------------------------------------- *-------------------------------* _______________________________________________ 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: Getting the PID
by Justin Walker 11 Dec '03

11 Dec '03

On Wednesday, December 10, 2003, at 03:45 PM, Matt Jaffa wrote: I am trying to get the PID of something. All I basically have to work with is a socket file descriptor. say when you do this int so = socket(....,....., ...., ... etc.); you get back the socket file descriptor, now how can i take that number and see which process created it? You really must enjoy poking yourself in the eye with sharp sticks. This is one of many reasons why doing "user-mode things" in "kernel mode" is not A Good Idea. In the first place, there is almost no reasonable way to determine, from a file descriptor, what process "created" it. Even in user-mode, this is indeterminate. For example: scenario 1: parent creates (open/socket/...) a file descriptor parent forks child execs parent exits Now, the creator of the descriptor is long gone, and the code in the child has no idea where the file descriptor came from, or, in general, what it describes. scenario 2: process A gets a file descriptor somehow process A passes this descriptor to process B process A exits process B execs Again, the code left running (the new code in process B) is left "holding the bag", without knowing much about the bag. What are you trying to do? Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | Some people have a mental | horizon of radius zero, and | call it their point of view. | -- David Hilbert *--------------------------------------*-------------------------------* _______________________________________________ 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: Getting the PID
by Brian Bergstrand 11 Dec '03

11 Dec '03

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 No. The number you get back from socket() or open() is no good without a corresponding PID. In the kernel, each process contains an array of pointers to kernel file structures. The number you get back from socket() or open() (the "file descriptor") is just the index into the process's array of file pointers. So without know what process you are dealing with, there is no way to correlate a descriptor (index) number to anything. And even knowing what process it is, you still can't guarantee anything, since the descriptor could be invalid or pointing to another file by the time you get around to querying for it. HTH. On Dec 10, 2003, at 5:45 PM, Matt Jaffa wrote: > Ok, > > I am trying to get the PID of something. > All I basically have to work with is > a socket file descriptor. > > say when you do this int so = socket(....,....., ...., ... etc.); > > you get back the socket file descriptor, now how can i take that > number and see which process created it? > > Thanks, > Matt > Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key: <http://www.bergstrand.org/brian/misc/public_key.txt> AIM: triryche206 It is impossible to enjoy idling thoroughly unless one has plenty of work to do. - Jerome K. Jerome -----BEGIN PGP SIGNATURE----- Version: PGP 8.0.3 iQA/AwUBP9ep5HnR2Fu2x7aiEQIdvACfUyEuG9NNvMylECAymDqoRkSC2BUAoKIU 0yeTdcSo4Tu4Sz6Vi+V+KAnk =fRYz -----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

Getting the PID
by Matt Jaffa 10 Dec '03

10 Dec '03

Ok, I am trying to get the PID of something. All I basically have to work with is a socket file descriptor. say when you do this int so = socket(....,....., ...., ... etc.); you get back the socket file descriptor, now how can i take that number and see which process created it? Thanks, Matt _______________________________________________ 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: Link/Load Errors
by Brian Bergstrand 10 Dec '03

10 Dec '03

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Looks like you are compiling with the C++ compiler (maybe an IOKIt extension), this is mangling the getsock() symbol name: > kld(): Undefined symbols: > __Z7getsockP8filedesciPP4file This means that the header that declares getsock() does not have ' extern "C" {} ' protection, so when you include the header in your code, you will have to provide that protection yourself. In general, this will be true for any non-IOKit kernel header. They expect the C compiler. extern "C" { #include <sys/socketvar.h> } HTH. On Dec 10, 2003, at 1:24 PM, Matt Jaffa wrote: > I have made a kernel Extension and i am using the getsock > function within in it. > > WebDAV uses it just fine, > > Mine compiles and everything but when i load it, I get this error > So obviously it is not linking right, In The Project in Expert view > am I suppose to include something or what is going on? > > Matt-Jaffas-Computer:~ matt$ sudo kextload -t > /System/Library/Extensions/KnowControl.kext/ > kextload: extension /System/Library/Extensions/KnowControl.kext/ > appears to be valid > kextload: notice: extension > /System/Library/Extensions/KnowControl.kext/ has debug properties set > kld(): Undefined symbols: > __Z7getsockP8filedesciPP4file > kextload: kld_load_from_memory() failed for module > /System/Library/Extensions/KnowControl.kext/Contents/MacOS/KnowControl > kextload: a link/load error occured for kernel extension > /System/Library/Extensions/KnowControl.kext/ > load failed for extension /System/Library/Extensions/KnowControl.kext/ > (run kextload with -t for diagnostic output) > > Thanks, > > Matt > Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key: <http://www.bergstrand.org/brian/misc/public_key.txt> AIM: triryche206 One advantage of talking to yourself is that you know at least somebody's listening. - Franklin P. Jones As of 02:21:40 PM, iTunes is playing "Dancing Nancies" from "Under The Table And Dreaming" by "Dave Matthews Band" -----BEGIN PGP SIGNATURE----- Version: PGP 8.0.3 iQA/AwUBP9d0AXnR2Fu2x7aiEQIBogCfdrhSV78K2fKxD1PEej+YmyxErz4AoM8H TuiNQrnf1eKIwN9c21ysYAjh =O3fZ -----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

Link/Load Errors
by Matt Jaffa 10 Dec '03

10 Dec '03

I have made a kernel Extension and i am using the getsock function within in it. WebDAV uses it just fine, Mine compiles and everything but when i load it, I get this error So obviously it is not linking right, In The Project in Expert view am I suppose to include something or what is going on? Matt-Jaffas-Computer:~ matt$ sudo kextload -t /System/Library/Extensions/KnowControl.kext/ kextload: extension /System/Library/Extensions/KnowControl.kext/ appears to be valid kextload: notice: extension /System/Library/Extensions/KnowControl.kext/ has debug properties set kld(): Undefined symbols: __Z7getsockP8filedesciPP4file kextload: kld_load_from_memory() failed for module /System/Library/Extensions/KnowControl.kext/Contents/MacOS/KnowControl kextload: a link/load error occured for kernel extension /System/Library/Extensions/KnowControl.kext/ load failed for extension /System/Library/Extensions/KnowControl.kext/ (run kextload with -t for diagnostic output) Thanks, Matt _______________________________________________ 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