February 2003 - Darwin-kernel

sample driver code ?
by vinai 15 Feb '03

15 Feb '03

Hi Folks, I'm new to these lists, and joined because I was interested in trying to code a driver for a PCMCIA card I used under classic, but is no longer supported under OS X. As far as my programming background goes, I've done a bit of C coding and a very minor amount of device driver work with the Linux kernel. I got some of Apple's documentation about Objective C and their notes on their IO Kit and kernel programming. What really helps me is looking at something that has traits similar to my final goal, and modifying and optimizing as I go along. The card I would like to try to get working is Irez's PCMCIA Capsure card. I know there's the Video4Linux API for the linux kernel, and the bttv driver for the Brook Tree series of digitizer chips. Is there something akin to these out there (under the GPL preferably) for OS X ? And if not - what's the closest thing people can recommend ? thanks much all. cheers vinai _______________________________________________ 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 14 Feb '03

14 Feb '03

On Friday, February 14, 2003, at 02:11 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 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. You can, of course, manage your pointers and make sure you do proper initialization to minimize the chance for violating segment boundaries, etc. (e.g. init to NULL and check if ptr == NULL before using...simple basic stuff). If you feel ambitious, you could always instrument the C RTL and gcc to do range checking and memory management for you. There are debugging tools (Purify, Insure++) that do these types of things. As far as getting a pointer passed to you from external entity: you just have to hope that the interface is being followed (i.e. someone doesn't pass you an un-init'd pointer when you expect a ptr that is init'd, for example. 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.

1 0

Mach Reference Materials/Tutorials?
by Brent Marykuca 14 Feb '03

14 Feb '03

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

Pointer Testing
by Brent Marykuca 14 Feb '03

14 Feb '03

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.

1 0

Re: Re (5): Baffled
by Eric Long 14 Feb '03

14 Feb '03

> This clears up lots of things. If you don't build clean and > transition a driver from a NKE to an IOKit module, then the file that > is generated by project builder may not get regenerated. Though I'm > still puzzled because that means that the system doesn't know that > your module contains IOKit classes. Do you subclass any IOKit > classes or do you just use them straight? I subclass IOService and IOUserClient. I didn't have any problems with it being loaded and called. Matching worked fine and my client app had no problem talking to it. Somehow, the unsupported mix was working. Dean's suggestion to call my NKE start/stop methods from init() and free() seems to work well. I am concerned about safety if the NKE fails to unload, but otherwise the process seems good. Eric _______________________________________________ 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: Re (5): Baffled
by Godfrey van der Linden 14 Feb '03

14 Feb '03

This clears up lots of things. If you don't build clean and transition a driver from a NKE to an IOKit module, then the file that is generated by project builder may not get regenerated. Though I'm still puzzled because that means that the system doesn't know that your module contains IOKit classes. Do you subclass any IOKit classes or do you just use them straight? If you need a function to be called at load time AND you are subclassing OSObject, directly or indirectly through IOService, then I suggest you use static class construction/destruction. There is a pretty clear cut example on how to do this in IOSerialFamily/IOSerialFamily.kextproj/IOSerialBSDClient.cpp Godfrey At 11:19 -0800 03-2-14, Eric Long wrote: > I think the problem is that you are trying to use the module start/stop entry points AND have an I/O Kit driver in the same kext. This isn't supported, because the C++ runtime hijacks the module start/stop entry points in order to setup your kext's classes. This is the primary difference between the "IOKit Driver" and "Generic Kernel Extension" project templates in ProjectBuilder. I suspected this was involved. The puzzlement is that I was able to use the extension for a long time without this problem appearing. Originally, the extension had no IOKit code. It was just the basic NKE. Later, I added code to interface with a client app and to allow me to use lots of helpful IOKit classes in parsing XML preference data passed to me. Somehow, it worked. Until I trashed the original build folder to build from scratch, things were peachy. 1) Split your kext into two - one I/O Kit driver, and one generic kext. 2) Relocate the code from your module start/stop routines into init() and free(), or someplace similar. Be aware that init & free will be called once for each instance of your class, rather than just once like module start/stop. This is probably the best choice if you can adapt your setup teardown code. 3) You may be able to provide a static initializer/destructor with your class that will be called at module start/stop time. I'll have to check to see if this is possible, as I've not tried to do it myself. I'll check that out. Thanks for clearing things up. _______________________________________________ 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(6): Baffled
by Peter Lovell 14 Feb '03

14 Feb '03

Woops. You had mentioned IOKit and I hadn't paid enough attention. IOKit is not my area of expertise (I deal mostly with non-IOKit NKE's) but I do seem to remember that startup and stop routines for IOKit are handled differently. I thought that start and stop entry points for IOKit modules were actually in IOKit code, and the IOKit start/stop then called your methods. Regards.....Peter On Friday, February 14, 2003, at 01:33 PM, Eric Long wrote: Did you make sure that the entry-point names are defined as extern "C" ?? They mustn't be mangled so if your source file is C++ then you must define them with regular C linkage. The module start/stop function definitions reside exclusively in a .c file. I think that particular message refers to the IOKit related code, which descends from IOKit classes. If the OS is looking for my start/stop function names and can't find them, I would expect to see some kind of error noted. It doesn't look to me like it is even trying to find them. It's like it doesn't think there's anything besides the IOKit related code in the kext. The problem occurs even when the kext is owned by root, in group wheel. In remote debugging, I can see all the symbols I expect. If I set breakpoints in my start/stop functions, they never get hit. I've looked at the data in the executable. It has the start/stop function names, unmangled, within it. Very strange. Eric _______________________________________________ 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: Finding one parent for a vnode born under X
by Umesh Vaishampayan 14 Feb '03

14 Feb '03

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? 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... 2. using volfs to get the vnode from the CNID. <http://developer.apple.com/qa/qa2001/qa1113.html> Grrrr..... *NO!*. Do not use volfs in a KEXT. --Umesh This second alternative is a significant binary compatibility liability. It's *definitely* not appropriate to put it in a product that you're going to ship to lots of users. OTOH, it's more compatible than going poking around in v_data. 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. -- Umesh Vaishampayan Mac OS X - Kernel _______________________________________________ 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: Re (5): Baffled
by Eric Long 14 Feb '03

14 Feb '03

> I think the problem is that you are trying to use the module start/stop > entry points AND have an I/O Kit driver in the same kext. This isn't > supported, because the C++ runtime hijacks the module start/stop entry > points in order to setup your kext's classes. This is the primary > difference between the "IOKit Driver" and "Generic Kernel Extension" > project templates in ProjectBuilder. I suspected this was involved. The puzzlement is that I was able to use the extension for a long time without this problem appearing. Originally, the extension had no IOKit code. It was just the basic NKE. Later, I added code to interface with a client app and to allow me to use lots of helpful IOKit classes in parsing XML preference data passed to me. Somehow, it worked. Until I trashed the original build folder to build from scratch, things were peachy. > 1) Split your kext into two - one I/O Kit driver, and one generic kext. > 2) Relocate the code from your module start/stop routines into init() > and free(), or someplace similar. Be aware that init & free will be > called once for each instance of your class, rather than just once like > module start/stop. This is probably the best choice if you can adapt > your setup teardown code. > 3) You may be able to provide a static initializer/destructor with your > class that will be called at module start/stop time. I'll have to > check to see if this is possible, as I've not tried to do it myself. I'll check that out. Thanks for clearing things up. Eric _______________________________________________ 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 (5): Baffled
by Dean Reece 14 Feb '03

14 Feb '03

Eric, I think the problem is that you are trying to use the module start/stop entry points AND have an I/O Kit driver in the same kext. This isn't supported, because the C++ runtime hijacks the module start/stop entry points in order to setup your kext's classes. This is the primary difference between the "IOKit Driver" and "Generic Kernel Extension" project templates in ProjectBuilder. There are a few options available to you: 1) Split your kext into two - one I/O Kit driver, and one generic kext. 2) Relocate the code from your module start/stop routines into init() and free(), or someplace similar. Be aware that init & free will be called once for each instance of your class, rather than just once like module start/stop. This is probably the best choice if you can adapt your setup teardown code. 3) You may be able to provide a static initializer/destructor with your class that will be called at module start/stop time. I'll have to check to see if this is possible, as I've not tried to do it myself. Cheers, - Dean On Friday, February 14, 2003, at 09:10 AM, Eric Long wrote: Mike, Peter, Stiphane, Thanks for the feedback. I'm not sure if I was clear about the state of things. To clarify, my kext appears to be loaded as if nothing is wrong. Attached is a log of the results from using kextload -v 6, followed by kextstat. I don't see anything that sheds light on things in that log. When the extension loads, I expect to see this: Feb 12 11:13:00 Gonzo mach_kernel: iCleanSniffer_module_start entered Feb 12 11:13:00 Gonzo mach_kernel: ALHI_URLParser_Init Feb 12 11:13:00 Gonzo mach_kernel: iCleanSniffer pffindproto found TCP Feb 12 11:13:00 Gonzo mach_kernel: ICSnDriverShim::init() Feb 12 11:13:00 Gonzo mach_kernel: ICSnDriverShim::probe() Feb 12 11:13:00 Gonzo mach_kernel: ICSnDriverShim::start() Instead, I'm seeing this: Feb 12 11:13:00 Gonzo mach_kernel: ICSnDriverShim::init() Feb 12 11:13:00 Gonzo mach_kernel: ICSnDriverShim::probe() Feb 12 11:13:00 Gonzo mach_kernel: ICSnDriverShim::start() The module start function never gets called. I can purposely put code in their to cause a panic, without effect. Regarding root/wheel, when the kext loads, it gets flagged for not having root/wheel as the owner/group, and an alert appears to warn about it, it asks whether to cancel, use, or fix. I select use. This has been my standard practice all along, without problems, until now. Here's some of the important project settings: buildSettings = { FRAMEWORK_SEARCH_PATHS = ""; HEADER_SEARCH_PATHS = "///src"; INSTALL_PATH = "$(SYSTEM_LIBRARY_DIR)/Extensions"; KERNEL_MODULE = YES; LIBRARY_SEARCH_PATHS = /usr/lib/gcc/darwin/3.1; MODULE_IOKIT = YES; MODULE_NAME = com.aladdinsys.nke.iCleanSniffer; MODULE_START = iCSniffer_start; MODULE_STOP = iCSniffer_stop; MODULE_VERSION = 1.0.0; OPTIMIZATION_CFLAGS = "-O0"; OTHER_REZFLAGS = ""; PREFIX_HEADER = src/MacKEXT.Debug.prefix.h; PRODUCT_NAME = iCleanSniffer_Debug; SECTORDER_FLAGS = ""; WARNING_CFLAGS = "-Wmost -Wno-four-char-constants -Wno-unknown-pragmas"; WRAPPER_EXTENSION = kext; }; <!DOCTYPE plist PUBLIC \"-//Apple Computer//DTD PLIST 1.0//EN\" \"http://www.apple.com/DTDs/PropertyList-1.0.dtd\"> <plist version=\"1.0\"> <dict> <key>CFBundleDevelopmentRegion</key> <string>English</string> <key>CFBundleExecutable</key> <string>iCleanSniffer_Debug</string> <key>CFBundleGetInfoString</key> <string>iClean version 6.0a3, Copyright ,) 2002-2003 Aladdin Systems.</string> <key>CFBundleIconFile</key> <string></string> <key>CFBundleIdentifier</key> <string>com.aladdinsys.nke.iCleanSniffer</string> <key>CFBundleInfoDictionaryVersion</key> <string>6.0</string> <key>CFBundleName</key> <string>iCleanSniffer</string> <key>CFBundlePackageType</key> <string>KEXT</string> <key>CFBundleShortVersionString</key> <string>6.0a3</string> <key>CFBundleSignature</key> <string>????</string> <key>CFBundleVersion</key> <string>1.0.0</string> <key>IOKitPersonalities</key> <dict> <key>ICSnClientInterface</key> <dict> <key>CFBundleIdentifier</key> <string>com.aladdinsys.nke.iCleanSniffer</string> <key>IOClass</key> <string>com_aladdinsys_ICSnDriverShim</string> <key>IOProviderClass</key> <string>IOResources</string> <key>IOResourceMatch</key> <string>IOKit</string> <key>IOUserClientClass</key> <string>com_aladdinsys_ICSnClientInterface</string> </dict> </dict> <key>OSBundleLibraries</key> <dict> <key>com.apple.kernel.bsd</key> <string>1.1</string> <key>com.apple.kernel.iokit</key> <string>1.1</string> <key>com.apple.kernel.libkern</key> <string>1.1</string> <key>com.apple.kernel.mach</key> <string>1.1</string> </dict> </dict> </plist> Thanks again for chiming in. I'll test to see if setting privileges to root/wheel has any effect, but I don't think it is a factor in this problem. Eric [demime 0.98b removed an attachment of type multipart/appledouble] _______________________________________________ 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