February 2005 - Darwin-kernel

Re: Callback from NKE
by Peter Lovell 17 Feb '05

17 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Feb 17, 2005, at 12:08 PM, Justin Walker wrote: Hi Justin, Regards.....Peter _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… Notable among them is W. Richard Stevens's "Unix Network Programming" (v. 1, 2nd ed). There is a third edition by a different author, and I'm not familiar with that. The third edition (Stevens/Fenner/Rodoff) continues the same high standard. The www.kohala.com website is no longer active; the example code is now available at www.unpbook.com This email sent to site_archiver(a)lists.apple.com

1 0

RE: Callback from NKE
by Carl Smith 17 Feb '05

17 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcUVE2Y2ypk2dfZiQwSHqBhFiWqOlQAADtkQ Thread-topic: Callback from NKE Great stuff Justin I really appreciate the information. Carl Sent: Thursday, February 17, 2005 12:09 PM To: darwinKernel Dev Subject: Re: Callback from NKE On Feb 17, 2005, at 8:50, Carl Smith wrote: > Excuse my igronance, being I am new to both Mac OS X and unix, from > what you wrote below, then you are saying that my select() or CFSocket > would be intiated and used in userland? Do you know of any example > code using this method? Yes, both select() and the CFSocket stuff are user-mode APIs. I don't know of any CF* documentation other than what is available at Apple's developer website (there may be other info out there), but for select(), there are several books that give a great intro to the use of select() and other BSD-style low-level APIs. Notable among them is W. Richard Stevens's "Unix Network Programming" (v. 1, 2nd ed). There is a third edition by a different author, and I'm not familiar with that. The Stevens website has some details plus a tarball of a bunch of example code that goes with the book (<http:www.kohala.com>). Stevens also has a book called "Advanced Programming in the Unix Environment" that is worth having (plus code on the site). He was an excellent expositor. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | When LuteFisk is outlawed | Only outlaws will have | LuteFisk *--------------------------------------*-------------------------------* _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/csmith%40fortresstech.… This email sent to csmith(a)fortresstech.com --- Incoming mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com) Version: 6.0.853 / Virus Database: 581 - Release Date: 2/1/2005 --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com) Version: 6.0.853 / Virus Database: 581 - Release Date: 2/1/2005 _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Callback from NKE
by Justin Walker 17 Feb '05

17 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Feb 17, 2005, at 8:50, Carl Smith wrote: Yes, both select() and the CFSocket stuff are user-mode APIs. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | When LuteFisk is outlawed | Only outlaws will have | LuteFisk *--------------------------------------*-------------------------------* _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… Excuse my igronance, being I am new to both Mac OS X and unix, from what you wrote below, then you are saying that my select() or CFSocket would be intiated and used in userland? Do you know of any example code using this method? I don't know of any CF* documentation other than what is available at Apple's developer website (there may be other info out there), but for select(), there are several books that give a great intro to the use of select() and other BSD-style low-level APIs. Notable among them is W. Richard Stevens's "Unix Network Programming" (v. 1, 2nd ed). There is a third edition by a different author, and I'm not familiar with that. The Stevens website has some details plus a tarball of a bunch of example code that goes with the book (<http:www.kohala.com>). Stevens also has a book called "Advanced Programming in the Unix Environment" that is worth having (plus code on the site). He was an excellent expositor. This email sent to site_archiver(a)lists.apple.com

1 0

FW: Callback from NKE
by Carl Smith 17 Feb '05

17 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcUUkMpUGM5Xnal8TeuI2sM570c7AgAf4ybw Thread-topic: Callback from NKE Vincent- Excuse my igronance, being I am new to both Mac OS X and unix, from what you wrote below, then you are saying that my select() or CFSocket would be intiated and used in userland? Do you know of any example code using this method? Thanks Carl -----Original Message----- From: Vincent Lubet Sent: Wednesday, February 16, 2005 8:33 PM To: Carl Cc: darwin-kernel(a)lists.apple.com Subject: Re: Callback from NKE Carl, Your userland program uses a kernel control socket file descriptor to communicate with your kernel control module. A kernel control socket file descriptor is above all a file descriptor so you can use a kernel control socket file descriptor userland API that work with socket file descriptors. To avoid to poll you can for example use select() or CFSocket for this with a call back function. Vincent On Feb 16, 2005, at 3:24 PM, Carl Smith wrote: > I have an NKE that I currently have registered using ctl_register > function and from my userland I connect to the NKE using sockets, > that is calling socket(PF_SYSTEM, SOCK_DGRM, SYSPROTO_CONTROL) then > on success connect(). > > > > What I would like to do is be able for the NKE to signal or notify > the app when it has some something to do or send, without the app > going into a constant loop waiting for data via a recv() function. > > > > Someone had mentioned the possibility of using CFSocket for this with > a call back function. > > > > I am not sure if this possible using a NKE and that is my question. > Can my idea be done using CFSocket and if yes are there any examples > of this using a kext? Of if not any other suggestions? > > > > Thanks > > Carl > > > > > > > --- > Outgoing mail is certified Virus Free. > Checked by AVG anti-virus system (http://www.grisoft.com) > Version: 6.0.853 / Virus Database: 581 - Release Date: 2/1/2005 > > _______________________________________________ > Do not post admin requests to the list. They will be ignored. > Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) > Help/Unsubscribe/Update your Subscription: > http://lists.apple.com/mailman/options/darwin-kernel/vlubet% > 40apple.com > > This email sent to vlubet(a)apple.com --- Incoming mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com) Version: 6.0.853 / Virus Database: 581 - Release Date: 2/1/2005 --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com) Version: 6.0.853 / Virus Database: 581 - Release Date: 2/1/2005 _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: problem of userclient for filter scheme driver
by Quinn 17 Feb '05

17 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com When i load the filter scheme driver it matches on the content hint and successfully expose the IOMedia object.I can read/write on this IOMedia object successfully by manually creating file or directory on this exposed IOMedia volume. But when i execute the sample user space application tool i am always getting the following error: Couldn't find any matching. This error comes when user space application tool looks for the serviceObjects and it didn't find one. Note that the "registered" flag appears inside the angle brackets. S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Technical Support * Networking, Communications, Hardware _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… At 18:59 +0530 16/2/05, Yogesh P wrote: Unfortunately you didn't post the code to your user space tool. My guess is that you're trying to location your I/O Kit service using IOServiceGetMatchingService[s]. If so, the most common error is that these routines will only find /registered/ services. You should use the ioreg tool to make sure that your service is register. $ ioreg -w 0 [...] +-o IOMediaBSDClient <class IOMediaBSDClient, registered, matched, active, busy 0, retain count 6> This email sent to site_archiver(a)lists.apple.com

1 0

Kernel extensions and code injection?
by bjk02＠hampshire.edu 17 Feb '05

17 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com User-agent: Internet Messaging Program (IMP) 3.1 I've been working on a kernel extension that adds functionality to the IOHIDSystem class. Specifically, I want to be able to intercept mouse and keyboard events, manipulate specific events, and then pass them along. The point of doing this in a kernel extension is so that I can do this before the events arrive at the target application. (Incidentally, if there is a way to do this without writing a kernel extension, perhaps at the point where the window server receives the events, please do tell! AFAIK, though, there isn't.) So the question is, is there a standard or approved way of injecting or hooking code into the kernel via a kernel extension? Or specifically, in IOHIDSystem? I'm familiar with the way programs like uControl and DoubleCommand do it, which is by creating a C++ faux-subclass for IOCommandGate, and replacing the instance of it in IOHIDSystem with a new version that has a different version of the runAction() virtual function. The problem, though, is that there are several input devices, notably some Kensington mice and all Wacom tablets, that post events to the kernel either from userspace or without utilizing IOCommandGate. Consequently, the kernel extension will miss these events. What I'm looking for is a safe way to add new functionality into the IOHIDSystem class, either by some kind of code injection, a hook, or anything really. However, I can't find any information on how one would do this in this part of the kernel. (Creating the same kind of faux-subclass for IOHIDSystem isn't feasible, since it would mean recompiling and re-releasing the kernel extension for each new subversion of Darwin.) Any ideas? - Brian bjk02(a)hampshire.edu _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Callback from NKE
by Vincent Lubet 17 Feb '05

17 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Carl, Vincent On Feb 16, 2005, at 3:24 PM, Carl Smith wrote: Thanks Carl --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com) Version: 6.0.853 / Virus Database: 581 - Release Date: 2/1/2005 _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… Your userland program uses a kernel control socket file descriptor to communicate with your kernel control module. A kernel control socket file descriptor is above all a file descriptor so you can use a kernel control socket file descriptor userland API that work with socket file descriptors. To avoid to poll you can for example use select() or CFSocket for this with a call back function. I have an NKE that I currently have registered using ctl_register function and from my userland I connect to the NKE using sockets, that is calling socket(PF_SYSTEM, SOCK_DGRM, SYSPROTO_CONTROL) then on success connect(). What I would like to do is be able for the NKE to signal or notify the app when it has some something to do or send, without the app going into a constant loop waiting for data via a recv() function. Someone had mentioned the possibility of using CFSocket for this with a call back function. I am not sure if this possible using a NKE and that is my question. Can my idea be done using CFSocket and if yes are there any examples of this using a kext? Of if not any other suggestions? _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/vlubet% 40apple.com This email sent to vlubet(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Callback from NKE
by Carl Smith 16 Feb '05

16 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcUUfrWchTp1V4OISuGIoky9+xp4Xw== Thread-topic: Callback from NKE _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: strange panic, debugging help wanted..
by Andrew Gallatin 16 Feb '05

16 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Andrew Gallatin writes: > > Godfrey van der Linden writes: > > I'd like to see the assembly that this kernel is running at 0x2CAE8. > > I'd be willing to bet an offset from a NULL pointer is being taken and > > that is why you are panicing. > > > > I think the 'r1' panic is a red herring, the first exception state is > > 'PC=0x0002CAE8; MSR=0x00001030; DAR=0x000000D4; DSISR=0x40000000; > > LR=0x0002 CAD8; R1=0x0CC33DB0; XCP=0x0000000C (0x300 - Data access)' > > Indicates that the r1 is valid at the time that the panic is taken. Do > > you have a symbolled kernel for the version that is taking the panic. > > If you can find out what routine was passed a NULL pointer you may have > > a suspect. > I've just found this crash waiting for me again this morning, and I managed to connect to it via remote gdb. This is with 10.3.7, though the PC is the same as it was for the much earlier version of MacOSX. The code seems to be in a totally different spot, which means maybe I totally botched converting the address -> symbols last time. (gdb) paniclog Unresolved kernel trap(cpu 1): 0x300 - Data access DAR=0x00000000000000D4 PC=0x000000000002CAE8 Latest crash info for cpu 1: Exception state (sv=0x22D43000) PC=0x0002CAE8; MSR=0x00001030; DAR=0x000000D4; DSISR=0x40000000; LR=0x0002CAD8; R1=0x0CC5BDB0; XCP=0x0000000C (0x300 - Data access) Backtrace: 0x0002CAD8 0x0002C8A8 0x0002C870 Proceeding back via exception chain: Exception state (sv=0x22D43000) previously dumped as "Latest" state. skipping... Exception state (sv=0x00A01500) PC=0x00000000; MSR=0x0000D030; DAR=0x00000000; DSISR=0x00000000; LR=0x00000000; R1=0x00000000; XCP=0x00000000 (Unknown) Kernel version: Darwin Kernel Version 7.7.0: Sun Nov 7 16:06:51 PST 2004; root:xnu/xnu-517.9.5.obj~1/RELEASE_PPC [ CPU 0 then panics because a simple lock acquisition times out] The symbolic bactrace is: (gdb) bt #0 0x0002cae8 in do_thread_scan () at /SourceCache/xnu/xnu-517.9.5/osfmk/kern/sched_prim.c:2790 #1 0x0002c8a8 in sched_tick_thread_continue () at /SourceCache/xnu/xnu-517.9.5/osfmk/kern/sched_prim.c:2671 Frame 0 works out to be: thread = processor->idle_thread; if (thread->sched_stamp != sched_tick) { if (stuck_count == MAX_STUCK_THREADS) { restart_needed = TRUE; break; } And 0xd4 is the offset of sched_stamp: (gdb) p ((thread_t) 0)->sched_stamp Cannot access memory at address 0xd4 So it looks like the idle thread got zeroed out somehow. This seems to be confirmed by other sources: (gdb) info locals restart_needed = 9012778 thread = 0x0 pset = 0x331800 processor = 0x1239948 s = 1 The processor struct looks like this. (gdb) p/x *processor $9 = {processor_queue = {next = 0x337478, prev = 0x331814}, state = 0x33705c, active_thread = 0x0, next_thread = 0x0, idle_thread = 0x0, processor_set = 0x1239948, current_pri = 0x80000000, quantum_timer = {q_link = { next = 0x0, prev = 0x0}, func = 0x0, param0 = 0x0, param1 = 0x0, deadline = 0x0, state = 0x266450}, quantum_end = 0x0, last_dispatch = 0xcb18000, timeslice = 0x104, deadline = 0x2400000080, runq = {highq = 0x80, bitmap = {0x5f, <...> This looks corrupt, and looking at the pset the processor address seems funny. Eg: (gdb) p *pset $13 = {idle_queue = {next = 0x331800, prev = 0x331800}, idle_count = 0, active_queue = {next = 0x337000, prev = 0x33748c}, processors = {next = 0x1239948, prev = 0x33748c}, processor_count = 2, sched_lock = {lock_data = 19103745}, runq = {highq = 0, bitmap = {0, 0, 0, 1}, count = 0, urgency = 0, queues = {{ next = 0x331840, prev = 0x331840}, {next = 0x331848, prev = 0x331848}, { <...> Eg, all the addresses here are in the 0x33xxxx range, so processor = 0x1239948 seems rather suspicious. The other one looks more reasonable: $15 = (struct processor *) 0x33748c (gdb) p *(processor_t)0x33748c $16 = {processor_queue = {next = 0x33180c, prev = 0x337000}, state = 1, active_thread = 0x1238000, next_thread = 0x0, idle_thread = 0x1238948, processor_set = 0x331800, current_pri = 95, quantum_timer = {q_link = { next = 0x135ad6c, prev = 0x320ed0}, func = 0x28738 <thread_quantum_expire>, param0 = 0x33748c, param1 = 0x1238000, deadline = 37553127975012, state = DELAYED}, quantum_end = 37553127975012, last_dispatch = 37553127641687, timeslice = 1, deadline = 18446744073709551615, runq = {highq = 0, bitmap = {0, 0, <...> But that 0x1238948 address is showing up here too (for the idle thread). I have no idea what is supposed to be there. Dumping the contents don't mean much to me: (gdb) x/32 0x1238948 0x1238948 <mhp.0+15366720>: 0x00000000 0x00000000 0x00000000 0x00000000 0x1238958 <mhp.0+15366736>: 0x00000000 0x00000000 0x01238948 0x80000000 0x1238968 <mhp.0+15366752>: 0x00000000 0x00000000 0x00000000 0x00000000 0x1238978 <mhp.0+15366768>: 0x00000008 0xffffffff 0x00000000 0x0002c490 0x1238988 <mhp.0+15366784>: 0x00000000 0x00000000 0x00000000 0x0cd30000 0x1238998 <mhp.0+15366800>: 0x00000184 0x00000004 0x00000000 0x00000000 0x12389a8 <mhp.0+15366816>: 0x0000005f 0x00000051 0x00000000 0x00000000 0x12389b8 <mhp.0+15366832>: 0x00000000 0x00000000 0x0000000e 0x00000000 If anybody has any ideas, I'll leave this in gdb for a while.. Thanks, Drew _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

problem of userclient for filter scheme driver
by Yogesh P 16 Feb '05

16 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi folks, I had written the sample filter scheme driver using Mass_storage.pdf and also written the sample userclient and user space application for that filter scheme driver. When i load the filter scheme driver it matches on the content hint and successfully expose the IOMedia object.I can read/write on this IOMedia object successfully by manually creating file or directory on this exposed IOMedia volume. But when i execute the sample user space application tool i am always getting the following error: Couldn't find any matching. This error comes when user space application tool looks for the serviceObjects and it didn't find one. If i have to add IOProviderClass and IOMatchCategory properties in my IOKitPersonality then what values these properties contain for filter scheme driver so that it exposes the interface for the userclient and in turn user space application tool???? I am specifying my Info.plist here. <?xml version="1.0" encoding="UTF-8"?> <!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>MyFilterScheme</string> <key>CFBundleIdentifier</key> <string>com.MySoftwareCompany.driver.MyFilterScheme</string> <key>CFBundleInfoDictionaryVersion</key> <string>6.0</string> <key>CFBundlePackageType</key> <string>KEXT</string> <key>CFBundleSignature</key> <string>????</string> <key>CFBundleVersion</key> <string>1.0d3</string> <key>IOKitPersonalities</key> <dict> <key>MyFilterScheme</key> <dict> <key>CFBundleIdentifier</key> <string>com.MySoftwareCompany.driver.MyFilterScheme</string> <key>Content Hint</key> <string>MySoftwareCompany_MyContent</string> <key>Content Mask</key> <string>MySoftwareCompany_MyContent</string> <key>IOClass</key> <string>com_MySoftwareCompany_driver_MyFilterScheme</string> <key>IOMatchCategory</key> <string>IOStorage</string> <key>IOProviderClass</key> <string>IOMedia</string> <key>IOUserClientClass</key> <string>com_MySoftwareCompany_MyUserClient</string> <key>IOkitDebug</key> <string>65535</string> </dict> </dict> <key>OSBundleLibraries</key> <dict> <key>com.apple.iokit.IOStorageFamily</key> <string>1.0b1</string> <key>com.apple.kernel.iokit</key> <string>1.0b1</string> <key>com.apple.kernel.libkern</key> <string>1.0b1</string> <key>com.apple.kernel.mach</key> <string>1.0b1</string> </dict> </dict> </plist> Does am i specifying the correct values for IOProviderClass and IOMatchCategory??? Please correct me if i am mistaking somewhere. Thanks in advance Yogesh Pahilwan _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0