February 2005 - Darwin-kernel

Re: NKE failing on ctl_register call
by Vincent Lubet 16 Feb '05

16 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com What version of Mac OS X is this happening on? Vincent On Feb 14, 2005, at 8:59 AM, Carl Smith wrote: here on this forum along with some code from the tcplog example. this NKE of mine it is failing, with an errno of 1. this means anything. my NKE. "kern_return_t ClientKern103_start (kmod_info_t * ki, void * d) { int nReturn = 0; struct kern_ctl_reg AFctl_reg; bzero(&AFctl_reg, sizeof(AFctl_reg)); AFctl_reg.ctl_id = AFMAC_KERN_ID; AFctl_reg.ctl_unit = 0; AFctl_reg.ctl_flags = CTL_FLAG_PRIVILEGED; AFctl_reg.ctl_sendsize = 0; AFctl_reg.ctl_recvsize = 0; // Dispatch functions // ================== AFctl_reg.ctl_connect = AFKernConnect; AFctl_reg.ctl_disconnect = AFKernDisconnect; AFctl_reg.ctl_write = AFClientKernXWrite; AFctl_reg.ctl_set = NULL; AFctl_reg.ctl_get = AFKernGet; nReturn = ctl_register(&AFctl_reg, 0, &ctlref); if(nReturn != 0) { if(nReturn == EINVAL) printf("Kernel Register failed"); else } return KERN_SUCCESS; } " going between my NKE and my client. Any help/suggestions/example code would greatly be appreciated. Thank You --- 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.… I am writing a NKE following some of the example/text given in the 'About Network Kernel Extensions' document, along with following some code from an example I found My problem is that on the call 'ctl_register' I am getting a return value of (22) from the ctl_register call and according to the documentation ctl_register should return a 0 on success. And in turn from my client app when I try to 'connect' to So I am assuming something is wrong with my ctl_register call. I load my kext using 'sudo kextload -v mynke.kext'. I am originally signed in as 'admin'. Not sure if Below is a portion of the code, from my NKE, that I am using to try and register Does the ctl_id have anything to do with it? I did register it but do I have to do anything special with/to the kernel? I am also using the same ID in my client side. static kern_ctl_ref ctlref = 0; /* Reference of the kernel controller */ printf("Kernel Register failed with the error (%d)\n", nReturn); I am not looking to anything fancy, yet, I am just trying to get the communication _______________________________________________ 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

RE: Change to root uid
by Matthew Jaffa 15 Feb '05

15 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Carl, AFctl_reg.ctl_flags = CTL_FLAG_PRIVILEGED; with this line: AFctl_reg.ctl_flags = 0; That should work. Matt From: "Carl Smith" <CSmith(a)fortresstech.com> To: <darwin-kernel(a)lists.apple.com> Subject: Change to root uid Date: Tue, 15 Feb 2005 10:23:14 -0500 I register my NKE as following: " struct kern_ctl_reg AFctl_reg; bzero(&AFctl_reg, sizeof(AFctl_reg)); AFctl_reg.ctl_id = AFMAC_KERN_ID; AFctl_reg.ctl_unit = 0; AFctl_reg.ctl_flags = CTL_FLAG_PRIVILEGED; AFctl_reg.ctl_sendsize = 0; AFctl_reg.ctl_recvsize = 0; AFctl_reg.ctl_connect = AFKernConnect; AFctl_reg.ctl_disconnect = AFKernDisconnect; AFctl_reg.ctl_write = AFClientKernXWrite; AFctl_reg.ctl_set = NULL; AFctl_reg.ctl_get = AFKernGet; static kern_ctl_ref ctlref = 0; /* Reference of the kernel controller */ nReturn = ctl_register(&AFctl_reg, 0, &ctlref); " Then in user land I make the connection to the NKE as follows: fd = socket(PF_SYSTEM, SOCK_DGRAM, SYSPROTO_CONTROL); ... and so on }" Thanks Carl -----Original Message----- From: Peter Sent: Monday, February 14, 2005 7:17 PM To: Carl Cc: darwin-kernel(a)lists.apple.com Subject: Re: Change to root uid On Feb 14, 2005, at 6:35 PM, Carl Smith wrote: > I have a NKE and a client that I to talk to each other. If the client > app is run at some users, that might not be root, I still want my > client/NKE to talk with each other, but it is my understanding that > NKEs need to be set with root user options only. In this case I need > to set the uid to root within my client application. > > Is this possible and if so could you point me in the right direction > as to what apis I need to be using? The ownership/permissions of the nke need to be root/0755 for security reasons but that doesn't mean that it "runs" as root. It runs in the kernel and any user can talk to it, using various mechanisms. It's up to you to make sure that the uid and/or application which talk to your nke are suitable and appropriate for what you want to do. Your client application does not need to be suid root, nor should you (for usual activity) restrict your nke to a connection which can only be made by someone logged in as root. You haven't given us enough information to provide much more help than this. The best thing would be to describe a bit more of what you're trying to do. Regards.....Peter p.s. remember also that nke interfaces are subject to change, and Apple presently discourages some of this development precisely because of the version-to-version compatibility issues --- 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/mtjaffa%40msn.com _______________________________________________ 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.… Sorry Carl that last message didn't go through. In order for any application root or non-root to talk with your system socket, replace this line: Thanks for the reply Peter. Yes I guess I was too general in my description. What I am trying to accomplish is to have my NKE sending packet statistic to a userland/client application. Seeing as how I want this userland application to be assessable by who ever signs on to the Mac I do not want to limit the ability of my userland application to talk to my NKE. " u_int32t unit = 0; int fd; int result = 1; Struct sockaddr_ctl addr: Bzero(&addr, sizeof(addr)); addr.sc_len = sizeof(addr); addr.sc_Family = AF_SYSTEM; addr.ss_sysaddr = AF_SYS_CONTROL; addr.sc_id = AFMAC_KERN_ID; // unique registered creator ID addr.sc_unit = unit; if(fd) { result = connect(fd, (struct sockaddr*)&addr, sizeof(struct sockaddr_ctl)); My socket routine works fine but I am failing on the connect with errno = 1 or EPERM. My understanding is the with the unit value set in both the NKE and the client that this is setting the ownership to root. You said there are various mechanisms for the client to communicate with the NKE, maybe I am not using the correct mechanism or I am setting some value incorrectly, do you see anything in error? I was basically following the example in the "About Network Kernel Extensions" material. This email sent to mtjaffa(a)msn.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: Change to root uid
by Matthew Jaffa 15 Feb '05

15 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com _______________________________________________ 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 <x-tad-bigger>AFctl_reg.ctl_flags = 0; </x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>Thanks for the reply Peter. Yes I guess I was too general in my description.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>What I am trying to accomplish is to have my NKE sending packet statistic to a userland/client application. Seeing as how I want this userland application to be assessable by who ever signs on to the Mac I do not want to limit the ability of my userland application to talk to my NKE.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>I register my NKE as following:</x-tad-bigger> <x-tad-bigger>"</x-tad-bigger> <x-tad-bigger>struct kern_ctl_reg AFctl_reg;</x-tad-bigger> <x-tad-bigger>bzero(&AFctl_reg, sizeof(AFctl_reg));</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>AFctl_reg.ctl_id = AFMAC_KERN_ID;</x-tad-bigger> <x-tad-bigger>AFctl_reg.ctl_unit = 0; </x-tad-bigger> <x-tad-bigger> AFctl_reg.ctl_flags = CTL_FLAG_PRIVILEGED;</x-tad-bigger> <x-tad-bigger>AFctl_reg.ctl_sendsize = 0;</x-tad-bigger> <x-tad-bigger>AFctl_reg.ctl_recvsize = 0;</x-tad-bigger> <x-tad-bigger>AFctl_reg.ctl_connect = AFKernConnect;</x-tad-bigger> <x-tad-bigger> AFctl_reg.ctl_disconnect = AFKernDisconnect; </x-tad-bigger> <x-tad-bigger> AFctl_reg.ctl_write = AFClientKernXWrite; </x-tad-bigger> <x-tad-bigger> AFctl_reg.ctl_set = NULL; </x-tad-bigger> <x-tad-bigger> AFctl_reg.ctl_get = AFKernGet; </x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger> static kern_ctl_ref ctlref = 0; /* Reference of the kernel controller */ </x-tad-bigger> <x-tad-bigger> nReturn = ctl_register(&AFctl_reg, 0, &ctlref);</x-tad-bigger> <x-tad-bigger> "</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>Then in user land I make the connection to the NKE as follows:</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>"</x-tad-bigger> <x-tad-bigger>u_int32t unit = 0;</x-tad-bigger> <x-tad-bigger>int fd;</x-tad-bigger> <x-tad-bigger>int result = 1;</x-tad-bigger> <x-tad-bigger>Struct sockaddr_ctl addr:</x-tad-bigger> <x-tad-bigger>Bzero(&addr, sizeof(addr));</x-tad-bigger> <x-tad-bigger>addr.sc_len = sizeof(addr);</x-tad-bigger> <x-tad-bigger>addr.sc_Family = AF_SYSTEM;</x-tad-bigger> <x-tad-bigger>addr.ss_sysaddr = AF_SYS_CONTROL;</x-tad-bigger> <x-tad-bigger>addr.sc_id = AFMAC_KERN_ID; // unique registered creator ID addr.sc_unit = unit;</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>fd = socket(PF_SYSTEM, SOCK_DGRAM, SYSPROTO_CONTROL);</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>if(fd)</x-tad-bigger> <x-tad-bigger>{</x-tad-bigger> <x-tad-bigger> result = connect(fd, (struct sockaddr*)&addr, sizeof(struct sockaddr_ctl));</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>... and so on</x-tad-bigger> <x-tad-bigger>}"</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>My socket routine works fine but I am failing on the connect with errno = 1 or EPERM.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>My understanding is the with the unit value set in both the NKE and the client that this is setting the ownership to root.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>You said there are various mechanisms for the client to communicate with the NKE, maybe I am not using the correct mechanism or I am setting some value incorrectly, do you see anything in error?</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>I was basically following the example in the "About Network Kernel Extensions" material.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>Thanks</x-tad-bigger> <x-tad-bigger>Carl</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>-----Original Message-----</x-tad-bigger> <x-tad-bigger>From: Peter</x-tad-bigger> <x-tad-bigger>Sent: Monday, February 14, 2005 7:17 PM</x-tad-bigger> <x-tad-bigger>To: Carl</x-tad-bigger> <x-tad-bigger>Cc: darwin-kernel(a)lists.apple.com</x-tad-bigger> <x-tad-bigger>Subject: Re: Change to root uid</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>On Feb 14, 2005, at 6:35 PM, Carl Smith wrote:</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>> I have a NKE and a client that I to talk to each other. If the client</x-tad-bigger> <x-tad-bigger>> app is run at some user’s, that might not be root, I still want my</x-tad-bigger> <x-tad-bigger> > client/NKE to talk with each other, but it is my understanding that</x-tad-bigger> <x-tad-bigger> > NKE’s need to be set with root user options only. In this case I need</x-tad-bigger> <x-tad-bigger> > to set the uid to root within my client application.</x-tad-bigger> <x-tad-bigger>></x-tad-bigger> <x-tad-bigger>> Is this possible and if so could you point me in the right direction</x-tad-bigger> <x-tad-bigger>> as to what api’s I need to be using?</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>The ownership/permissions of the nke need to be root/0755 for security</x-tad-bigger> <x-tad-bigger> reasons but that doesn't mean that it "runs" as root. It runs in the</x-tad-bigger> <x-tad-bigger> kernel and any user can talk to it, using various mechanisms. It's up</x-tad-bigger> <x-tad-bigger> to you to make sure that the uid and/or application which talk to your</x-tad-bigger> <x-tad-bigger> nke are suitable and appropriate for what you want to do.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>Your client application does not need to be suid root, nor should you</x-tad-bigger> <x-tad-bigger> (for usual activity) restrict your nke to a connection which can only</x-tad-bigger> <x-tad-bigger> be made by someone logged in as root.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>You haven't given us enough information to provide much more help than</x-tad-bigger> <x-tad-bigger> this. The best thing would be to describe a bit more of what you're</x-tad-bigger> <x-tad-bigger> trying to do.</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>Regards.....Peter</x-tad-bigger> <x-tad-bigger> </x-tad-bigger> <x-tad-bigger>p.s. remember also that nke interfaces are subject to change, and Apple</x-tad-bigger> <x-tad-bigger> presently discourages some of this development precisely because of the</x-tad-bigger> <x-tad-bigger> version-to-version compatibility issues</x-tad-bigger>

1 0

Change to root uid
by Carl Smith 15 Feb '05

15 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcUTckRMz44jcYqOTYi+S7Ds7OL3Rg== Thread-topic: Change to root uid _______________________________________________ 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: Change to root uid
by Peter Lovell 15 Feb '05

15 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Feb 14, 2005, at 6:35 PM, Carl Smith wrote: 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.… I have a NKE and a client that I to talk to each other. If the client app is run at some user’s, that might not be root, I still want my client/NKE to talk with each other, but it is my understanding that NKE’s need to be set with root user options only. In this case I need to set the uid to root within my client application. Is this possible and if so could you point me in the right direction as to what api’s I need to be using? The ownership/permissions of the nke need to be root/0755 for security reasons but that doesn't mean that it "runs" as root. It runs in the kernel and any user can talk to it, using various mechanisms. It's up to you to make sure that the uid and/or application which talk to your nke are suitable and appropriate for what you want to do. Your client application does not need to be suid root, nor should you (for usual activity) restrict your nke to a connection which can only be made by someone logged in as root. You haven't given us enough information to provide much more help than this. The best thing would be to describe a bit more of what you're trying to do. p.s. remember also that nke interfaces are subject to change, and Apple presently discourages some of this development precisely because of the version-to-version compatibility issues This email sent to site_archiver(a)lists.apple.com

1 0

Change to root uid
by Carl Smith 14 Feb '05

14 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcUS7eZMnftK9lmSTGSeIkbczIrJbA== Thread-topic: Change to root uid _______________________________________________ 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: xnu-517.11.1 won't compile.
by Kevin Van Vechten 14 Feb '05

14 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com - Kevin On Feb 13, 2005, at 11:17 PM, Kevin Van Vechten wrote: - Kevin On Feb 10, 2005, at 3:26 PM, Josh Ellithorpe wrote: Anyone gotten this to work... or have a hint for me =) Josh Ellithorpe quest(a)mac.com _______________________________________________ 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.… A new tarball been posted both at Apple's site and at OpenDarwin. I verified it builds OK on my system. Ben Herrenschmidt correctly diagnosed the problem as an incomplete version of osfmk/ppc/machine_routines_asm.s. This is indeed a problem with the source tarball. The osfmk/ppc/machine_routines_asm.s source file is incomplete. I'll try to post a fix with the complete version of this file tomorrow. I have compiled every version of xnu since 10.3 came out however xnu-517.11.1 just won't build for me. All my dependencies are up to date however I get the following error. If someone can help me with this I would really apperciate it. /usr/bin/strip -S -x /usr/local/src/xnu-517.11.1/BUILD/obj/RELEASE_PPC/./mach_kernel.sys -o /usr/local/src/xnu-517.11.1/BUILD/obj/RELEASE_PPC/./mach_kernel; /usr/bin/ld: Undefined symbols: _cacheDisable _cacheInit _clock_get_uptime _cpu_number _mach_absolute_time _machine_idle_ret _ml_get_timebase _ml_read_temp _ml_scom_read _ml_scom_write _ml_set_processor_speed_dfs _ml_set_processor_speed_dpll _ml_set_processor_speed_powertune _ml_thrm_init _set_machine_current_act _current_act _current_thread make[2]: *** [do_build_mach_kernel] Error 1 make[1]: *** [build_all] Error 2 make: *** [all] Error 2 I even grabbed xnu-517.9.5 from the 10.3.7 repository and that built fine... the only dependency that I saw changed between 10.3.7 and 10.3.8 was IOKitUser-184 which I already installed. _______________________________________________ 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/kevin% 40opendarwin.org This email sent to kevin(a)opendarwin.org This email sent to site_archiver(a)lists.apple.com

1 0

NKE failing on ctl_register call
by Carl Smith 14 Feb '05

14 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcUStpMROgShRqBNRSaPMy/ANtOYag== Thread-topic: NKE failing on ctl_register call _______________________________________________ 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: Cannot mprotect() dylib code?
by Lassi A.Tuura 14 Feb '05

14 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, Lassi -- Treat a person as he is, and he will remain as he is. Treat him as he could be, and he will become what he should be. --Jimmy Johnson _______________________________________________ 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.… Thanks for the clear reply. The DYLD_IMAGE_SUFFIX works on libraries that have _debug version, but not on others. At least it gets me to hook into libSystem. DYLD_NEW_LOCAL_SHARED_REGIONS didn't work, at least not with mprotect(), didn't yet try vm_protect(). On VM_PROT_COPY sounds exactly what I need. I'll try out something, but just to clarify my understanding: Another thing you can try is setting VM_PROT_COPY in your call to vm_protect. This forces the system to make a copy-on-write copy of that section of the shared region, which you can then manipulate. In this case, when you call vm_protect, don't specify VM_PROT_EXECUTE. The max protection on the shared region submap is VM_PROT_READ, so specify VM_PROT_EXECUTE in your vm_protect call triggers an error, even if you also specify VM_PROT_COPY. I am not sure I followed this entirely, but with a few trials this is how I understood it. I start with page protected with VM_PROT_READ | VM_PROT_EXECUTE (from dylib). To modify that page, I start with VM_PROT_READ | VM_PROT_COPY, which automatically makes it writable, then modify the page, and then set protections back to VM_PROT_READ | VM_PROT_EXECUTE. I have had partial success with this, but some things are not right -- I get further, but not as far as I would like to. I'll get back on this when I have had a chance to debug things further. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Cannot mprotect() dylib code?
by Quinn 14 Feb '05

14 Feb '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com 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 23:32 +0100 11/2/05, Lassi A.Tuura wrote: We have a debugging type library that parses and modifies the machine code on the fly for certain functions, mainly to insert a jump to "capture" functions which do some special stuff then invoke the original code. The run-time patching appears to work on functions from the main executable, but not on functions from dylibs -- mprotect() on a page of code from the latter always fails, errno is set to 13 (EACCESS, Permission denied). I also tried vm_protect(mach_task_self(), ...), it also fails. I suspect that this fails because shared library code is all stored in a system-wide shared map (0x90000000..0x9FFFFFFF). You can't change the protection on these areas because it would affect all users of the shared map (that is, everyone). One trivial way to work around this is to use _debug variant of the system libraries (by setting the DYLD_IMAGE_SUFFIX environmont variable to _debug). The debug libraries don't load into the shared map, and thus their protection can be modified by you. Another thing to explore might be DYLD_NEW_LOCAL_SHARED_REGIONS. This forces your process (and any of its children) to pick up its own shared region, rather than sharing the system one. I don't know if you'll be able to modify the protection of this shared region, but it's worth a try. Another thing you can try is setting VM_PROT_COPY in your call to vm_protect. This forces the system to make a copy-on-write copy of that section of the shared region, which you can then manipulate. In this case, when you call vm_protect, don't specify VM_PROT_EXECUTE. The max protection on the shared region submap is VM_PROT_READ, so specify VM_PROT_EXECUTE in your vm_protect call triggers an error, even if you also specify VM_PROT_COPY. This email sent to site_archiver(a)lists.apple.com

1 0