May 2007 - Darwin-dev - lists.apple.com

RE: kext - calling ctl_register - from cpp - Undefined symbols
by Erez Kaplan 31 May '07

31 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: Acei1L/EdVwLX3pOQS+cqMDAGLRppAAeNuBO Thread-topic: kext - calling ctl_register - from cpp - Undefined symbols _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: kext - calling ctl_register - from cpp - Undefined symbols
by Josh Graessley 30 May '07

30 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

kext - calling ctl_register - from cpp - Undefined symbols
by Erez Kaplan 30 May '07

30 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: Aceilxwn56B0yvneRtySunjCgT6N0w== Thread-topic: kext - calling ctl_register - from cpp - Undefined symbols _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Half-closing bidirectional pipes
by Dave Zarzycki 29 May '07

29 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com davez On May 29, 2007, at 12:45 PM, Tomasz Kukielka wrote: Hello, TIA, Tom _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/zarzycki%40apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Pipes, are by definition, uni-directional. You should not rely on any implementation (Mac OS X or otherwise) that uses bi-directional pipes. In fact, pipes in the next version of Mac OS X are uni-directional. If you want bi-directional behavior, please use a local domain socket (also known as Unix domain socket). Then you can use the shutdown() API to close one direction of the socket. It is not very clear to me but it looks like pipes in Mac OS 10.4 are bidirectional. If yes, how can I half-close such pipe, sending EOF when I finish writing but keep it open for reading? What am I trying to do? popen(cmd, "r+") of course. So I want to send some data to standard input of an executed tool, close the writing part and read the result. I cannot just close everything when I finish writing because then I will not get the result in standard output. And if I don't send the EOF when writing, some tools will not start processing it (e.g. grep). This email sent to zarzycki(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Half-closing bidirectional pipes
by Tomasz Kukielka 29 May '07

29 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Thread-index: AceiKfa4NVuS7g4dEdysTgAX8jPHcQ== Thread-topic: Half-closing bidirectional pipes User-agent: Microsoft-Entourage/11.3.3.061214 Hello, It is not very clear to me but it looks like pipes in Mac OS 10.4 are bidirectional. If yes, how can I half-close such pipe, sending EOF when I finish writing but keep it open for reading? What am I trying to do? popen(cmd, "r+") of course. So I want to send some data to standard input of an executed tool, close the writing part and read the result. I cannot just close everything when I finish writing because then I will not get the result in standard output. And if I don't send the EOF when writing, some tools will not start processing it (e.g. grep). TIA, Tom _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: shared libraries and undefined environ
by Jack Howarth 29 May '07

29 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.4.2.2i Peter, This is exactly the sort of problem fink developers using openmpi have been running into... ----------------------------------------------------------------------- I tried several patches to the paraview build scripts in order to get rid of this undefined symbol, but I was not successful, one reason being that the build process uses cmake which I find much less transparent than the usual configure/libtool combination. While looking through the build script, I see now that in fact there is a difference between the linker command for this library libicet_mpi.dylib and many other dylibs built in paraview, for example libvtkParallel.pv2.6.dylib. Those use an explicit -Wl,-flat_namespace,-U,_environ linker flag. Could it be that this is there in order to avoid exactly this problem? I don't know. I'll look some more where this comes from. Here the error: cd /sw_unstable/src/fink.build/paraview-mpi-openmpi-2.6.1-1001/paraview-mpi-openmpi-darwin/Utilities/IceT/src/communication && /sw/bin/cmake -E cmake_link_script CMakeFiles/icet_mpi.dir/link.txt --verbose=1 /sw/bin/gcc -O3 -DNDEBUG -dynamiclib -headerpad_max_install_names -Wl,-search_paths_first -o ../../../../bin/libicet_mpi.dylib -install_name /sw/lib/paraview-2.6/libicet_mpi.dylib "CMakeFiles/icet_mpi.dir/mpi.o" -L/sw_unstable/src/fink.build/paraview-mpi-openmpi-2.6.1-1001/paraview-mpi-openmpi-darwin/bin -L/usr/X11R6/lib -lGLU -L/usr/X11R6/lib -lGL -L/sw/lib/openmpi -lmpi -lmpi_cxx -licet -L/usr/X11R6/lib -lGLU -L/usr/X11R6/lib -lGL -L/sw/lib/openmpi -lmpi -lmpi_cxx ld: Undefined symbols: _environ /usr/bin/libtool: internal link edit command failed make[2]: *** [bin/libicet_mpi.dylib] Error 1 ----------------------------------------------------------------------------- So the question becomes which is worse, forcing the openmpi developers to not leave undefined environ variables in their shared libraries or off loading the complexities of linking shared libraries using those with the undefined symbols. My view is, since openmpi is already wrappering those declarations of environ for _WINDOWS, that expanding that to __APPLE__ is trivial. In the process of creating such a patch I discovered that they weren't catching all of the declarations of environ for Windows anyway. Jack _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

shmget() and shmat() problem
by Ritesh Nadhani 29 May '07

29 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: a=rsa-sha1; c=relaxed/relaxed; d=gmail.com; s=beta; h=domainkey-signature:received:received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=BV3TsX178jznhzprghrgfJNtaVyIYv8eIe6yq2SUInjJaXx+UDwU3BiiJMFiiWOAWFI0H26So2ez0QjRfzUvIXmnx37eyE1gspNQiAkn1dhqanS3i2ymu4nauSBm0Q5Kf8jLpXJueQbfORZtn20MHN1AukXKFFtq26axz9oxLZ8= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=beta; h=received:message-id:date:from:to:subject:mime-version:content-type:content-transfer-encoding:content-disposition; b=OX+enGBcp0nRbG4bAelQleE9I+UFRSWFer4kgLx65k/ipztmls6R6DZ2QHp1z6NPSUg2iaK7JMi8J2vW4kHBes8P8Cih+Kd1q0f3xjpq1YgNSbl61/zYtxF58QYkFSOaF4PVhDi3bhnSfWYAZ4/GXJ8InAu09hO2NExSVfwsmfY= Hello I hope this is the correct list for my question. If not please let me know which one would be more appropriate for me. I am trying to compile http://www.dbmail.org/ on my MacBook. I have successfuly compiled it on FreeBSD and Ubuntu so it should work on Mac OS X too. One of the file has the following code: void scoreboard_new(serverConfig_t * conf) { int serr; if ((shmid = shmget(IPC_PRIVATE, (sizeof(child_state_t) * HARD_MAX_CHILDREN), 0644 | IPC_CREAT)) == -1) { serr = errno; TRACE(TRACE_FATAL, "shmget failed [%s]", strerror(serr)); } scoreboard = shmat(shmid, (void *) 0, 0); serr=errno; if (scoreboard == (Scoreboard_t *) (-1)) { TRACE(TRACE_FATAL, "scoreboard init failed [%s]", strerror(serr)); scoreboard_delete(); } scoreboard_lock_new(); scoreboard->conf = conf; scoreboard_setup(); scoreboard_conf_check(); /* Make sure that we clean up our shared memory segments when we exit * normally (i.e. not by kill -9, if you do that, you get to clean this * up yourself!) * */ atexit(scoreboard_delete); } The code compiles correctly but shmget() and shmat() fails with error: shmget failed [No space left on device] scoreboard init failed [Permission denied] I tried to follow the steps given at http://www.spy-hill.net/help/apple/SharedMemory.html to solve the shmget() issue but without success. For the second issue, I found: http://developer.apple.com/technotes/tn2002/tn2053.html A problem where shmat would fail with a permission denied error when it was attempted by root has been corrected. (r. 2910913). but apparently it still fails on my Mac OS X Tiger even though I am running the app as *sudo*. Any ideas? -- Ritesh _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: shared libraries and undefined environ
by Peter O'Gorman 29 May '07

29 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Tue, 2007-05-29 at 09:08 -0400, Jack Howarth wrote: > Is there an Apple policy on the presence of undefined > symbols in shared libraries? Specifically, the libraries > created by openmpi currently create undefined symbols for > environ because the following wrapper is not being used... > > #if !defined(__WINDOWS__) > +# ifdef __APPLE__ > +# include <crt_externs.h> > +# define environ (*_NSGetEnviron()) > +# else > extern char **environ; > +# endif > #endif /* !defined(__WINDOWS__) */ > > The upstream openmpi developers are resisting this > change since it is target specific (which I find > amusing since they already wrapper these lines > for Windows). They claim this can be handled > by the application developer through including > the symbol there. My argument is that this is > bad form since certain cases can't be handled > so easily. For instance, if a python module were > developed that used openmpi's shared libraries > the system python would have to be replaced with > on that provided the environ symbol. > I would appreciate any advice on Apple tech > notes addressing this issue or other arguments > we can make to the openmpi developers for > accepting such patches. The environ symbol (as well as the other symbols that you see in crt_externs.h are available to every executable, they are in /usr/lib/crt1.o that is included by gcc in the link for every executable. Shared libraries, at link time, do not see crt1.o, so environ is marked as undefined, it will always be available at runtime. So, python, ruby and so on all have environ symbols in their executable, it is harmless to leave it as an undefined, however, with two level namespace, there is a startup speed advantage to having no undefined symbols in your shared libraries, if there is only one undefined symbol and the definition is available in the main executable image (as is the case with environ), then the speed advantage is quite likely not even measurable. Hope this helps, Peter _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

shared libraries and undefined environ
by Jack Howarth 29 May '07

29 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.4.2.2i Is there an Apple policy on the presence of undefined symbols in shared libraries? Specifically, the libraries created by openmpi currently create undefined symbols for environ because the following wrapper is not being used... #if !defined(__WINDOWS__) +# ifdef __APPLE__ +# include <crt_externs.h> +# define environ (*_NSGetEnviron()) +# else extern char **environ; +# endif #endif /* !defined(__WINDOWS__) */ The upstream openmpi developers are resisting this change since it is target specific (which I find amusing since they already wrapper these lines for Windows). They claim this can be handled by the application developer through including the symbol there. My argument is that this is bad form since certain cases can't be handled so easily. For instance, if a python module were developed that used openmpi's shared libraries the system python would have to be replaced with on that provided the environ symbol. I would appreciate any advice on Apple tech notes addressing this issue or other arguments we can make to the openmpi developers for accepting such patches. Jack _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Passing parameters to a kext on load
by Nathan Toone 28 May '07

28 May '07

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -Nathan On May 25, 2007, at 11:34 PM, Jim Guyton wrote: Your dictionary is passed in with the init() call, then you can do things like I suspect most of the drivers in the darwin source have some set of parameters they look for. Cheers, -- Jim On May 25, 2007, at 9:54 PM, Nathan Toone wrote: That's exactly what I was wondering! Perfect! -Nathan On May 25, 2007, at 9:15 PM, Jim Guyton wrote: The usual way to pass info to a driver on load is via the dictionary stored in the Info.plist file. The kext loader will copy the data items to the registry and the driver can read it from there. For debugging you can look at the registry with ioreg and/or the developer tools applicatio IORegistryExplorer. -- Jim Guyton On May 25, 2007, at 8:04 PM, Nathan Toone wrote: Thanks! _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/guyton%40apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com Will this work as well for an NKE? Most of where I see documentation for this is for IO kit drivers... I know that using the kernel control mechanisms are the "approved" way of getting parameters to an NKE, but I'd like to set a couple of parameters which can be used in a failure condition - when the userspace application is *not* available. const OSNumber *numObject = OSDynamicCast(OSNumber, dict->getObject ("parameterName")); if (numObject) value = numObject->unsigned32BitValue(); Does anyone have sample code which reads from the registry? (I'm pretty new at kernel development...) I have noticed that kernel modules (I'm specifically working on a socket filter NKE) have parameters passed to their MODULE_START and MODULE_STOP functions. a kmod_info_t * structure and a void *. What do these parameters represent? Is there a way for me to specify data that gets passed to my module on load? If so, how do I set that up? Pointers to documentation would be appreciated as well. I wasn't able to find what I was looking for. -Nathan _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/guyton% 40apple.com This email sent to guyton(a)apple.com This email sent to guyton(a)apple.com smime.p7s

1 0