February 2003 - Darwin-kernel

Re: Mach RPC
by David Blanton 26 Feb '03

26 Feb '03

On 2/25/03 5:13 PM, "Marek Kozubal" <marek(a)portents.com> wrote: > Look at: > http://www-2.cs.cmu.edu/afs/cs/project/mach/public/www/doc/tutorials.html#p… > mig > > Its the reference I used in doing my MIG RPC work. > > And your use of header files is wrong I believe. You never want to add > header files to /usr/include/sys (or > /System/Library/Frameworks/Kernel.Framekwork/Headers/sys either). > > You make a .defs file which describes your function calls and then the > "migger" converts them to C code for your client and server portions, > write some incantations that links it all together into your RunLoop and > off you go. > > --- > Marek Kozubal > marek(a)portents.com > _______________________________________________ Thanks for the reference Marek. I do have a Mach Server and Client running. I choose random as the subsystem as it was used in a CMU example. I will look for more on subsystems in the above reference. Peace. David Blanton Elisar Software Corp _______________________________________________ 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: How to block write access to a file.
by Marek Kozubal 26 Feb '03

26 Feb '03

On Tue, 25 Feb 2003, Mr. Wong wrote: > Can someone give me some pointers on this? What should I be using? Has > anyone done this before? I am aware that there will be inconsistancies and > errors introduced but I'm willing to tolerate / fix / learn about the issues > involved. There is a way to do this. What you want to do is get the source for the nullfs filesystem and modify it to cover the mount vnode instead of mapping a vnode from another part of the file system. Then *all* file system accesses go though your layer before hitting the underlying HFS or whatever file system. You can add code to VOP_WRITE, VOP_ACCESS, VOP_RMDIR, VOP_CREATE, etc to prevent writing to the underlying file system. nullfs already does this if the mount flags say the nullfs mount layer is a read-only file system. You should read up about the VFS stack and stacked file systems to undertsand whats going on. Also look at the nullfs code. If you're serious about doing this, I would highly recommend you setup a 2 computer Kernel debug setup as well so you can figure out the ref-count and locking kernel panics you'll probably see during development. I've also been using the URL http://www.gsp.com/support/man/ which has man page entries for the various VOP/VFS functions, even though its FreeBSD oriented the ideas still work. Also the file vnode_if.src in the open source tree (in xnu i think) is very helpful in the locking protocols. All of this is of course not offically supported by Apple as they are not promising not to break this in a future release. --- Marek Kozubal marek(a)portents.com _______________________________________________ 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: Mach RPC
by Marek Kozubal 26 Feb '03

26 Feb '03

Look at: http://www-2.cs.cmu.edu/afs/cs/project/mach/public/www/doc/tutorials.html#p… Its the reference I used in doing my MIG RPC work. And your use of header files is wrong I believe. You never want to add header files to /usr/include/sys (or /System/Library/Frameworks/Kernel.Framekwork/Headers/sys either). You make a .defs file which describes your function calls and then the "migger" converts them to C code for your client and server portions, write some incantations that links it all together into your RunLoop and off you go. --- Marek Kozubal marek(a)portents.com _______________________________________________ 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

How to block write access to a file.
by Mr. Wong 25 Feb '03

25 Feb '03

Hello, I'm interested in writing a Kernel Extension that allows me to block write access to files. Essentially I want to be able to block ALL users, including ROOT from being able to write to files from within applications such as MS WORD, or any of the web browsers. I'd like to be able to do this in kernel level meaning that its not just a permission but its more like impossble. Can someone give me some pointers on this? What should I be using? Has anyone done this before? I am aware that there will be inconsistancies and errors introduced but I'm willing to tolerate / fix / learn about the issues involved. --Matthew K. Wong _________________________________________________________________ Add photos to your e-mail with MSN 8. Get 2 months FREE*. http://join.msn.com/?page=features/featuredemail _______________________________________________ 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

Maximum number of user processes
by William Kucharski 25 Feb '03

25 Feb '03

>From: "Mr. Wong" <keilinw(a)hotmail.com> >Subject: 1. Maximum number of user processes (William Kucharski) > >Hello, > >I'm trying to build custom kernels and have succeeded in making some >changes. The recent email about the Maximum number of users processes >intrigues me. I can recompile the kernel and change the max number of user >processes. But, is there any way to determine if my changes actually took >effect? How about determining if there are new bugs introduced? > >--Matthew K. Wong The easiest way is to check using the getrlimit() system call along with RLIMIT_NPROC, or via the "ulimit -u" built-in in bash or csh. William Kucharski kucharsk(a)mac.com _______________________________________________ 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

1. Maximum number of user processes (William Kucharski)
by Mr. Wong 24 Feb '03

24 Feb '03

Hello, I'm trying to build custom kernels and have succeeded in making some changes. The recent email about the Maximum number of users processes intrigues me. I can recompile the kernel and change the max number of user processes. But, is there any way to determine if my changes actually took effect? How about determining if there are new bugs introduced? --Matthew K. Wong Message: 1 Date: Sat, 22 Feb 2003 00:09:54 -0700 From: "William Kucharski" <kucharsk(a)mac.com> To: "Cameron Knowlton" <cameronk(a)igods.com> Cc: darwin-kernel(a)lists.apple.com, macos-x-server(a)lists.apple.com Subject: Maximum number of user processes For root, you can indeed set the maximum number of processes up to NPROC. Unfortunately, the number of non-root user processes is hard coded to be MAXUPRC, which is #defined in xnu/bsd/sys/param.h to be CHILD_MAX, which is currently 100 (see xnu/bsd/sys/syslimits.h.) The initialization of limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur to MAXUPRC can be found in xnu/bsd/kern/bsd_init.c, and the limit to CHILD_MAX can be found in the dosetrlimit() code in xnu/bsd/kerne_resource.c. So in short, yes, if you want to increase the maximum number of normal user processes, you'll need a custom kernel. William Kucharski kucharsk(a)mac.com --__--__-- _______________________________________________ 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. End of darwin-kernel Digest _________________________________________________________________ Help STOP SPAM with the new MSN 8 and get 2 months FREE* http://join.msn.com/?page=features/junkmail _______________________________________________ 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 RPC
by David Blanton 24 Feb '03

24 Feb '03

In the statement subsystem random 500 I understand that there is a file, random.h found in /usr/include/sys/ random.h is referenced in the code implementing a Mach server. If this is true can I add foo,h to /usr/include/sys/ ------ foo.h #ifndef __SYS_FOO_H__ #define __SYS_FOO_H__ #endif /* __SYS_FOO_H__ */ ------ And then write subsystem foo 101 And reference foo.h in the code implementing a Mach server ? _______________________________________________ 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

Maximum number of user processes
by William Kucharski 22 Feb '03

22 Feb '03

For root, you can indeed set the maximum number of processes up to NPROC. Unfortunately, the number of non-root user processes is hard coded to be MAXUPRC, which is #defined in xnu/bsd/sys/param.h to be CHILD_MAX, which is currently 100 (see xnu/bsd/sys/syslimits.h.) The initialization of limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur to MAXUPRC can be found in xnu/bsd/kern/bsd_init.c, and the limit to CHILD_MAX can be found in the dosetrlimit() code in xnu/bsd/kerne_resource.c. So in short, yes, if you want to increase the maximum number of normal user processes, you'll need a custom kernel. William Kucharski kucharsk(a)mac.com _______________________________________________ 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: NPROC / MAXUPRC
by Brian Tabone 22 Feb '03

22 Feb '03

You can set the maximum number of processes on OS X by using sysctl For example sudo sysctl -w kern.maxproc=1000 Do a man on sysctl for more info. As for Max number of proc by user, I'm not sure how that would be done under OS X. Hope that helps, Brian Tabone On Friday, February 21, 2003, at 01:22 PM, Cameron Knowlton wrote: Can anyone tell me how I'd change the NPROC and/or the MAXUPRC kernel parameters under OS X? I'm running out of processes, apparently. I am running a lot of statistical applications on the system, of course they'll run out of resources eventually. I was hoping to get further than I have, though. The story so far... ideas? thanks in advance, gurus. Cameron Knowlton cameronk(a)igods.com ================= This is usually one of two things. There is a tunable kernel parameter, NPROC, which determines the maximum number of processes that may be running at any time. You may have exceeded this limit. The usual method of solving this is to increase it a fair bit and see if the problem goes away. If you are running on OSR5, this is unlikely to be the reason, as NPROC is allocated dynamically. There is another tunable kernel parameter, MAXUPRC, which determines the maximum number of processes any one user may have running at one time. Under Unix, for example, a large number of mail messages being processed at once may cause this to be exceeded by MMDF, usually resulting in "uux failure - pipe broke" or similar messages. Once again, increase it and see if the problem goes away. Also, have a look at the console and/or /usr/adm/messages for any system messages which appeared at the same time the user got this message. They may point to another potential reason, such as being out of swap space or exceeding NREGION (see the previous topic). ------ I didn't even know those were available - sysctl -a doesn't list 'em - if it did, it could set them. On my OS X: bash-2.05a$ sysctl -a | grep -i proc kern.maxproc = 532 On a Free BSD system: bash-2.05a$ sysctl -a | grep -i proc kern.maxproc: 8212 kern.maxfilesperproc: 16424 kern.maxprocperuid: 8211 kern.ps_showallprocs: 1 kern.shutdown.kproc_shutdown_wait: 60 vm.max_proc_mmap: 43690 PROC: 416, 0, 268, 173, 182379 debug.sizeof.proc: 412 jail.limits.rlimit_nproc.hard: 8211 jail.limits.rlimit_nproc.current: 17 Tony Lawrence ------ Well, NPROC is defined in xnu/bsd/conf/param.c in the darwin sources as (20 + 16 * MAXUSERS), but I'm not sure where MAXUSERS comes from. NPROC is a compile-time parameter so to change it you'd have to build your own kernel. That's possible, but not for the feint of heart, and you'd run into maintenance headaches like crazy. It's probably a better strategy to figure out a way to have fewer processes running on your system at once. It's a sure bet that they're not all actually running at once (otherwise your system would be unusable), so perhaps you can structure things so that unused processes are terminated until they're needed. It's a pretty unusual problem. I've never heard of it before. Let me know if you find a solution. Cheers, Brent ------ -- thanks 8^) Cameron Knowlton cameronk(a)igods.com _______________________________________________ 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

NPROC / MAXUPRC
by Cameron Knowlton 21 Feb '03

21 Feb '03

Can anyone tell me how I'd change the NPROC and/or the MAXUPRC kernel parameters under OS X? I'm running out of processes, apparently. I am running a lot of statistical applications on the system, of course they'll run out of resources eventually. I was hoping to get further than I have, though. The story so far... ideas? thanks in advance, gurus. Cameron Knowlton cameronk(a)igods.com ================= This is usually one of two things. There is a tunable kernel parameter, NPROC, which determines the maximum number of processes that may be running at any time. You may have exceeded this limit. The usual method of solving this is to increase it a fair bit and see if the problem goes away. If you are running on OSR5, this is unlikely to be the reason, as NPROC is allocated dynamically. There is another tunable kernel parameter, MAXUPRC, which determines the maximum number of processes any one user may have running at one time. Under Unix, for example, a large number of mail messages being processed at once may cause this to be exceeded by MMDF, usually resulting in "uux failure - pipe broke" or similar messages. Once again, increase it and see if the problem goes away. Also, have a look at the console and/or /usr/adm/messages for any system messages which appeared at the same time the user got this message. They may point to another potential reason, such as being out of swap space or exceeding NREGION (see the previous topic). ------ I didn't even know those were available - sysctl -a doesn't list 'em - if it did, it could set them. On my OS X: bash-2.05a$ sysctl -a | grep -i proc kern.maxproc = 532 On a Free BSD system: bash-2.05a$ sysctl -a | grep -i proc kern.maxproc: 8212 kern.maxfilesperproc: 16424 kern.maxprocperuid: 8211 kern.ps_showallprocs: 1 kern.shutdown.kproc_shutdown_wait: 60 vm.max_proc_mmap: 43690 PROC: 416, 0, 268, 173, 182379 debug.sizeof.proc: 412 jail.limits.rlimit_nproc.hard: 8211 jail.limits.rlimit_nproc.current: 17 Tony Lawrence ------ Well, NPROC is defined in xnu/bsd/conf/param.c in the darwin sources as (20 + 16 * MAXUSERS), but I'm not sure where MAXUSERS comes from. NPROC is a compile-time parameter so to change it you'd have to build your own kernel. That's possible, but not for the feint of heart, and you'd run into maintenance headaches like crazy. It's probably a better strategy to figure out a way to have fewer processes running on your system at once. It's a sure bet that they're not all actually running at once (otherwise your system would be unusable), so perhaps you can structure things so that unused processes are terminated until they're needed. It's a pretty unusual problem. I've never heard of it before. Let me know if you find a solution. Cheers, Brent ------ -- thanks 8^) Cameron Knowlton cameronk(a)igods.com _______________________________________________ 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