September 2004 - Darwin-dev

Re: Who takes care of a crashing process?
by Justin Walker 22 Sep '04

22 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Sep 22, 2004, at 8:13, Stephane Sudre wrote: Stupid question probably: When a process is properly exited, exit is called in the Kernel. When a process crashes, exit is not called. 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-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… Is there a specific syscall called when a process crash? Are you sure that exit() in the kernel is not called? If this were not the case, we'd be leaking file entries, memory, and other resources. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Mach RPC and SecToken
by Jim Magee 22 Sep '04

22 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com --Jim _______________________________________________ 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… On Sep 16, 2004, at 4:17 AM, Ryan McGann wrote: Does anybody out there know what SecToken in MIG definition files are used for, and where I can find out more about them? I have some MIG definitions as part of some sample code that use SecToken and AuditToken parameters, but then the source does nothing with them (except the MIG generated source). I'm curious as to how they would be used in "the real world". Requesting the Mach security token be appended to a Mach message trailer, and then exposing it through MIG RPC interfaces by means of these keywords, is the Mach equivalent of calling the local-socket getpeereid(3) interface - but on a message-by-message basis. That is, a structure containing the effective uid, and effective gid of the sending process, is filled in by the kernel in a "trusted path" (unspoofable) way for each message that arrives. MIG determines the presence of this structure in the trailer, and provides it as a "magic" parameter to the RPC call. The SecToken keyword is the "symmetrical" option. Both the client and the server see the security token of the other. Other options are asymmetrical - and provide the token only on one side of the interface or the other (prototypes for client and server don't match in these cases). As we started down the path of providing a full-featured auditing system to Darwin/Mac OS X (a work still in progress), we realized that we needed additional information be added to the trailers in order to cut a meaningful audit record during such an RPC. Because of the limited size of the data available in the existing security token structure - and because people have discerned the meaning of the fields in the opaque structure definition over time - we needed to add a new (expanded) token to contain the audit information (rather than alter the old). Just as originally intended for the security token, the structure is meant to be opaque. But this time we are adding library routines to the audit library in order to access the contents from the very beginning. Of course, neither are finalized yet until the audit mechanism "goes live." So, you really can't meaningfully use this mechanism yet. As identities become even more complex (e.g. ACL system already announced for Tiger), we will likely be providing even more options for managing identities through Mach IPC. Future options will likely return counted references to full credentials - allowing much more robust authorizations to be requested. Stay tuned for that. PS: One trick: You have to make sure your runloop/serverloop provides the proper options to receive the Mach message trailers that contain these tokens. These options are identified in <mach/message.h>. This email sent to site_archiver(a)lists.apple.com

1 0

Who takes care of a crashing process?
by Stephane Sudre 22 Sep '04

22 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Stupid question probably: When a process is properly exited, exit is called in the Kernel. When a process crashes, exit is not called. Is there a specific syscall called when a process crash? _______________________________________________ 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

CPU reset GPIO & latest powerbooks
by Benjamin Herrenschmidt 22 Sep '04

22 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi ! I know it's an odd question but I would really appreciate if somebody from Apple could find an answer to it ;) So basically, earlier machine models used to have the CPU soft reset line connected to a KL GPIO (used on SMP boxes to bring CPUs up, but fortunately for me also wired on UP boxes). This was a very useful kernel debugging tool as I could use a firewire controller to remotely toggle the GPIO and thus get a 0x100 exception even when the CPU was locked up with interrupts disabled. I'm currently chasing a lockup of the latest albooks (in linux) but couldn't seem to trigger a 0x100 with any of those GPIOs (nor a couple of other random ones I tried). Is there a way to do the same thing at all on those machines models ? Thanks in advance for any reply, Regards, Ben. _______________________________________________ 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: diskutil createRAID question
by Soren Spies 21 Sep '04

21 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.4i Now that I have installed a second identical SATA drive I am interested in trying to get a multi-hfs RAID-1 mirror set up on -- Soren Spies DiskImages, etc; I/O Kit team Apple Computer, Inc _______________________________________________ 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… On Sat, Sep 18, 2004 at 02:09:37AM -0400, Jack Howarth wrote: The Help for Disk Utility is misleading. In Panther, only entire devices can be used as RAID members. This email sent to site_archiver(a)lists.apple.com

1 0

Re: diskutil createRAID question
by Brian Bechtel 20 Sep '04

20 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Mon, 20 Sep 2004 09:36:50 -0400 (EDT), Jack Howarth <howarth(a)bromo.msbb.uc.edu> wrote: > ps At the very least the manpage for diskutil should be updated > to reflect this situation since it currently leads the user to > believe that one can use the disk identifier form (disk0s3) > with diskutil anywhere it takes a device name (disk0). Jack, please file a bug with <http://bugreporter.apple.com>. That is the only way bugs get tracked and fixed at Apple. (You can get a free developer account if you don't already have one.) _______________________________________________ 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: diskutil createRAID question
by howarth＠bromo.msbb.uc.edu 20 Sep '04

20 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com To answer my own question, it appears to be currently impossible to use the device identifiers rather than the device itself to create a RAID. If one has two identical SATA drives (each partitioned with a single extended HFS partition) and boots off of a third drive or a boot CD, the following command will fail under MacOS X 10.3.5... diskutil createRAID mirror PantherMirror disk0s3 disk1s3 ...where s3 is the slice containing the extended HFS partition on each SATA drive. Diskutil fails with a message that "Error -9983 encountered updating or creating raid on disk". However the device form of this command succeeds... diskutil createRAID mirror PantherMirror disk0 disk1 I would note that the failure of the first form of the command using the device identifiers with the slice number fails in a way that mangles the partition tables on those two drives so you have to repartition them after the error in diskutil. The reason I mention all of this is that this bug rules out the possibility of having a set of mirrored 2-partition drives under the current MacOS X releases. Jack ps At the very least the manpage for diskutil should be updated to reflect this situation since it currently leads the user to believe that one can use the disk identifier form (disk0s3) with diskutil anywhere it takes a device name (disk0). _______________________________________________ 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: how to test atomicity?
by Sean McBride 19 Sep '04

19 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Matt Watson (mwatson(a)apple.com) on Sun, Sep 19, 2004 19:30 said: >Unfortunately, that's not available to user space. One could, however, >peek at the kernel sources to look at the implementation... For user space there are functions in DriverSynchronization.h and OpenTransport.h, both part of the CoreServices umbrella framework. But I guess that's not really part of Darwin, which may or may not matter to the OP. _______________________________________________ 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: how to test atomicity?
by Matt Watson 19 Sep '04

19 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com matt. On Sep 19, 2004, at 4:25 PM, Sean McBride wrote: _______________________________________________ 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… Unfortunately, that's not available to user space. One could, however, peek at the kernel sources to look at the implementation... If you haven't discovered it you yet, you might also look at OSAtomic.h. This email sent to site_archiver(a)lists.apple.com

1 0

Re: how to test atomicity?
by Sean McBride 19 Sep '04

19 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com jmzorko(a)mac.com (jmzorko(a)mac.com) on Sun, Sep 19, 2004 11:22 said: >I'm porting a class library / framework to OSX 10.3 / Darwin PPC. One >of the things needed is atomic increments, decrements and swaps. With >the help of others here, and various web sites, I found the lwarx / >stwcx pair and some assembly doing exactly what I want. Cool! If you haven't discovered it you yet, you might also look at OSAtomic.h. _______________________________________________ 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