September 2004 - Darwin-dev

Re: New SMU chip in iMac G5
by Benjamin Herrenschmidt 18 Sep '04

18 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Sat, 2004-09-18 at 19:09, Finlay Dobbie wrote: > I don't know personally, but I'd say it's highly unlikely, given Apple > classes its thermal management as important IP for a feature that > distinguishes itself from the competition (and the recent fan control & > power management kexts haven't been open-sourced). > > Don't hold your breath. Perhaps Kevin could confirm this? :-) Well, the core thermal management is open sourced as it's part of the platform plugins... What is not are the low level drivers that interface to the various HW chips (mostly i2c sensors etc...), most of them having available specs, it makes little sense imho.... SMU is important for more than just thermal management anyway (and probably doesn't contain any thermal management policy per-se, just the ability to send messages to the fan control unit I suppose). However, it's also the required interface to machine restart/shutdown, real rime clock, sleep mode, etc... 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: New SMU chip in iMac G5
by Finlay Dobbie 18 Sep '04

18 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 18 Sep 2004, at 06:21, Benjamin Herrenschmidt wrote: Is there any chance for the Darwin driver for this one to ever be open sourced ? Don't hold your breath. Perhaps Kevin could confirm this? :-) -- Finlay _______________________________________________ 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… I don't know personally, but I'd say it's highly unlikely, given Apple classes its thermal management as important IP for a feature that distinguishes itself from the competition (and the recent fan control & power management kexts haven't been open-sourced). This email sent to site_archiver(a)lists.apple.com

1 0

Re: Using MIG RPC between 2 threads inside a process
by Steve Checkoway 18 Sep '04

18 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com #include <pthread.h> pthread_mutex_t lock; pthread_cond_t cond; pthread_cond_init( &cond, NULL ); pthread_mutex_init( &lock, NULL ); server: create client thread pthread_mutex_lock( &lock ); while(true) { pthread_cond_wait( &cond, &lock ); do something what you need } It should be something like this: while (ServerShouldKeepGoing()) { DoSomeStuff(); pthread_mutex_lock(&lock); while (!CheckMyCondition()) pthread_cond_wait(&cond, &lock); DoStuffRequiringMutualExclusion(); pthread_mutex_unlock(&lock); PossiblyDoMoreStuff(); } client: while( true ) { pthread_mutex_lock( &lock ); Put some data in a global varibale or so. pthread_cond_signal( &cond ); } - Steve _______________________________________________ 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… I saw this a few days back and I meant to point out that this is not the correct way to use conditions. On Sep 16, 2004, at 1:12 AM, Dario wrote: That inner while loop is absolutely essential. What if the reason the condition cond had been signaled was no longer true, you do not wish to proceed in this case. What if instead of signaling, it was broadcast and multiple threads woke up in response to it? A client might or might not loop over and over but one thing you forgot was to unlock after you signal. pthread_mutex_lock(&lock); DoSomeWork(); pthread_cond_signal(&cond); pthread_mutex_unlock(&lock); It might be better to signal after unlocking but since pthread_cond_signal doesn't wake up the other threads right away, I'm not sure it makes a difference. I'll bet someone here could say which way is best. Or can be resolved by using a queue structure with synchronized access and verification (still with mutexes), I have an example if you need it. with this code as you can see you can send only one job and the client will block until the end of the previous on the server thread This is usually easier to work with if you can get away with it. Some times you find that you need to manipulate multiple data structures atomically and this just doesn't quite work in that case. This email sent to site_archiver(a)lists.apple.com

1 0

diskutil createRAID question
by howarth＠bromo.msbb.uc.edu 18 Sep '04

18 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com 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 my dual G5 under 10.3.5. I noticed that the DiskUtility program graphical interface won't work. While its help page on "Create a RAID set" indicates that either disks or volumes can be dragged to the RAID disks window to create a RAID in reality only disks can be used. Clicking on a volume from the disk removes the RAID tab in the DiskUtility window. My question is the following. It would seem my only remaining option is the diskutil command line program from Darwin. However the manpage for diskutil is unclear on the Create a RAID option. It says.... Create a RAID diskutil createRAID mirror MirrorDisk HFS+ disk disk2 and later says... DEVICES A device parameter to any of the above commands (except where explicitly defined otherwise) is any of the following The device node entry. Any entry of the form of /dev/disk*. ex. /dev/disk2 The disk identifier. Any entry of the form of disk*. ex. disk1s9 The volume mount point. Any entry of the form of /Volumes/Mountpoint. ex. /Volumes/Untitled which I would assume implies that I could in theory create two different RAID-1 partitions from a set of drives identically partitioned using commands like... diskutil createRAID mirror MirrorPanther HFS+ disk0s9 disk1s9 diskutil createRAID mirror MirrorTiger HFS+ disk0s10 disk1s10 where the s9 and s10 slices are existing partitions on the drives. Does anyone know if this has any chance of working? I would rather not waste my time if diskutil is known to not work in this manner. 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

New SMU chip in iMac G5
by Benjamin Herrenschmidt 18 Sep '04

18 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi ! Is there any chance for the Darwin driver for this one to ever be open sourced ? 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: bind
by Damien Bobillot 17 Sep '04

17 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Le 17 sept. 04, à 19:56, Marc Majka a écrit : -- Damien Bobillot _______________________________________________ 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… Originally, we had intended to leave the BIND-8 library in libSystem and remove the BIND-8 headers (resolv.h and nameser.h) but we ended up keeping the old headers and adding the BIND_8_COMPAT switch so that developers that really, really, had no other choice except to keep building with the BIND-8 version could do so. If at all possible, please avoid using BIND-8. It will eventually go away. It's important to keep the BIND 8 version, not for developers (they can rewritte their code) but for administrators who want to compile and use applications not updated to BIND 9 (administrators are often not able to rewritte code). This email sent to site_archiver(a)lists.apple.com

1 0

Trick to compiling Apple3Com905x?
by Walt Pawley 17 Sep '04

17 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I happened to have a 3Com 3C905B-TX ethernet card and wanted to plug it into a Beige G3 using Mac OS 10.2.6. I tried compiling the Darwin distributions of both the -3 and -7 drivers but both failed in the same manner on several different systems. I know it has been successfully compiled because I found a working version of a kext which I'm using now. I confess I'm extremely naive about Project Builder, etc. but it seems to me failing with a "Internal compiler error in instantiate_virtual_regs_1" message on several different compiler versions is something would not have gone unnoticed by now. I suspect I'm simply doing something that would be considered stupid by someone who knows what they're doing. Or, should I be submitting bug reports? -- perl -e 'foreach(qw/13 10 60 47 98 117 115 104 62 13 10 10/){print chr($_)};' Walter M. Pawley <walt(a)wump.org> Wump Research & Company 676 River Bend Road, Roseburg, OR 97470 541-672-8975 _______________________________________________ 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: bind
by Marc Majka 17 Sep '04

17 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -- Marc Majka yes it does, would this be better than -DBIND_8_COMPAT=1 ? S a m _______________________________________________ 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… There are actually two versions of the BIND resolver client library co-resident on the system. We used to just have a BIND-8 derived library (Mac OS X 10.2 "Jaguar" and before), and it was found in libSystem. In 10.3 "Panther", we brought in a BIND-9 resolver client, but we kept the old BIND-8 version in libSystem for binary compatibility. The new BIND-9 based library is in /usr/lib/libresolv.dylib. Originally, we had intended to leave the BIND-8 library in libSystem and remove the BIND-8 headers (resolv.h and nameser.h) but we ended up keeping the old headers and adding the BIND_8_COMPAT switch so that developers that really, really, had no other choice except to keep building with the BIND-8 version could do so. If at all possible, please avoid using BIND-8. It will eventually go away. We highly encourage developers to use API at the highest level possible. This will give you the best level of service that we can provide to your application. It also means that your application will behave the same way as other high-level applications available to users. If you limit your application to using very low-level API, then your users may find that your application has annoying limitations, or that it will obtain different results than other applications. If you can, you should take advantage of Core Foundation Networking APIs. If you are developing code at the "BSD" level, then consider using API like getaddrinfo(), getnameinfo(), getipnodebyname(), getipnodebyaddr(), and so on. There are many places where the system may go to find this information. DNS is one source, but these APIs may also consult a per-thread ("Level 1") cache, a system-wide ("Level 2") cache, the /etc/hosts file, as well as directory services like NetInfo and LDAP-based services. If you really need to do direct to DNS, then we highly encourage you to consider using the API in /usr/include/dns.h and dns_util.h. This is a DNS query-routing API that supports multiple DNS resolver clients. The API is very similar to the basic BIND search and query APIs, but this layer routes queries to a "default" resolver (what you would expect to find in /etc/resolv.conf), as well as to multicast DNS (".local"), and to private DNS resolvers. We are taking increasing advantage of this layer for new Apple features in Mac OS X 10.4. Some of them were described at WWDC. DNS query-routing is becoming increasingly dynamic, and DNS configuration management is becoming increasingly automatic. If you really need to use DNS services directly, this is the API to choose. Be careful about using the BIND APIs directly. Name resolution on Darwin / Mac OS has a lot more going on than just calling res_search. A lot of effort has been made to improve performance through the use of caching. A lot of extra search functionality has been added to support Rendezvous, private DNS servers, and DNS over VPN links. Sending your query to a single resolver might get you the answer you want, but it will probably be slower and may fail to find answers that applications using higher-level API will find. On 17 Sep, 2004, at 07:56, Sam Hart wrote: On 17 Sep 2004, at 15:46, Matt Watson wrote: Does adding -lresolv to your link line help? This email sent to site_archiver(a)lists.apple.com

1 0

Re: Portmap daemon on Mac OS X 10.3 going into unstable state.
by Mike Mackovitch 17 Sep '04

17 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Mutt/1.4i On Fri, Sep 17, 2004 at 07:52:37AM +0000, GirisH wrote: > > Portmap service must be running in order to make RPC calls. More importantly, portmap needs to be running before any RPC services are started so that the servers can register their ports with portmap so clients can be told what ports to connect to. > After creating , modifying , deleting NFS share we restart mountd service using > > killall -HUP mountd > > When we try to access NFS share through NFS client sometimes portmap daemon > goes into unstable state automatically. Can you be more specific about the details of this "unstable state"? > Because of that when user try to mount > NFS share from client machine mount command hangs up.Even if we try to restart > portmap service NFS client can not mount the daemon, it gets hanged. Restarting portmap is NOT a good idea, since that will cause all the current port registrations to be lost. > Most of > the times after rebooting machine (Mac OS 10.3) portmap works properly and NFS > client can mount NFS share. That's probably because the startup scripts know how to start all the NFS server services in the correct order: portmap rpc.statd rpc.lockd mountd nfsd > Can you please suggest us why portmap service on Mac OS X 10.3 behaves like > this? > > What we can do so that portmap service should not get into unstable state? Well, you already know that rebooting after making NFS export changes works. Yes, it's a big hammer... but it works. If the system was booted with NFS exports defined, all the services should be started properly. If you change your existing set of exports, then a "sudo killall -HUP mountd" should just work. "rpcinfo -p" can be used to check the state of portmap "showmount -e" can be used to check the state of your exports If everything looks OK and it's still not working, then perhaps it's time to start grabbing network packet traces with tcpdump. --macko _______________________________________________ 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: bind
by Matt Watson 17 Sep '04

17 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com matt. On Sep 17, 2004, at 7:56 AM, Sam Hart 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… I'd think so, since you'd be using the bind9 library, but I can't say whether you'd notice the difference. -DBIND_8_COMPAT lets your code use the compatibility routines built in to libSystem. In Darwin 7.0, the resolver code was updated to bind9 and pulled out into its own dylib. On 17 Sep 2004, at 15:46, Matt Watson wrote: Does adding -lresolv to your link line help? This email sent to site_archiver(a)lists.apple.com

1 0