June 2002 - Darwin-kernel - lists.apple.com

Admin notice: New archive search engine public beta
by Kevin 28 Jun '02

28 Jun '02

This message is being sent to all subscribers of Apple's Darwin mailing lists. We apologize if you receive multiple copies of this message. ************* * Public beta availability of new lists.apple.com archive search engine ************* Apple is excited to announce the public beta release of its new mailing list archive search functionality for lists.apple.com. This beta release will only include the archives of the Darwin mailing lists, so we hope you will try out the new search engine and give us your feedback. You can access the new search engine by visiting: http://search.lists.apple.com/index-beta.html Access to these archives still requires the archives/archives username and password combination. This simple authentication prevents spambots and other automated e-mail address harvesters from obtaining e-mail addresses from the archives. We are currently developing a new system that will remove this restriction in the future while still protecting your privacy. We welcome any and all feedback! Please send your comments to search.feedback(a)lists.apple.com. We know a lot of you have waited very patiently for improved archive search functionality. We thank for your patience and hope you will like this update. Thanks to all of you for subscribing to Apple's mailing lists. We look forward to hearing your feedback! Kevin (The new) Administrator, Lists.apple.com listadmin(a)lists.apple.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: Two Kext Locking Question
by Scott Taggart 28 Jun '02

28 Jun '02

I'm a bit lost on something below... (thanks in advance) At 11:03 AM 6/28/2002, Godfrey van der Linden wrote: >The IOCommandGate is designed to do this, check out the header doc. > >However I wrote a paper on designing a PCI device quite a while ago which >should still be available on RobB's How too site. > >In simple explanation is that the workloop is the master lock for one >driver stack. In your case kext A will be registering with your >provider's workloop. Now a work loop has lots of event sources registered >with it and each event source has an associated Action. It is guaranteed >that one and only one Action routine can run on a particular workloop at >any time. > >So in your example if you find your workloop by just calling >provider->getWorkLoop() in KextA you will automatically be able to >synchronise with the USBs work loop and all completion routines come out >on that context. Then if you need to protect your own driver against >these completion routines and you addEventSource() them to your providers >work loop. Note it is also OK to add IOTimerEventSources if you have >times outs. (Note however the IOTimerEventSource doesn't currently have a >reliable cancel path so be very careful when you are trying to unload your >driver.) > >Then In the APIs from your client you wish to synchronise you use the >fGate->runAction(...) function. > >Pseudo code follows (check header doc for details). > >in your start routine in Kext A > >fGate = IOCommandGate::commandGate(this); >IOWorkLoop *wl = getWorkLoop(); // Find my providers work loop. >if (!fGate || !wl) > return false; >if (kIOReturnSuccess != wl->addEventSource(fGate)) > return false; >// That's it your now have an IOCommandGate set up in member variable fGate > >IOReturn KextA::myClientsAPI(args) >{ > return fGate->runAction(&KextA::myClientsAPIGated, args); >} In my comments, I am assuming that Kext B calls your below static function named myClientsAPIGated.... Two things I don't understand about what you show in the above myClientsAPI()... First, the way you show the myClientsAPI() method above, it looks like it's a non-static method being called by some other member-function of KextA:: -- while this may be useful somewhere else, I need KextB to be able to call into the static function you show below, grab the gate and then call KextA::xxx members. It seems like you are showing this the other way around, or do I not understand what "runAction" is doing? Here's what I envision (is this right??): // this is static... IOReturn KextA::KextBEntry(IOService *servSelf,...) { KextA *self = (KextA *) servSelf; self->GoFromB(...); } // next 2 are non-static IOReturn KextA::GoFromB(...) { return fGate->runAction(&KextA::DoSomethingFinally, ...); } IOReturn KextA::DoSomethingFinally(...) { // do gate-protected stuff here } >// Make sure this function is declared as a 'static' function in the header >IOReturn KextA::myClientsAPIGated(IOService *servSelf, void *vArg1, void >*vArg2, /*etc ... */ ) >{ > KextA *self = (KextA *) servSelf; > UInt32 arg1 = (UInt32) vArg1; > > // You can see where this is going I'm sure > // Just remember to either call a member function here or prefix > // all class access with self-> > return result; >} > >FInally I recommend only cleaning up the fGate at free time and no >earlier. A lot of tear down problems have been due to premature tear down >of the synchroniser. > >free() >{ > if (fGate) > { > IOWorkLoop *wl = fGate->getWorkLoop(); > if (wl) > wl->removeEventSource(fGate); > fGate->release(); > fGate = 0; > } >} > >Godfrey > >At 8:20 -0700 02-6-28, Scott Taggart wrote: >>Hi, >> >>I have two kexts: kext A is a USB driver sub-classed from IOService and >>Kext B sub-classed from IOEthernetController. B "calls" into "A" at >>various times to perform functions such as delivering outbound frames. I >>want to "lock" B from entering into A if A has already been entered (via >>USB callback, timer, etc.) (i.e., its "gate/lock" is busy). Can someone >>please give me the basics of how I make a given function entry point into >>A and have it play the "IO Gate" game (i.e., to make sure it's the only >>one with the lock). Function calls and a pointer to some sample code >>would be a big help. >> >>Thanks, >> >>Scott _______________________________________________ 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: Two Kext Locking Question
by Godfrey van der Linden 28 Jun '02

28 Jun '02

The IOCommandGate is designed to do this, check out the header doc. However I wrote a paper on designing a PCI device quite a while ago which should still be available on RobB's How too site. In simple explanation is that the workloop is the master lock for one driver stack. In your case kext A will be registering with your provider's workloop. Now a work loop has lots of event sources registered with it and each event source has an associated Action. It is guaranteed that one and only one Action routine can run on a particular workloop at any time. So in your example if you find your workloop by just calling provider->getWorkLoop() in KextA you will automatically be able to synchronise with the USBs work loop and all completion routines come out on that context. Then if you need to protect your own driver against these completion routines and you addEventSource() them to your providers work loop. Note it is also OK to add IOTimerEventSources if you have times outs. (Note however the IOTimerEventSource doesn't currently have a reliable cancel path so be very careful when you are trying to unload your driver.) Then In the APIs from your client you wish to synchronise you use the fGate->runAction(...) function. Pseudo code follows (check header doc for details). in your start routine in Kext A fGate = IOCommandGate::commandGate(this); IOWorkLoop *wl = getWorkLoop(); // Find my providers work loop. if (!fGate || !wl) return false; if (kIOReturnSuccess != wl->addEventSource(fGate)) return false; // That's it your now have an IOCommandGate set up in member variable fGate IOReturn KextA::myClientsAPI(args) { return fGate->runAction(&KextA::myClientsAPIGated, args); } // Make sure this function is declared as a 'static' function in the header IOReturn KextA::myClientsAPIGated(IOService *servSelf, void *vArg1, void *vArg2, /*etc ... */ ) { KextA *self = (KextA *) servSelf; UInt32 arg1 = (UInt32) vArg1; // You can see where this is going I'm sure // Just remember to either call a member function here or prefix // all class access with self-> return result; } FInally I recommend only cleaning up the fGate at free time and no earlier. A lot of tear down problems have been due to premature tear down of the synchroniser. free() { if (fGate) { IOWorkLoop *wl = fGate->getWorkLoop(); if (wl) wl->removeEventSource(fGate); fGate->release(); fGate = 0; } } Godfrey At 8:20 -0700 02-6-28, Scott Taggart wrote: Hi, I have two kexts: kext A is a USB driver sub-classed from IOService and Kext B sub-classed from IOEthernetController. B "calls" into "A" at various times to perform functions such as delivering outbound frames. I want to "lock" B from entering into A if A has already been entered (via USB callback, timer, etc.) (i.e., its "gate/lock" is busy). Can someone please give me the basics of how I make a given function entry point into A and have it play the "IO Gate" game (i.e., to make sure it's the only one with the lock). Function calls and a pointer to some sample code would be a big help. Thanks, Scott _______________________________________________ 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

Two Kext Locking Question
by Scott Taggart 28 Jun '02

28 Jun '02

Hi, I have two kexts: kext A is a USB driver sub-classed from IOService and Kext B sub-classed from IOEthernetController. B "calls" into "A" at various times to perform functions such as delivering outbound frames. I want to "lock" B from entering into A if A has already been entered (via USB callback, timer, etc.) (i.e., its "gate/lock" is busy). Can someone please give me the basics of how I make a given function entry point into A and have it play the "IO Gate" game (i.e., to make sure it's the only one with the lock). Function calls and a pointer to some sample code would be a big help. Thanks, Scott P.S., sorry for the cross-post: I'm not sure which is the best list for this queston. _______________________________________________ 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: context switching and profiling
by Jonathan Hess 21 Jun '02

21 Jun '02

At 1:29 PM -0400 6/21/02, Jim Magee wrote: No real-time notifications. But there is a trace mechanism that gives you the raw information to post-process. That's the same data that feeds the sc_usage, fs_usage, and latency performance utility tools. Looking at the man page for sc_usage gave me another idea. I saw the CPU_TIME, WAIT_TIME, and CUR_WAIT_TIME column headers. If the active (function) profiler code on the user side had fast read access to such information at the thread level, and if the resolution of that information was at RTC register granularity, then I think the in/out time differentials could be calculated. All without having to mess with any callbacks/signals etc. Where would I look for more information on the thread specific cpu accounting information? -- Jonathan Hess ViviStar Consulting http://www.vivistar.com/ mailto:jahess@vivistar dot 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: context switching and profiling
by Jim Magee 21 Jun '02

21 Jun '02

On Friday, June 21, 2002, at 12:50 PM, Jonathan Hess wrote: How would one receive notification (signal or otherwise) that the current thread is about to be or was just swapped out and that the current thread is about to be or was just swapped back in? By "just about to be swapped out" I assume you mean "just about to lose control/ownership of a processor." If so, the answer is "no way, no how." That is about the most time critical path in the system. We can't put notifications/callbacks in there. I guess if you wanted to waste your own cycles on processing a "just got control back" kind of notification, that would be a different thing. But I don't think it would be very useful, and there is no such mechanism today. With present compilers and the PPC RTC register "active" profilers are a real possibility. But, to make them robust on a preemptive multi-cpu OS the question above must be resolved. I've thought about instrumenting semaphore and mutex abstractions and finding and instrumenting the time slice expired signal. To be generally useful, all calls that would likely block the current thread (read, write, accept, glFlush, etc, etc) would have to be instrumented. But these all just seem like problematic hacks that avoid finding the real solution. Is there an exposed relatively hi-level way in the shipping kernel to do this? If not, any pointers on where to look in the kernel? No real-time notifications. But there is a trace mechanism that gives you the raw information to post-process. That's the same data that feeds the sc_usage, fs_usage, and latency performance utility tools. --Jim _______________________________________________ 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

context switching and profiling
by Jonathan Hess 21 Jun '02

21 Jun '02

How would one receive notification (signal or otherwise) that the current thread is about to be or was just swapped out and that the current thread is about to be or was just swapped back in? With present compilers and the PPC RTC register "active" profilers are a real possibility. But, to make them robust on a preemptive multi-cpu OS the question above must be resolved. I've thought about instrumenting semaphore and mutex abstractions and finding and instrumenting the time slice expired signal. To be generally useful, all calls that would likely block the current thread (read, write, accept, glFlush, etc, etc) would have to be instrumented. But these all just seem like problematic hacks that avoid finding the real solution. Is there an exposed relatively hi-level way in the shipping kernel to do this? If not, any pointers on where to look in the kernel? -- Jonathan Hess ViviStar Consulting http://www.vivistar.com/ mailto:jahess@vivistar dot 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

Newbie gdb (*nix?) Question
by Richard Jackson 21 Jun '02

21 Jun '02

This is probably obvious, but I've scanned the gdb manual and can't seem to find any answer... If I have tcsh environment variables set, for example: $MYPROJ=~/Projects/MyProject, or $MYTARGET=mytarget.mycompany.com is there a way to use these inside gdb? I seem to be able to start gdb and use these as command variables, for example this works fine: gdb -s $MYPROJECT/mysyms.sym but if I launch gdb and try to use the environment variables in a gdb command, gdb doesn't seem to know how to parse them: (gdb) add-symbol-file $MYPROJ/mysyms.sym $MYPROJ/mysyms.sym: No such file or directory Yet if I type out the full path it works fine: (gdb) add-symbol-file ~/Projects/MyProject/mysyms.sym I'd like to be able to use my environment variables to write gdb command scripts: what am I doing wrong? TIA, - Richard Jackson _______________________________________________ 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

kmodload failing
by Heiko Panther 19 Jun '02

19 Jun '02

Hi All, I'm trying to do two-machine debugging, and I follow the kext tutorial "Debugging Device Drivers with GDB". As I try to load my kext, the following happens. #kmodload myAudioDriver kmodload: Can't find superclass for _28myAudioDevice : _13IOAudioDevice I guess something with the name mangling is wrong? What can I do about it? Heiko Panther _______________________________________________ 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

External USB modem makes kernel panics
by AM.URIARTE 19 Jun '02

19 Jun '02

Hi guys, I really have a problem with a kernel extension. Could you help me with the provided information? I know this question has been discussed in this mailing lists most probably a lot of times but following some conversations I dont get how to proceed. My computer is an iMac G3/400 DV with 512MB RAM (2 x 256MB), under Mac OS X 10.1.4. My internal modem was broken and instead of repair it I bought a 3Com OfficeConnect Business Modem, external USB and compliant with MacOS as its built by USR Robotics but not specifically indicated for Mac OS X, but its not a windmodem. It doesnt matter if the modem is connected directly to the free USB port or through an 7-port Keyspan USB hub connected to the free USB. I get two or three very good connections and after that I loose the control of the modem. This loose of control always occurs when the line has dropped and the Internet Application thinks is connected then I click Cancel (Disconnect) and the Internet Application says "Disconnecting..." when the modem has hang up minutes ago and I can not connect again. Then when after just selecting restart I get this kernel panic (two records from different days and with two different versions of Mac OS X, 10.1.3 and 10.1.4, indeed the problems appear since update to 10.1.3 where IOSerialFamily.kext was modified to the current version 5.3.0d14). Sometimes switching off and on the modem, or unplug / plug, makes the iMac sees it again and connect to Internet but in most of cases the only solution is to restart and get the kernel trap. Once is plugged in the 7-port Keyspan USB Hub is shown in the Active Ports in Network System Preferences, not added driver, that's the reason I am thinking the Mac OS X kernel drivers are being used for my modem but I don't understand the reason IOSerialFamily is used when the modem is an USB device. Indeed an HP DeskJet 950C printer is connected to the same USB hub and no problems but it's configured as USB interface (what are the differences between USB device and USB interface?) with the specific driver HP provides. Looking at Kernel Extensions I got IOSerialFamily.kext is 2001-07-30 5.3 (5.3d14) and IOUSBFamily.kext is 1.8.7. Are they OK for 10.1.4? Unresolved kernel trap(cpu 0): 0x300 - Data access DAR=0x00000060 PC=0x135cdc40 Latest crash info for cpu 0: Exception state (sv=0x135d4000) PC=0x135CDC40; MSR=0x00009030; DAR=0x00000060; DSISR=0x40000000; LR=0x135cc30c; R1=0x0d203be0; XCP=0x0000000c (0x300 - Data access) Backtrace: 0x135cd96c 0x135cc30c 0x000b81d8 0x000b5484 0x00191bc8 0x00092448 0x0008e600 0x00000000 Kernel loadable modules in backtrace: com.apple.iokit.IOSerialFamily(5.3.0d14)@0x135c9000 Proceeding back via exception chain: Exception state (sv=0x135d4000) previously dumped as "Latest" state. skipping... Exception state (sv=0x135f0000) PC=0x700035DC; MSR=0x0000d030; DAR=0x000cf000; DSISR=0x40000000; LR=0x700035a4; R1=0xbffff450; XCP=0x00000030 (0xC00 - System call) Kernel version: Darwin Kernel Version 5.3: Thu Jan 24 22:06:02 PST 2002; root:xnu/xnu-201.19.obj~1/RELEASE_PPC Memory access exception (1,0,0) Waiting for remote debugger connection. Options..... Type ------------ ---- continue.... 'c' reboot...... 'r' Unresolved kernel trap(cpu 0): 0x300 - Data access DAR=0x00000060 PC=0x135cec40 Latest crash info for cpu 0: Exception state (sv=0x136de000) PC=0x135CEC40; MSR=0x00009030; DAR=0x00000060; DSISR=0x40000000; LR=0x135cd30c; R1=0x0d18bbe0; XCP=0x0000000c (0x300 - Data access) Backtrace: 0x135ce96c 0x135cd30c 0x000b8c98 0x000b5f44 0x00192b84 0x00092f08 0x0008f0c0 0x5f313953 Kernel loadable modules in backtrace: com.apple.iokit.IOSerialFamily(5.3.0d14)@0x135ca000 Proceeding back via exception chain: Exception state (sv=0x136de000) previously dumped as "Latest" state. skipping... Exception state (sv=0x13823000) PC=0x700035DC; MSR=0x0000d030; DAR=0x0002b90; DSISR=0x40000000; LR=0x700035a4; R1=0xbffff7a0; XCP=0x00000030 (0xC00 - System call) Kernel version: Darwin Kernel Version 5.4: Wed Apr 16 09:27:47 PST 2002; root:xnu/xnu-201.19.3.obj~1/RELEASE_PPC Memory access exception (1,0,0) Waiting for remote debugger connection. Options..... Type ------------ ---- continue.... 'c' reboot...... 'r' Thanks in advance, Alberto am.uriarte(a)terra.es _______________________________________________ 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