November 2002 - Darwin-kernel

Re: Performance problems under 10.2
by Jonas Maebe 12 Nov '02

12 Nov '02

On Tue, 12 Nov 2002, Jonas Maebe wrote: > The distributed.net client has the same problem. Here's some sc_usage > info: Maybe one extra tidbit of info: dnetc process runs at priority 20, or "idle" priority according to the man page of nice (although there is no such thing under Darwin and there will never be one it seems from the "operates according to specifications" closure of my enhancement request Radar on this :( ) Jonas _______________________________________________ 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: About APPLE_PRIVATE code...
by Vincent Lubet 12 Nov '02

12 Nov '02

Sure, bug fixes are always warmly welcomed. Vincent On Tuesday, November 12, 2002, at 09:54 AM, Stiphane Sudre wrote: Is it possible to submit fixes for code in Darwin which is embedded in a #define APPLE_PRIVATE portion? Because I'm seeing a panic call which could be avoided. _______________________________________________ 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: About APPLE_PRIVATE code...
by Justin C. Walker 12 Nov '02

12 Nov '02

On Tuesday, November 12, 2002, at 09:54 AM, Stiphane Sudre wrote: Is it possible to submit fixes for code in Darwin which is embedded in a #define APPLE_PRIVATE portion? Sure; the #define doesn't mean you have to pretend you've never seen that code :-}. Because I'm seeing a panic call which could be avoided. That would be a good thing to fix. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | It's not whether you win or lose... | It's whether *I* win or lose. *--------------------------------------*-------------------------------* _______________________________________________ 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: Performance problems under 10.2
by Jonas Maebe 12 Nov '02

12 Nov '02

On zaterdag, nov 9, 2002, at 20:13 Europe/Brussels, Tyler Ward wrote: Has anyone noticed strange performance problems under 10.2? I have noticed that things which were once fast now use inordinate amounts of system time. For instance, setiathome (the unix command line version) uses almost 25% system time, whereas before it used only a few percent system time. Monitoring it with "top -es 3" shows that it is performing almost 1,000 context switches per second. For instance, here's the setiathome line from my top output.... The distributed.net client has the same problem. Here's some sc_usage info: *** dnetc 575 preemptions 577 context switches 2 threads 16:59:32 0 faults 15 system calls 0:00:15 TYPE NUMBER CPU_TIME WAIT_TIME ------------------------------------------------------------------------ ------ System Idle 0:00.220( 0:00.013) System Busy 0:02.031( 0:00.095) dnetc Usermode 0:12.254 clock_sleep_trap 14(1) 0:00.000 0:10.648( 0:00.863) W access 1 0:00.000 gettimeofday 115(6) 0:00.000 __sysctl 115(6) 0:00.001 0:00.000 mach_msg_trap 14(1) 0:00.000 clock_get_time 14(1) 0:00.000 *** Sample info (I've snipped a lot, this one was the only one that seemed relevant). *** Call graph: 300 Thread_0e03 300 0x24f0 300 0x26c0 300 0x2bd18 300 0x2bc0c 300 0x2eacc 300 0x1b768 300 sleep 300 nanosleep 300 clock_sleep_trap 300 clock_sleep_trap [STACK TOP] *** Previously, dnetc only used negligable numbers of system time, but since 10.2 it's usually between 20% and 50%, though sometimes it jumps to 100% (and what's also strange, is that the system usage numbers are almost always rounded to a multiple of 5). Here are some "top -ud" snaps: *** Processes: 53 total, 3 running, 50 sleeping... 122 threads 19:12:48 Load Avg: 1.24, 1.20, 1.04 CPU usage: 80.0% user, 20.0% sys, 0.0% idle Networks: 2 ipkts/1K 0 opkts /0K Disks: 0 reads/0K 0 writes/0K VM: 0 pageins 0 pageouts PID COMMAND %CPU TIME FAULTS PGINS/COWS MSENT/MRCVD BSD/MACH CSW 404 dnetc 87.7% 59:39.93 0 0/0 1/1 20/2 357 525 UniversalA 5.6% 0:36.59 0 0/0 100/50 1/225 79 787 top 2.8% 0:00.28 121 0/1 583/581 80/689 12 293 Window Man 0.9% 1:57.23 2 0/0 158/7 0/262 137 529 Mail 0.0% 6:07.00 0 0/0 0/0 0/0 0 532 Navigator 0.0% 3:46.00 0 0/0 0/0 0/408 211 0 kernel_tas 0.0% 1:51.49 0 0/0 930/695 0/0 503 *** Here's one with 100% sys time! *** Processes: 53 total, 3 running, 50 sleeping... 122 threads 19:12:55 Load Avg: 1.45, 1.26, 1.08 CPU usage: 0.0% user, 100.0% sys, 0.0% idle Networks: 14 ipkts/2K 0 opkts /0K Disks: 0 reads/0K 0 writes/0K VM: 0 pageins 0 pageouts PID COMMAND %CPU TIME FAULTS PGINS/COWS MSENT/MRCVD BSD/MACH CSW 404 dnetc 98.2% 59:44.10 0 0/0 1/1 28/2 370 787 top 1.9% 0:00.39 115 0/1 583/581 77/689 9 529 Mail 0.0% 6:08.01 0 0/0 0/0 0/0 0 532 Navigator 0.0% 3:46.00 0 0/0 0/0 0/406 206 293 Window Man 0.0% 1:57.68 0 0/0 0/0 0/93 3 0 kernel_tas 0.0% 1:51.55 0 0/0 794/588 0/0 385 *** And one with 50%: *** Processes: 53 total, 3 running, 50 sleeping... 122 threads 19:12:56 Load Avg: 1.36, 1.25, 1.07 CPU usage: 50.0% user, 50.0% sys, 0.0% idle Networks: 3 ipkts/1K 0 opkts /0K Disks: 0 reads/0K 0 writes/0K VM: 0 pageins 0 pageouts PID COMMAND %CPU TIME FAULTS PGINS/COWS MSENT/MRCVD BSD/MACH CSW 404 dnetc 98.2% 59:45.10 0 0/0 1/1 20/2 359 787 top 1.9% 0:00.41 115 0/1 583/581 76/689 7 529 Mail 0.0% 6:08.01 0 0/0 0/0 0/0 0 532 Navigator 0.0% 3:46.00 0 0/0 0/0 0/408 208 293 Window Man 0.0% 1:57.68 0 0/0 0/0 0/122 2 0 kernel_tas 0.0% 1:51.55 0 0/0 796/588 0/0 366 *** And again 100%: *** Processes: 53 total, 3 running, 50 sleeping... 122 threads 19:12:57 Load Avg: 1.29, 1.23, 1.07 CPU usage: 0.0% user, 100.0% sys, 0.0% idle Networks: 12 ipkts/1K 0 opkts /0K Disks: 0 reads/0K 0 writes/0K VM: 0 pageins 0 pageouts PID COMMAND %CPU TIME FAULTS PGINS/COWS MSENT/MRCVD BSD/MACH CSW 404 dnetc 98.2% 59:46.10 0 0/0 1/1 27/2 353 787 top 0.9% 0:00.42 115 0/1 583/581 76/689 7 529 Mail 0.0% 6:08.01 0 0/0 0/0 0/0 0 532 Navigator 0.0% 3:46.00 0 0/0 0/0 0/408 208 293 Window Man 0.0% 1:57.68 0 0/0 0/0 0/116 3 0 kernel_tas 0.0% 1:51.55 0 0/0 793/588 0/0 365 *** Jonas _______________________________________________ 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

About APPLE_PRIVATE code...
by Stéphane Sudre 12 Nov '02

12 Nov '02

Is it possible to submit fixes for code in Darwin which is embedded in a #define APPLE_PRIVATE portion? Because I'm seeing a panic call which could be avoided. _______________________________________________ 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: kextunload problem
by chuck remes 12 Nov '02

12 Nov '02

On Monday, November 11, 2002, at 08:22 PM, chuck remes wrote: On Sunday, November 10, 2002, at 10:54 AM, chuck remes wrote: I'm trying to get my kext to successfully unload so I can test it without going through a reboot cycle. I'm having problems. After running kextload, my kext goes through its init() and start() sequences. I have engineered it so start() fails. When I run kextunload, I get this in system.log. : Can't unload com.chuck.iokit.tulip due to - : com.chuck.iokit.tulip: com_cremes_eeprom has 1 instance(s) : com.chuck.iokit.tulip: com_chuck_iokit_tulip has 1 instance(s) : com.chuck.iokit.tulip: com_cremes_eeprom_sio has 1 instance(s) : com.chuck.iokit.tulip: com_cremes_eeprom_sio_dec2114x has 1 instance(s) : com.chuck.iokit.tulip: com_chuck_iokit_tulip_dec2114x has 1 instance(s) : com.chuck.iokit.tulip: com_chuck_iokit_tulip_pnic2 has 1 instance(s) <snip> Okay, here's a little more information. I'm cross-posting this in darwin-kernel since it deals with some kernel internals. In the terminal window the command generates: [12-247-46-43:projects/tulip/build] cremes% sudo kextunload tulip.kext/ unload kext tulip.kext/ failed (result code 0xdc004002) I don't mean to SPAM this list, but here's a teeny bit more info. I changed the start() routine in my com_chuck_iokit_tulip_pnic2 subclass (which now directly inherits from com_chuck_iokit_tulip to get the dec2114x subclass out of the way). The start() routine now exits with a failure immediately and doesn't do *any* object allocation, so this eliminates the com_cremes_eeprom classes from the picture. As soon as start() exits, the system calls free(). I always check for the existence of each object before calling release on it, so this should essentially skip to the end where it calls IOEthernetController::free(). I still get this output: : Can't unload com.chuck.iokit.tulip due to - : com.chuck.iokit.tulip: com_chuck_iokit_tulip has 1 instance(s) : com.chuck.iokit.tulip: com_chuck_iokit_tulip_pnic2 has 1 instance(s) I'm rapidly losing my hair because this doesn't make sense! It seems like IOKit is keeping around some references to my classes and not cleaning them up. cr _______________________________________________ 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: kextunload problem
by chuck remes 12 Nov '02

12 Nov '02

On Sunday, November 10, 2002, at 10:54 AM, chuck remes wrote: I'm trying to get my kext to successfully unload so I can test it without going through a reboot cycle. I'm having problems. After running kextload, my kext goes through its init() and start() sequences. I have engineered it so start() fails. When I run kextunload, I get this in system.log. : Can't unload com.chuck.iokit.tulip due to - : com.chuck.iokit.tulip: com_cremes_eeprom has 1 instance(s) : com.chuck.iokit.tulip: com_chuck_iokit_tulip has 1 instance(s) : com.chuck.iokit.tulip: com_cremes_eeprom_sio has 1 instance(s) : com.chuck.iokit.tulip: com_cremes_eeprom_sio_dec2114x has 1 instance(s) : com.chuck.iokit.tulip: com_chuck_iokit_tulip_dec2114x has 1 instance(s) : com.chuck.iokit.tulip: com_chuck_iokit_tulip_pnic2 has 1 instance(s) <snip> Okay, here's a little more information. I'm cross-posting this in darwin-kernel since it deals with some kernel internals. In the terminal window the command generates: [12-247-46-43:projects/tulip/build] cremes% sudo kextunload tulip.kext/ unload kext tulip.kext/ failed (result code 0xdc004002) I did a little digging in the sources for kextunload. It's successfully completing two of three loops to perform the unload and failing on the last loop. The first two loops unload the class names and then the bundle identifiers. The final loop tries to unload the actual kext module. It's calling IOCatalogueTerminate() which is returning the hex value printed above. Does anyone know what this kernel result means? I dug all the way into xnu sources. In IOCatalogue.cpp I traced the logic to the method IOCatalogue::unloadModule(). The method verifies there are ZERO references left on the module; if so, it calls the stop() function and unloads it from kernel memory. If there is any reference to the module, which in my case there is, it returns kIOReturnBadArgument (0x2c2). So now I understand why the stop() function isn't being called (there is still a reference) but I don't know where that reference is coming from! I'm probably missing something obvious here. Can anyone help? cr _______________________________________________ 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

Open and read a file in kernel space
by Gerhard Ahuis 11 Nov '02

11 Nov '02

Hi All, How can I open and read a file from the local filesystem in the kernel space from Darwin ? I need to do this in a driver. Best regards, Gerhard. -- Gerhard Ahuis pe1rlz(a)amsat.org Unsolicited advertisements subject to $1000 consulting fee. _______________________________________________ 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: devfs user client
by Godfrey van der Linden 11 Nov '02

11 Nov '02

Darwin supports the use of dynamic devfs entry creation. A kext and MODULE_START time can allocate a devfs entry and name iself anything it wants. An example of this can be found in IOSerialFamily/IOSerialBSDClient.cpp Godfrey At 21:28 +0100 02-11-11, Matthias Ringwald wrote: Hi Andrew, thanks for your answer. Could somebody help me clarify things a bit more ? For writing a working /dev/... device (driver)... (e.g. a device that could echo data or just provides the current date.. ) Do I have to re-compile the kernel ? Do I need to write a kernel extension ? Do I need the darwin kernel sources to write a user land driver ? I tried "Howto Add a Character Device to Darwin". http://www.opendarwin.org/documentation/chardev.php gcc failed because of the missing header <kernel/thread.h> In this LIRC Linux project, they create a device node with mknod (doesn't work on darwin) and then create unix socket and then bind it to the /dev/lircd node. looks pretty easy... on mac os x, the bind fails 'operation not supported' (I already failed making the dev node properly) I just tried mkfifo. Looks good, but I can't make a fifo in /dev/ .. when opening the fifo I get 'device is not configured' Any ideas for this ? thanks, Matthias Ringwald On Montag, November 11, 2002, at 08:16 Uhr, Andrew Gallatin wrote: Matthias Ringwald writes: hi I found a discussion on providing /dev/.. devices by a user client on this list from nov 2001. Is it possible to write a pseudo device driver that can be used by other applications over open/read/write/ioctl POSIX calls ? I'm trying to port the lirc linux infrared control to mac os x and it uses /dev/.. devices for their communication. Sure. Using ioctls and devices is much more portable when you've got code that runs on other UNIXes. Note that MacOS X uses devfs (the older, uglier one, I think), so you should make the special files from inside the driver. Also note that unlike linux & solaris, MacOS & *BSD do not support factory devices. Drew PS: Here's working sample code ripped from my (real hardware) driver. The idea is to make several /dev/sample* devices, with different permissions based on device number. _______________________________________________ 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

Re: allocate huge amounts of kernel memory
by Umesh Vaishampayan 11 Nov '02

11 Nov '02

On Friday, November 8, 2002, at 02:20 AM, Heiko Panther wrote: > Hi Umesh, > > > IOMalloc() is just a wrapper around kalloc(). kalloc() allocates > > memory from a sub map which is limited to 16MB. > > Out of curiosity: Why do you use that sub map, and don't allocate > straight from kernel_map? (Which is, by the way, my workaround.) Avoid fragmentation. Avoid lock contention.... --Umesh -- Umesh Vaishampayan Apple Computer, Inc. Mac OS X Kernel Ph: (408) 974 0229 _______________________________________________ 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