October 2004 - Darwin-kernel

Re: pmap_unnest: Other processor did not see interruption request
by David Gatwood 11 Oct '04

11 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Oct 9, 2004, at 6:57 AM, Parav Pandit wrote: Let me give you more idea about my driver during loading phase: It does the following steps in sequence: David _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… It first enables the memory mapped io for PCI card's register file.( I have 256 registers) Enables the PCI bus mastering, disables the Option ROM on the card. ( I have to do that) Some memory mapped read/write operation on the PCI card to reset the card and processor on the Card. All memory allocation goes here. I can see IOLog message printed here. (Up to here- all works fine) After this message I get this kernel panic on pmap_unnest. If you have your driver initialization stop right here, do you still get a panic? This email sent to site_archiver(a)lists.apple.com

1 0

Re: __private_extern__
by stephane sudre 11 Oct '04

11 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Oct 11, 2004, at 5:13 PM, Brian Bergstrand wrote: -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Stephane, Are you assigning an initial value to the var definition in your.c file? If not, then the compiler will treat the definition as a declaration, and no storage will be allocated. .h __private_extern__ int foo; .c __private_extern__ int foo = 0; // Allocate storage for foo. Thanks. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… That's definitely my case (no initial value). I would never have suspected this. This email sent to site_archiver(a)lists.apple.com

1 0

Re: __private_extern__
by Brian Bergstrand 11 Oct '04

11 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Stephane, Are you assigning an initial value to the var definition in your.c file? If not, then the compiler will treat the definition as a declaration, and no storage will be allocated. .h __private_extern__ int foo; .c __private_extern__ int foo = 0; // Allocate storage for foo. HTH. On Oct 11, 2004, at 10:04 AM, stephane sudre wrote: > What could be the explanation for a __private_extern__ "local global" > var not being seen as defined? > > I have a local global var being defined as a __private_extern__ (both > in the .h and .c) > > When I try to load the Kernel Extension using kextload, I'm being told > the var is not defined. > > The problem I have to understand the reason behind this is that I have > other vars defined the same way which are seen as being defined. > > Any clue? > Brian Bergstrand <http://www.bergstrand.org/brian/>, AIM: triryche206 PGP Key: <http://www.bergstrand.org/brian/misc/public_key.txt> reality (n.) 1. The illusion resulting from lack of alcohol. As of 10:09:55 AM, iTunes is playing "Burning Man" from "Q2K" by "Queensrÿche" -----BEGIN PGP SIGNATURE----- Version: PGP 8.1 iQA/AwUBQWqU+HnR2Fu2x7aiEQIatgCgnBJYJQ8aOinQeZkCmJmZdbWlqCYAoLVb L/dSCZ7QyeyEOeXC1P6hrYRz =ggpF -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

__private_extern__
by stephane sudre 11 Oct '04

11 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Any clue? _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… What could be the explanation for a __private_extern__ "local global" var not being seen as defined? I have a local global var being defined as a __private_extern__ (both in the .h and .c) When I try to load the Kernel Extension using kextload, I'm being told the var is not defined. The problem I have to understand the reason behind this is that I have other vars defined the same way which are seen as being defined. This email sent to site_archiver(a)lists.apple.com

1 0

NFDescriptor nf_handle
by Mark Woollard 11 Oct '04

11 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thanks Mark _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… I've been looking at NKEs and the examples I've seen set a value for the NFDescriptor nf_handle member. I can't find any docs on this but assume its a globally unique identifier for the NKE. If so, do I request a suitable value through the http://developer.apple.com/datatype/creatorcode.html page? Or am I off track here? This email sent to site_archiver(a)lists.apple.com

1 0

Re: pmap_unnest: Other processor did not see interruption request
by Parav Pandit 09 Oct '04

09 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Importance: Normal Hi, You suggested some testing to do. Here are the results. 1. I tested my hardware using Apple Protection Plan test CD which I got with G5. And all the tests are passed. BTW we have two G5, and on both the system I get the same result, kernel panic also and hardware test results ok also. 2. from CHUD One processor is disabled, and I tried loading driver for 200 times.( script did that :-) ) And it doesn't get kernel panic in that testing. But When I enabled both the CPUs - same kernel panic happens on 11th load. I looked in the xnu code- pmap.c for pmap_unnest() comments says - removes the shared map. I don't understand that I have nothing like that in my driver, what it actually code/pmap it unnest from other processor. What does kext loader put as shared map during driver load that had to be unnested ? 3. This panic is not fixed on cpu 1 only, it randomly happen on cpu 0 also. 4. [David Gatwood] There's some really, really bizarre bug in the kernel code that triggers the interrupt (and all other interrupts are routed only to CPU 1 for some equally bizarre reason), by "kernel code" do you xnu kernel code or our driver kernel code that we need to debug again? ------------------------------------------------------ Let me give you more idea about my driver during loading phase: It does the following steps in sequence: It first enables the memory mapped io for PCI card's register file.( I have 256 registers) Enables the PCI bus mastering, disables the Option ROM on the card. ( I have to do that) Some memory mapped read/write operation on the PCI card to reset the card and processor on the Card. All memory allocation goes here. I can see IOLog message printed here. (Up to here- all works fine) After this message I get this kernel panic on pmap_unnest. Then after this I set the PCI-X card's cache line to 0x10h. (required by hardware) Then I download the firmware using my PCI card's bus mastering. Even till this point interrupts are not enabled. as there is no address that falls in to my driver's program memory area, I could not figure what cause panic. Regards Parav Pandit -----O riginal Message----- From: David Gatwood [mailto:dgatwood@apple.com] Sent: Thursday, October 07, 2004 1:28 AM To: darwin-kernel(a)lists.apple.com Cc: parav.pandit(a)einfochips.com Subject: Re: pmap_unnest: Other processor did not see interruption request On Oct 6, 2004, at 12:05 PM, parav.pandit(a)einfochips.com wrote: My IOKit driver has provider Class IOPCIDevice. and Driver class is derived from IOSCSIParallelInteraceController. Running my driver on G5 PCI-X dual 1.8GHz system. When I load the driver using kextload -v myDriver.kext Randomly I get kernel panic at this point: ("pmap_unnest: Other processor (%d) did not see interruption request\n", i); . . . panic(cpu 1): pmap_unnest: Other processor (0) did not see interruption request That's a little bizarre. One CPU failed to acknowledge the interrupt from the other CPU (or any other interrupt, for that matter) for over half a second. That really shouldn't be possible. The obvious possibilities are either: 1. CPU 0 hung in an interrupt handler (infinite loop in an I/O Filterpossibly caused by some race conditionnot likely to be your driver, though, since I don't think it would be running yet), 2. There's some really, really bizarre bug in the kernel code that triggers the interrupt (and all other interrupts are routed only to CPU 1 for some equally bizarre reason), or 3. You have a defective CPU and/or main logic board. The third one makes the most sense. It would be sporadic because the code sometimes executes on CPU 0, and that CPU may be able to reliably signal CPU 1, but not vice-versa. Other drivers wouldn't see the same behavior because pmap-based hardware address mapping doesn't occur for any peripherals except for PCI/AGP (and a few built-in devices), and those drivers are probably loaded before the second CPU is enabled, which means that it would always be CPU 0 signaling CPU 1, while the panic is the reverse of that. Try running the hardware test CD and see if it turns anything up. Try unloading and loading another PCi driver and see if the same behavior occurs. Finally, try disabling one CPU with CHUD. Let us know what happens. David _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: swap space
by Brian O'Neill 09 Oct '04

09 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thanks, Brian T. O'Neill _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On Oct 8, 2004, at 4:57 PM, Justin Walker wrote: The latter. Pages are added to available swap files according to available free space, which (obviously) comes and goes as the system runs. As the demand for paging space decreases, generally by having processes exit, more space gets freed up. Depending on other mysterious algorithms, the pager will "clean up" by moving pages from later, lightly-used, swap files to earlier swap files, and eventually remove unused swap files. This is what I was hoping was the correct answer, thats what the dynamic_pager code seemed to point to, but that code is within all the mach message handling stuff which I'm still not that familiar with. So the best way I can see to get my code to return the total amount of swap space currently configured would be to do an opendir/stat loop on the /private/var/vm/ directory and just add up the space. Now I just need to figure out how much of the space is actually allocated. This email sent to site_archiver(a)lists.apple.com

1 0

Re: swap space
by Justin Walker 08 Oct '04

08 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Oct 8, 2004, at 13:41, Brian O_Neill wrote: True. Don't know the answer to this one. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | Some people have a mental | horizon of radius zero, and | call it their point of view. | -- David Hilbert *--------------------------------------*-------------------------------* _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… Please correct me if I'm wrong, but it's my understanding that in Darwin 7.x dynamic_pager will create brand new swap files of increasingly larger sizes as swap demand increases. When it does create a larger file, When needed; you can review the code in the dynamic pager, which handles this, to see what "when needed" really means. I think it's a function of available free space and recent demand, but it's been a while... does it transfer the swapped pages to the new file, or does it keep all files it has created in use? The latter. Pages are added to available swap files according to available free space, which (obviously) comes and goes as the system runs. As the demand for paging space decreases, generally by having processes exit, more space gets freed up. Depending on other mysterious algorithms, the pager will "clean up" by moving pages from later, lightly-used, swap files to earlier swap files, and eventually remove unused swap files. Also, is there any function that will return, or a kernel symbol you can lookup, to determine exactly how many pages have been swapped to disk. The pstat command is supposed to have a -s option to list this data, but it is currently non-functional. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Update frequency of Time Base reg
by Josh de Cesare 08 Oct '04

08 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thanks much, -Arrigo _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/decesare%40apple.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… Arrigo, It is currently about 33 MHz on all G5 systems. To get the real answer programmatically call sysctlbyname with "hw.tbfrequency". Josh At 2:31 PM -0700 10/8/04, Arrigo Benedetti wrote: I'm looking for information about the update frequency of the Time Base register for the latest dual G5 Powermac and Xserve models. This email sent to decesare(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Update frequency of Time Base reg
by Arrigo Benedetti 08 Oct '04

08 Oct '04

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com User-agent: Mozilla Thunderbird 0.8 (Windows/20040913) Thanks much, -Arrigo _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… I'm looking for information about the update frequency of the Time Base register for the latest dual G5 Powermac and Xserve models. This email sent to site_archiver(a)lists.apple.com

1 0