February 2002 - Darwin-kernel

bug in headers (how to file a bug for such things?)
by Francis Bouchard 07 Feb '02

07 Feb '02

There is a bug in the Kernel headers: 1- create a new kext project (not IOKit). 2- this will not compile: #include <mach/mach_types.h> #include <IOKit/IOLib.h> #include <sys/systm.h> kern_return_t test_start (kmod_info_t * ki, void * d) { return KERN_SUCCESS; } kern_return_t test_stop (kmod_info_t * ki, void * d) { return KERN_SUCCESS; } but this will COMPILE: #include <mach/mach_types.h> #include <sys/systm.h> // SWAPPED THE INCLUDES #include <IOKit/IOLib.h> // SWAPPED THE INCLUDES kern_return_t test_start (kmod_info_t * ki, void * d) { return KERN_SUCCESS; } kern_return_t test_stop (kmod_info_t * ki, void * d) { return KERN_SUCCESS; } Franics _______________________________________________ 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

User Client Question
by Frank McGirt 07 Feb '02

07 Feb '02

Hi, I hope this is an easy question but I have been having trouble getting the parameters of the IOConnectMethods from user space to match up with my methods in my User Client in my kernel driver. I am writing a driver for a custom PCI card with an onboard large (8MB) memory buffer which connects to external hardware. On command from user space, the PCI card will either get a block of data from the external hardware to the onboard buffer of put a block of data from the buffer to the external hardware. Several parameters must be passed on either the get or the put to setup the hardware for the transfer. These are in addition to the user space buffer addresses and lengths. So I have defined two user client methods, getBlock as StructIStructO, and putBlock as ScalarIStructI and in the UserClient::sMethods these appear as { // kGetBlock method kMethodObjectThis, (IOMethod) &PCIFMDriverUserClient::getBlock, kIOUCStructIStructO, // struct input, struct output sizeof(MapRegisterStruct), // input struct size 0xffffffff // variable size of the output structure }, { // kPutBlock method kMethodObjectThis, (IOMethod) &PCIFMDriverUserClient::putBlock, kIOUCScalarIStructI, // scalar input, struct input 0, // no scalar input 0xffffffff // variable size of the input structure }, where I intend to use the variable size structures to pass the data back and forth to user space. The MapRegisterStruct structure is defined as typedef struct MapRegisterStruct { UInt32 *userAddress; // user space buffer address UInt32 vmeAddress; UInt32 numberBytes; UInt16 addressModifier; UInt16 addressSpace; UInt8 accessWidth; } MapRegisterStruct; and contains the parameters needed as well as the user space buffer address and byte length of transfer. My user client methods are then defined as virtual IOReturn getBlock(void *vInData,void *vOutData, void *, void *, void *); virtual IOReturn putVmeLongData(void *vInData, void *, void *, void *, void *, void *); I guess I am really confused about how to pass a variable size user buffer address to/from a user client running in the kernel. I have looked at several examples but none seem to match exactly what I want o do or if I try them, they don't work. I have also tried seperating the transfer parameters into their own structure (StructI) and passing the user space buffer address and length as individual arguments - haven't gotten this to work either. I would appreciate any suggestions or links to appropriate documentation, sample code, etc. And if this is the wrong list, please let me know. Frank McGirt fmcgirt(a)mac.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: IOPanic ans assert not working ? [partly solved!]
by Justin C. Walker 06 Feb '02

06 Feb '02

On Wednesday, February 6, 2002, at 10:46 AM, Francis Bouchard wrote: Ok here is my investigation results: After swapping almost anything swappable with another machine, I found that it might be related to the Apple ADC connector. Why ? If I plug a standard vga monitor into my GeForce3 it DOES panic properly. By the way I did remove the NDRV and I cleared the ROM of my second monitor board. This makes sure OS X does not see it as a display. This was the first problem: the panic was going the the second crt which was not able to ahndle it... So I will contact DTS for this strange behavior since I don't know if it is the NVidia driver or Darwin! But now is I use the VGA port of my GeForce(AGP) I get my panic... Glad you got it figured out. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | Men are from Earth. | Women are from Earth. | Deal with it. *--------------------------------------*-------------------------------* _______________________________________________ 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: IOPanic ans assert not working ?
by David A. Gatwood 06 Feb '02

06 Feb '02

On Tue, 5 Feb 2002, Justin C. Walker wrote: > On Tuesday, February 5, 2002, at 05:38 PM, David A. Gatwood wrote: > > > On the macs, they quiesce the OF, which means that you can't use OF > > drivers in the kernel. I can only think of one OS that doesn't do this > > (NetBSD). It's just a lot of effort to keep OF around. > > And, of course, Solaris, where OF was first perpetrated on an > unsuspecting world. Sorry, should have been more clear. I meant to say that I can only think of one OS that doesn't do this on the Mac/PPC platform. :-) David --------------------------------------------------------------------- Check out my weekly web comic: http://www.techmagazine.org

1 0

Re: IOPanic ans assert not working ?
by Justin C. Walker 06 Feb '02

06 Feb '02

On Tuesday, February 5, 2002, at 05:38 PM, David A. Gatwood wrote: On Tue, 5 Feb 2002, Andrew Gallatin wrote: David A. Gatwood writes: You can't use OF for video by the time the Mac OS X kernel starts. The mach console is legacy MkLinux code that still manages to scratch its way to the surface every so often. ;-) Ahh. I see.. I'm used to alphas where you can generally switch back to the SRM console firmware for I/O if you're in a bad state.. On the macs, they quiesce the OF, which means that you can't use OF drivers in the kernel. I can only think of one OS that doesn't do this (NetBSD). It's just a lot of effort to keep OF around. And, of course, Solaris, where OF was first perpetrated on an unsuspecting world. Speaking of consoles. Does anybody happen to know if one of the 2 or more serial port products for G4s can be used as a serial console by both OFW and Darwin? I'm talking about the things that plug into the modem socket on the motherboard and give you a Mini Din 9 connector. They should. They basically just route the lines off the motherboard SCC to a port. As long as the serial console code works correctly, it should be pretty uncaring about what's attached to the SCC's output. We've used these (e.g., the so-called 'stealth' cards) to debug bringup in the past. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | If you're not confused, | You're not paying attention *--------------------------------------*-------------------------------*

1 0

Re: IOPanic ans assert not working ?
by David A. Gatwood 06 Feb '02

06 Feb '02

On Tue, 5 Feb 2002, Andrew Gallatin wrote: > David A. Gatwood writes: > > > > You can't use OF for video by the time the Mac OS X kernel starts. The > > mach console is legacy MkLinux code that still manages to scratch its way > > to the surface every so often. ;-) > > Ahh. I see.. I'm used to alphas where you can generally switch back > to the SRM console firmware for I/O if you're in a bad state.. On the macs, they quiesce the OF, which means that you can't use OF drivers in the kernel. I can only think of one OS that doesn't do this (NetBSD). It's just a lot of effort to keep OF around. > Speaking of consoles. Does anybody happen to know if one of the 2 or > more serial port products for G4s can be used as a serial console by > both OFW and Darwin? I'm talking about the things that plug into the > modem socket on the motherboard and give you a Mini Din 9 connector. They should. They basically just route the lines off the motherboard SCC to a port. As long as the serial console code works correctly, it should be pretty uncaring about what's attached to the SCC's output. Later, David --------------------------------------------------------------------- Check out my weekly web comic: http://www.techmagazine.org

1 0

Re: IOPanic ans assert not working ?
by Andrew Gallatin 06 Feb '02

06 Feb '02

David A. Gatwood writes: > > You can't use OF for video by the time the Mac OS X kernel starts. The > mach console is legacy MkLinux code that still manages to scratch its way > to the surface every so often. ;-) Ahh. I see.. I'm used to alphas where you can generally switch back to the SRM console firmware for I/O if you're in a bad state.. Speaking of consoles. Does anybody happen to know if one of the 2 or more serial port products for G4s can be used as a serial console by both OFW and Darwin? I'm talking about the things that plug into the modem socket on the motherboard and give you a Mini Din 9 connector. Thanks, Drew

1 0

Re: IOPanic ans assert not working ?
by David A. Gatwood 06 Feb '02

06 Feb '02

On Tue, 5 Feb 2002, Andrew Gallatin wrote: > Given the author is from matrox.com, it is possible he's working on a > video driver... > > What does OS X do if the video driver is hosed? Does it switch to > using Open Firmware for I/O at panic time? Or does it call back into > the possibly hosed video driver which may no longer be able to write > things onto the framebuffer? You can't use OF for video by the time the Mac OS X kernel starts. The mach console is legacy MkLinux code that still manages to scratch its way to the surface every so often. ;-) Panic messages, AFAICT, take a different path through video_console.c than other text, and I would hope that they don't do any callbacks to the video driver. If it did, you could get something like: 1. Driver init routine panics 2. Mach starts panic display. 3. Oh, driver isn't initted. Call init routine. 4. Go to step 1. Which is a bad thing. I don't see any call-backs, though, so I kind of doubt that's what's happening. At least I hope not. :-) Later, David --------------------------------------------------------------------- Check out my weekly web comic: http://www.techmagazine.org

1 0

Re: IOPanic ans assert not working ?
by David A. Gatwood 06 Feb '02

06 Feb '02

On Tue, 5 Feb 2002, Justin C. Walker wrote: > Given that, the call, per se, does not 'hose the kernel'; it just stops > processing. It should print the character string you pass it, and then > grind ot a halt. The fact that it doesn't is a bit odd. Maybe you (he) > mentioned this, but I've lost it: is this call in a kext? Is it just > supposed to call IOPanic()? And another (gross) thought: what happens if you trigger a panic indirectly, e.g. mypanic(char *message) { float k; printf(message); k /= 0; } Does that panic (more) correctly for you? David --------------------------------------------------------------------- Check out my weekly web comic: http://www.techmagazine.org

1 0

Re: IOPanic ans assert not working ?
by Andrew Gallatin 06 Feb '02

06 Feb '02

Given the author is from matrox.com, it is possible he's working on a video driver... What does OS X do if the video driver is hosed? Does it switch to using Open Firmware for I/O at panic time? Or does it call back into the possibly hosed video driver which may no longer be able to write things onto the framebuffer? Drew

1 0