April 2005 - Darwin-kernel - lists.apple.com

Re: Using ctl_enqueuembuf
by Justin Walker 20 Apr '05

20 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Apr 20, 2005, at 13:01, Carl Smith wrote: Oh sorry. In my kernel I use the function 'int ctl_register(struct kern_ctl_reg *userctl, void *userdata, kern_ctl_ref *ctlref)'. In my code it looks like Ctl_register(My_kern_ctl_struct, 0, &my_kern_ctl_ref); Now in kernel_control calls I would use the returned value in my_kern_ctl_ref in subsequent calls to functions like ctl_enqueuembuf, as follows: ctl_enqueuembuf(my_kern_ctl_ref, my_mbuf, myFlags); so I am asking seeing as how ctl_enqueuembuf can obtain the struct sockbuf pointer from my_kern_ctl_ref, can I also obtain a socket descriptor from my_kern_ctl_ref? In short how can I find/obtain the socket descriptor(int) that the kernel is currently talking/connected to? Regards, Justin -- Justin C. Walker, Curmudgeon at Large Institute for General Semantics ----------- My wife 'n kids 'n dog are gone, I can't get Jesus on the phone, But Ol' Milwaukee's Best is my best friend. ----------- _______________________________________________ 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.… [snip] You are slightly confused. Socket descriptors are user-mode gizmos, namely, file descriptors that are returned by the socket() call. Nothing in the kernel uses file descriptors (save for those calls that manipulate them on behalf of the user). I'm still not clear on what you want. For a start, review the kern_control source (xnu/bsd/kern/kernel_control.c) to see how things are set up. I don't know that there is any source that currently uses this stuff (to use as an example); someone else will have to answer that. This email sent to site_archiver(a)lists.apple.com

1 0

RE: Using ctl_enqueuembuf
by Carl Smith 20 Apr '05

20 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcVF3k+DOnsCHH3GS9WbN7DIakENhgAAbIWg Thread-topic: Using ctl_enqueuembuf Oh sorry. In my kernel I use the function 'int ctl_register(struct kern_ctl_reg *userctl, void *userdata, kern_ctl_ref *ctlref)'. In my code it looks like Ctl_register(My_kern_ctl_struct, 0, &my_kern_ctl_ref); Now in kernel_control calls I would use the returned value in my_kern_ctl_ref in subsequent calls to functions like ctl_enqueuembuf, as follows: ctl_enqueuembuf(my_kern_ctl_ref, my_mbuf, myFlags); Looking at the ctl_enqueuembuf source code, I see that ctl_enqueuembuf uses sbappend(), and to get sbappend to work ctl_enqueuembuf does this: ctl_enqueuembuf (void *ctlref, struct mbuf *m, u_int32_t flags) { Struct ctl *ctl = (struct ctl *)ctlref; Struct socket *so = (struct socket*)ctl->skt; ...... sbappend(&so->so_rcv, m); ..... return 0; } so I am asking seeing as how ctl_enqueuembuf can obtain the struct sockbuf pointer from my_kern_ctl_ref, can I also obtain a socket descriptor from my_kern_ctl_ref? In short how can I find/obtain the socket descriptor(int) that the kernel is currently talking/connected to? Thanks Carl -----Original Message----- From: darwin-kernel-bounces+csmith=fortresstech.com(a)lists.apple.com [mailto:darwin-kernel-bounces+csmith=fortresstech.com@lists.apple.com] On Behalf Of Justin Walker Sent: Wednesday, April 20, 2005 3:22 PM To: darwin-kernel Dev Subject: Re: Using ctl_enqueuembuf On Apr 20, 2005, at 12:01, Carl Smith wrote: [snip] > So say I want to use a sendto or write, I used 'ctl_register()' in my > kernel to register my Kernel Controller, is there some way to obtain > the > socket fd that sendto() or write() require using the kern_ctl_ref > return > by the call to ctl_register()? Sorry, Carl. I can't parse this sentence. Can you rephrase? Regards, Justin -- Justin C. Walker, Curmudgeon at Large Institute for General Semantics ----------- My wife 'n kids 'n dog are gone, I can't get Jesus on the phone, But Ol' Milwaukee's Best is my best friend. ----------- _______________________________________________ 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/csmith%40fortresste ch.com This email sent to csmith(a)fortresstech.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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Using ctl_enqueuembuf
by Justin Walker 20 Apr '05

20 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Apr 20, 2005, at 12:01, Carl Smith wrote: Sorry, Carl. I can't parse this sentence. Can you rephrase? Regards, Justin -- Justin C. Walker, Curmudgeon at Large Institute for General Semantics ----------- My wife 'n kids 'n dog are gone, I can't get Jesus on the phone, But Ol' Milwaukee's Best is my best friend. ----------- _______________________________________________ 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.… [snip] So say I want to use a sendto or write, I used 'ctl_register()' in my kernel to register my Kernel Controller, is there some way to obtain the socket fd that sendto() or write() require using the kern_ctl_ref return by the call to ctl_register()? This email sent to site_archiver(a)lists.apple.com

1 0

RE: Using ctl_enqueuembuf
by Carl Smith 20 Apr '05

20 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcVFX5QdXb/F6+b4S5+WP8/3kbPyXgAe6S2w Thread-topic: Using ctl_enqueuembuf Thanks Justin, things make a little more sense. I think I was confusing myself looking at the TCPlogger example, and I guess I was assuming mbufs where doing something 'magical' on my structure. So say I want to use a sendto or write, I used 'ctl_register()' in my kernel to register my Kernel Controller, is there some way to obtain the socket fd that sendto() or write() require using the kern_ctl_ref return by the call to ctl_register()? Thanks Carl -----Original Message----- From: darwin-kernel-bounces+csmith=fortresstech.com(a)lists.apple.com [mailto:darwin-kernel-bounces+csmith=fortresstech.com@lists.apple.com] On Behalf Of Justin Walker Sent: Wednesday, April 20, 2005 12:15 AM To: darwin-kernel Dev Subject: Re: Using ctl_enqueuembuf On Apr 19, 2005, at 11:26, Carl Smith wrote: > Ok thinking.... Always a good sign :-} > I am coping the sizeof the structure. Now only because I was looking at > the TCPlogger example, which passed a rather large structure 'struct > TCPLogEntry', in which it had sub structures and in turn had buffers > with log data, that I thought the mbuf code took care of the structure > and everything in it. Keep in mind that mbufs are kernel structures that describe data buffers. There's no magic in them, and the semantics that you apply to their content is a complete mystery to the support routines that deal with the mbufs. In particular, if you have a buffer pointer pointing off into the great beyond, ctl_enqueuembuf won't be aware of it. > But the total data, would be the sizeof the struct plus the sizeof > buf1. That's what you want to transfer to user mode, but to the kernel code looking at the mbuf, the total data is the size of the structure. > I guess I am not clear on how to get buf1 up to userland using > ctl_enqueuembuf. > > Do I need to make two calls to ctl_enqueuembuf, with the struct then > the > buf1? Of is there a way to get both buffers into my mbuf? buf1 points to a malloc'd structure, so you can't treat that as an mbuf. Also, don't forget to free that buffer when you are done with it. See below. > Then if I have to make two calls to ctl_enqueuembuf, in the kernel > side, > I am not certain how 'catch' it in user land? I would take an alternate approach. For a start, review the man page for recv(), and the Wright/Stevens book for background. There may be something in there that strikes a chord with you. In short, you will have to turn all your data containers into mbufs if you want to use mbuf-handling routines to deal with them. Regards, Justin -- Justin C. Walker, Curmudgeon-at-Large /~\ The ASCII \ / Ribbon Campaign X Help Cure HTML Email / \ _______________________________________________ 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/csmith%40fortresste ch.com This email sent to csmith(a)fortresstech.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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Using ctl_enqueuembuf
by Justin Walker 20 Apr '05

20 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Apr 19, 2005, at 11:26, Carl Smith wrote: Always a good sign :-} I am coping the sizeof the structure. Now only because I was looking at the TCPlogger example, which passed a rather large structure 'struct TCPLogEntry', in which it had sub structures and in turn had buffers with log data, that I thought the mbuf code took care of the structure and everything in it. I guess I am not clear on how to get buf1 up to userland using ctl_enqueuembuf. Regards, Justin -- Justin C. Walker, Curmudgeon-at-Large /~\ The ASCII \ / Ribbon Campaign X Help Cure HTML Email / \ _______________________________________________ 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.… Ok thinking.... Keep in mind that mbufs are kernel structures that describe data buffers. There's no magic in them, and the semantics that you apply to their content is a complete mystery to the support routines that deal with the mbufs. In particular, if you have a buffer pointer pointing off into the great beyond, ctl_enqueuembuf won't be aware of it. But the total data, would be the sizeof the struct plus the sizeof buf1. That's what you want to transfer to user mode, but to the kernel code looking at the mbuf, the total data is the size of the structure. Do I need to make two calls to ctl_enqueuembuf, with the struct then the buf1? Of is there a way to get both buffers into my mbuf? buf1 points to a malloc'd structure, so you can't treat that as an mbuf. Also, don't forget to free that buffer when you are done with it. See below. Then if I have to make two calls to ctl_enqueuembuf, in the kernel side, I am not certain how 'catch' it in user land? I would take an alternate approach. For a start, review the man page for recv(), and the Wright/Stevens book for background. There may be something in there that strikes a chord with you. In short, you will have to turn all your data containers into mbufs if you want to use mbuf-handling routines to deal with them. This email sent to site_archiver(a)lists.apple.com

1 0

RE: Using ctl_enqueuembuf
by Carl Smith 19 Apr '05

19 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcVE+ZiCeJb5zamsRlK/Y1su+nF9sgAEfFFQ Thread-topic: Using ctl_enqueuembuf Ok thinking.... I am coping the sizeof the structure. Now only because I was looking at the TCPlogger example, which passed a rather large structure 'struct TCPLogEntry', in which it had sub structures and in turn had buffers with log data, that I thought the mbuf code took care of the structure and everything in it. But the total data, would be the sizeof the struct plus the sizeof buf1. I guess I am not clear on how to get buf1 up to userland using ctl_enqueuembuf. Do I need to make two calls to ctl_enqueuembuf, with the struct then the buf1? Of is there a way to get both buffers into my mbuf? Then if I have to make two calls to ctl_enqueuembuf, in the kernel side, I am not certain how 'catch' it in user land? Thanks Carl -----Original Message----- From: darwin-kernel-bounces+csmith=fortresstech.com(a)lists.apple.com [mailto:darwin-kernel-bounces+csmith=fortresstech.com@lists.apple.com] On Behalf Of Justin Walker Sent: Tuesday, April 19, 2005 12:01 PM To: darwin-kernel Dev Subject: Re: Using ctl_enqueuembuf On Apr 19, 2005, at 8:17, Carl Smith wrote: > I am using the function ctl_enqueuembuf to pass data up to my userland > process from my NKE. My problem lies in that I am passing a structure > that has some pointers to buffers within this structure and these > buffers are always showing empty/zeros when I receive them on the > userland side. I am using the example from TCPlogger, in how TCPlogger > is passing a structure up to it's user land, but for some reason for > me it is not working. Here's a hint: how much data are you copying up to user space, and what is in that data? Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for General Semantics -------- If you're not confused, You're not paying attention -------- _______________________________________________ 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/csmith%40fortresste ch.com This email sent to csmith(a)fortresstech.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.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: passing buffer from User Client to Driver
by Godfrey van der Linden 19 Apr '05

19 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com G'day Imran, I only have a couple of observations. Can you change "com_MySoftwareCompany" to "com_spsoftindia"? This is what I do when I debug code that I'm sure works. Godfrey On 04/18/2005, at 23:21 , Imran wrote: Hi GodFrey, IOMemoryDescriptor address value in driver = 0x11b2680 Please take your time to go through the code segment of my driver. IOReturn retDrWrite = kIOReturnSuccess; IOStorageCompletion completion; // If the user client's 'open' method was never called, return kIOReturnNotOpen. if (!isOpen()) /* if-else Added - 15 Apr 2005 */ { IOLog("\n Failed to open userclient for the driver....\n"); return kIOReturnNotOpen; } else { IOLog("\n Userclient is open for the driver....\n"); } IOByteCount readBytes; UInt64 byteStart = 0; IOByteCount bufLength = inBuffer->getLength(); IOLog("\n bufLength = %lu\n", bufLength); //char *myBuff = new char[bufLength]; char myBuff[512]; IOLog("\n Calling readbytes.......in driver\n"); readBytes = inBuffer->readBytes(0, myBuff, bufLength); getProvider()->write(this, byteStart, inBuffer, completion); return retWrite; } _______________________________________________ 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.… can assure you that the IOMemoryDescriptor works and readBytes does return data from user land as expected. So now it is just an exercise in debugging. This is what I'd do, first check the arguments to the memory descriptor, do you have the correct task and virtual address? Does the user land task indeed have the expected data? Have you prepared() the descriptor? Is it still prepared when the I/O is performed, i.e. if the I/O is asynch did you complete() the descriptor prematurely? As per your reply, I had eliminated the mapping calls and change variables (vInStruct to InStruct)in my userclient. But still I am facing the same problem, my IOMemoryDescriptor buffer is not getting passed from userclient to driver. When I print the IOMemoryDescriptor address in driver it shows the following output: The Output of readBytes is as follows. MyFilterScheme:: readBytes = 512 byteStart = 0 bufLength = 512 buffer = Or do we need to use vm_allocate for achieving the buffer passing from user space application to the driver through user client. *********************************************************** // kMyDoWrite IOReturn com_MySoftwareCompany_driver_MyFilterScheme::DoDriverWrite (IOMemoryDescripto r * inBuffer) { IOLog("\n Inside com_MySoftwareCompany_driver_MyFilterScheme::myDoWrite call\n"); IOLog("%s[%p]::%s\n", getName(), this, __FUNCTION__); IOLog("\n IOMemoryDescriptor value in driver = %p\n", inBuffer); IOLog("\n IOMemoryDescriptor buffer in driver = %s\n", inBuffer); IOLog("MyFilterScheme:: readBytes = %lu byteStart = %lu bufLength = %lu buffer = %s \n", (long)readBytes, (long)byteStart, (long)bufLength, myBuff); The Outputs are: IOMemoryDescriptor value in driver = 0x11b2680 IOMemoryDescriptor buffer in driver = bufLength = 512 MyFilterScheme:: readBytes = 512 byteStart = 0 bufLength = 512 buffer = This email sent to site_archiver(a)lists.apple.com

1 0

Re: Cann't make connection to a graphics IOService
by Jason Seng 19 Apr '05

19 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi Amit, Please make sure you reply to the list, so all can benefit. :) Cheers, -- J-Sun Seng Networking & Communications / Hardware Developer Technical Support Apple Computer, Inc. #1 Infinite Loop M/S: 303-2T Cupertino, CA 95014 _______________________________________________ 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.… At 9:16 +0530 4/19/05, Amit Gupta wrote: Hi Jason, Thanks for ur reply. Actually I am trying to call IOFBGetFramebufferInformationForAperture() fn. This fn requires io_connect_t handle obtained from IOServiceOpen for IOFramebuffer Service. If Window Server has opened it exclusively, then how can I call IOFBGetFramebufferInformationForAperture fn. The short answer is you can't. IOFramebufferUserClient only supports one client. Perhaps if you were to explain what your goal is, or why you need the information from IOFBGetFramebufferInformationForAperture, we may be able to come up with an alternate solution. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Determine if user can admin
by Quinn 19 Apr '05

19 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com For "sit" read "site". S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Technical Support * Networking, Communications, Hardware _______________________________________________ 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.… At 17:20 +0100 19/4/05, Quinn wrote: And, if the user, or a sit admin, doesn't like your default right specification, they can change it. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Determine if user can admin
by Quinn 19 Apr '05

19 Apr '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com I'm trying to set things up in a manner that is more convenient for a user who is allowed to administer the computer. Sort of like how the OS used to require you to authorize constantly to do just about anything, then things were made easier for users who are designated administrators, so they only have to authenticate to perform tasks that actually alter the system. It sounds like you should integrate with Authorization Services. <http://developer.apple.com/technotes/tn2002/tn2095.html> <http://developer.apple.com/samplecode/AuthForAll/AuthForAll.html> S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Technical Support * Networking, Communications, Hardware _______________________________________________ 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.… At 9:04 -0700 19/4/05, Eric Long wrote: Because the OS caches the users credentials when they log in, you can create a custom right specification that requires the user be part of the admin group, and that right specification will benefit from the cached credentials. And, if the user, or a sit admin, doesn't like your default right specification, they can change it. This email sent to site_archiver(a)lists.apple.com

1 0