March 2006 - Darwin-kernel - lists.apple.com

Re: What to do if no protocol family?
by Brian Doyle 02 Mar '06

02 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thanks Josh, -b On Mar 2, 2006, at 10:15 AM, Josh Graessley wrote: Hi Brian, -josh On Mar 2, 2006, at 12:56 AM, Brian Doyle wrote: Hello, Any help will be greatly appreciated, thanks!! -b _______________________________________________ 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 will be perfect. I tried PF_NDRV and it didn't even attach (EEXIST) so fortunately that's not something that can even be done in error. You can use a dynamically allocated large value. I believe the best way to get one of these values is to use kev_vendor_code_find. This will give you a unique 32bit value that you can use for the protocol family. There should be a better way to coordinate protocol families, but there isn't. The protocol family is just used for intermediate matching. The packet comes it, is it matched against ethernet demux descriptors. When it finds one that matches, the protocol family is returned. The interface then looks for the input function that was registered for that protocol family, the one you specified when attaching your protocol to the interface. Don't use PF_NDRV, that will really screw things up. A specific protocol family can only attach once, if you use PF_NDRV, that will prevent everything else that uses PF_NDRV from working. I'm writing a driver that needs to read and write raw ethernet frames. After reading source code all day I've come to the conclusion that hooking in to the dlil with ifnet_attach_protocol () would be great, except that ether_demux() dispatches to protocol handlers by looking at both the ethernet type field and the protocol family that was specified when the protocol was registered. For my purposes I want to send and receive frames that have a specific ethernet type (IANA registered but non- standard). There are no registered protocol families for this type however. My question is, what should I use for the protocol family (if in fact hooking in via ifnet_attach_protocol() is the correct thing to do)? I could use PF_NDRV which seems to be for raw frames, but if I do I suspect that will conflict with userland processes that wish to use PF_NDRV at the socket level. I could also just use a value greater than AF_MAX, but that seems a bit dodgy as it might collide later with something else. If I'm thinking about this all wrong I'd love to hear some other ideas. I *do* need to be in the kernel as I'm interfacing with other drivers but all I need is to send and receive ethernet frames with a specific type while not interfering with the rest of the network stack. _______________________________________________ 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/jgraessley% 40apple.com This email sent to jgraessley(a)apple.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: What to do if no protocol family?
by Josh Graessley 02 Mar '06

02 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi Brian, -josh On Mar 2, 2006, at 12:56 AM, Brian Doyle wrote: Hello, Any help will be greatly appreciated, thanks!! -b _______________________________________________ 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 You can use a dynamically allocated large value. I believe the best way to get one of these values is to use kev_vendor_code_find. This will give you a unique 32bit value that you can use for the protocol family. There should be a better way to coordinate protocol families, but there isn't. The protocol family is just used for intermediate matching. The packet comes it, is it matched against ethernet demux descriptors. When it finds one that matches, the protocol family is returned. The interface then looks for the input function that was registered for that protocol family, the one you specified when attaching your protocol to the interface. Don't use PF_NDRV, that will really screw things up. A specific protocol family can only attach once, if you use PF_NDRV, that will prevent everything else that uses PF_NDRV from working. I'm writing a driver that needs to read and write raw ethernet frames. After reading source code all day I've come to the conclusion that hooking in to the dlil with ifnet_attach_protocol() would be great, except that ether_demux() dispatches to protocol handlers by looking at both the ethernet type field and the protocol family that was specified when the protocol was registered. For my purposes I want to send and receive frames that have a specific ethernet type (IANA registered but non-standard). There are no registered protocol families for this type however. My question is, what should I use for the protocol family (if in fact hooking in via ifnet_attach_protocol() is the correct thing to do)? I could use PF_NDRV which seems to be for raw frames, but if I do I suspect that will conflict with userland processes that wish to use PF_NDRV at the socket level. I could also just use a value greater than AF_MAX, but that seems a bit dodgy as it might collide later with something else. If I'm thinking about this all wrong I'd love to hear some other ideas. I *do* need to be in the kernel as I'm interfacing with other drivers but all I need is to send and receive ethernet frames with a specific type while not interfering with the rest of the network stack. _______________________________________________ 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/jgraessley% 40apple.com This email sent to jgraessley(a)apple.com smime.p7s

1 0

What to do if no protocol family?
by Brian Doyle 02 Mar '06

02 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hello, Any help will be greatly appreciated, thanks!! -b _______________________________________________ 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 writing a driver that needs to read and write raw ethernet frames. After reading source code all day I've come to the conclusion that hooking in to the dlil with ifnet_attach_protocol() would be great, except that ether_demux() dispatches to protocol handlers by looking at both the ethernet type field and the protocol family that was specified when the protocol was registered. For my purposes I want to send and receive frames that have a specific ethernet type (IANA registered but non-standard). There are no registered protocol families for this type however. My question is, what should I use for the protocol family (if in fact hooking in via ifnet_attach_protocol() is the correct thing to do)? I could use PF_NDRV which seems to be for raw frames, but if I do I suspect that will conflict with userland processes that wish to use PF_NDRV at the socket level. I could also just use a value greater than AF_MAX, but that seems a bit dodgy as it might collide later with something else. If I'm thinking about this all wrong I'd love to hear some other ideas. I *do* need to be in the kernel as I'm interfacing with other drivers but all I need is to send and receive ethernet frames with a specific type while not interfering with the rest of the network stack. This email sent to site_archiver(a)lists.apple.com

1 0

Re: faults in a kext
by Mike Smith 02 Mar '06

02 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Mar 1, 2006, at 7:22 AM, Joshua LeVasseur wrote: Not in any sort of straightforward or supported manner. = Mike _______________________________________________ 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 experimenting with kext development on x86, and I'd like to know how Darwin handles faults that occur due to instructions executed by my kext, particularly the general protection fault. I don't have an x86 Apple machine, and am using a friend's machine, which I'd like to avoid crashing. As a general rule, kext development is accompanied by a great deal of crashing. Does Darwin halt the kernel if a kext raises a GP? GP faults are actually less common than you might expect; you're more likely to generate un-resolvable pagefaults or destroy data belonging to other parts of the kernel. Does Darwin survive, but terminate a kext if the kext raises a GP? No, kexts run in the same context as the rest of the kernel. As noted above, quite often the damage you cause will not directly affect any thread in your kext. Is it possible to tell Darwin that a particular instruction may raise a GP, and if it does raise a GP, then Darwin should call my dedicated error handler that cleans up and returns the system to a nominal state? This email sent to site_archiver(a)lists.apple.com

1 0

LOCKLEAF is not used in Darwin8?
by Lex Dejavu 02 Mar '06

02 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=R0ZvLpZDhVR6mBvEQbjYsDSzKUksF0FrMc2YLinTtI7SLURCrJ9poK+rsTLt+MUiOlU6L21NEOwFkF4CMIdK3AZyuBWrHwnziF2T5FKNLNhAUknTVcbnqqCy4A8GxfPy6eWEaWW940US2rnwbXLBu06evX2rRvo6IvPW4XHc8a4= ; _______________________________________________ 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: faults in a kext
by Kim-Thomas Moeller 01 Mar '06

01 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com User-agent: Debian Thunderbird 1.0.7 (X11/20051017) Thanks, Kim -Josh _______________________________________________ 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/kim%40kim-com.de _______________________________________________ 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.… There is support for GP exceptions from within the kernel (see xnu-792.6.2/osfmk/i386/trap.c, line 371). The recovery table maps fault addresses to recovery addresses. Entries for the recovery table are allocated statically by the macro RECOVER (in xnu-792.6.2/osfmk/i386/locore.s). In addition to the recovery table mechanism, each thread can specify its own recovery address, which is called when there is no recovery table entry. If neither the kernel nor the kernel debugger handle the GP, a panic will occur. Joshua LeVasseur wrote: I'm experimenting with kext development on x86, and I'd like to know how Darwin handles faults that occur due to instructions executed by my kext, particularly the general protection fault. I don't have an x86 Apple machine, and am using a friend's machine, which I'd like to avoid crashing. Does Darwin halt the kernel if a kext raises a GP? Does Darwin survive, but terminate a kext if the kext raises a GP? Is it possible to tell Darwin that a particular instruction may raise a GP, and if it does raise a GP, then Darwin should call my dedicated error handler that cleans up and returns the system to a nominal state? This email sent to kim(a)kim-com.de This email sent to site_archiver(a)lists.apple.com

1 0

faults in a kext
by Joshua LeVasseur 01 Mar '06

01 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -Josh _______________________________________________ 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 experimenting with kext development on x86, and I'd like to know how Darwin handles faults that occur due to instructions executed by my kext, particularly the general protection fault. I don't have an x86 Apple machine, and am using a friend's machine, which I'd like to avoid crashing. Does Darwin halt the kernel if a kext raises a GP? Does Darwin survive, but terminate a kext if the kext raises a GP? Is it possible to tell Darwin that a particular instruction may raise a GP, and if it does raise a GP, then Darwin should call my dedicated error handler that cleans up and returns the system to a nominal state? This email sent to site_archiver(a)lists.apple.com

1 0

Re: VFS plug-in Mac OS X 10.4
by Justin C. Walker 01 Mar '06

01 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Feb 28, 2006, at 19:26 , Lex Dejavu wrote: Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large Institute for the Enhancement of the Director's Income -------- Experience is what you get when you don't get what you want. -------- _______________________________________________ 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.… Well, there is no complete document about KPI yet, source code of xnu can help you a lot(also you can read source package of ntfs etc...). btw: I heard rumors like Apple will NOT release source code of xnu and drivers in the future, anyone can give more information? Apple has not said they will not release the source for xnu and drivers. You likely won't hear anything on this list (other than the rumors you mention) until either the sources appear, or Apple makes an announcement (presumably through their website). This email sent to site_archiver(a)lists.apple.com

1 0

Re: VFS plug-in Mac OS X 10.4
by Lex Dejavu 01 Mar '06

01 Mar '06

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=Message-ID:Received:Date:From:Subject:To:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding; b=xmomFYTKiGMvwfPbC1SU/izvSKh4bHkrMuI0PcVgCbyARQgQMfSzYek5iFTGW7QkosPu5IjkjtiZ0HyE4Pa0R/FB8zKVLEFiD+duSO18n0p1F7ylfHuCBWr6AYz/7UTzJQ0tmTzHn7IxcwKhCsf9C0/RR+IyZC98j9Jhf0WjsyU= ; _______________________________________________ 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