September 2004 - Darwin-dev

Re: Thermal scheduling
by Sam Hart 29 Sep '04

29 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com John wrote: In the scheduler ? I doubt it What are you trying to achieve ? S a m ------------------------------------------------------- Drawing on my fine command of language, I said nothing. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… Does the Darwin 7.4 (or later) kernel scheduler take CPU temperature into account at all? I've been poring over the xnu source trying to find a place where CPU temp is involved in some sort of thread priority or scheduling calculation, but have not found one yet. I'm hoping it'll be faster to just ask here :) The temperature of the CPU(s) is not gleaned by reading a register, at least not on PowerPC. You would set ( in supervisor mode ) a threshold watchpoint, and when the threshold is breached an interrupt is generated. It is CPU model dependent, in fact it might even be optional for PowerPC. Different strategies can be employed to cool down the CPU(s), spinning up a fan isn't a bad one :) How temperature shifts are handled by Apple could well be considered a competitive advantage. This is no word of authority but my guess would be that it is handled by AppleThermal.kext which appears to be closed source. This email sent to site_archiver(a)lists.apple.com

1 0

ipfw Divert Sockets and PPTP
by Jeff Cruikshank 28 Sep '04

28 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi list, Normal (working) scenario: Traffic from mail client: Before: 192.168.1.50:50299 -> 213.213.213.213:143 Modified: 192.168.1.50:50299 -> 127.0.0.1:10143 Traffic to mail client: Before: 127.0.0.1:10143 -> 192.168.1.50:50299 Modified: 213.213.213.213:143 -> 192.168.1.50:50299 PPTP (broken) scenario: Traffic from mail client: Before: 10.214.2.34:50299 -> 213.213.213.213:143 Modified: 10.214.2.34:50299 -> 127.0.0.1:10143 BTW, here are the ipfw rules that I'm using to divert mail traffic: ipfw add 100 divert 15000 tcp from any to any 143 out ipfw add 101 divert 15000 tcp from any 143 to any in ipfw add 102 divert 15000 tcp from any to any 10143 ipfw add 103 divert 15000 tcp from any 10143 to any Thanks for any help! Jeff Cruikshank -- Jeff Cruikshank <jeff(a)pgp.com> Senior Software Engineer, PGP Corporation _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… I have a daemon which performs a sort of aliasing (not compltely unlike natd), to transparently proxy mail traffic between a mail client and a mailserver. For the most part, things work great, but PPTP seems to complicate matters in such a way that my packets don't get diverted to the right interface. The nature of my app is to act as a man in the middle (I know, sounds scary, but I promise it's not malicious), such that IMAP connections from the mail client automatically get redirected to my proxy app (which listens on 127.0.0.1:10143). The proxy app, in turn, connects to the mail server and proxies the traffic between the two connections, as if they were one. When things are running well, packets sent/rcvd from the mail client get processed in the following manner: The problem I'm having with PPTP would seem to be that the address translation that I perform when sending traffic back to the mail client, doesn't cause the packet to get routed correctly (this is a best guess). In my testing, the connect() call times out because the 3-way handshake can't be completed. Traffic to mail client: Before: 127.0.0.1:10143 -> 10.214.2.34:50299 After: 213.213.213.213:143 -> 10.214.2.34:50299 // Packet never received by PPTP interface with addresss 10.214.2.34 After perusing the kernel source (ip_divert.c), I thought that perhaps I could inspect (and modify) the sin_zero field of the struct sockaddr_in to control which interface the packet gets sent to (ppp0). Though I was able to get the interface name this way, changing it seemed to have no effect. I was interested to find that Apple's natd (as configured via the Internet Sharing panel) doesn't seem to work when the same PPTP connection is in place. To test this, I had another machine connect through the natd gateway to make sure it could connect to the outside world, then I had the gateway make its PPTP connection. The result was that the machine behind nat could no longer communicate with the outside world. Is there a way that I can cause the divert socket machinery to redirect a packet to a specific interface? This email sent to site_archiver(a)lists.apple.com

1 0

Re: Trick to compiling Apple3Com905x?
by Felix Kronlage 28 Sep '04

28 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Sep 27, 2004, at 23:25, Walt Pawley wrote: iD8DBQFBWTGGDTGb09msdNARAk3fAJwPimPXVcVNhrPZF9QKJMWGJswpAwCffS1B TK47lVqZB9EzesTFox/BpD0= =mwWw -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… FWIW: I did find the PearPC material. It did work OK ... except that it was a 1.4MB kext (a tad large for an ethernet driver I suspect) and it was VERY busy at writing entries into the system log about every twitch that took place. [...] The logging activity is a result of having logging turned on in the 3C90xDebug.h file. Changing the debugging level from 2 to 0 results in no logging activity that I've seen so far ... and I'm presently running over a GB through it (would have logged several Meg of message text in the system log). this sounds very interesting. Great that you looked at it and report it working. I'll see that we definitly bundle the ppc part as well with the next OpenDarwin release, so it might be that if people see that it does not cause stability problems, that it goes back into the Darwin tarball for this driver as well (one at least would asume so). - -fkr - -- gpg-fingerprint: 076E 1E87 3E05 1C7F B1A0 8A48 0D31 9BD3 D9AC 74D0 http://fkr.hazardous.org/ | http://opendarwin.org/~fkr/ | FKR-RIPE -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (Darwin) This email sent to site_archiver(a)lists.apple.com

1 0

Thermal scheduling
by John Siracusa 27 Sep '04

27 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com User-agent: Microsoft-Entourage/11.0.0.040405 Does the Darwin 7.4 (or later) kernel scheduler take CPU temperature into account at all? I've been poring over the xnu source trying to find a place where CPU temp is involved in some sort of thread priority or scheduling calculation, but have not found one yet. I'm hoping it'll be faster to just ask here :) -John _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Trick to compiling Apple3Com905x?
by Walt Pawley 27 Sep '04

27 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com >On Sep 18, 2004, at 23:44, Terry Simons wrote: > >> The PearPC project has a working pre-compiled PPC KEXT that I have >> used a bit - Possibly the same one you found Walt? >> They also have sources. I don't know if they've modified them or not, >> but the binary works for me. >> http://pearpc.sourceforge.net/apsl.html > On 9/19/04 11:19 AM +0200, Felix Kronlage wrote: >I started looking at that a while ago, but running the kext on my >quicksilver with 10.3.4 caused it to crash :/ I did not swap cards back >then, I will do this and report back. Felix, How'd that test go? ;-) FWIW: I did find the PearPC material. It did work OK ... except that it was a 1.4MB kext (a tad large for an ethernet driver I suspect) and it was VERY busy at writing entries into the system log about every twitch that took place. It turns out that the size is largely a function of having all the symbols available for debugging even though the "deployment" option is chosen in PB. Perhaps this is a nominal behavior - I'm pretty new at diddling with such things. I had not have expected that and didn't even know where to look to find out. This dropped the size from 2MB to 112K. The 2MB is what I was getting from compiling - not the size of the distributed binary. What the discrepancy is eludes me but I suspect it's a matter of library symbol differences. The logging activity is a result of having logging turned on in the 3C90xDebug.h file. Changing the debugging level from 2 to 0 results in no logging activity that I've seen so far ... and I'm presently running over a GB through it (would have logged several Meg of message text in the system log). Given my experience working with this, I would imagine a driver guru could do much better ... judging from the size of other such kext files. But I'm "happy" with this result so I'll probably stop mucking with it and just use it. ;-) I guess one thing still puzzles me some - why isn't the Darwin distribution of this driver both x86 and PPC capable? -- perl -e 'foreach(qw/13 10 60 47 98 117 115 104 62 13 10 10/){print chr($_)};' Walter M. Pawley <walt(a)wump.org> Wump Research & Company 676 River Bend Road, Roseburg, OR 97470 541-672-8975 _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: root tool acting on behalf of a user - what to set and how
by Dan Bernstein 25 Sep '04

25 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Sat, 25 Sep 2004 23:13:17 +0100, Finlay Dobbie <finlay(a)dobbie.net> wrote: > > > > On 25 Sep 2004, at 23:00, Finlay Dobbie wrote: > > > On 25 Sep 2004, at 10:16, Dan Bernstein wrote: > > > >> 1. Given the user's uid, what's the best way to determine the user's > >> gid and groups list? > > > > getpwuid? > > And for the groups list, getgrouplist (too hasty on the send :-) Oops (same problem here :-) Thanks again! _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: root tool acting on behalf of a user - what to set and how
by Dan Bernstein 25 Sep '04

25 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Sat, 25 Sep 2004 23:00:00 +0100, Finlay Dobbie <finlay(a)dobbie.net> wrote: > > On 25 Sep 2004, at 10:16, Dan Bernstein wrote: > > > 1. Given the user's uid, what's the best way to determine the user's > > gid and groups list? > > getpwuid? Thanks, but where in the passwd struct do I find the groups list? (And as long as we're discussing this structure, is it safe to assume (as finger does, IIRC) that pw_gecos points to a UTF8 string with the user's full name?) _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

Re: root tool acting on behalf of a user - what to set and how
by Finlay Dobbie 25 Sep '04

25 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 25 Sep 2004, at 23:00, Finlay Dobbie wrote: On 25 Sep 2004, at 10:16, Dan Bernstein wrote: 1. Given the user's uid, what's the best way to determine the user's gid and groups list? And for the groups list, getgrouplist (too hasty on the send :-) -- Finlay _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… getpwuid? This email sent to site_archiver(a)lists.apple.com

1 0

Re: root tool acting on behalf of a user - what to set and how
by Finlay Dobbie 25 Sep '04

25 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On 25 Sep 2004, at 10:16, Dan Bernstein wrote: 1. Given the user's uid, what's the best way to determine the user's gid and groups list? getpwuid? -- Finlay _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… This email sent to site_archiver(a)lists.apple.com

1 0

quick Open Directory question
by Finlay Dobbie 25 Sep '04

25 Sep '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Resent-date: Sat, 25 Sep 2004 22:59:06 +0100 Resent-from: Finlay Dobbie <finlay(a)dobbie.net> Resent-message-id: <C9FB32A8-0F3D-11D9-9BFC-000A959E2138(a)dobbie.net> Resent-to: darwin-dev(a)lists.apple.com -- Finlay _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.app… For dsOpenDirProxy, you're supposed to provide a tDataNode and two tDataBuffers for authentication purposes. How are you meant to allocate those without a dir ref? Or am I going to have to create a throw away local dir reference to allocate those structures, open the remote directory proxy, then close the local dir reference? This email sent to site_archiver(a)lists.apple.com

1 0