March 2004 - Darwin-kernel - lists.apple.com

Asynchronous I/O
by Lamont Cranston 26 Mar '04

26 Mar '04

I'm developing a utility that measures AIO disk performance in 10.3.3. I'm using aio_read(), aio_write(), etc. If I do AIO to, say, two drives (rdisks) with a queue depth of 5, everything works fine. If I increase the queue depth to 10 (I didn't try anything in between), I start getting failures for aio_read and aio_write. errno = EAGAIN. I've determined that the copyin call is failing (result = 0x23) in the aio_queue_async_request call. I'm using one buffer for all reads and one for writes (the data is not important). Also, once I get this error, I continue to get it until I reboot - even if I just do one request to one drive. What am I doing wrong or what are the system limitations? IMHO, 20 outstanding requests is not extravagant. Yahoo! Finance Tax Center - File online. File on time. _______________________________________________ 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

Using Altivec for TCP checksum calculations
by Brian Tabone 26 Mar '04

26 Mar '04

I was reading up on the PowerPC 7455, and happened upon an interesting document (http://e-www.motorola.com/files/if/cnb/ALTIVECTCPIPWP.pdf) on Motorola's website. This document covered the advantages of doing TCP checksum calculations in the Altivec SIMD unit versus doing it in the scalar integer units. Digging around in the xnu source, I found the checksum assembly code for Darwin (in bsd/dev/ppc/xsumas.s) and found that it was doing the checksum calculations in the integer units. My question is , beyond having to support the G3, is there any reason why the Darwin kernel does the checksum calculations in scalar code versus doing it in vector code? Is there reason to avoid Altivec instructions in the kernel? I'm assuming that there should be some way to test for altivec and use the scalar code when the altivec is not available so that the kernel remains compatible with the G3. If Motorola's numbers are to be believed, it would seem that this tweak would open up a potential bottleneck in the TCP stack, leaving more CPU time to do stuff like serve web pages and query databases. This would be a fine tweak for a web server or some other network busy server type. -Brian _______________________________________________ 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: A problem with ftp
by Adrian Bool 26 Mar '04

26 Mar '04

Hi Noemi, Check out 'man ftpd.conf'. It seems you want something like, umask real 022 in /etc/ftpd.conf on the server. Just tried it here and it seemed to work fine. Regards, aid On 26 Mar 2004, at 14:28, noemi.gil(a)uvic.es wrote: Hello, I have a problem using FTPd. I use an eMAC with OS X Jaguar version; The FTP Server is Lukemftpd 1.1. I open the System Preferences and click on Sharing under Internet & Network. I check the box marked Allow FTP Access and Shared Web >From Terminal, I create ftpchroot file with the users. I go to the System Preferences, Users, New User and I create a new user. If I prove to connect it to the web, I can. I see the Apache server saying "It Worked! The Apache Web Server is Installed on this Web I connect to the FTP with the user name and password. I upload a file (i.e. index.html) If I try to connect to this page (index.html) using a internet tool I cant see the page, it say me, Not Found. The requested URL /~user/index.html was not found on this server. I go to the Terminal and I write ls la and I see my file without read permission, exactly  -rw-r----- How can I do to automatically change the uploaded files permissions? Thank You, Noemm Gil _______________________________________________ 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. _______________________________________________ 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

A problem with ftp
by noemi.gil＠uvic.es 26 Mar '04

26 Mar '04

Hello, I have a problem using FTPd. I use an eMAC with OS X Jaguar version; The FTP Server is Lukemftpd 1.1. I open the System Preferences and click on Sharing under Internet & Network. I check the box marked Allow FTP Access and Shared Web >From Terminal, I create ftpchroot file with the users. I go to the System Preferences, Users, New User and I create a new user. If I prove to connect it to the web, I can. I see the Apache server saying "It Worked! The Apache Web Server is Installed on this Web I connect to the FTP with the user name and password. I upload a file (i.e. index.html) If I try to connect to this page (index.html) using a internet tool I cant see the page, it say me, Not Found. The requested URL /~user/index.html was not found on this server. I go to the Terminal and I write ls la and I see my file without read permission, exactly  -rw-r----- How can I do to automatically change the uploaded files permissions? Thank You, Noemm Gil _______________________________________________ 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: task emulation removal?
by Jim Magee 24 Mar '04

24 Mar '04

On Mar 24, 2004, at 10:18 AM, Morgan Winer wrote: > My question is simply this: why remove the task_{s,g}et_emulation > functionality from the Mach kernel? Besides the implementation penalties on "normal" traps that implementing the emulation vectors can introduce (as I mentioned in an earlier email), emulation vectors also lost favor with most implementors of system-call emulations. For instance, OSF AD/1 and MkLinux used a completely different mechanism (syscall exceptions). Why? I can tell you a few reasons: Emulating a system call within the task itself (as the emulation vector code does), it is really difficult to provide proper handing of things like EFAULT vs actually taking a fault. By implementing the handler in another task (as is better facilitated by the syscall exception mechanism), you can handle this much more correctly. The emulation vector code requires the Mach syscalls numbers to never conflict with the syscall numbers you are trying to emulate (because you're going to need to make Mach syscalls in the implementation of the emulation vector). This is well and good if you can compile your own kernel to assure these numbers are in disjoint sets. But for a binary released system like Mac OS X, this can be difficult to arrange. The syscall exception mechanism is controlled by the task_set_exception_ports() interface (EXC_MASK_SYSCALL and EXC_MASK_MACH_SYSCALL handlers). But before you get all excited, this mechanism was last used before the BSD and Mach kernels were reunited. So, the ability to override the in-kernel BSD syscall numbers is not glued up yet. But if you are looking for a mechanism that has a better chance of getting re-added to the kernel, I would put my bets on this mechanism - not the emulation vector stuff. --Jim [demime 0.98b removed an attachment of type application/pkcs7-signature which had a name of smime.p7s] _______________________________________________ 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: task emulation removal?
by Jim Magee 24 Mar '04

24 Mar '04

On Mar 24, 2004, at 10:18 AM, Morgan Winer wrote: > My question is simply this: why remove the task_{s,g}et_emulation > functionality from the Mach kernel? > > While Apple may not need these functions, there are many other > independent developers, myself included, that could make use of task > emulation, perhaps as open-source projects or otherwise. The main reason was that it added complexity to every trap - at a very low level where such complexity can really add to the cost of a trap (because we may not otherwise have to be in a mode where we could detect the existence of emulations - like having to be in virtual mode on PowerPC processors). There were also extensions to the trap mechanisms that didn't fit into the emulation model and nobody had time to make them fit. Surely, the emulation vector logic could haven been re-written to have a larger machine-dependent component to avoid having the performance implications of the current implementation. You could also add meaningful handling of the additional traps that weren't part of the pre-fast-trap design. But, as there were no known clients of such a facility on the horizon, it got scheduled at the bottom of the pile. I don't think there is any philosophical argument against it getting re-implemented in such a "more efficient" manner. It's just that nobody has signed up to do it. --Jim [demime 0.98b removed an attachment of type application/pkcs7-signature which had a name of smime.p7s] _______________________________________________ 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

task emulation removal?
by Morgan Winer 24 Mar '04

24 Mar '04

I have been looking into the possibility of binary compatibility for *nux and possibly other programs on Mac OS X, as well as other interesting or potentially useful programs to play with OS interoperability. Invariably, I have been lead to the task emulation functions, and immediately dove into learning about them (task_set_emulation, task_set_emulation_vector, and task_get_emulation_vector). In browsing and searching the mailing lists, I've run across posts from other people trying to make use of the task emulation functions as well, only to be met with the news that these functions have been unhooked and are therefore rendered useless. My question is simply this: why remove the task_{s,g}et_emulation functionality from the Mach kernel? While Apple may not need these functions, there are many other independent developers, myself included, that could make use of task emulation, perhaps as open-source projects or otherwise. - ardent developer _______________________________________________ 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: Error codes
by Justin Walker 23 Mar '04

23 Mar '04

On Mar 23, 2004, at 11:15, Vincent Lubet wrote: On Mar 22, 2004, at 9:08 AM, Justin Walker wrote: On Mar 22, 2004, at 7:59, Matt Jaffa wrote: I was wondering what error code I should throw. [snip] - Classic, and OT in particular, was not built with Unix-style errno codes in mind. The deeper recesses of OT do use the same error code names, though, but they use a different version of the header <sys/errno.h>. In particular, it's possible that EACCES (on Mac OS X) and EAGAIN (on Classic/OT/Mentat) (for example) share the same numeric code. That would be a bummer, of course. An OT expert will have to share his understanding of this one. In fact OT was built with Unix-style errno codes because it is a STREAMS stack after all. The errno codes known by OT can be found in <OpenTransport.h> I thought that OT was built with good ole Mac-style error codes (-XYZW), but that the STREAMS innards continued to use the unix-style error codes. In any case, comparing <sys/errno.h> from Mac OS X with <OpenTransport.h> may indicate how close the match is. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | It's not whether you win or lose... | It's whether *I* win or lose. *--------------------------------------*-------------------------------* _______________________________________________ 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: Error codes
by Vincent Lubet 23 Mar '04

23 Mar '04

On Mar 22, 2004, at 9:08 AM, Justin Walker wrote: On Mar 22, 2004, at 7:59, Matt Jaffa wrote: I was wondering what error code I should throw. You may not be alone :-} I am intercepting TCP traffic through my KEXT, after the result comes back from the daemon and I want to block the traffic, What is the best error to throw that will let internet explorer in classic not try the URL again. It works for Safari and IE(non classic version) if I use ENETDOWN, EACCES, and they don't try it again. But when I am using those errors for Classic's IE it blocks but keeps trying the URL over and over again. There could be many reasons for this behavior, and you'll need an OT expert to unravel them... - Classic, and OT in particular, was not built with Unix-style errno codes in mind. The deeper recesses of OT do use the same error code names, though, but they use a different version of the header <sys/errno.h>. In particular, it's possible that EACCES (on Mac OS X) and EAGAIN (on Classic/OT/Mentat) (for example) share the same numeric code. That would be a bummer, of course. An OT expert will have to share his understanding of this one. In fact OT was built with Unix-style errno codes because it is a STREAMS stack after all. The errno codes known by OT can be found in <OpenTransport.h> - The Classic version of IE doesn't know what to do with EACCES. Again, an OT expert may have the answer. That's correct. - The error code returned by your NKE gets lost somewhere between your NKE and IE; if it doesn't get lost, it gets translated; see the previous two comments. Yes, because the error code comes from the driver modules, it is likely to be transformed into a generic code (I did not check) I think you get the drift. It's not a problem that can easily be solved, since the two systems (Mac OS X and Classic/OT) were never intended to merge smoothly together. Hope that helps clarify things (and, perhaps, prompts someone with Knowledge to pipe up). Instead of returning an error code in the interface filter, I would simply drop the outgoing packet, and instead send up a packet with TCP RST and/or an ICMP port (or host) unreachable. That way both the OT stack and the application will see the connection cannot be made. Vincent _______________________________________________ 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: Error codes
by Sam Hart 23 Mar '04

23 Mar '04

Matt wrote: But when I am using those errors for Classic's IE it blocks but keeps trying the URL over and over again. I may be wrong, but I remember that being Classic IE's default behavior. S a m ------------------------------------------------------------------------ ------ Drawing on my fine command of language, I said nothing. _______________________________________________ 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