November 2004 - Darwin-dev - lists.apple.com

Re: trying to find right list
by Matt Watson 18 Nov '04

18 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Nov 16, 2004, at 5:46 AM, Steven J Abner wrote: On Tuesday, November 16, 2004, at 01:38 AM, Matt Watson wrote: Thanks for making me think!!! I found the problem. The compiler. times: first code / second code 110 / 55 Now who should get this this affects systems 10.1.5 - 10.3.6? Do you need a code snippet? Even better would be a "recipe" to reproduce. Something like: 1) Install 10.1.5 on a 400MHz PowerMac G3 2) Install the 10.1.5 developer tools (December 2002) 3) Compile the following program with cc -Os -o foo foo.c: #include <stdio.h> int main(void) { printf("Hello, World!\n"); return 0; } 4) Run the program with /usr/bin/time ./foo Something like that... _______________________________________________ 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 Can you rephrase the question? If you post a compilable, self-contained program that exhibits the issue you're trying to illustrate, it would be easier to respond. A good problem report will include: On a PPC two different machines and OS(s) using GCC 2.95.2 NOT 3.3 the code: "do *--strPtr = ((val_ul & 7) + 48); while ((val_ul >>= 3));" runs twice as slow as "do *--strPtr = _lowerhexstr[(val_ul & 7)]; while ((val_ul >>= 3));" Yes. In addition, the answers to the below questions would be very useful: 1) OS versions being used/compared (uname -a will tell you this) 2) Compiler versions being used/compared (cc -v will tell you this) 3) Machine configurations being used/compared (/usr/sbin/system_profiler SPHardwareDataType will tell you this) 4) Self-contained compilable code example (a single .c file with main()) 5) The compiler invocation being used (cc -Os -mdynamic-no-pic -prebind -o foo foo.c) 6) What you observed/expected ("took 125 seconds on 400MHz G4, took 99 seconds on 300MHz G3. I would have expected the G4 to be faster...") For your case, it's not clear how to measure the timing you're seeing, since we don't have a self-contained example. matt. smime.p7s

1 0

Re: simultaneous debuggers?
by Jim Ingham 18 Nov '04

18 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Jim On Nov 18, 2004, at 12:03 PM, darwin-dev-request(a)lists.apple.com wrote: Message: 1 Date: Wed, 17 Nov 2004 16:08:39 -0800 From: Eric Ocean <subscriber(a)ampede.com> Subject: simultaneous debuggers? To: darwin-dev(a)lists.apple.com Message-ID: <FFABBC41-38F5-11D9-8A90-000A959C51DC(a)ampede.com> Content-Type: text/plain; charset=US-ASCII; format=flowed Is it possible for multiple debuggers to attach themselves simultaneously to a single, running executable? (I'm not familiar enough with command-line GDB to try this myself.) Best, Eric _______________________________________________ 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… No, the system only lets one process attach with ptrace to another process - which is (part of) what gdb uses. You can use Sample & friends on a process that is running under gdb, but not another gdb. This email sent to site_archiver(a)lists.apple.com

1 0

Re: trying to find right list
by Michael Maibaum 18 Nov '04

18 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Nov 16, 2004, at 05:39, Steven J Abner wrote: You might want to try the Apple performance optimisation list at <http://lists.apple.com/mailman/listinfo/perfoptimization-dev> Michael _______________________________________________ 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… Hi, I'm looking for the correct list for a processor-specific problem. This is code ran on MacOSX, 10.1.5 and 10.3.6, a PCI G4 computer. The reason it is processor- specific is that a G3 building the same code runs faster than the faster(Mhz) G4. Matt's advice regarding a self contained test is of course extremely useful. This email sent to site_archiver(a)lists.apple.com

1 0

simultaneous debuggers?
by Eric Ocean 18 Nov '04

18 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Best, Eric _______________________________________________ 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… Is it possible for multiple debuggers to attach themselves simultaneously to a single, running executable? (I'm not familiar enough with command-line GDB to try this myself.) This email sent to site_archiver(a)lists.apple.com

1 0

Re: processing IOService in kext
by Mike Smith 17 Nov '04

17 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com gc, I am implementing this over a disk image. Then I copy ANY application into the disk image in order to cache / optimize / re route it's requests. (say its looking for a certain file on the disk, which is not there, but I can get it from else where) What you're describing here is basically an HSM. You can do this in two different ways: = Mike _______________________________________________ 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… On Nov 17, 2004, at 11:26 AM, calcmach(a)actcom.net.il wrote: Why do you want to intercept file I/O calls? The system will generally do a better job of this, unless you know something that it doesn't. 1) Write a new filesystem, which implements sideways lookup. You will end up with a large synthetic volume that behaves like it's full of symbolic links. If the files are just somewhere else on the system, this will work fairly well. 2) Implement this by subverting library functions in userland. Here you are protected by the kernel's security model, since you're doing work as the same user that's running the program, and you can control things cleanly on a per-application basis. You also don't run the risk of running the kernel out of memory. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Intercepting file system calls (read, write, open, close, etc)
by Mike Smith 17 Nov '04

17 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On Nov 17, 2004, at 11:26 AM, darwin-dev-request(a)lists.apple.com wrote: Andrew Downs <andrew(a)downs.ws> wrote: On Nov 16, 2004, at 2:05 PM, Al Ameen Shah wrote: Was support for VFS stacking removed? I was warned of such a year or two ago, and stopped my VFS efforts at that point (also trying to intercept file system calls). Stacking filesystems in the BSD model doesn't work very well. It gets expensive very quickly, and locking can become a nightmare. It's also often not really what you want; the above is a good example of that. One problem was my module did not get triggered when the user manipulated files in the Finder, which I wanted to capture. But the BSD side worked okay. = Mike _______________________________________________ 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… 1. Try VFS Stacking ? Or over-riding V-node ops ? You don't want a stacking VFS module to intercept FS calls at the application level; it should be done in userland. If you want to catch Carbon at it, you need to sit on top of getattrlist, setattrlist and getdirentriesattr as well. This email sent to site_archiver(a)lists.apple.com

1 0

Re: -fno-leading-underscore
by David Thompson 17 Nov '04

17 Nov '04

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com pass -Wl,-isymbol:_symbol on your compile/link line David Anyone else have any suggestions? - Mike -- ............................................................................. David L. Thompson Visualization and Imagery Solutions, Inc. mailto:dthompsn@vizsolutions.com 5515 Skyway Drive, Missoula, MT 59804 Phone : (406)756-7472 _______________________________________________ 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 will create a duplicate symbol that points to _symbol so external processes can deal with it. This email sent to site_archiver(a)lists.apple.com

1 0

Re: processing IOService in kext
by calcmach＠actcom.net.il 17 Nov '04

17 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Garth, -DM- Hi DM, On Nov 17, 2004, at 9:22 AM, calcmach(a)actcom.net.il wrote: gc, Ah, thanks. --gc Hello DM, On Nov 17, 2004, at 2:04 AM, calcmach(a)actcom.net.il wrote: -- Hi, I am trying to further implement the following Apple Kernel sample. How do I parse / filter / the IOService which is handed to me? Why do you want to intercept file I/O calls? --gc Note: this is a kext with KERNEL = 1, so not all api's are accessible. Thanks, DM getProvider()->read(this, byteStart, buffer, completion); } _______________________________________________ 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/gcummings%40apple.com This email sent to gcummings(a)apple.com __________________________________________________________________ Garth Cummings Apple Developer Technical Support gcummings(a)apple.com http://developer.apple.com/technicalsupport -- =========================== Erez Kaplan Collector and Researcher Mechanical Calculating Machines Virtual Museum: http://www.webcom.com/calc =========================== -- =========================== Erez Kaplan Collector and Researcher Mechanical Calculating Machines Virtual Museum: http://www.webcom.com/calc =========================== _______________________________________________ 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… Thanks for your explanation, this gives me a more clear understanding of the situation. At 09:58 -0800 17/11/2004, Garth Cummings wrote: Mime-Version: 1.0 (Apple Message framework v619) Content-Type: multipart/signed; micalg=sha1; boundary=Apple-Mail-2-519140; protocol="application/pkcs7-signature" Why do you want to intercept file I/O calls? I am implementing this over a disk image. Then I copy ANY application into the disk image in order to cache / optimize / re route it's requests. (say its looking for a certain file on the disk, which is not there, but I can get it from else where) Again, at this layer you have no information about files. In your hypothetical case, the open call from the application would end up in the filesystem name lookup code. The filesystem plugin would determine that the file didn't exist and your filter scheme wouldn't even be called. (The filter scheme would be called when the filesystem plugin read in the filesystem blocks from disk, but the filter scheme would have no way to tell that the read calls came from the filesystem plugin without basically reimplementing the filesystem code in the filter scheme, which is a really bad idea.) I don't have a simple answer for you. Perhaps someone on the list will, or you can always submit a paid tech support incident to DTS if you'd like us to help you research possible solutions. -DM- Mime-Version: 1.0 (Apple Message framework v619) Content-Type: multipart/signed; micalg=sha1; boundary=Apple-Mail-41--2092227; protocol="application/pkcs7-signature" I have tried several options, but am getting nowhere. I need to know the name of the class (open, create, etc) and the specific params for each call client->serviceMatching( "open", table ); A filter scheme is located well below the filesystem, so it has no knowledge of UNIX syscalls or files. A filter scheme only knows about reading/writing a buffer at a particular location on an IOMedia object. An IOMedia object is typically either a partition or the entire medium. The IOService handed to you is the IOMedia object published by the filter scheme. P.S. Please do not cross-post to multiple lists. If you do, people on one list don't know anything about the responses on the other list. I'm replying to the darwin-dev list only. ======= Apple Code ======= void com_apple_dts_driver_SampleFilterScheme::read(IOService* client , UInt64 byteStart, IOMemoryDescriptor* buffer, IOStorageCompletion completion) { // Read data from the storage object at the specified byte offset into the // specified buffer, asynchronously. When the read completes, the caller // will be notified via the specified completion action. // // The buffer will be retained for the duration of the read. // // For simple partition schemes, the default behavior is to simply pass the // read through to the provider media. More complex partition schemes such // as RAID will need to do extra processing here. // Attachment converted: ek disk:smime 1.p7s ( / ) (001184FC) __________________________________________________________________ Garth Cummings 1 Infinite Loop, MS 303-2T Apple Developer Technical Support Cupertino, CA 95014 "The last good thing written in C was 408-974-7711 phone Franz Schubert's Symphony No. 9." 408-862-7602 fax --Erwin Dieterich gcummings(a)apple.com Attachment converted: ek disk:smime 2.p7s ( / ) (00118516) This email sent to site_archiver(a)lists.apple.com

1 0

Re: -fno-leading-underscore
by Godfrey van der Linden 17 Nov '04

17 Nov '04

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Godfrey On Nov 16, , at 13:53, Michael Hogsett wrote: The -fno-leading-underscore to GCC appears to have no effect. We have some lisp code which calls foreign functions. The foreign function calls fail because all of the symbols have an underscore prepended to their name. e.g. [dh-7-13] [hogsett] [~] --> gcc -fno-leading-underscore -c -o foo.o foo.c [dh-7-13] [hogsett] [~] --> nm foo.o 0000003c b _b.0 00000000 T _foo [dh-7-13] [hogsett] [~] --> gcc -c -o foo.o foo.c [dh-7-13] [hogsett] [~] --> nm foo.o 0000003c b _b.0 00000000 T _foo How do I resolve this other than rewriting our lisp code? This does not occur on Solaris, Linux, or FreeBSD. _______________________________________________ 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'm missing something. Wouldn't it be a whole lot easier to change the lisp code that calls a foreign function to prepend an '_' than it is to recompile every thing you ever attempt to call. wouldn't that mean that you can never access any system libraries? That seems somewhat limiting. - Mike This email sent to site_archiver(a)lists.apple.com

1 0

Re: processing IOService in kext
by Garth Cummings 17 Nov '04

17 Nov '04

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi DM, On Nov 17, 2004, at 9:22 AM, calcmach(a)actcom.net.il wrote: gc, Ah, thanks. --gc Hello DM, On Nov 17, 2004, at 2:04 AM, calcmach(a)actcom.net.il wrote: -- Hi, I am trying to further implement the following Apple Kernel sample. How do I parse / filter / the IOService which is handed to me? Why do you want to intercept file I/O calls? --gc Thanks, DM getProvider()->read(this, byteStart, buffer, completion); } This email sent to gcummings(a)apple.com __________________________________________________________________ Garth Cummings Apple Developer Technical Support gcummings(a)apple.com http://developer.apple.com/technicalsupport -- =========================== Erez Kaplan Collector and Researcher Mechanical Calculating Machines Virtual Museum: http://www.webcom.com/calc =========================== __________________________________________________________________ Garth Cummings 1 Infinite Loop, MS 303-2T Apple Developer Technical Support Cupertino, CA 95014 "The last good thing written in C was 408-974-7711 phone Franz Schubert's Symphony No. 9." 408-862-7602 fax --Erwin Dieterich gcummings(a)apple.com _______________________________________________ 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 Why do you want to intercept file I/O calls? I am implementing this over a disk image. Then I copy ANY application into the disk image in order to cache / optimize / re route it's requests. (say its looking for a certain file on the disk, which is not there, but I can get it from else where) Again, at this layer you have no information about files. In your hypothetical case, the open call from the application would end up in the filesystem name lookup code. The filesystem plugin would determine that the file didn't exist and your filter scheme wouldn't even be called. (The filter scheme would be called when the filesystem plugin read in the filesystem blocks from disk, but the filter scheme would have no way to tell that the read calls came from the filesystem plugin without basically reimplementing the filesystem code in the filter scheme, which is a really bad idea.) I don't have a simple answer for you. Perhaps someone on the list will, or you can always submit a paid tech support incident to DTS if you'd like us to help you research possible solutions. -DM- Mime-Version: 1.0 (Apple Message framework v619) Content-Type: multipart/signed; micalg=sha1; boundary=Apple-Mail-41--2092227; protocol="application/pkcs7-signature" I have tried several options, but am getting nowhere. I need to know the name of the class (open, create, etc) and the specific params for each call client->serviceMatching( "open", table ); A filter scheme is located well below the filesystem, so it has no knowledge of UNIX syscalls or files. A filter scheme only knows about reading/writing a buffer at a particular location on an IOMedia object. An IOMedia object is typically either a partition or the entire medium. The IOService handed to you is the IOMedia object published by the filter scheme. P.S. Please do not cross-post to multiple lists. If you do, people on one list don't know anything about the responses on the other list. I'm replying to the darwin-dev list only. Note: this is a kext with KERNEL = 1, so not all api's are accessible. ======= Apple Code ======= void com_apple_dts_driver_SampleFilterScheme::read(IOService* client , UInt64 byteStart, IOMemoryDescriptor* buffer, IOStorageCompletion completion) { // Read data from the storage object at the specified byte offset into the // specified buffer, asynchronously. When the read completes, the caller // will be notified via the specified completion action. // // The buffer will be retained for the duration of the read. // // For simple partition schemes, the default behavior is to simply pass the // read through to the provider media. More complex partition schemes such // as RAID will need to do extra processing here. // _______________________________________________ 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/ gcummings%40apple.com Attachment converted: ek disk:smime 1.p7s ( / ) (001184FC) smime.p7s

1 0