May 2005 - Darwin-dev - lists.apple.com

Re: calling open() on an older cd device returns EBUSY??
by Sam Vaughan 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com David, Sam On 05/05/2005, at 9:39 PM, darwin-dev-request(a)lists.apple.com wrote: so, opendev() gives you the char device, while open() gives you the block device? I was unsuccessful doing open("disk2"), I *was* successful doing open("rdisk2"). Are you saying it's better to call opendev("rdisk2")? what exactly does prepending "r" in front of the "disk2" do? _______________________________________________ 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 see you've already had a few replies, but I thought I'd point out where you can look if you want to read about how the devices in /dev get there and what they do. If you download the IOStorageFamily code from the Darwin section of developer.apple.com, you'll find a file inside called IOMediaBSDClient.cpp. This file is responsible for making IOKit devices such as disks available to BSD, which it does by adding them to the devfs file system mounted on /dev. (run "mount" to see all your currently mounted file systems) If you run /Developer/Applications/Utilities/IORegistryExplorer.app, open the Find window and search for "IOMediaBSDClient", you can inspect these IOKit objects and their parents. You'll see the BSD Name entries for each of your /dev/diskN and /dev/rdiskN devices in there, along with a lot of other interesting properties. So as you can see from the calls to devfs in IOMediaBSDClient.cpp, the names given to these devices are completely arbitrary. The devfs entries, along with their BSD Major and BSD Minor numbers visible with `ls -l /dev/*disk*`, are really just a bridge to allow you to communicate with the relevant IOKit driver for a particular device. This email sent to site_archiver(a)lists.apple.com

1 0

Re: vmmap
by Christopher D. Lewis 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com me too! after upgrade: % ls /usr/bin/vmmap ls: /usr/bin/vmmap: No such file or directory Best regards, Chris On May 3, 2005, at 3:27 PM, Spencer Nielsen wrote: Spencer Nielsen On May 2, 2005, at 10:48 PM, Spencer Nielsen wrote: Spencer Nielsen On May 2, 2005, at 9:27 PM, Joshua Durham wrote: - Josh On May 3, 2005, at 12:05 AM, Spencer Nielsen wrote: This email sent to joshdurham(a)mac.com This email sent to sea_dragons(a)mac.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… Now that I know running the .pkg to reinstall it runs a risk of b0rking the syst3m, I will steer clear :-) Just a quick update, running Essentials.pkg borked my system upon reboot. Half of my apps didn't run anymore and gave a shared library error with Carbon. I tried a fresh clean install of 10.4 and now vmmap is where it should be. I have filed a report about the upgrading issue. I loaded up the Essentials.pkg and ran it. It reported that there were errors in installation which was kind of unnerving. However, vmmap is now back again which make me happy. I will keep an eye out for any side effects. I have not yet tried a fresh install of 10.4 on any machine yet, but I plan on doing one soon. If anyone else has seen a similar behavior in upgrading from 10.3 I will probably file a report. Thanks for the help! On 10.4, it seems to be in Essentials.pkg: hc652a8d4:~ jdurham$ lsbom /Library/Receipts/Essentials.pkg/ Contents/Archive.bom | grep vmmap ./usr/bin/vmmap 102555 0/2 40980 1597189142 ./usr/share/man/man1/vmmap.1 100644 0/0 6705 1971577080 This was a clean install, so maybe something happened during the upgrade? One more quick question. I may be missing something here, but after upgrading I can find no sign of vmmap on either of the machines that I have upgraded to 10.4. I checked on a 10.3 machine and it normally resides in /usr/bin. I can imagine that there are some updates that have yet to be made to it for Tiger and that we might see it back soon. Anybody have any info on this? Is the source available anywhere? Thanks. Spencer Nielsen _______________________________________________ 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/joshdurham% 40mac.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/sea_dragons% 40mac.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: calling open() on an older cd device returns EBUSY??
by Dan Markarian 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.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

1 0

Re: calling open() on an older cd device returns EBUSY??
by Kevin Elliott 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com iChatAV/AIM: kelliott(a)mac.com (video chat available) __________________________________________ _______________________________________________ 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… At 19:06 -0700 on 5/4/05, David M. Cotter wrote: You can use opendev() to open the character device, that is, "/dev/rdisk2" so, opendev() gives you the char device, while open() gives you the block device? I was unsuccessful doing open("disk2"), I *was* successful doing open("rdisk2"). Are you saying it's better to call opendev("rdisk2")? what exactly does prepending "r" in front of the "disk2" do? The rdisk2 is a raw dev node. It provides block level passthrough to the underlying device- in other words, all I/O is handled directly with no caching. If you read the same block 5 times in a row, you will get 5 reads in a row on the bus. The only restriction is that all I/O must be block aligned, so be sure you know the block size of your media before using it (their are a variety of API's you can use to get the block size). My recommendation is to simply avoid the char dev node (i.e. disk2) whenever possible- if you need block level access you generally don't need caching, and because the filesystem opens the char dev node, their is always the potential for weird things to happen. correct, i just want to rip the stuff off and be done with it. i'm not going to read it more than once. and my apologies to anyone on the list who saw "anyone gotn'y clue here" as something like "this list is clueless", totally honestly and humbly, i state for the record I was feeling that *I* was the utterly clueless one, just hoping for Obi-Wan Kenobi to save me. Use the raw dev node young padawan. The ways of the char dev node are dark and mysterious, for the system does own it and shall block your path at every turn. Verily I say, the raw dev node is the path of light and shall not fail you when all is dark.... -- __________________________________________ Kevin Elliott <mailto:kevine@speedtools.com> Software Engineer, Intech Software Corporation www.intechUSA.com This email sent to site_archiver(a)lists.apple.com

1 0

Re: while (1) { dashboard();}
by Andrew Gallatin 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Ian Lister writes: > > Thankfully the Darwin source releases have been very punctual for some > time now (thanks Kevin!) so for each OS update I build a custom kernel > with CHILD_MAX (xnu/bsd/sys/syslimits.h) increased from 100 to 500 and > hack in a "ulimit -u" invocation in /etc/rc or somewhere (there being no > login.conf or the like to be able to set it for login sessions). > Doesn't that just change the default rlim_cur? BTW, does anybody know if it is possible to patch values into executables with GDB? On old, closed-source systems, you used to be able to do: # dbx -k /vmunix /dev/mem (dbx) patch maxproc=8192 8192 (dbx) This would make the value stick after reboot. If you could do this in macosx, you could avoid having to rebuild the kernel yourself. I copied the kernel to /tmp, and tried gdb -write /tmp/mach_kernel, but it didn't want to let me write it: (gdb) p ((int *) &maxproc)[0] $1 = 532 (gdb) set ((int *) &maxproc)[0] = (1024 + 20) Cannot access memory at address 0x37ca10 Drew _______________________________________________ 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: writing a new GUI environment
by chuck remes 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com On May 4, 2005, at 3:59 PM, Erich Ocean wrote: Good luck. cr [1] http://www.opendarwin.org/ _______________________________________________ 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 anyone know how Apple's X implementation is done? It's not a Carbon or Cocoa app AFAIK... I'd skip looking at Apple's implementation and instead look at the one used by OpenDarwin [1]. By definition OpenDarwin does NOT include any of the higher-level closed source libraries (Carbon, Cocoa, Quartz, etc.). Version 4.5.0 of XFree86 has been built for darwin8 and is available at <http://www.opendarwin.org/~ssen/XFree86-4.5.0-Darwin- x86-8.0.1.tar.bz2>. This is probably binary-only, but I bet the source is available there too or there's a pointer to it somewhere. This email sent to site_archiver(a)lists.apple.com

1 0

Re: while (1) { dashboard();}
by Ian Lister 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Cheers, Ian _______________________________________________ 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 Thu, 5 May 2005, Axel Andersson wrote: There's always /etc/sysctl.conf, which is intended for setting custom sysctl values at boot time, though it's not documented/included by default in Mac OS X. Yep, that's what I use for setting the sysctls, but you still need something in rc.common or the like to change the ulimit. This email sent to site_archiver(a)lists.apple.com

1 0

Re: while (1) { dashboard();}
by Axel Andersson 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com On May 5, 2005, at 00:24, robert delius royar wrote: -- Axel Andersson axel(a)zankasoftware.com http://www.zankasoftware.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… Since 10.2 I have included the following at the bottom of /etc/ rc.common. It works, or it has for me. There's always /etc/sysctl.conf, which is intended for setting custom sysctl values at boot time, though it's not documented/included by default in Mac OS X. It should work identically to this, however: http://www.openbsd.org/cgi-bin/man.cgi? query=sysctl.conf&apropos=0&sektion=0&manpath=OpenBSD +Current&arch=i386&format=html This email sent to site_archiver(a)lists.apple.com

1 0

Re: Darwin 8.0 Source Code Available
by Kevin Van Vechten 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com - Kevin On May 4, 2005, at 6:05 AM, Jack Howarth wrote: Does the fact that the released Darwin 8.0 source code requires gcc 3.3 to build imply that the kernel in Tiger itself is currently being built with gcc 3.3 as well rather than gcc 4.0? If so is there a plan to transition to building the Tiger kernel with gcc 4.0 in the 10.4 development cycle or will that be put off until 10.5? Jack _______________________________________________ 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… Almost everything is built with gcc-3.3, hence the Darwin project's name "gcc_os". Projects with ppc64 code are built using "gcc_os_35". Nothing is built with gcc-4.0. In fact, I'd be surprised if you had succeed building xnu with gcc-4.0. I don't think you'll ever get any official comment about when a gcc-4.0 transition will take place, if ever. This email sent to site_archiver(a)lists.apple.com

1 0

Re: calling open() on an older cd device returns EBUSY??
by David M. Cotter 05 May '05

05 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.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… You can use opendev() to open the character device, that is, "/dev/rdisk2" so, opendev() gives you the char device, while open() gives you the block device? I was unsuccessful doing open("disk2"), I *was* successful doing open("rdisk2"). Are you saying it's better to call opendev("rdisk2")? what exactly does prepending "r" in front of the "disk2" do? It seems unusual to us that the open on the block device was permitted since an open from a file system is exclusive, so for an Audio CD, "/dev/disk2" is opened by cddafs and hence cannot be opened by someone else. We are somewhat at a loss to explain how someone else can open it, at the moment. BSD enforces it. well, i can assure you, i have no problems at all calling open("disk2") on most systems. at least pre-tiger anyway, i don't know if tiger has better BSD enforcement We are looking into it. The character device is the correct device to use though, especially that I do not believe your intent was to have the system cache all the data your application reads from the Audio CD, which could be up to 700 MB of data to have to remember. correct, i just want to rip the stuff off and be done with it. i'm not going to read it more than once. and my apologies to anyone on the list who saw "anyone gotn'y clue here" as something like "this list is clueless", totally honestly and humbly, i state for the record I was feeling that *I* was the utterly clueless one, just hoping for Obi-Wan Kenobi to save me. This email sent to site_archiver(a)lists.apple.com

1 0