September 2008 - Darwin-dev

Re: CFFileDescriptor and stdin
by Nicolas Roussel 06 Sep '08

06 Sep '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com I agree. But I'm not using stdio. I shouldn't have said "stdin"... http://lists.apple.com/archives/Unix-porting/2006/Jan/msg00072.html http://lists.apple.com/archives/Unix-porting/2006/Jan/msg00084.html Could this be related to my problem with the CFFileDescriptor? Nicolas _______________________________________________ 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… In general, you should avoid multiple layers of buffering. I would suggest turning buffering off at the stdio layer using setbuf(3). I'm simply passing 0 as the file descriptor parameter to CFFileDescriptorCreate. And I'm not using any Carbon or Coca stuff for the moment (check the code in my first message). It seems others have had problems try to detect user input on the terminal with kqueue: This email sent to site_archiver(a)lists.apple.com

1 0

Re: CFFileDescriptor and stdin
by Terry Lambert 06 Sep '08

06 Sep '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi, -- Terry _______________________________________________ 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 Sep 6, 2008, at 9:01 AM, Nicolas Roussel <Nicolas.Roussel(a)lri.fr> wrote: I'm trying to use CFFileDescriptors to know when opened files or network sockets are readable. Although the following code works fine with a file it opens itself, it fails with stdin. It also fails with a dupped version of stdin (tried this thinking the CF machinery might close fd #0 at some point...). Does anybody know why this doesn't work? This is something I first ran into with X Window events when using certain widget toolkits (*cough* Motif *cough*) in the late 1980's. When you wrap a buffered I/O source in a buffeted I/O source, you have to check for data availability in the intermediate buffer before you look to the underlying source. It's possible (and reasonable) for the intermediate source to buffer data in its own buffers, such that there is data available in the intermediate layer buffers without data being available at the underlying data source (i.e. it has already been buffered up). This leads to having a "do you have data available?" API at each layer that's hard to integrate into a "notify me when data is available" paradigm like "select" or "poll" or "kevent" or "RunLoop", without a wrapper at each layer to peek into layer specific buffer state data. This is going to be a bigger for stdio, since we change it for efficiency reasons all the time. In general, you should avoid multiple layers of buffering. I would suggest turning buffering off at the stdio layer using setbuf(3). Alternately, create a subclass that can peek into the stdio buffer and tell if it is empty before looking at the real source and deciding there is no data available (or see that the intermediate buffer is not empty and report "data available" regardless of the state of the primary source; this is what Xselect was invented to do). CF is not cocoa, so it's likely that would end in failure in this case. My recommenation would be the former (do not use multiple buffering layers). This email sent to site_archiver(a)lists.apple.com

1 0

CFFileDescriptor and stdin
by Nicolas Roussel 06 Sep '08

06 Sep '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Hi, Does anybody know why this doesn't work? Nicolas _______________________________________________ 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 I'm trying to use CFFileDescriptors to know when opened files or network sockets are readable. Although the following code works fine with a file it opens itself, it fails with stdin. It also fails with a dupped version of stdin (tried this thinking the CF machinery might close fd #0 at some point...). test-stdin.c

1 0

Re: DiskArbitration, unmounting busy due to fseventsd?
by Terry Lambert 05 Sep '08

05 Sep '08

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com -- Terry On Sep 5, 2008, at 6:40 AM, Tim Schooley <tim(a)sbdev.net> wrote: Hi Terry, Many thanks for any help! Kind regards, Tim Good day! So here's the setup: 1) I have a kext that creates an IOBlockStorageDriver. I setup and tear down block devices from userspace with an application. 3) Using an application, and the DiskArbitration calls (namely DADiskMountWithArguments and DADiskUnmount), I mount these devices to a directory on the filesystem. The problem I have is this: 2.1) I try to unmount a mounted device using DADiskUnmount. 2.2) This fails because the device is "busy". - There are no IO operations happening. - There are no files open by users. - The only file open on the mount point is (output from lsof): fseventsd 34 root 12u REG 14,4 0 999999999 <mountpoint>/.fseventsd/0000000000006478 2.3) If I use the kDADiskUnmountOptionForce option in the call to DADiskUnmount, it corrupts the filesystem. So my questions are: Yes. 3.1) What is this fseventsd doing sticking its nose into my mountpoint? It is looking for file creations/changes on the device so it can report them to Spotlight for (re)indexing, or to other clients of fseventsd (like TimeMachine backup). It sticks it's nose in your mount point to create an event log file, which it then holds open until the disk is unmounted. The intent of the log file is to ensure that you don't have to rescan the entire disk due to last minute changes (or that last minute changes will not be indexed or backed up). Typically such changes represent events that have occurred, but for which fseventd clients have either been stopped or are not reading and reacting to events (e.g. at shutdown time, due to non-ordered shutdown of processes). It technically doesn't actually need to keep this log file open if it's not actively logging something, but it does so anyway. Yes. The mount option MNT_NOBROWSE ("-o nobrowse" on the mount command line) will prevent fseventd from listening for events on the volume. It will also disable TimeMachine, Spotlight indexing, and other clients of fseventd, so this is generally not a good idea. -- _______________________________________________ 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… Are you maybe using command line mounts and unmounts other than diskutil, which talks to DA? DA will run your command line tools under the covers after doing the right things. As long as you go through DA to do your mounts and unmount rather than explicit command line utilities or programs not going through DA, you should not need the nobrowse. As I said before, DA will notify fseventsd to stop looking. If Spolight and Time Machine are undesirable, you can explicitly disable them programattically by updating settings for "don't look here". Pretty sure this is well documented. fseventsd seems to have caused a lot of problems for our application. We have a feature allowing our device to be ejected automatically after it is idle for a specified time. Obviously, with fseventsd poking around, the timeout is not what the user would expect, and can be inaccurate by many minutes. Regarding the unmount issue - I have successfully registered an unmount approval callback function, which I can see gets called. However, more often than not, the unmount will still be prevented with a "busy" return code. Using the "nobrowse" option then seems to be the way to go. Time Machine and Spotlight are in fact undesirable features for our devices. However, if using this mount option, the Finder does not update to show the newly mounted device, nor does the Desktop (I guess Desktop and Finder are one and the same). Our application provides the user with the choice of mounting in a "standard" location (i.e. letting DA decide the mount point) and also to a specific location. The nobrowse option works great for specific locations. But for "standard" locations, we really need the mount to show in Finder and the Desktop as a volume. And again, we need fseventsd out of the way for the idle timeout functionality. Is there a way to tell the Finder/Desktop to "notice" the new disks with the "nobrowse" option? I've tried using NSNotifyByPath() with no luck. On Fri, 6 Jun 2008 14:17:59 -0700, Terry Lambert wrote: On Jun 6, 2008, at 1:44 AM, Tim Schooley wrote: 2) I format these devices by calling standard format commands, such as newfs_{hfs,msdos,...}. 3.0) Is the fseventsd causing this busy mount point issue? fseventsd registers a callback with DARegisterDiskUnmountApprovalCallback to permit it to get notification of the pending unmount. If you have failed to register an unmount completion callback, a direct call to the unmount is considered synchronous instead of asynchronous, and the unmount will fail, since it does not give fseventd time to close down the log. 3.2) Can I stop the fseventsd from doing this? 3.3) Am I missing some sort of call to the fseventsd? No, but you are missing registering an unmount completion callback with DiskArbitration so that the DADiskUnmount is not synchrnously answered with the only answer available at the time it returns ("busy"). -- Terry This email sent to site_archiver(a)lists.apple.com

1 0

Re: hanging bluetooth driver? (well, user progs opening it)
by Brad Parker 05 Sep '08

05 Sep '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Organization: Heeltoe Consulting Eli Bach wrote: > >You could also use Instruments to generate a stack trace, so you could >possibly find out what call is blocking in the kernel, which is >preventing your tool from exitting. I tried using gdb, but it would not attach to the pid. (I suppose I could try and attach *before* it hung, but I suspect once it's hung I won't be able to get anything) I'm not familiar with "Instruments" (or xray). Should I assume it has more powerful kungfu than gdb? Time to do some homework... thanks! -brad _______________________________________________ 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: hanging bluetooth driver? (well, user progs opening it)
by Eli Bach 05 Sep '08

05 Sep '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:from:to :in-reply-to:content-type:content-transfer-encoding:mime-version :subject:date:references:x-mailer; bh=OnD30Vn4la4qNN50TLfVCgpZiPgbBEsJ4+2FJGeYqgo=; b=VjvIXoYys9dL482EuAFsgx9IH8JXNBHI5iqpKv5467q+bxeV/4qb0lQ+WjoV5cpfNE G+q16OieRML5906f53auwCebTqim5ksLeYo4QCBQtBFNUMSSe5x+MuJ+pb8Ll8TPbszD GwlcW6Kq0KzOEdmzT3xxU6INJnJi1tkw05brk= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:from:to:in-reply-to:content-type :content-transfer-encoding:mime-version:subject:date:references :x-mailer; b=CwZt/VNI6tfRCF0GiSMuyxvTgSnK3cCyqB01wp54ZekoCg1yPOZ5x/i+whgkoC58jI oqeY7PgBlpTkJwFjW5Byq3QMxeOA17atcc7f9ST8fO9A0pOKxkSQoDkWD2bEqEidiUn0 n9OEN7ejFMi0glNLpjkTSCtEkk+3cgKKyTFo8= On Sep 5, 2008, at 6:03 AM, Brad Parker wrote: On Sep 4, 2008, at 5:15 PM, Brad Parker wrote: The only way I have found to clear this state is to reboot. Any ideas? Any way to kill the zombie? You could try turning BlueTooth off, then back on again in System Preferences. I tried that (didn't I say so? if not, I did). Surprisingly it had no effect. Makes me wonder what the bluetooth tty head looks like. Is that code I could just download and look at? (he said, sheepisly) _______________________________________________ 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… Eli Bach wrote: I have not tried this in 10.5, but I will at some point. -brad I don't think the BlueTooth driver is open-sourced, but you could dig through the darwin web page and look for it. You could also use Instruments to generate a stack trace, so you could possibly find out what call is blocking in the kernel, which is preventing your tool from exitting. This email sent to site_archiver(a)lists.apple.com

1 0

Re: DiskArbitration, unmounting busy due to fseventsd?
by Tim Schooley 05 Sep '08

05 Sep '08

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Hi again, On Fri, 5 Sep 2008 14:40:05 +0100, Tim Schooley wrote: > Using the "nobrowse" option then seems to be the way to go. Time Machine and > Spotlight are in fact undesirable features for our devices. However, if > using this mount option, the Finder does not update to show the newly > mounted device, nor does the Desktop (I guess Desktop and Finder are one > and the same). So it appears that the (undocumented) nobrowse option is actually intended to hide mounted Volumes, rather than stop things like fseventsd sniffing it. So I'm back to the question: how can I disable fseventsd (with the added twist: without hiding the device from Finder) on my device? Thanks for any help. Kind regards, Tim _______________________________________________ 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: DiskArbitration, unmounting busy due to fseventsd?
by Tim Schooley 05 Sep '08

05 Sep '08

site_archiver(a)lists.apple.com Delivered-To: Darwin-dev(a)lists.apple.com Hi Terry, fseventsd seems to have caused a lot of problems for our application. We have a feature allowing our device to be ejected automatically after it is idle for a specified time. Obviously, with fseventsd poking around, the timeout is not what the user would expect, and can be inaccurate by many minutes. Regarding the unmount issue - I have successfully registered an unmount approval callback function, which I can see gets called. However, more often than not, the unmount will still be prevented with a "busy" return code. Using the "nobrowse" option then seems to be the way to go. Time Machine and Spotlight are in fact undesirable features for our devices. However, if using this mount option, the Finder does not update to show the newly mounted device, nor does the Desktop (I guess Desktop and Finder are one and the same). Our application provides the user with the choice of mounting in a "standard" location (i.e. letting DA decide the mount point) and also to a specific location. The nobrowse option works great for specific locations. But for "standard" locations, we really need the mount to show in Finder and the Desktop as a volume. And again, we need fseventsd out of the way for the idle timeout functionality. Is there a way to tell the Finder/Desktop to "notice" the new disks with the "nobrowse" option? I've tried using NSNotifyByPath() with no luck. Many thanks for any help! Kind regards, Tim On Fri, 6 Jun 2008 14:17:59 -0700, Terry Lambert wrote: > On Jun 6, 2008, at 1:44 AM, Tim Schooley wrote: > > Good day! > > > > So here's the setup: > > > > 1) I have a kext that creates an IOBlockStorageDriver. I setup and > > tear > > down block devices from userspace with an application. > > > > 2) I format these devices by calling standard format commands, such as > > newfs_{hfs,msdos,...}. > > > > 3) Using an application, and the DiskArbitration calls (namely > > DADiskMountWithArguments and DADiskUnmount), I mount these devices > > to a > > directory on the filesystem. > > > > > > The problem I have is this: > > > > 2.1) I try to unmount a mounted device using DADiskUnmount. > > > > 2.2) This fails because the device is "busy". > > > > - There are no IO operations happening. > > - There are no files open by users. > > - The only file open on the mount point is (output from lsof): > > > > fseventsd 34 root 12u REG 14,4 0 999999999 > > <mountpoint>/.fseventsd/0000000000006478 > > > > 2.3) If I use the kDADiskUnmountOptionForce option in the call to > > DADiskUnmount, it corrupts the filesystem. > > > > So my questions are: > > > > 3.0) Is the fseventsd causing this busy mount point issue? > > Yes. > > > 3.1) What is this fseventsd doing sticking its nose into my > > mountpoint? > > It is looking for file creations/changes on the device so it can > report them to Spotlight for (re)indexing, or to other clients of > fseventsd (like TimeMachine backup). > > It sticks it's nose in your mount point to create an event log file, > which it then holds open until the disk is unmounted. The intent of > the log file is to ensure that you don't have to rescan the entire > disk due to last minute changes (or that last minute changes will not > be indexed or backed up). Typically such changes represent events > that have occurred, but for which fseventd clients have either been > stopped or are not reading and reacting to events (e.g. at shutdown > time, due to non-ordered shutdown of processes). It technically > doesn't actually need to keep this log file open if it's not actively > logging something, but it does so anyway. > > fseventsd registers a callback with > DARegisterDiskUnmountApprovalCallback to permit it to get notification > of the pending unmount. If you have failed to register an unmount > completion callback, a direct call to the unmount is considered > synchronous instead of asynchronous, and the unmount will fail, since > it does not give fseventd time to close down the log. > > > 3.2) Can I stop the fseventsd from doing this? > > Yes. The mount option MNT_NOBROWSE ("-o nobrowse" on the mount > command line) will prevent fseventd from listening for events on the > volume. It will also disable TimeMachine, Spotlight indexing, and > other clients of fseventd, so this is generally not a good idea. > > > 3.3) Am I missing some sort of call to the fseventsd? > > No, but you are missing registering an unmount completion callback > with DiskArbitration so that the DADiskUnmount is not synchrnously > answered with the only answer available at the time it returns ("busy"). > > -- Terry -- _______________________________________________ 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: hanging bluetooth driver? (well, user progs opening it)
by Brad Parker 05 Sep '08

05 Sep '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Organization: Heeltoe Consulting Eli Bach wrote: > >On Sep 4, 2008, at 5:15 PM, Brad Parker wrote: > >> The only way I have found to clear this state is to reboot. >> >> Any ideas? Any way to kill the zombie? >> >> I have not tried this in 10.5, but I will at some point. > >You could try turning BlueTooth off, then back on again in System >Preferences. I tried that (didn't I say so? if not, I did). Surprisingly it had no effect. Makes me wonder what the bluetooth tty head looks like. Is that code I could just download and look at? (he said, sheepisly) -brad _______________________________________________ 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: Darwin / OSX USB sniffer source?
by Jean-Daniel Dupas 05 Sep '08

05 Sep '08

site_archiver(a)lists.apple.com Delivered-To: darwin-dev(a)lists.apple.com Le 5 sept. 08 à 02:47, John Zorko a écrit : Hello, all ... Regards, John Falling You - exploring the beauty of voice and sound http://www.fallingyou.com I never try it myself, but it may do what you need. _______________________________________________ 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 interested in sniffing the USB communications between my Mac and various devices connected to it. Is there source (even if in basic "example" form) that demonstrates USB sniffing on Darwin / OSX? I think there is an USB Logger into Xcode Tools (in /Developer/ Applications/Utilities): USB Probe. This email sent to site_archiver(a)lists.apple.com

1 0