January 2005 - Darwin-kernel

Re: IOServiceGetMatchingServices() doesn't find driver
by Godfrey van der Linden 31 Jan '05

31 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Godfrey On Jan 31, , at 14:12, Brian Jorgage wrote: I am following the example tutorial in "Writing an I/O Kit Device Driver". I create a matching dictionary via IOServiceMatching (myDriver). When I verify the contents of the dictionary the (only) key-value pair is "IOProviderClass:<myDriver>". Shouldn't the key part of this be "IOClass" rather than "IOProviderClass" ? By using IOServiceMatching() it seems I am locking myself in to using the "IOProviderClass" key. If instead I use the call IOServiceMatching("AppleUSBComposite") to find an Apple driver then IOServiceGetMatchingServices() still doesn't find anything. However I know there are instances of AppleUSBComposite which I can verify via IORegistryExplorer. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… If you use ioreg -w0. Does the device you are looking for have the "registered" flag? This email sent to site_archiver(a)lists.apple.com

1 0

IOServiceGetMatchingServices() doesn't find driver
by Brian Jorgage 31 Jan '05

31 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com I am following the example tutorial in "Writing an I/O Kit Device Driver". I create a matching dictionary via IOServiceMatching (myDriver). When I verify the contents of the dictionary the (only) key-value pair is "IOProviderClass:<myDriver>". Shouldn't the key part of this be "IOClass" rather than "IOProviderClass" ? By using IOServiceMatching() it seems I am locking myself in to using the "IOProviderClass" key. If instead I use the call IOServiceMatching("AppleUSBComposite") to find an Apple driver then IOServiceGetMatchingServices() still doesn't find anything. However I know there are instances of AppleUSBComposite which I can verify via IORegistryExplorer. So what gives here ? Brian Jorgage _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: current_task() in start() , stop(), thread differs
by Parav Pandit 28 Jan '05

28 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, > > Hi, > > Please grant me to ask this silly question. > > > > I am creating one thread using IOCreateThread(). > > I want to do syncronization between this thread > and my > > interrupt handler, which is in the workloop > thread. > > > > Currently I create one semaphore using > > semaphore_create(current_task()) in the start(). > > Sounds like you have two kernel only threads. If > this is the case, you > will get undefined behavior from using > current_task(), since you are > tying the semaphore to a (possibly) transient task. > When the task > disappears (or gets a signal) the semaphore will > make it's way back out > to userland and disappear with it. If you are > dealing with strictly > kernel threads, use the kernel global kernel_task to > create the sem. > tied to the kernel. I got your point -when we want to sem_wait(), signal() from kernel only, kernel_task should be owner of this. This global kernel_task, we use for memory mapping from user space to kernel space, via IOMemoryDescriptor. my doubt is, as kextload command is issued, code for start() will be executed in the task context of that kextload ( transient task.) Even though is is kernel code, context of execution will be user initiated task (transient one) and not the kernel task context. Is my understanding correct? Regards, Parav ________________________________________________________________________ Yahoo! India Matrimony: Find your life partner online Go to: http://yahoo.shaadi.com/india-matrimony _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: current_task() in start() , stop(), thread differs
by Brian Bergstrand 28 Jan '05

28 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Jan 28, 2005, at 11:44 AM, Parav Pandit wrote: > Hi, > Please grant me to ask this silly question. > > I am creating one thread using IOCreateThread(). > I want to do syncronization between this thread and my > interrupt handler, which is in the workloop thread. > > Currently I create one semaphore using > semaphore_create(current_task()) in the start(). Sounds like you have two kernel only threads. If this is the case, you will get undefined behavior from using current_task(), since you are tying the semaphore to a (possibly) transient task. When the task disappears (or gets a signal) the semaphore will make it's way back out to userland and disappear with it. If you are dealing with strictly kernel threads, use the kernel global kernel_task to create the sem. tied to the kernel. HTH. Brian Bergstrand <http://www.bergstrand.org/brian/>, AIM: triryche206 PGP Key: <http://www.bergstrand.org/brian/misc/public_key.txt> MCSE - Must Consult Someone Experienced As of 01:29:00 PM, iTunes is playing "Guilt On Skin" from "The Dying Daylights" by "Charon" -----BEGIN PGP SIGNATURE----- Version: PGP 8.1 iQA/AwUBQfqFzHnR2Fu2x7aiEQKoJgCfSJrVKnVmjU20JhLm9XvbxbnZAv8AoP2X nKwJijjcOy99p3m1nhO5MHbN =jy+U -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

current_task() in start() , stop(), thread differs
by Parav Pandit 28 Jan '05

28 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, Please grant me to ask this silly question. I am creating one thread using IOCreateThread(). I want to do syncronization between this thread and my interrupt handler, which is in the workloop thread. Currently I create one semaphore using semaphore_create(current_task()) in the start(). My CreateThread sleeps on this semaphore, when isr completes it's job, it does semaphore_signal(). And than I do semaphore_destroy(current_task(), ..) in the stop(). This has memory leak, I believe. I did IOLog("current_task = %p\n", current_task()); at (1) start() (2) Inside my created kernel thread (3) stop(). At all the 3 places current_task() is different. So when i do semaphore_destory() from the stop(), it won't get destoryed. start() stop() code is initiated by user command kextload/unload but still executed from the kernel task, isn't it? Why all the 3 places shows different? Does it mean that I have to create those semaphores inside my thread and when I come out of this thread using IOExitThread(), before that I should destroy them. like IOCreateThread (MyThread,..) MyThread() { semaphore_create(current_task(), ...) ; while(1) { semaphore_wait() .......... ......... do the work() } signal received to comeout. semaphore_destroy(current_task(), ) IOExitThread(); } Or should I use some different mechanism? Regards, Parav Pandit ________________________________________________________________________ Yahoo! India Matrimony: Find your life partner online Go to: http://yahoo.shaadi.com/india-matrimony _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Question on Filter scheme driver example
by Yogesh P 28 Jan '05

28 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi Dan, I have some concerns about the content hint property in IOMedia object.As u mention in your mail that i have to create partition type of com_spsoftindia_HFS on the specific volumes so that my driver gets loaded for that partitions. My concern is if i create logical volumes using diskutility it puts the Content Hint value as "Apple_HFS" by default.If i have to load my driver for this logical volume i have to override the Content Hint.But i dont know how to override this content hint using diskutility. Is diskutility provides some functionality to override the Content Hint of the partitions??? Is that i have to provide some functionality in my filter driver to override the Content Hint of the partition by reading the Content Hint of all the partitions by using function getContentHint() and override the value of this Content Hint by using setProperty() function???? Is this a kind of thing i have to do to achieve this??? I also had to write userclient for this driver.Is there any documentation or sample code for writing userclient for filter scheme driver???? Any helps certainly helps me to solve this problem. Thanks in advance. YOGESH PAHILWAN (SOFTWARE ENGINNER) S P Software Technologies (I) Pvt. Ltd. (An ISO 9001:2000 Company) Pune, Maharashtra, India - 411 004. Tel : 91-20-4006154, 4006983, 4007503. Fax : 91-20-4006153. Email : pahilwan.yogesh(a)spsoftindia.com Visit us at : www.spsoftindia.com On Tue, 2005-01-18 at 19:40, Dan Markarian wrote: > Hi Yogesh, > > If you match Apple_HFS content, then produce Apple_HFS content, then > you would match recursively forever. It is advisable to either change > the partition type to com_spsoftindia_HFS on the specific volumes the > user is interested in enabling your software on, or if it needs to be > transparent to all volumes, then add a property on the media object you > produce, such that you can tell when it recurses in your probe() by > checking for the property, and failing the match if you see it. > > Dan > > On 15 Jan 2005, at 8:21 AM, Yogesh P wrote: > > > Hi folks, > > > > I had gone through the Mass storage driver pdf and according to the > > example given in the pdf for Developing filter scheme driver write > > sample filter scheme driver and it works fine for the diskimage which i > > had created through hdiutil > > > > In this example i specify "Content Hint" property as > > MySoftwareCompany_MyContent in the IOKit personality for driver > > matching > > and read/write calls in my filter scheme driver works fine when i write > > or read on this partition. > > > > But when i specify "Content Hint" property as Apple_HFS which i think > > is > > the default content hint for the partitions created on MAC OS X and > > load > > my filter scheme driver which also has this property with the value > > Apple_HFS i doesn't get the read/write IOLogs on the console and sees > > only start and init function IOLogs repeatedly. > > > > I have four partitons with Apple_HFS content hint which i had created > > during installation of MAC OS X. > > > > I want to know if i specify the "Content Hint" as Apple_HFS will my > > filter scheme driver work for all the four partitions other than the > > diskimage partition i have created using hdiutil??? > > > > > > Thanks in advance, > > YOGESH PAHILWAN > > (SOFTWARE ENGINNER) > > > > S P Software Technologies (I) Pvt. Ltd. > > (An ISO 9001:2000 Company) > > Pune, Maharashtra, India - 411 004. > > > > Tel : 91-20-4006154, 4006983, 4007503. > > Fax : 91-20-4006153. > > > > Email : pahilwan.yogesh(a)spsoftindia.com > > Visit us at : www.spsoftindia.com > > > > > > > > _______________________________________________ > > Do not post admin requests to the list. They will be ignored. > > Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) > > Help/Unsubscribe/Update your Subscription: > > http://lists.apple.com/mailman/options/darwin-kernel/ > > markarian%40apple.com > > > > This email sent to markarian(a)apple.com > _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: non blocking memory allocation api
by Brian Bergstrand 27 Jan '05

27 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 #include <kern/kalloc.h> kalloc_noblock() This can (and will) return NULL, so you have to be prepared for that. In particular it will always fail for allocations larger than 8KB. You have to match this with a call to kfree(), which can and will block, and there isn't a noblock version of it. So you cannot release memory while under a spinlock. kfree() also requires the size of the memory that you are releasing. The kalloc functions sit on top of the zalloc functions. Everything other memory api in the system (IOKit, BSD) uses kalloc, so kalloc can be called from any driver type. HTH. On Jan 27, 2005, at 1:53 AM, Parav Pandit wrote: > Hi, > I am seaching for API in IOKit which provides non > blocking memory allocation functionallity. > But I think IOkit has only IOMalloc one, > > is there any non blocking api available through i can > allocate memory, when i held the spin lock? > I searched the xnu code and there was one called > zalloc_noblock() > header : /osfmk/kern/zalloc.h > > but it uses some zone ... > is it the one which i should use? > if yes any idea how to use this zone in the code? > Can i use it at IOkit driver level? > any sample availble or this is for xnu kernel only? > Brian Bergstrand <http://www.bergstrand.org/brian/>, AIM: triryche206 PGP Key: <http://www.bergstrand.org/brian/misc/public_key.txt> It is easy to convert a bug to a feature; document it. - The Microsoft Employee Handbook, p.215, section B, article VII As of 09:56:07 AM, iTunes is playing "Here To Stay" from "Untouchables" by "Korn" -----BEGIN PGP SIGNATURE----- Version: PGP 8.1 iQA/AwUBQfkEKHnR2Fu2x7aiEQK/SgCfQpXmD+z9yEgy4YYDpX8sEPMekUMAnAsG naAotDW0AXG2+pVhBIBvbeKS =zf+K -----END PGP SIGNATURE----- _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

non blocking memory allocation api
by Parav Pandit 27 Jan '05

27 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, I am seaching for API in IOKit which provides non blocking memory allocation functionallity. But I think IOkit has only IOMalloc one, is there any non blocking api available through i can allocate memory, when i held the spin lock? I searched the xnu code and there was one called zalloc_noblock() header : /osfmk/kern/zalloc.h but it uses some zone ... is it the one which i should use? if yes any idea how to use this zone in the code? Can i use it at IOkit driver level? any sample availble or this is for xnu kernel only? Regards, Parav Pandit ________________________________________________________________________ Yahoo! India Matrimony: Find your life partner online Go to: http://yahoo.shaadi.com/india-matrimony _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: booting in console mode
by Peter Lovell 26 Jan '05

26 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Jan 25, 2005, at 5:40 PM, ravi anand wrote: Regards.....Peter _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… On 10.3.7 kernel ">console" takes you to text mode but quickly it switches back to the GUI mode. Seems like there is some issue ? Any idea of how to disable the GUI (Aqua) completely and boot in the text mode ? Thanx Ravi On my 10.3.7 system, logging in to console as "root" seems to stay that way. It will go back to GUI mode when you exit -- perhaps your session terminates for some reason? This email sent to site_archiver(a)lists.apple.com

1 0

Kern Control: ctl_enqueuedata & Kernel Panic
by Stephane Sudre 25 Jan '05

25 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com When calling: I'm seeing a Kernel Panic (in ctl_enqueuedata call). Yet: ------------ * Does the data need to be copyout before calling ctl_enqueuedata? _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… ctl_enqueuedata(kern_ctl_ref ctlref, void *data, size_t len, u_int32_t flags); /* sys/kern_control.h */ The ctl_ref looks OK, no ENOTCONN error. The len looks OK, no EMSGSIZE error. flags is set to 0. And data is pointing to a local struct with no indirection within it. And I didn't get an EINVAL error. * The Kern_control does not have a ctl_write (or equivalent) callback. Does it need to have one to make this work? * ctl_sendsize and ctl_recvsize (or equivalents) sizes are set to 0 (default values) but since the len is ok, I don't think the error is coming from this. This email sent to site_archiver(a)lists.apple.com

1 0