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

Re: VNOP_GETATTR
by Mike Smith 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On May 12, 2005, at 11:10 AM, Pezeshgi, Shahriar wrote: struct vnode *vnode; struct vnode_attr vattr; If (VNOP_GETATTR (vnode, &vattr, vfs_context_create(NULL) == 0) { size = vattr.va_size; } VATTR_INIT(&va); VATTR_WANTED(&va, va_data_size); err = vnode_getattr(vp, &va, ctx); if (!err && VATTR_IS_SUPPORTED(&va, va_data_size)) my_file_size = va.va_data_size; = Mike _______________________________________________ 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.… I'm trying to read a file size from my KEXT under Tiger but I'm getting 0 only, is there anyone with more experience on this? Do not call VNOP_GETATTR, use vnode_getattr. Avoid using the VNOP functions entirely; they are not as a general rule meant to be called directly. You must initialise the vnode_attr structure with VATTR_INIT, and set the requested field(s) appropriately: vnode_t vp; /* use vnode_t, to you it is not a structure, it's a handle */ struct vnode_attr va; vnode_context_t ctx; Note that there are a number of size values associated with a file; pick the one that best suits your needs. Also note that vnode_getattr will synthesise values not supported by the underlying filesystem; you don't need to handle those cases yourself. This email sent to site_archiver(a)lists.apple.com

1 0

VNOP_GETATTR
by Pezeshgi, Shahriar 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcVXHe7Qr/YEzPfIQA2qJQyuhjganA== Thread-topic: VNOP_GETATTR Hi I'm trying to read a file size from my KEXT under Tiger but I'm getting 0 only, is there anyone with more experience on this? struct vnode *vnode; struct vnode_attr vattr; If (VNOP_GETATTR (vnode, &vattr, vfs_context_create(NULL) == 0) { size = vattr.va_size; } Shahriar _______________________________________________ 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: Tiger kextcache command
by Garth Cummings 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hello Gilles, On May 12, 2005, at 1:23 AM, Gilles Vayssié (TRI-EDRE) wrote: Hi, i encounter a problem with the kextcache command under Tiger (8A428). The supported way to do this is: sudo touch /Volumes/ExternalDisk/System/Library/Extensions Please help or give advices. Thanks. __________________________________________________________________ Garth Cummings Apple Developer Technical Support gcummings(a)apple.com http://developer.apple.com/technicalsupport Spotlight on Innovation Worldwide Developers Conference 2005 June 6-10 in San Francisco http://developer.apple.com/wwdc _______________________________________________ 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 I would like to update cache on an external volume. Why? Because on startup, the volume is read online, so the caches can't be updated. So i need to update the cache before booting on this volume, from an "alive" system. Here is how i did under Panther, and it worked well: sudo kextcache -l -m /Volumes/ExternalDisk/System/Library/ Extensions.mkext /Volumes/ExternalDisk/System/Library/Extensions When the volume is next booted, the boot loader will notice that the KEXT caches are out of date and rebuild them. The problem with manually tweaking the caches is that each version of Mac OS X implements a different set of caches, and how they work is considered an implementation detail that developers shouldn't care about. Touching the Extensions directory will assure that the proper caches are rebuilt on all versions of Mac OS X. Gilles smime.p7s

1 0

Re: Read through Filter Scheme Driver
by Dan Markarian 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi Imran, IOLog("\n Calling read routine of driver.....\n"); getProvider()->read(client, 0, OutBuffer, completion); IOLog("\n Called read routine of driver.....\n"); Dan On 11 May 2005, at 8:31 AM, Imran Kazi wrote: Dear Mike, Please go through the below code for our Read issue /***************User Space Application **************************/ /* Purpose: This calls the 'read' method of user client using IOConnectMethodStructureIStructureO */ static kern_return_t MyDoRead(io_connect_t connect) kern_return_t kernResult; MySampleStruct sampleStructIn; // Input structure MySampleStruct sampleStructOut; //Output structure kernResult = vm_allocate(mach_task_self(), (vm_address_t*) &sampleStructIn.buff, vm_page_size, TRUE); if(sampleStructIn.buff) { char* buff = (char*) sampleStructOut.buff; strcpy(buff, ""); } IOByteCount structSizeIn = sizeof(MySampleStruct); IOByteCount structSizeOut = sizeof(MySampleStruct); if(retRead == kIOReturnSuccess) { IOLog("\n prepare for buff Succeded....\n"); // mapping the IOMemoryDescriptr buffer retRead = ReadBuffer->map(); retRead = UCProvider->CoreSnapDriverDoRead(this, 512, ReadBuffer, completion); IOByteCount readBytes; IOByteCount requestByteCount = 0; IOReturn status = 0; IOByteCount bufLength = OutBuffer->getLength(); char *myBuff = new char[bufLength]; readBytes = OutBuffer->readBytes(0, myBuff, bufLength); // Make "filter" changes to the buffer here. IOLog("\n Calling read routine of driver.....\n"); getProvider()->read(client, 0, OutBuffer, completion); IOLog("\n Called read routine of driver.....\n"); IOLog("\n Calling readbytes.......in driver\n"); readBytes = OutBuffer->readBytes(0, myBuff, bufLength); IOLog("\n Calling complete routine of driver......\n"); getProvider()->complete(completion, status, requestByteCount); IOLog("\n Called complete routine of driver......\n"); delete[] myBuff; return retDrRead; } Do we need to call withAddress() and prepare() and map() the IOMemoryDescriptor before passing it to standard read() method. Thanks Imran 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.… You made an asynchronous request. You must wait until the completion routine you pass in fires before you may access the result. You are welcome to make a synchronous request in case that was your intention. You do not require a completion routine in that case. if (connect) { kernResult = IOConnectMethodStructureIStructureO (connect, // connection handle returned from IOServiceOpen(). kCoreDoRead, // an index to the function in the kernel. structSizeIn, // the number of scalar parameters. &structSizeOut, &sampleStructIn, &sampleStructOut ); } return kernResult; } /***************User client **************************/ getTargetAndMethodForIndex(IOService ** target, UInt32 index) { { // kCoreDoRead NULL, (IOMethod) &com_CoreSnap_UserClient::CoreSnapUserClientDoRead, kIOUCStructIStructO, sizeof(MySampleStruct), sizeof(MySampleStruct) }, } IOReturn com_CoreSnap_UserClient::CoreSnapUserClientDoRead (MySampleStruct *InStruct, IOByteCount InStructSize, MySampleStruct *OutStruct, IOByteCount *OutStructSize) { IOMemoryDescriptor *ReadBuffer = NULL; // Constructing memory descriptor for out put structure ReadBuffer = IOMemoryDescriptor::withAddress((vm_address_t) OutStruct->buff, (IOByteCount) OutStruct->size, kIODirectionIn, fTask); if(ReadBuffer) //Created IOMemoryDescriptor for buff Successfully.... { retRead = ReadBuffer->prepare(); IOLog("\n address of prepare buffer = %p\n", retRead); if(retRead == kIOReturnSuccess) { IOLog("map successful\n"); // get the provider for our userclient and call driver read routine UCProvider = OSDynamicCast(com_CoreSnap_driver, fProvider); IOLog("\n UCProvider = %p\n", UCProvider); IOStorageCompletion completion; // calling driver read call by passing IOMemoryDescriptor buffer if(retRead == kIOReturnSuccess) { IOLog("\n CoreSnapDriverDoRead of Driver Succeded...\n"); //return retRead; } }// end of prepare } ReadBuffer->complete(); ReadBuffer->release(); if (kIOReturnSuccess != retRead) { IOLog("%s[%p]::%s - returning err 0x%08x", getName(), this, __FUNCTION__, retRead); } fProvider->release(); return retRead; } /********************************Driver Read routine ***************************/ // kCoreDoRead IOReturn com_CoreSnap_driver::CoreSnapDriverDoRead(IOService* client, UInt64 byteStart, IOMemoryDescriptor * OutBuffer, IOStorageCompletion completion) { IOReturn retDrRead = kIOReturnSuccess; IOLog("com_CoreSnap_driver::read BEFORE **** readBytes = %lu byteStart = %lu \ bufLength = %lu buffer = %s ***\n", (long)readBytes, (long)byteStart, (long)bufLength, myBuff); IOLog("com_CoreSnap_driver::***AFTER::readBytes = %lu byteStart = %lu bufLength = %lu buffer = %s \n", (long)readBytes, (long)byteStart, (long)bufLength, myBuff); We are able to prepare() and map() IOMemoryDescriptor successfully and pass this buffer to driver's read routine(we also have interchange the prepare and map). My problem is while debugging it prints all the IOLogs before driver's standard read() method, but IOLogs after read() does not get printed in system.logs. Are we following correct steps for reading the buffer from disk and passing it to user space application? _______________________________________________ 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 site_archiver(a)lists.apple.com

1 0

Re: Write on specific IOMedia through filter driver
by Dan Markarian 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi Imran, Dan On 10 May 2005, at 9:16 AM, Imran Kazi wrote: Dear Dan where this -> IOService* client, byteStart -> byte offset, WriteBuffer -> IOMemoryDescriptor*, completion -> IOStorageCompletion. Thanks Imran _______________________________________________ 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.… But still my driver writes the buffer on all the media objects. It is entirely under the control of your driver. You need to avoid multiplying those I/Os. We do not multiply I/Os on our own. Are the two partitions in question not separate locations on the disk? Does your stack not match the picture I drew? As per your previous mail, we have change the partition name of the volumes, and use media->getName() function which actually outputs the various partition names. My driver write call syntax is as follows- getProvider->write(this, byteStart, WriteBuffer, completion); But still my driver writes the buffer on all the media objects. How should we specify the write call of the driver to write the buffer on the specific media object? Or what parameters we have to pass to the write call, so that it is work for specific media object only? In our case the write call takes IOService* as input and the GetName () is IOMedia Method, so even if we check the name and fires the write call it will write on all the media objects. How should we give IOService* specific media object? This email sent to site_archiver(a)lists.apple.com

1 0

Tiger kextcache command
by Gilles Vayssié (TRI-EDRE) 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, i encounter a problem with the kextcache command under Tiger (8A428). # Copy it to the righ place: sudo mv Extensions.mkext /Volumes/ExternalDisk/System/Library/ # Repair permissions sudo chown root /Volumes/ExternalDisk/System/Library/Extensions.mkext sudo chgrp wheel /Volumes/ExternalDisk/System/Library/Extensions.mkext sudo chmod 644 /Volumes/ExternalDisk/System/Library/Extensions.mkext Please help or give advices. Thanks. Gilles _______________________________________________ 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.… I would like to update cache on an external volume. Why? Because on startup, the volume is read online, so the caches can't be updated. So i need to update the cache before booting on this volume, from an "alive" system. Here is how i did under Panther, and it worked well: sudo kextcache -l -m /Volumes/ExternalDisk/System/Library/ Extensions.mkext /Volumes/ExternalDisk/System/Library/Extensions This command created a file "Extensions.mkext" on the external volume "ExternalDisk". Now, under Tiger, if i do this i have the following error: "can't create file /Volumes/ExternalDisk/System/Library/ Extensions.mkext: Cross-device link" To find a work around, i tried this (shell syntax): # Create cache locally sudo kextcache -l -m Extensions.mkext /Volumes/ExternalDisk/System/ Library/Extensions But i can't boot on ExternalDisk... The Systems stop, with a message "Still waiting for root device", so the cache seems corrupted... This email sent to site_archiver(a)lists.apple.com

1 0

proc_ucred() temporary? What's permanent?
by Kevin Brock 12 May '05

12 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thanks. Kevin Brock _______________________________________________ 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.… The comment in sys/proc.h above the declaration of proc_ucred() says that it's a temporary API. Getting this information is pretty important to what we're doing... Does anyone know if this is really a temporary API? If it is, does anyone know what the permanent API will look like? This email sent to site_archiver(a)lists.apple.com

1 0

Darwin 8: Missing symbols
by Brian Bergstrand 11 May '05

11 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I can't seem to find the following symbols in any of the KPI's (including Unsupported): current_debugger // This is just used in our debug builds, so no big deal host_info ml_get_interrupts_enabled // ml_set_interrupts_enabled is exported by libkern thread_policy_get klogwakeup Is my only choice linking against the whole kernel? Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2 -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.0.1 (Build 2130) iQA/AwUBQoKZNnnR2Fu2x7aiEQLKAACg/sjVdACd7PBdTzzapdwfFPwvcYcAoLeR lYMsaux2rPVOdnLoFyB4TMTp =bjM0 -----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

Darwin 8: assert_wait_timeout & thread_cancel_timer
by Brian Bergstrand 11 May '05

11 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 We had code in Darwin 7.x that did the following assert_wait_timeout() wresult = thread_block() if (THREAD_TIMED_OUT != wresult) thread_cancel_timer(); thread_cancel_timer has been removed from the 8.x headers and is listed as Unsupported (even though it's still exported by the kernel - -- but not he KPI's). Is there a replacement? Or should we not even worry about the (THREAD_TIMED_OUT != wresult) case in 8.x? TIA. Brian Bergstrand <http://www.bergstrand.org/brian/> PGP Key ID: 0xB6C7B6A2 -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.0.1 (Build 2130) iQA/AwUBQoJ5BXnR2Fu2x7aiEQLTJwCeNZNmUpZz3hySat7XGI55sO6XoYwAnRJd OXVWqoD4oV6WG5mlO/JO7JkN =Ie7d -----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

Re: Where is the "REAL" kernel of my G5?
by Kevin Van Vechten 11 May '05

11 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com The solution is to either: a) $ sudo touch /System/Library/Extensions $ sudo reboot or b) manually rebuild the cache using kextcache(8) $ sudo kextcache -e - Kevin On May 11, 2005, at 10:29 AM, Ramble Leon wrote: I have two G5, one is 1.6G UP, another is 1.8G DP, I've made my own kernel and put xnu/BUILD/dst/* in a pkg, by install the package, the mach_kernel will be copied to /mach_kernel, kexts and frameworks will be copied to /System/Library/Extensions & Frameworks. I installed the package in both my G5, the 1.6UP is working fine, the 1.8DP is also bootable, but the booting kernel is not my kernel. Although I check the /mach_kernel in both machines by md5, they are exactly same kernel, but uname of these two machines are different, uname output of the 1.8DP is information of the old kernel. Even I removed the old one and reboot, still the same result. Where is the REAL kernel in the 1.8DP G5? How can I replace it? Anybody can help me a little bit? Thanks. _______________________________________________ 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.… Mac OS X pre-links the mach_kernel and Extensions, and the results are cached in the /System/Library/Extensions.mkext. The system will only regenerate this cache if it notices that the /System/Library/ Extensions directory is newer than the mach_kernel. My guess is that you copied them over in such a way that the mach_kernel still has a newer mtime than the Extensions directory, and therefore the old uname is still cached in the mkext. This email sent to site_archiver(a)lists.apple.com

1 0