July 2012 - Darwin-kernel - lists.apple.com

Re: About partition scheme.
by Phil Jordan 04 Jul '12

04 Jul '12

On Wed, Jul 4, 2012 at 8:18 AM, Li Flost.Dexiong <havefan2012(a)gmail.com> wrote: > I add some log on the read function of my class. > and notice that if i do some read on disk1s1, my read function will be called. > if i do some read on disks1, my read function will not be called. The IOMedia object corresponding to disk1s1 is created by your partition scheme. The IOMedia object corresponding to disk1 is NOT created by your partition scheme - it is the provider media of your partition scheme and created by the driver for the underlying disk device. HTH phil _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: About partition scheme.
by Phil Jordan 04 Jul '12

04 Jul '12

On Wed, Jul 4, 2012 at 8:14 AM, Li Flost.Dexiong <havefan2012(a)gmail.com> wrote: > /dev/disk1 > #: TYPE NAME SIZE IDENTIFIER > 0: *16.1 GB disk1 > 1: new_testdisk 16.1 GB disk1s1 > > > I'm wandering why there is a #0 on my disk, when and how does it create? "disk 0" is the whole disk that the partition scheme sits inside. Look at the IOMedia hierarchy in IORegistryExplorer and it'll make more sense. > Also, are there any reference about how the "disk1s1" string was generate and record on my system? The "1" is generated from the location you set on your newly created IOMedia object like so: newMedia->setLocation("1"); This only happens if the IOMedia object isn't "whole". If it's whole, a new diskX node is generated. HTH phil _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: problem in using SampleFilterScheme
by Phil Jordan 04 Jul '12

04 Jul '12

On Wed, Jul 4, 2012 at 4:57 AM, Li Flost.Dexiong <havefan2012(a)gmail.com> wrote: > My change like this: > > 1) add a probe function to SampleFilterScheme; this function implement as description in my last mail. Without seeing the code for probe() I can still only guess. Specifically, it'd be interesting how you're allocating the read buffer and calling read() inside probe(). > 2) modify read function declaration to > virtual void read(IOService* client, > UInt64 byteStart, > IOMemoryDescriptor* buffer, > IOStorageAttributes* attributes, > IOStorageCompletion * completion); > > and its implementation is: > > void com_apple_dts_driver_SampleFilterScheme::read(IOService* __attribute__ ((unused)) client, > UInt64 byteStart, > IOMemoryDescriptor* buffer, > IOStorageAttributes* attributes, > IOStorageCompletion *completion) > { > > char pname[4096] = {'\0'}; > proc_selfname(pname, sizeof(pname)); > DEBUG_LOG("%s[%p]::%s(%p, %llu, %p, %p, %p)\n", getName(), this, __FUNCTION__, client, byteStart, buffer, completion); > printf("%s->%s[%p]::%s(%p, %llu, %p, %p)\n",pname, getName(), this, __FUNCTION__, client, byteStart, buffer, completion); > > getProvider()->read(this, byteStart, buffer, attributes, completion); > } You're allocating 4KB on the stack in this function. Don't. You're only using it for printing anyway, so limit it to 64 characters or so. Even better, do the debug output in a helper function so the stack space is returned before you get to the provider's read() method. Allocate larger buffers (more than a few dozen bytes) dynamically on the "heap". The typical storage driver stack on OSX can already produce some very deep stacks - between PCI, the AHCI driver, various IOStorage family stuff, the partition scheme, the file system, and VFS this can easily add up to 10-12K before you add your own layer. It's therefore really important to keep your stack usage to a minimum, especially in "forwarding" methods like read(). > I'm curious about > 1) why the GetProvider() will turn back to my class? It won't. It's probably something else. In particular, it looks to me from the stack trace (first argument to your methods) like your 'this' pointer is something different every time. Could it be that you're ending up with mulitple instances of your filter scheme stacked on top of each other? I strongly recommend capturing firewire/serial port kprintf() and gdb via firewire/ethernet from a second mac. It saves an enormous amount of time stumbling in the dark. > 2) if the stack is small to 16KB, but in my probe function, i just using a array of 4096B. how can this be overflow? Look at your bottom stack pointer: 0x30ca3fc8 - you can round this up to the next page boundary, i.e. 0x30ca4000. and the top stack frame: 0x30ca0d18 the difference between these is 0x32E8 or 13032. As com_apple_dts_driver_SampleFilterScheme::read() tries to allocate 4K, this pushes it over 16K, thus blowing the stack. Hope that helps, phil _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: About partition scheme.
by Li Flost.Dexiong 04 Jul '12

04 Jul '12

more information about my class and my test. I add some log on the read function of my class. and notice that if i do some read on disk1s1, my read function will be called. if i do some read on disks1, my read function will not be called. my read operations like this:(using terminal) #dd if=/dev/disk1s1 of=/dev/null bs=4096 count=256 this will result my read function be called. If i do #dd if=/dev/disk1 of=/dev/null bs=4096 count=256 my read will not be called. 在 2012-7-4，下午2:14， Li Flost.Dexiong 写道： > > These days I implement a class based on IOPartitionScheme, > my child_media init with an content hint of "new_testdisk"; > and use "new_testdisk" to set the childmedia name, > > when I insert my disk, and my class was probed, and could start normally, > > but when I use "diskutil list" in a terminal, it shown as below: > > /dev/disk0 > #: TYPE NAME SIZE IDENTIFIER > 0: GUID_partition_scheme *160.0 GB disk0 > 1: EFI 209.7 MB disk0s1 > 2: Apple_HFS Macintosh HD 159.7 GB disk0s2 > /dev/disk1 > #: TYPE NAME SIZE IDENTIFIER > 0: *16.1 GB disk1 > 1: new_testdisk 16.1 GB disk1s1 > > > I'm wandering why there is a #0 on my disk, when and how does it create? > Also, are there any reference about how the "disk1s1" string was generate and record on my system? > > > Thanks. > > > Best regards, > Flost.D.Li > havefan2012(a)gmail.com > > > Best regards, Flost.D.Li havefan2012(a)gmail.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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

About partition scheme.
by Li Flost.Dexiong 04 Jul '12

04 Jul '12

These days I implement a class based on IOPartitionScheme, my child_media init with an content hint of "new_testdisk"; and use "new_testdisk" to set the childmedia name, when I insert my disk, and my class was probed, and could start normally, but when I use "diskutil list" in a terminal, it shown as below: /dev/disk0 #: TYPE NAME SIZE IDENTIFIER 0: GUID_partition_scheme *160.0 GB disk0 1: EFI 209.7 MB disk0s1 2: Apple_HFS Macintosh HD 159.7 GB disk0s2 /dev/disk1 #: TYPE NAME SIZE IDENTIFIER 0: *16.1 GB disk1 1: new_testdisk 16.1 GB disk1s1 I'm wandering why there is a #0 on my disk, when and how does it create? Also, are there any reference about how the "disk1s1" string was generate and record on my system? Thanks. Best regards, Flost.D.Li havefan2012(a)gmail.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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: problem in using SampleFilterScheme
by Li Flost.Dexiong 04 Jul '12

04 Jul '12

thanks Phil, I did call read by using getProvider()--> read(xxx). I have another try this morning, but got another panic stack, this stack was quit small, but still have "Double fault".and this time, the probe function could not finish and got panic. the stack like this: Wed Jul 4 10:37:15 2012 panic(cpu 0 caller 0x2aaf41): Double fault at 0x2a17c65e, thread:0x6adeb7c, trapno:0x8, err:0x0,registers: CR0: 0x8001003b, CR2: 0x30c9fcc8, CR3: 0x00100000, CR4: 0x00000660 EAX: 0x00020000, EBX: 0x07fd3100, ECX: 0x2a17c64a, EDX: 0x2a17cd40 ESP: 0x30c9fcd0, EBP: 0x30ca0d18, ESI: 0x2a4bb2d2, EDI: 0x07fd5200 EFL: 0x00010202, EIP: 0x2a17c65e Backtrace (CPU 0), Frame : Return Address (4 potential args on stack) 0x819e18 : 0x21b837 (0x5dd7fc 0x819e4c 0x223ce1 0x0) 0x819e68 : 0x2aaf41 (0x59dfe4 0x59e0df 0x2a17c65e 0x6adeb7c) 0x819f58 : 0x2a29d2 (0x0 0x0 0x0 0x0) 0x30ca0d18 : 0x2a17c766 (0x7fd3e80 0x7fd2400 0x20000 0x0) com_apple_dts_driver_SampleFilterScheme::read(IOService*, unsigned long long, IOMemoryDescriptor*, IOStorageAttributes*, IOStorageCompletion*)+284>: add $0x1040,%esp 0x30ca1d68 : 0x2a17c766 (0x7fd3100 0x7fd2500 0x20000 0x0) com_apple_dts_driver_SampleFilterScheme::read(IOService*, unsigned long long, IOMemoryDescriptor*, IOStorageAttributes*, IOStorageCompletion*)+284 0x30ca2db8 : 0x2a4c111e (0x7fd5f00 0x7fd4f00 0x20000 0x0) IOStorage::read(IOService*, unsigned long long, IOMemoryDescriptor*, unsigned long long*)+88 0x30ca2e08 : 0x2a17c9b1 (0x7fd4f00 0x7fd5200 0x20000 0x0) com_apple_dts_driver_SampleFilterScheme::probe(IOService*, long*)+407 0x30ca3e78 : 0x541e8f (0x7fd5200 0x7fd4f00 0x30ca3f0c 0x70e6000) 0x30ca3f28 : 0x5426e9 (0x7fd4f00 0x709d7c0 0x0 0xffffffff) 0x30ca3f78 : 0x5443d5 (0x7fd4f00 0x0 0x30ca3fac 0x1) 0x30ca3fc8 : 0x2a179c (0x6834470 0x0 0x2a17ab 0x401aee4) Kernel Extensions in backtrace (with dependencies): com.apple.dts.driver.SampleFilterScheme(1.2.0d1)@0x2a17b000->0x2a17dfff dependency: com.apple.iokit.IOStorageFamily(1.6.3)@0x2a4af000 com.apple.iokit.IOStorageFamily(1.6.3)@0x2a4af000->0x2a4c7fff My change like this: 1) add a probe function to SampleFilterScheme; this function implement as description in my last mail. 2) modify read function declaration to virtual void read(IOService* client, UInt64 byteStart, IOMemoryDescriptor* buffer, IOStorageAttributes* attributes, IOStorageCompletion * completion); and its implementation is: void com_apple_dts_driver_SampleFilterScheme::read(IOService* __attribute__ ((unused)) client, UInt64 byteStart, IOMemoryDescriptor* buffer, IOStorageAttributes* attributes, IOStorageCompletion *completion) { char pname[4096] = {'\0'}; proc_selfname(pname, sizeof(pname)); DEBUG_LOG("%s[%p]::%s(%p, %llu, %p, %p, %p)\n", getName(), this, __FUNCTION__, client, byteStart, buffer, completion); printf("%s->%s[%p]::%s(%p, %llu, %p, %p)\n",pname, getName(), this, __FUNCTION__, client, byteStart, buffer, completion); getProvider()->read(this, byteStart, buffer, attributes, completion); } I'm curious about 1) why the GetProvider() will turn back to my class? 2) if the stack is small to 16KB, but in my probe function, i just using a array of 4096B. how can this be overflow? 在 2012-7-3，下午9:05， Phil Jordan 写道： > On Mon, Jul 2, 2012 at 5:53 PM, Flost Li <havefan2012(a)gmail.com> wrote: >> Hi all, >> >> I download the SampleFilterScheme example code from >> https://developer.apple.com/library/mac/#samplecode/SampleFilterScheme/Intr… >> But I have some issue on making study on it. >> >> I modify the code downloaded, and add the "probe" method. This probe >> function will scan a scsi disk and check whether it contains certain string >> on certain offset. If the probe function got the string, the function will >> return a score of 9999. >> >> I build the project, insert this kext, and than i connect my device. >> from the log, i notice that the disk was recognize by the kext, and the kext >> started. >> but soon after the driver start, the system got panic. when i restart the >> machine, and use gdb/kextutils to decode the panic stack sounds like below: >> (as the first several level are the same, i only print the first 7 level of >> the stack.) > > I can only guess, as you haven't provided any code, but: > > You need to call the read() method *on the provider IOMedia object* > (your underlying disk/partition) in your read() method. It looks like > you're calling this->read() or IOStorage::read() instead. > >> I'm wandering why the iokit got recursive status? does the panic called by >> the over stack, when the recursive for much time? > > Notice the "Double fault" - this almost always means you've hit a > stack overflow. Kernel stacks are only a little under 16KiB on OSX, so > they're very easy to blow, not that a bigger stack would have helped > in this case, as the recursion would probably have carried on forever. > > Hope that helps, > phil Best regards, Flost.D.Li havefan2012(a)gmail.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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: problem in using SampleFilterScheme
by Phil Jordan 03 Jul '12

03 Jul '12

On Mon, Jul 2, 2012 at 5:53 PM, Flost Li <havefan2012(a)gmail.com> wrote: > Hi all, > > I download the SampleFilterScheme example code from > https://developer.apple.com/library/mac/#samplecode/SampleFilterScheme/Intr… > But I have some issue on making study on it. > > I modify the code downloaded, and add the "probe" method. This probe > function will scan a scsi disk and check whether it contains certain string > on certain offset. If the probe function got the string, the function will > return a score of 9999. > > I build the project, insert this kext, and than i connect my device. > from the log, i notice that the disk was recognize by the kext, and the kext > started. > but soon after the driver start, the system got panic. when i restart the > machine, and use gdb/kextutils to decode the panic stack sounds like below: > (as the first several level are the same, i only print the first 7 level of > the stack.) I can only guess, as you haven't provided any code, but: You need to call the read() method *on the provider IOMedia object* (your underlying disk/partition) in your read() method. It looks like you're calling this->read() or IOStorage::read() instead. > I'm wandering why the iokit got recursive status? does the panic called by > the over stack, when the recursive for much time? Notice the "Double fault" - this almost always means you've hit a stack overflow. Kernel stacks are only a little under 16KiB on OSX, so they're very easy to blow, not that a bigger stack would have helped in this case, as the recursion would probably have carried on forever. Hope that helps, phil _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Expand relative path in KEXT...
by Justin C. Walker 02 Jul '12

02 Jul '12

On Jul 2, 2012, at 05:35 , simran charley wrote: > Hi, > > Is it possible to expand relative path in kext? In one sense, your question doesn't even have much meaning for a Kext. A Kext is a component of the kernel, and much of its code can't assume that it will run in a process context (so it won't have a "current working directory", and therefore, can't make sense of a relative path name). A more basic issue, though, is that you should not be trying to access files in kexts except in rare circumstances. There are other mechanisms available to do that (user-client interface) should it be necessary. What are you trying to accomplish? Justin -- Justin C. Walker, Curmudgeon-At-Large, Director Institute for the Enhancement of the Director's Income -------- The path of least resistance: it's not just for electricity any more. -------- _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

problem in using SampleFilterScheme
by Flost Li 02 Jul '12

02 Jul '12

_______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Is Darwin 11.4.0 capable of NFSv4?
by Andreas Boesen 02 Jul '12

02 Jul '12

Hi, Thanks Andreas _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com does OS X 10.7 support NFSv4 as a client. I couldn't find any documentation on developer.apple.com or Googol. There is only the man-Page[1] and a thread in this list[2] indicating that OS X 10.6 as NFSv4.0alpha support. Could someone point me to some kind of "changelog" or Documentation about the status of the NFS client implementation in OS X 10.7 please? [1] http://developer.apple.com/library/mac/#documentation/Darwin/Reference/ManP… [2] http://lists.apple.com/archives/darwin-kernel/2009/Aug/msg00048.html

1 0