June 2012 - Darwin-dev - lists.apple.com

openx_np() et al
by Alastair Houghton 21 Jun '12

21 Jun '12

Hi all, Does anyone know whether openx_np() et al, and the filesec_t functions on which they depend, are supposed to be undocumented? Right now, the only documented way to manipulate file ACLs is acl_get/set_*() or the not very well documented ATTR_CMN_EXTENDED_SECURITY attribute on get/setattrlist(), and the only documented way to manipulate the UUID-based ownership data is ATTR_CMN_UUID/ATTR_CMN_GRPUUID (again from get/setattrlist() et al). The odd part is that one of the pieces of sample code uses the filesec_t type and openx_np, even though there appears to be no documentation whatsoever for those functions. So: 1. Are there *supposed* to be man pages for filesec_init() et al, openx_np() et al? Are they just MIA for some reason? Is this supposed to be a documented API? (It appears to be a reasonable design, aside from the fact that chmodx_np() could do with a way to allow the caller to specify that it should set *as much* as possible and then return what it has set; setattrlist() also suffers from a similar issue.) 2. Why does the man page for getattrlist() not mention that the data returned by ATTR_CMN_EXTENDED_SECURITY is supposed to be opaque (it actually says it's a kauth_filesec struct)? Or that you can get at the ACL by using acl_copy_int_native()? i.e. that it's the expected "external representation"? I note also that various pieces of Open Source software whose sources are published on opensource.apple.com are using the filesec functions... Kind regards, Alastair. -- http://alastairs-place.net _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

Invalid pointers being passed to free() upon CFAllocatorSetDefault()
by Alexander Potapenko 20 Jun '12

20 Jun '12

Hi everyone, (my apologies to those who've seen this email in the Google-only mailing list) The following program: ==================================== $ cat t.mm #import <Foundation/Foundation.h> int main() { #ifdef REPLACE CFAllocatorSetDefault(kCFAllocatorMallocZone); #endif NSURL *base = [[NSURL alloc] initWithString:@"file://localhost/Users/glider/Library/"]; NSURL *u = [[NSURL alloc] initWithString:@"Saved Application State" relativeToURL:base]; return 0; } ==================================== behaves incorrectly if CFAllocatorSetDefault is called: $ clang++ t.mm -o t -DREPLACE -framework Foundation -g && ./t t(47457) malloc: *** error for object 0x10ba14348: pointer being freed was not allocated *** set a breakpoint in malloc_error_break to debug Abort trap: 6 This happens because _CFRuntimeCreateInstance (which is called from CFURLAlloc) checks whether the supplied allocator is kCFAllocatorSystemDefault and, if it is not, stores the allocator reference at the beginning of the allocated memory and returns the pointer to the allocated memory plus sizeof(CFAllocatorRef). See http://www.opensource.apple.com/source/CF/CF-550/CFRuntime.c for the reference. When the object is destroyed, nobody actually checks whether the pointer was skewed and corrects it: Breakpoint 1, 0x00007fff8f4e86c0 in malloc_error_break () (gdb) bt #0 0x00007fff8f4e86c0 in malloc_error_break () #1 0x00007fff8f4e8805 in free () #2 0x00007fff9375ed53 in object_dispose () #3 0x00007fff8e295086 in -[NSObject dealloc] () #4 0x00007fff8e854d4f in -[NSURL(NSURL) initWithString:relativeToURL:] () #5 0x0000000100000ded in main () at t.mm:36 To the best of my knowledge, _CFRelease (see CFRuntime.c again) is the only function that checks for the presence of the allocator reference at the beginning of the memory block, but it isn't called while destroying the object. My two questions are: 1. Is this correct that kCFAllocatorSystemDefault, not kCFAllocatorDefault (i.e. NULL), is treated specially? IIUC most of the time the libraries just pass NULL as the allocator. 2. How do I work around this? Looking for a CFAllocator pointer before each freed block within free() sounds lame (albeit efficient). Interposing _CFRuntimeCreateInstance with my own implementation that doesn't do dirty pointer tricks may be more correct, but it's too easy to break down next time CFRuntime.c changes. TIA, Alexander Potapenko Software Engineer Google Moscow _______________________________________________ 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: https://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.ap… This email sent to site_archiver(a)lists.apple.com

1 0

DVI startup/wakeup problems
by Ian Smith-Heisters 13 Jun '12

13 Jun '12

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

1 0