Related
" PWNED " :-D
As you know, Archos bootloaders check digital signatures of init and recovery kernels, so you need to install SDE to use custom kernels, and it somehow "watermarks" the device.
Good news everyone! I've disassembled both bootloaders, found the code which checks signature, and replaced it (first instructions of verify_hash function) with "return 0" which is "mov r0, #0; bx lr" in ARM assembly. It's much the same hack as on Archos 5, thanks EiNSTeiN from archos.g3nius.org for reverse engineering previous generation.
Archos gen8 boots using OMAP boot ROM from internal eMMC card. Primary bootloader ("boot0") is in 0x20000 bytes after the first sector of internal flash (i.e. at 0x200) and secondary bootloader is written into rawfs, /mnt/rawfs/avboot. boot0 contains image size and loading address in first 8 bytes.
So, here is the patch:
1) boot0: replace 8 bytes at 0x7520 from the beginning of mmcblk0 from 7F402DE9003091E5 to 0000A0E31EFF2FE1.
2) avboot: replace 8 bytes at 0x14424 in avboot from 7F402DE9003091E5 to 0000A0E31EFF2FE1 (same patch). 0x14424 from avboot beginning is usually 0x14824 from the beginning of mmcblk0p1 (avboot comes first in rawfs, just after 2 blocks of header).
Of course you need root to do it. I've done it on my Archos 101, then changed 1 byte in recovery image - it boots into recovery without problem (before the hack it didn't boot into this 1-byte changed recovery).
And of course do it with caution and at your own risk DO NOT replace the bytes if you find other original data at these offsets! Bad boot0 or avboot means bricked Archos. There must be some sort of test point (something connected to OMAP SYS_BOOT5 pin) to boot from USB, or a boot UART interface, so debricking the device must be possible, but it would require some effort to find it, find a proper bootloader and use it.
If someone wants to see IDA database, I'll send my.
P.S: I do not have enough messages to post inside Development subforum, so I'm posting here.
Great work! With this base, can yout get something like CW to run?
I'm so waiting for him to come back and say April fools.
I'm gonna screw him up if this was an april fool
First, if this is an April fools, I will find you and hurt you.
Second, what does all that mean anyway? Does that mean Cyanogen on Gen8 is near? Does it have anything to do with roms?
vitalif said:
P.S: I do not have enough messages to post inside Development subforum, so I'm posting here.
Click to expand...
Click to collapse
Maybe you should increase that number of post by explaining how you did this.
)))) No it isn't an April fool, my device now really has a modified recovery. Ridiculously modified (1 byte changed), but that's the proof!
Check the patch by yourself )) all you need to write to mmcblk0 is a standard linux dd tool... which is included into standard Archos busybox...
wdl1908 said:
Maybe you should increase that number of post by explaining how you did this.
Click to expand...
Click to collapse
In fact, it was not hard, and if I knew ARM assembly language before, it would be even easier... All I had to do is to find bootloader on the flash (boot0 is obviously in its beginning, and avboot is on /mnt/rawfs), copy it to computer, download IDA, feed bootloader to it and find functions similar to ones described on archos.g3nius.org (BigInteger_ModulusEnter, RSADecipher, etc). It also could be simpler, as BigInteger_ModulusEnter is mentioned inside an ASCII string inside data section... But I've found them by text search also there is a magic "ZMfX" in first 4 bytes of avboot and some other magic inside init and recovery... One also could use them to find interesting points in bootloader.
At first I've started disassembling with the wrong base address, but bootloader has code which copies itself to the correct one in the very beginning, so I've changed it and started over. In fact, it has size and address in first 8 bytes, so this also could be simpler...
So the hack is done, what needs to be done by now - utilize it and create some custom ROM or simply flash urukdroid without SDE...
chulri said:
Great work! With this base, can you get something like CW to run?
Click to expand...
Click to collapse
CW == ClockWorkMod recovery? I don't have any experience with CWM porting yet, but in theory yes, the hack gives us the ability to run custom recovery images.
Don't know alot about the bootloader, but what advantage does this have?
SWFlyerUK said:
Don't know alot about the bootloader, but what advantage does this have?
Click to expand...
Click to collapse
Hm. I'll explain... Bootloader is the program which starts up the device, similar to bootloader on your PC signature check in bootloader prevents us installing modified Linux kernel, initial ramdisk and recovery images. So, for example, we can't have netfilter in kernel without installing SDE, we can't have ClockWorkMod recovery on Archos at all, and we can't, for example, change MMC card splitting into 512M mmcblk0 for system + remaining for "internal SD" with data.
With signature check removed, all this is possible.
The underlying idea of all this signature checking is probably protecting f**king DRM... I HATE IT !!!!!! And hate companies promoting it =) When you install SDE on previous generation archos (5it), it removes drm keys from device memory (this is the "watermarking" mentioned on Archos site). It makes device unable to play the content buyed for it anymore... Not a big deal, but unpleasant. I don't know if this is the same on gen8.
In detail: Archos 101 has OMAP3630 processor. The "0-stage" (very-very first stage) bootloader, i.e. program which gains control after processor power-up, is hard-coded into one-time programmable area on the processor itself and is named "OMAP boot ROM" (similar to PC BIOS). The boot ROM can continue device booting process from different devices including SD/MMC card, NAND flash, UART (serial port) or USB interfaces. The boot sequence is determined from physical pin connection configuration. Our Archos boots from internal eMMC card.
So, OMAP boot ROM loads primary Archos bootloader, without checking any signatures or checksums, and simply transmits control to it. Primary bootloader sets up some processor configuration and then reads secondary bootloader (avboot) from flash. Then, it checks its MD5-RSA digital signature using Archos public key. If signature is incorrect, it hangs the device (goes to infinite loop). So if we modify avboot without removing signature check from boot0, device would be bricked. If signature is correct, control is transmitted to avboot. Avboot determines what system we want to start by pressing different keys, loads it, checks signature if system is init (normal system) or recovery, sets up configuration for Linux kernel and transmit control to Linux.
Interesting facts:
* According to the code, boot0 can use rawfs or FAT filesystems for boot partition.
* During boot process, various messages are printed to serial console. avboot even has some code for receiving commands over serial connections.
* OMAP processor boot sequence can be configured via special memory area which remains unchanged after soft reset, and this configuration will override one determined by physical pin configuration. This does not give us much profit, but is also interesting...
Thanks for the explanation, so is it worth doing for a noticable difference in performance etc?
SWFlyerUK said:
Thanks for the explanation, so is it worth doing for a noticable difference in performance etc?
Click to expand...
Click to collapse
Whats being done will have no affect on performance of the device. It will however, allow a lot of work that can contribute to better performance on the device. That is assuming that we can put on a modified clockworkmod recovery on these devices without bricking them.
He says the only way to do this is with root but in order to have root with r/w access at this point is SDE....right? Don't get me wrong custom recovery with the ability to make backups would be awesome but it seems SDE will still be necessary unless a new rooting option comes along.
*on a side note about root has anyone tried using psneuter to gain temp root through ADB? I really am not super knowledgeable about this stuff but this was used on the thunderbolt to aid in getting full root and s-off.
Sent from my ADR6400L using XDA App
JBO1018 said:
He says the only way to do this is with root but in order to have root with r/w access at this point is SDE....right? Don't get me wrong custom recovery with the ability to make backups would be awesome but it seems SDE will still be necessary unless a new rooting option comes along.
*on a side note about root has anyone tried using psneuter to gain temp root through ADB? I really am not super knowledgeable about this stuff but this was used on the thunderbolt to aid in getting full root and s-off.
Sent from my ADR6400L using XDA App
Click to expand...
Click to collapse
Archangel will give you temp root without using SDE.
He said root with r/w access. Archangel won't do that, the file system is still protected.
pbarrett said:
He said root with r/w access. Archangel won't do that, the file system is still protected.
Click to expand...
Click to collapse
Nope r/w access is not needed the only changes to be made are on /dev/mmcblk0p1 which is mounted on /mnt/rawfs the read-only is on the root file system so they are seperate. Archangel will do just fine for this.
wdl1908 said:
Nope r/w access is not needed the only changes to be made are on /dev/mmcblk0p1 which is mounted on /mnt/rawfs the read-only is on the root file system so they are seperate. Archangel will do just fine for this.
Click to expand...
Click to collapse
To be correct, there is no write protection on internal MMC at all, there is readonly rootfs which is mounted from a squashfs archive (squashfs is compressed readonly filesystem commonly used on Linux Live CDs), so you can't modify _files_ on it while it is mounted. But, nothing stops you from updating it as a whole.
Urukdroid
Someone should give a shout out ro $auron, creator of the Urukdroid project about this, he might find it useful.
So, if your hack is confirmed, that would give us the possibility to port CW recovery and Cyanogen to Gen8 devices... am I right ?
shrewdlove said:
Someone should give a shout out ro $auron, creator of the Urukdroid project about this, he might find it useful.
Click to expand...
Click to collapse
I think he has already seen this thread but you can ask him
lechuckthepirate said:
So, if your hack is confirmed, that would give us the possibility to port CW recovery and Cyanogen to Gen8 devices... am I right ?
Click to expand...
Click to collapse
Yes you are^^ but the thing is you have to port cyanogen to our gen8^^ and this must be done by a or more devs
i heard the biggest problem is that our touchscreen is connected by an usb controller inside the archos thats why the honeycomb port by luisivan is not recognize our touchscreen ( but when the source code is released, finally, we will get a hc port )
Lennb said:
i heard the biggest problem is that our touchscreen is connected by an usb controller inside the archos thats why the honeycomb port by luisivan is not recognize our touchscreen ( but when the source code is released, finally, we will get a hc port )
Click to expand...
Click to collapse
this isn't a problem for cyanogen (v7 = Android 2.3.3) because we have the source.
No, I'm not going to do anything with these, just posting for other devs.
https://review.tizen.org/git/?p=kernel/u-boot.git
Board codename is "trats", this seems like full build chain that can replace Sbl and things.
HF with it.
U-Boot alternate bootloader for Galaxy S2? Wow! Great news. This bootloader is much more functional than Samsung SBL.
good stuff
most of us are at almost end of warranty so i ll flash it when available
So Fast boot? And other benefits? Anyone can tell?
Sent from my GT-I9100 using Tapatalk 2
< here be reserved for dragons >
PS. Board codename is "trats" = "start" backwards!
@Rebellos:
Great! But...
1) What make you think it is the "trats" board that is used and not another one?
2) Where are the XOM registers read? (I cant find it in there..)
3) Can you have a look at this post, using the official Uboot repository code which seem of a later edition and very different...
E:V:A said:
@Rebellos:
Great! But...
1) What make you think it is the "trats" board that is used and not another one?
2) Where are the XOM registers read? (I cant find it in there..)
3) Can you have a look at this post, using the official Uboot repository code which seem of a later edition and very different...
Click to expand...
Click to collapse
1) I can't be sure it's exactly same board, thus there's naming
Device: I9100XX
Board: SLP_U1
perhaps it doesn't cover in 100% with available SGS2 devices, but I'm pretty certain if someone finds way to load it into memory it will init properly, parse PIT partition and try to find&boot kernel.
2) Nowhere, it's not needed on this booting stage. Though in other sources there are OM parsing procedures, rather for informative/debugging purposes.
3) I've been analysing these things for pretty long time with no unambigous results. I might just point your attention at that BIC instruction, it's masking out every other bits but 1,2,3,4,5 (counting from 0), but when you look into iROM disassembly, OM there is parsed multiple times, taken from different sources and in different ways. Even bits as high as 22 (AFAIR) of OM register are used for something (It might be HW hack made in later revisions by Samsung cuz they had some wakeup issues)
Rebellos pointed this protocol out earlier today.
https://review.tizen.org/git/?p=kernel/u-boot.git;a=blob;f=common/cmd_usbd.c
USB download mode...
Code:
1370 static const char dl_key[] = "THOR";
1371 static const char dl_ack[] = "ROHT";
Thor, Son of Odin, brother of Heimdall and tamer of Loki(Loke)...
Loke is the Odin3 Daemon. I bet this can be used for heimdall development.
https://review.tizen.org/git/?p=kernel/u-boot.git;a=blob;f=common/cmd_usbd.c#l1307
I'm a little late to this thread, but this is the s5pc110 goodies:
https://review.tizen.org/git/?p=kernel/u-boot.git;a=tree;f=board/samsung/universal_c110
Also interesting is that this is 2011.03. I've noticed that hw vendors usually use the older 1.1.x or 1.3.x versions of u-boot.
For a hw vendor, this is fairly recent. (unless your freescale or other more progressive vendors)
what is this, a kernel? a recovery mode, i don't get it
LastStandingDroid said:
what is this, a kernel? a recovery mode, i don't get it
Click to expand...
Click to collapse
A bootloader?
http://forum.xda-developers.com/showthread.php?t=1680898
It's a bootloader that you flash to kernel partition, not sbl
Sent from my GT-I9100 using Tapatalk 2
Hi all,
Has anyone followed Rayman's excellent article the-inner-workings-of-secure-boot-key-and-nvflash and fully recovered a N7 from APX only mode?
I have this situation which I think resulted from the battery dying during the 4.4.2 update - Doh I know, but thought I had enough juice to complete the update.
Rayman says the required files will be made available but I cannot find them anywhere
Since every motherboard has a unique key, there is no generic blob. To be able to recover your N7, you will need a backup of it, but it's impossible to make if your device is dead.
Try to send it to Asus/Google.
Erovia said:
Since every motherboard has a unique key, there is no generic blob. To be able to recover your N7, you will need a backup of it, but it's impossible to make if your device is dead.
Try to send it to Asus/Google.
Click to expand...
Click to collapse
Did you read the article? Sounds like you can use the sbk which is a hash of the cpuid...
Nope, but why don't you ask around in the flatline topic?
Erovia said:
Nope, but why don't you ask around in the flatline topic?
Click to expand...
Click to collapse
too much of a noob to post on the forum, but thanks for the pointer.
FYI Raymans article. It does sound possible to bring it back, but there was no follow up with the required files;
What is Secure Boot Key and how does it work?
I've been getting lots of questions about this, so here is some simple background:
The secure boot key is an AES128 encryption key that can used to encrypt various data on the flash memory. It's a generic nvidia tegra2 thing, that the manufacturer can optionally use to make their device more "secure".
When the SBK is set, it's stored in a one-time-programmable "fuse". This also means that now that the key is out, they can't change it on already released devices, only new devices.
When the tegra2 starts up, the AES key is available to the hardware AES engine only. E.g. not even the bootloader can read it back! However, the bootloader can *use* the key to encrypt whatever data it wants through the hardware AES engine. And here is the explanation why the blob flashing method actually works! The bootloader checks for the blob in the staging partition and encrypts and flashes it as needed.
Once the bootloader is done, it clear the key from the AES engine which makes it impossible to encrypt or decrypt things from within the OS.
So what happens when it boots into APX/Nvflash mode?
The basic APX mode is stored in the BootROM and hence can never be changed. It appears to accept only a very limited range of commands, and each command needs to be encrypted using the SBK to be accepted. If it receives a command that's not properly encrypted, it disconnects the USB and appears to be off. This is the dreaded "0x4" error that people have been getting when attempting to get nvflash working.
It should be noted, that even with the SBK inputted into nvflash, most regular nvflash commands won't be available. I'm still not entirely sure why (and I can't rule out it will change).
What *is* available, is the nvflash --create command. What this command does is repartition and format all partitions, set bct and odmdata and send over all needed partitions to the device (and encrypt them as needed). This means a full recovery is possible, but regular ability to flash e.g. just boot.img or read partitions off of the device is not possible at this point.
So what do we need for nvflash?
In order to get a working (e.g. --create) nvflash, we need a few bits of information as well as some files:
◦Secure Boot Key
◦BCT file (boot device setup, ram configuration and a bit more)
◦ODM data (board-specific bit-field specifying various board settings. *Needs* to be correct
◦flash.cfg (e.g. list of settings and names/identifiers of partitions.
On top of these files, we also need all the partitions, e.g. bootloader.bin, boot.img, recovery.img and system.img. Luckily, these partition files are available in official ASUS updates and can be extracted from the blob file using my blob tools
The first four peices aren't readily available, but through lots of effort and a good deal of luck, we have managed to recreate the needed files. Secure Boot Key has already been released (note that this was by far the hardest!) and the rest will most likely follow over the weekend. Keep in mind that we want to keep this legal, so don't expect us to release any ready-made packs for unbricking! We will however make the recreated files available. Since these are recreated and not actual ASUS files, there should be no problems with them.
I hope this helps give a better understanding of how and what secure boot key is and what it gives us.
LG G5 VERIZON VS987
Unofficial bootloader unlock
In-Progress
This is a project to disassemble and rebuild an unlocked aboot that passes the sbl loader test, thus allowing installation of custom kernels, read/write access to system, TWRP recovery, and custom ROMs. You can follow the progress or ask questions about it here, or offer your help to make my life a little easier. Donations help, too, especially to maintain the machine these VMs and tools run on.
PHASE 1 - RAMDISK EXTRACTION: DONE
(Attachment 2 & 3): I have extracted the ramdisks from the KDZ bin data. I have the ramdisks and accompneying kernel files in tar files.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
PHASE 2 - TARGET DISASSEMBLY: DONE
(Attachment 1): I have disassembled the ELF binaries. objdump is sloppy, but I was able to get to the branch instructions and find my offsets for each subroutine. I also located the factory signature data for verification by sbl.
PHASE 3 - CERTIFICATE EXTRACTION: IN-PROGRESS
Extracting and parsing the PCS data. It's needed for verification by the sbl to avoid security fail when booting a modified aboot.
PHASE 4 - UNLOCKING NEEDS DONE
Re-coding aboot to accept a bootloader unlock, and access to fastboot.
PHASE 5 - RECOMPILATION AND TESTING (ABOOT)NEEDS DONE
Re-building the aboot binary and testing it.
PHASE 5 - LITTLEKERNEL AND SYSTEM ACCESS (BOOT)NEEDS DONE
Puling apart stock kernel, grabbing needed modules, disabling SELinux and dmverify.
Analysing system image, removing boot-time LG security checks, applying root.
PHASE 6 - TWRP BUILD (RECOVERY) NEEDS DONE
Building user-friendly recovery image.
PHASE 7 - BUILD TOT FILE NEEDS DONE
Package it all into an LGUp Flashable TOT and release.
If you would like to help with this project, please make yourself known.
If you would like to donate to show thanks, the button is underneath my name or in my profile.
If you have a comment or can leave documentation or advice, do so below
Nice.. Hoping for the best
LupineDream said:
LG G5 VERIZON VS987
Unofficial bootloader unlock
In-Progress
This is a project to disassemble and rebuild an unlocked aboot that passes the sbl loader test, thus allowing installation of custom kernels, read/write access to system, TWRP recovery, and custom ROMs. You can follow the progress or ask questions about it here, or offer your help to make my life a little easier. Donations help, too, especially to maintain the machine these VMs and tools run on.
PHASE 1 - RAMDISK EXTRACTION: DONE
(Attachment 2 & 3): I have extracted the ramdisks from the KDZ bin data. I have the ramdisks and accompneying kernel files in tar files.
View attachment 4016279
View attachment 4016280
PHASE 2 - TARGET DISASSEMBLY: DONE
(Attachment 1): I have disassembled the ELF binaries. objdump is sloppy, but I was able to get to the branch instructions and find my offsets for each subroutine. I also located the factory signature data for verification by sbl.
View attachment 4016278
PHASE 3 - CERTIFICATE EXTRACTION: IN-PROGRESS
Extracting and parsing the PCS data. It's needed for verification by the sbl to avoid security fail when booting a modified aboot.
PHASE 4 - UNLOCKING NEEDS DONE
Re-coding aboot to accept a bootloader unlock, and access to fastboot.
PHASE 5 - RECOMPILATION AND TESTING (ABOOT)NEEDS DONE
Re-building the aboot binary and testing it.
PHASE 5 - LITTLEKERNEL AND SYSTEM ACCESS (BOOT)NEEDS DONE
Puling apart stock kernel, grabbing needed modules, disabling SELinux and dmverify.
Analysing system image, removing boot-time LG security checks, applying root.
PHASE 6 - TWRP BUILD (RECOVERY) NEEDS DONE
Building user-friendly recovery image.
PHASE 7 - BUILD TOT FILE NEEDS DONE
Package it all into an LGUp Flashable TOT and release.
If you would like to help with this project, please make yourself known.
If you would like to donate to show thanks, the button is underneath my name or in my profile.
If you have a comment or can leave documentation or advice, do so below
Click to expand...
Click to collapse
Nice to know there is people working on an unofficial bootloader unlock for the lg g5 good job !
I would also like to ask a question if you don't mind : once done, does this variant of the g5 is similar enough to some variants to be able to "port" the unlock or the approach is too specific to the variant ?
Im currently grabbing a more android build-friendly release of Xenial. Google does not support building from a x86 OS. When I tried to git the "official" google toolchain in consideration of doing things 100% correct in a VM, I ran into the dreaded "x86 build host unsupported" problem. The VM host I am grabbing is here: https://forum.xda-developers.com/chef-central/android/guide-how-to-setup-ubuntu-16-04-lts-t3363669 I am going to need the google build source for the kernel, plus the makefiles for the generic arm v7a neon arcitecture. Best do it all the right way instead of having to re-do it all later.
In the meantime, I am cramming some thumb-2 tutorials. La la la~
nalf3in said:
Nice to know there is people working on an unofficial bootloader unlock for the lg g5 good job !
I would also like to ask a question if you don't mind : once done, does this variant of the g5 is similar enough to some variants to be able to "port" the unlock or the approach is too specific to the variant ?
Click to expand...
Click to collapse
I'd have to check the aboot and boot / kernel to see if there's any major differences. Since this is a direct aboot disassembly, it's varient specific, but I will surely be checking if the approach can be applied to different varients' booters via scripting or some such thing. It would mean coding of a patchfile and a TOT builder, adding an entire extra level of complexity, but I'll get to your answer as soon as I have finished with this model.
Howdy there,
Looks like you're attempting to modify an element in the Qualcomm Secure Boot chain (aboot in this case) - unfortunately what you're trying to do won't work.
I don't want to rain on your parade here, or discourage you from learning something for the sake of learning it but simply modifying a component of the boot stack by recompiling it and slapping the same signatures on it isn't possible.
I'd recommend you take a look at any of the awesome posts on XDA about how the Secure Boot 3 (is there a newer version?) process works as you'll see that each step in the boot process (pbl -> xbl -> {tz, rpm} -> aboot) is verified using RSA public/private key signatures.
You won't be able to simply extract the certificate chain from the old aboot (LK) image and append it to your source-built LK, as the hash embedded in the certificate chain won't match the computed hash of the LK application.
Your next thought would be "what if I could find where the validation function is called and replace it?", but you'll find that each step in the boot process calls TrustZone's secureboot verification functions which rely on data stored in qFuses.
Even then, the function TrustZone itself calls is actually provided by PBL and is burned into the chip itself.
Essentially, the path you're trying to take is one that Qualcomm has thoroughly hardened the boot process to protect against.
I'm not trying to discourage you from learning, since doing this sort of reverse engineering is very fascinating and a fun learning exercise - I just want to let you know that if you try to flash a modified aboot image that you'll end up with a brick (and no way that I'm aware of to recover short of soldering up some UFS lines and reflashing like that since iirc even JTAG is disabled now and we lack signed programmers for USB recovery).
Check out one of the Secure Boot PDFs that are available on Google/XDA and you should see what I mean.
Best of luck to you, and happy learning!
LupineDream said:
(Attachment 2 & 3): I have extracted the ramdisks from the KDZ bin data. I have the ramdisks and accompneying kernel files in tar files.
View attachment 4016279
View attachment 4016280
PHASE 2 - TARGET DISASSEMBLY: DONE
(Attachment 1): I have disassembled the ELF binaries. objdump is sloppy, but I was able to get to the branch instructions and find my offsets for each subroutine. I also located the factory signature data for verification by sbl.
View attachment 4016278
Click to expand...
Click to collapse
Like thecubed said, by flashing a modified aboot you'll brick the device, but the fact that you figured out how to extract the aboot image is EXTREMELY impressive! If you wouldn't mind it would be very helpful if you make a guide on how you did that or if you could post your extracted aboot files that would be much much appreciated An insight on how aboot for the G5 works would be something very nice to have
thecubed said:
Howdy there,
Looks like you're attempting to modify an element in the Qualcomm Secure Boot chain (aboot in this case) - unfortunately what you're trying to do won't work.
I don't want to rain on your parade here, or discourage you from learning something for the sake of learning it but simply modifying a component of the boot stack by recompiling it and slapping the same signatures on it isn't possible.
...
[ removed in quote ]
...
Best of luck to you, and happy learning!
Click to expand...
Click to collapse
Thank you for the inspiration. The task has been a thorough crash-course in arm v7a NEON Cortex technology and thumb-2 assembler. I was concerned when I saw the TrustZone documentation from ARM and the ACEP (Arbitrary Code Execution Prevention) stack. Your mention of qfuse brings back memories of the Samsung "Golden Routines" folly.
Unfortunetly I wasn't a backer of the Raspberry PI, but there are other PCBs with GPIO pins. My reason for mention this is the JTAG connundrum you mentioned. It is null and void to a soldering gun, a few resisters ($2-$3, and the on-chip ARM debugging interface included in most integrated ARM chipsets. PCBs that run Linux with a GPIO pin array are the cheapest and most flexable solution to an expensive piece of lab / R & D equipment (JTAG), and the impact on the bank account is a fraction of what it would cost the end-consumer.
Hash collisions and backprop neural nets. What? OK, I'll explain. Most reverse engineering begins with monitoring the usual operation of the device as you are aware. Approach 1 involves training a backpropping neural network to recognize (within a reasonably less amount of computation) a hash collision. Of course, you aren't going to be getting to these pre-coded hashes without breaking the RSA chain (thats 100 zeros O__O ) , this is fixed by monitoring the debug interface we both mentioned earlier, at power-on. ARM processors work such that they cannot do DMAC (direct memory access computation), so the values need to be pushed into registers and acted on as plain byte, then pushed back to memory. Sniff, find out what's coming to/from the qfuse memory bus area into registers, dump the data, and you'll have your root CA with luck. You'll be looking for the in operand, that is, the one that is to be acted on in the cipher that is incoming from qfuse. Cycle-timing to ignore the rest of the computations for this byte, you're basically waiting until qfuse pushes the next byte of its keys into a register and grabbing it in the same cycle the instruction is working on it.
That said, with a $100 and a soldering gun and a laptop, you have your RSA key direct from CPU memory via the debug interface. After you've got your root CA, would it be feasable to start decrypting the intermediate CA chain for each component of boot until you get the hash? In theory this should work without brute-force, but a lot of time in a code window and with an oscilliscope. Once you've had the decrypted hash, you can begin to look for a hash collision. There are open source tools that will find hash collisions in an hour or less using neural nets. Write an aboot that complies to this hash collision, flash it, and qfuse will not be able to tell the difference.
Thats my theory on exploiting the fact ARM needs to do its computations in an area that is accessable to the debug interface. Of course my theory falls apart if the TrustZone chipset has its own processor.
Honestly Annoying said:
Like thecubed said, by flashing a modified aboot you'll brick the device, but the fact that you figured out how to extract the aboot image is EXTREMELY impressive! If you wouldn't mind it would be very helpful if you make a guide on how you did that or if you could post your extracted aboot files that would be much much appreciated An insight on how aboot for the G5 works would be something very nice to have
Click to expand...
Click to collapse
I'd be okay with writing such guide as long as it followed community guidelines. I should not be posting disassembled code and analysis if it's against the rules to do so. If a mod could clear this for me....
After searching around I located the true specs for the G5. It's a Snapdragon 820 SoC. It runs the new ARM v8A instruction set.
More detailed information here -> http://www.tomshardware.com/reviews/snapdragon-820-performance-preview,4389-2.html.
More specifically, it's on the msm8996 platform: http://system-on-a-chip.specout.com/l/1170/Qualcomm-Snapdragon-820-MSM8996
Phones with this arcitechture: https://www.kimovil.com/en/list-smartphones-by-processor/qualcomm-snapdragon-820-msm8996
Xiaomi Mi5 has official bootloader unlock via a tool (reverse engineer to work with our platform?)
HTC 10 bootloader on certain models can be oem-unlocked.
Lenovo ZUK Z2 bootloader can be oem-unlocked
LG V20 bootloader can be oem-unlocked, and is the closest in hardware to our G5!
Perhaps stealing a kernel and boot from a similar varient that allows it, signing it with our key, hash-matching it? It all runs on the same platform with the same instruction set. Most of these phones have unlockable bootloaders.
I'd say the best target would be that unlock tool by Xaoimi. Its usually the third party vendor's tools you can find the biggest security holes in. xD From their documentation, it isn't an "enable OEM unlock" switch in Developer Options used to enable the unlock, the tool itself actually works on the bootloader. Perhaps the tool uses a feature of the MSM8996 we aren't aware of. It's worth looking into.
LupineDream said:
Thank you for the inspiration. The task has been a thorough crash-course in arm v7a NEON Cortex technology and thumb-2 assembler. I was concerned when I saw the TrustZone documentation from ARM and the ACEP (Arbitrary Code Execution Prevention) stack. Your mention of qfuse brings back memories of the Samsung "Golden Routines" folly.
Unfortunetly I wasn't a backer of the Raspberry PI, but there are other PCBs with GPIO pins. My reason for mention this is the JTAG connundrum you mentioned. It is null and void to a soldering gun, a few resisters ($2-$3, and the on-chip ARM debugging interface included in most integrated ARM chipsets. PCBs that run Linux with a GPIO pin array are the cheapest and most flexable solution to an expensive piece of lab / R & D equipment (JTAG), and the impact on the bank account is a fraction of what it would cost the end-consumer.
Hash collisions and backprop neural nets. What? OK, I'll explain. Most reverse engineering begins with monitoring the usual operation of the device as you are aware. Approach 1 involves training a backpropping neural network to recognize (within a reasonably less amount of computation) a hash collision. Of course, you aren't going to be getting to these pre-coded hashes without breaking the RSA chain (thats 100 zeros O__O ) , this is fixed by monitoring the debug interface we both mentioned earlier, at power-on. ARM processors work such that they cannot do DMAC (direct memory access computation), so the values need to be pushed into registers and acted on as plain byte, then pushed back to memory. Sniff, find out what's coming to/from the qfuse memory bus area into registers, dump the data, and you'll have your root CA with luck. You'll be looking for the in operand, that is, the one that is to be acted on in the cipher that is incoming from qfuse. Cycle-timing to ignore the rest of the computations for this byte, you're basically waiting until qfuse pushes the next byte of its keys into a register and grabbing it in the same cycle the instruction is working on it.
That said, with a $100 and a soldering gun and a laptop, you have your RSA key direct from CPU memory via the debug interface. After you've got your root CA, would it be feasable to start decrypting the intermediate CA chain for each component of boot until you get the hash? In theory this should work without brute-force, but a lot of time in a code window and with an oscilliscope. Once you've had the decrypted hash, you can begin to look for a hash collision. There are open source tools that will find hash collisions in an hour or less using neural nets. Write an aboot that complies to this hash collision, flash it, and qfuse will not be able to tell the difference.
Thats my theory on exploiting the fact ARM needs to do its computations in an area that is accessable to the debug interface. Of course my theory falls apart if the TrustZone chipset has its own processor.
I'd be okay with writing such guide as long as it followed community guidelines. I should not be posting disassembled code and analysis if it's against the rules to do so. If a mod could clear this for me....
Click to expand...
Click to collapse
Hello again!
Regarding your points:
JTAG is disabled on most production devices, as in 100% inoperable. No amount of soldering will enable it. I'm sure there are some exceptions to the rule, but in this case I'm fairly confident in saying that the JTAG interface on the G5 is unusable.
...even if JTAG was enabled, the Qualcomm Secure Boot stack is designed to protect itself from the exact type of attack you're describing here. The component of TrustZone that is doing the verification of boot images is not running in anything that is JTAG-accessible, and to my understanding it's not even running in the main ARM core.
Re: hash collisions and neural nets... What you're describing sounds neat in theory, but in application (again, to my knowledge) won't work. I suggest you read up on how RSA works.
RSA is public/private key cryptography - in this case, the certificates contained on the phone are the public component, and LG/Qualcomm are the only ones with the private component. "Extracting the CA" will yield the public portion that is only useful for verifying signatures, not signing them.
The only way you'd be able to sign anything and have the phone trust it is to either a) replace the CA on the phone (not possible, it's burned into qFuses) or b) obtain the private component of the CA or any of it's subsidiaries (also burned into qFuses)
CA chains aren't encrypted, they're just a list of things that the device will accept a signature for. In our case it's Qualcomm as the Root CA, then LG as an intermediate (and possibly a few other intermediates to allow OTA updates to come from differing builders/engineering teams within LG). Again, extracting the 'hash' of a CA will do no good here, as there's no meaningful collisions that can be generated and still be a valid boot image. I'm not the best person to explain RSA in depth, so I'd really recommend doing some further research on it.
LupineDream said:
After searching around I located the true specs for the G5. It's a Snapdragon 820 SoC. It runs the new ARM v8A instruction set.
More detailed information here -> http://www.tomshardware.com/reviews/snapdragon-820-performance-preview,4389-2.html.
More specifically, it's on the msm8996 platform: http://system-on-a-chip.specout.com/l/1170/Qualcomm-Snapdragon-820-MSM8996
Phones with this arcitechture: https://www.kimovil.com/en/list-smartphones-by-processor/qualcomm-snapdragon-820-msm8996
Xiaomi Mi5 has official bootloader unlock via a tool (reverse engineer to work with our platform?)
HTC 10 bootloader on certain models can be oem-unlocked.
Lenovo ZUK Z2 bootloader can be oem-unlocked
LG V20 bootloader can be oem-unlocked, and is the closest in hardware to our G5!
Perhaps stealing a kernel and boot from a similar varient that allows it, signing it with our key, hash-matching it? It all runs on the same platform with the same instruction set. Most of these phones have unlockable bootloaders.
I'd say the best target would be that unlock tool by Xaoimi. Its usually the third party vendor's tools you can find the biggest security holes in. xD From their documentation, it isn't an "enable OEM unlock" switch in Developer Options used to enable the unlock, the tool itself actually works on the bootloader. Perhaps the tool uses a feature of the MSM8996 we aren't aware of. It's worth looking into.
Click to expand...
Click to collapse
Bootloaders for other phones will not work on our phone, as of course it'll fail the Secure Boot check.
Bootloader unlock tools for other phones will not work because they're relying on manufacturer specific unlock code that's compiled into aboot (LK). Qualcomm's CAF version of LK doesn't include any unlock code checking functionality, so most manufacturers add that themselves.
LG, in this case, is not even including the code for a bootloader unlock in the US model bootloaders. If you're familiar with C, essentially LG has `#ifdef`'d the entire section of code out (including fastboot).
The V20 bootloader does indeed contain oem-unlock code, but Secure Boot will prevent you from flashing the V20's bootloader (even if it magically was code-compatible with the G5) because Secure Boot checks the hardware ID that's burned into qFuses.
This means, to add a bootloader unlock, you'd have to modify aboot, which can't be done because of Secure Boot. Secure Boot components can't be modified because of RSA, and the RSA verification can't be altered because the keys are burned into qFuses.
thecubed said:
Hello again!
Regarding your points:
JTAG is disabled on most production devices, as in 100% inoperable. No amount of soldering will enable it. I'm sure there are some exceptions to the rule, but in this case I'm fairly confident in saying that the JTAG interface on the G5 is unusable.
...even if JTAG was enabled, the Qualcomm Secure Boot stack is designed to protect itself from the exact type of attack you're describing here. The component of TrustZone that is doing the verification of boot images is not running in anything that is JTAG-accessible, and to my understanding it's not even running in the main ARM core.
Re: hash collisions and neural nets... What you're describing sounds neat in theory, but in application (again, to my knowledge) won't work. I suggest you read up on how RSA works.
RSA is public/private key cryptography - in this case, the certificates contained on the phone are the public component, and LG/Qualcomm are the only ones with the private component. "Extracting the CA" will yield the public portion that is only useful for verifying signatures, not signing them.
The only way you'd be able to sign anything and have the phone trust it is to either a) replace the CA on the phone (not possible, it's burned into qFuses) or b) obtain the private component of the CA or any of it's subsidiaries (also burned into qFuses)
CA chains aren't encrypted, they're just a list of things that the device will accept a signature for. In our case it's Qualcomm as the Root CA, then LG as an intermediate (and possibly a few other intermediates to allow OTA updates to come from differing builders/engineering teams within LG). Again, extracting the 'hash' of a CA will do no good here, as there's no meaningful collisions that can be generated and still be a valid boot image. I'm not the best person to explain RSA in depth, so I'd really recommend doing some further research on it.
Bootloaders for other phones will not work on our phone, as of course it'll fail the Secure Boot check.
Bootloader unlock tools for other phones will not work because they're relying on manufacturer specific unlock code that's compiled into aboot (LK). Qualcomm's CAF version of LK doesn't include any unlock code checking functionality, so most manufacturers add that themselves.
LG, in this case, is not even including the code for a bootloader unlock in the US model bootloaders. If you're familiar with C, essentially LG has `#ifdef`'d the entire section of code out (including fastboot).
The V20 bootloader does indeed contain oem-unlock code, but Secure Boot will prevent you from flashing the V20's bootloader (even if it magically was code-compatible with the G5) because Secure Boot checks the hardware ID that's burned into qFuses.
This means, to add a bootloader unlock, you'd have to modify aboot, which can't be done because of Secure Boot. Secure Boot components can't be modified because of RSA, and the RSA verification can't be altered because the keys are burned into qFuses.
Click to expand...
Click to collapse
So in this case, what would be the next image in the stack that would allow any kind of modifications? Should I be looking at boot.img instead? Would there be a method of tricking the signed and secure boot chain into believing what it sees isn't executing as root?
had been looking through some old methods of root, like causing a boot error by zeroing laf or flashing over it, causing the bootloader to drop you into fastboot, booted securely from there you could call a kcal perameter of the stock kernel that allowed a sort of debugging mode with systemwide root. That exploit was in the G2 era, and how that device obtained root.
I've seen a method that causes Knox to lock up on some Samsung devices by overloading its memory addresses or repeatedly zeroing certain bits of RAM.
I'd really need to find out what method works. If you can't make any modification, maybe there's a workaround to make it THINK everything is legit beyond the boot chain.
aboot decription
Ok I have been working on an lg g5 vs987 myself and I got to the recoding the aboot part and was totally lost its all encrypted and I have no idea where to start to even see what the code is really saying I am new to this website and I am also new to coding on android if you can guide me in the right direction I might be able to help. I have always dreamed of being a part of a development like this and now I might have a shot. Thank you so much for your work and I hope to hear from you soon!
alphawolves said:
Ok I have been working on an lg g5 vs987 myself and I got to the recoding the aboot part and was totally lost its all encrypted and I have no idea where to start to even see what the code is really saying I am new to this website and I am also new to coding on android if you can guide me in the right direction I might be able to help. I have always dreamed of being a part of a development like this and now I might have a shot. Thank you so much for your work and I hope to hear from you soon!
Click to expand...
Click to collapse
Dissembling and recoding aboot wont work no matter even if it is not encrypted..
It is already mentioned above by @thecubed ..
The secure boot will always verify the signature and dissassembly will generate no signature.
We need to find the private keys used for generating the public hash.. but that is not possible unless leaked from lg which is also very unlikely..
What a hacker can do is find a bug/vulnerability in aboot that can bypass secureboot..or a hardware loophole..
Plus there is a trustzone that itself secures the secureboot process.. so we have to find ways to exploit the bootchain so that we can somehow make the bootloader load unsigned ramdisks/recovery and such..
experienced people could do it.. but i guess there has been a lot of fuss between devs and wannabees..
Also if we can get debug build of aboot/ramdisk we can flash them to get unlock or at the very least root..
I am not expert on this but that how its simply put
A brave member sent me a very useful PDF specification on the TSF. The TSF is the official name for the portion of the integrated circuitry that controls the storage of the secure boot chain. The hardware routines are there to allow root to happen, but as stated before by @thecubed, LG commented out the entire section of a boot at compile time that allows anything to occur. Now the approach is to disassemble HOW the code that LG commented out works in the international variant that has an OEM unlock. Hardware routines exist that place the TSF into non-commercial mode, disabling functionality of certain enterprise software and deauth-ing the root CA in qfuse, allowing the user to flash their own signing certificates to the TSF (TrustZone) when a request to switch off CC mode is sent. There are all kinds of virtual memory protections, ACE protections, malicious code protections, and other things that the TSF handles. Trying any kind of unauthorised write to protected areas of memory results in blow of qfuse write fuse. It's actually a physical microscopic fuse that when pushed a specific voltage pops at a microscopic level kind of like a big fuse blowing only at a very tiny level. At this point write access anywhere is gone. This happens at the same time a full wipe of the system happens. That is why they say once qfuse blows your phone becomes a very expensive cup holder. Because after the qfuse blows there is no way of software recovery.
The only way to be able to disable commercial (CC mode) and be allowed to do anything to put your own boot chain in is to place the device in debug mode. The international varient contains code how to accomplish this, our devices don't. You would need to compile an app that runs in normal mode, causes a flag to be set that places the device into debug mode, then reboot. While in debug boot, you should be able to execute a CC unlock manually. The PDF I got says it's very specifically timed when this can happen, what parts of boot the TSF allows it to happen, and a rough explanation if what disabling CC mode means. The only way of getting root is to use the TSF-approved method but all this code is removed. The TSF does not stop you from executing code if booted into factory debug mode. The new approach I propose is to find an exploit to get into userdebug, and manually write an unlock routine with disassembled information from the international varient., pushing it directly into execution memory while in userdebug, being absolutely sure to give the TSF what it asks for, when it asks for it, at the exact timing for it.
LupineDream said:
A brave member sent me a very useful PDF specification on the TSF. The TSF is the official name for the portion of the integrated circuitry that controls the storage of the secure boot chain. The hardware routines are there to allow root to happen, but as stated before by @thecubed, LG commented out the entire section of a boot at compile time that allows anything to occur. Now the approach is to disassemble HOW the code that LG commented out works in the international variant that has an OEM unlock. Hardware routines exist that place the TSF into non-commercial mode, disabling functionality of certain enterprise software and deauth-ing the root CA in qfuse, allowing the user to flash their own signing certificates to the TSF (TrustZone) when a request to switch off CC mode is sent. There are all kinds of virtual memory protections, ACE protections, malicious code protections, and other things that the TSF handles. Trying any kind of unauthorised write to protected areas of memory results in blow of qfuse write fuse. It's actually a physical microscopic fuse that when pushed a specific voltage pops at a microscopic level kind of like a big fuse blowing only at a very tiny level. At this point write access anywhere is gone. This happens at the same time a full wipe of the system happens. That is why they say once qfuse blows your phone becomes a very expensive cup holder. Because after the qfuse blows there is no way of software recovery.
The only way to be able to disable commercial (CC mode) and be allowed to do anything to put your own boot chain in is to place the device in debug mode. The international varient contains code how to accomplish this, our devices don't. You would need to compile an app that runs in normal mode, causes a flag to be set that places the device into debug mode, then reboot. While in debug boot, you should be able to execute a CC unlock manually. The PDF I got says it's very specifically timed when this can happen, what parts of boot the TSF allows it to happen, and a rough explanation if what disabling CC mode means. The only way of getting root is to use the TSF-approved method but all this code is removed. The TSF does not stop you from executing code if booted into factory debug mode. The new approach I propose is to find an exploit to get into userdebug, and manually write an unlock routine with disassembled information from the international varient., pushing it directly into execution memory while in userdebug, being absolutely sure to give the TSF what it asks for, when it asks for it, at the exact timing for it.
Click to expand...
Click to collapse
Thanks for that post @LupineDream ! I unfortunately doesn't understand most of the things you explained but perhaps I can save you a bit of time (your last post looks like you aren't aware of that exploit but thats maybe only me misundertanding) ; as far as I know the adb root method develloped by @HonestlyAnnoying include a userdebug kernel so (afaik) we already know an exploit to get into userdebug on marshmallow(dirty santa) and the fact that the users needs to run marshmallow shouldn't matter as (afaik) once the bootloader unlocked, the userdebug kernel is no longer needed and it is possible to change it.
Edit: just saw your post on the adb root thread, so I guess you are now aware of the exploit Sorry for the post, I just wanted to be sure you didn't missed it.
nalf3in said:
Thanks for that post @LupineDream ! I unfortunately doesn't understand most of the things you explained but perhaps I can save you a bit of time (your last post looks like you aren't aware of that exploit but thats maybe only me misundertanding) ; as far as I know the adb root method develloped by @HonestlyAnnoying include a userdebug kernel so (afaik) we already know an exploit to get into userdebug on marshmallow(dirty santa) and the fact that the users needs to run marshmallow shouldn't matter as (afaik) once the bootloader unlocked, the userdebug kernel is no longer needed and it is possible to change it.
Edit: just saw your post on the adb root thread, so I guess you are now aware of the exploit Sorry for the post, I just wanted to be sure you didn't missed the exploit.
Click to expand...
Click to collapse
Yes thank you for affirming that, but not to worry, I got it. LG has an sbl module called anti-rollback that prevents flashing older software. If we look at the boot chain:
Code:
recovery
| /------------------ laf (fastboot)
__________________|__|_________________________________________ laf (security fail screen)
/ / / /
pbl -> sbl > aboot > boot (kernel/ramdisk) > system
^ ^
| |
| \-- IS_UNLOCKED and sig_Check()
Anti-rollback
I beleive anti-rollback was updated to a new version that prevents this on Nugout. Correct me if I'm wrong. I've tried every LG hidden menu code I could find to see, but can't even seem to get the hidden menu working... And the reason they did this is because of the worldwide alerts about dirtycow, which affects not only Android, but the whole of Linux, so we need a nogout kernel. A marshmallow kernel with the new anti-rollback would theoretically end up in the red triangle of death.
We need someone that has an engineering model with a userdebug kernel. LG makes you apply. Their program is called "LG GATE", and they are very picky. I think that's what helped out the Sprint community. Someone got a hold of a developer / engineering model with a marshmallow kernel.
Well, I already heard of that strange issue with the hidden menu and I always though it was a code 18 but after googling, it looks like it some carrier potentially disabled it.. Anyway, I can confirm you that the anti-rollback version is still 0 on my h831 running the latest nougat unless the hidden menu is "lying". (And most variant I know except sprint, which did triggered the counter with the update, stayed at their rollback version). I still encounter the same issues most people on nougat experienced with the adb root (wasn't able to get past reboot and needed to flash with uppercut a fresh 7.0 ) so I guess the issue is somewhere else.
Also, if you need any information that is on the hidden menu, feel free to ask me
http://m.imgur.com/gallery/fTwgSUF
Very exciting, gl
finally they are working on it
What about this line in the aboot.bin
(from canadian aboot)
LOAD:0F960100 aBootVerificati DCB " : boot verification skip ",0xA,0
Is there any way to disable boot verification? no need to go with full bootloader unlock.. if it's possible to just disable boot verification, right?
I`m sorry if I created new thread that already exist .
I think, there is the way to unlock bootloader on this device. laginimaineb posted awesome information on blogger a couple years ago and I`m wondering ,if can somebody apply this on xt1528.
I don`t know how to insert link on that post. Try to google first sentence .It is better to read original post.
And i am sorry for my English
Unlocking the Motorola Bootloader
In this blog post, we'll explore the Motorola bootloader on recent Qualcomm Snapdragon devices. Our goal will be to unlock the bootloader of a Moto X (2nd Gen), by using the TrustZone kernel code execution vulnerability from the previous blog posts. Note that although we will show the complete unlocking process for this specific device, it should be general enough to work at-least for most modern Motorola devices.
Why Motorola?
After reporting the previous TrustZone kernel privilege escalation to Qualcomm, I was gifted a shiny new Moto X. However... There was one little snag - they accidentally sent me a locked device. This was a completely honest mistake, and they did offer many times to unlock the device - but where's the fun in that? So without further ado, let's dive into the Motorola bootloader and see what it takes to unlock it.
Setting the Stage
Before we start our research, let's begin with a short introduction to the boot process - starting right at the point at which a device is powered on.
First - the PBL (Primary Boot Loader), also known as the "BootROM" is executed. Since the PBL is stored within an internal mask ROM, it cannot be modified or provisioned, and is therefore an intrinsic part of the device. As such, it only serves the very minimal purpose of allowing the device to boot, and authenticating and loading the next part of the boot-chain.
Then, two secondary bootloaders are loaded, SBL1 (Secondary Boot Loader), followed by SBL2. Their main responsibility is to boot up the various processors on the SoC and configure them so that they're ready to operate.
Next up in the boot-chain, the third and last secondary bootloader, SBL3, is loaded. This bootloader, among other tasks, verifies and loads the Android Bootloader - "aboot".
Now this is where we get to the part relevant for our unlocking endeavours; the Android Bootloader is the piece of software whose responsibility is, as its name suggests, to load the Android operating system and trigger its execution.
This is also the piece of boot-chain that OEMs tend to customize the most, mainly because while the first part of the boot-chain is written by Qualcomm and deals with SoC specifics, the Android bootloader can be used to configure the way the Android OS is loaded.
Among the features controlled by aboot is the "bootloader lock" - in other words, aboot is the first piece of the boot-chain which can opt to break the chain of trust (in which each bootloader stage verifies the next) and load an unsigned operating system.
For devices with an unlockable bootloader, the unlocking process is usually performed by rebooting the device into a special ("bootloader") mode, and issuing the relevant fastboot command. However, as we will later see, this interface is also handled by aboot. This means that not only does aboot query the lock status during the regular boot process, but it also houses the code responsible for the actual unlocking process.
As you may know, different OEMs take different stances on this issue. In short, "Nexus" devices always ship with an "unlockable" bootloader. In contrast, Samsung doesn't allow bootloader unlocking for most of its devices. Other OEMs, Motorola included, ship their devices locked, but certain devices deemed "eligible" can be unlocked using a "magic" (signed) token supplied by the OEM (although this also voids the warranty for most devices).
So... it's all very complex, but also irrelevant. That's because we're going to do the whole process manually - if aboot can control the lock status of the device, this means we should probably be able to do so as well, given an elevated enough set of privileges.
Getting Started
Now that we have a general grasp of the components involved and of our goal, the next stage is to analyse the actual aboot code.
Since the binaries for all stages of the boot-chain are contained within the factory firmware image, that would naturally be a good place to start. There are several download links available - here are a few. In case you would like to follow along with me, I'm going to refer to the symbols in the version "ATT_XT1097_4.4.4_KXE21.187-38".
After downloading the firmware image, we are faced with our first challenge - the images are all packed using a proprietary format, in a file called "motoboot.img". However, opening the file up in a hex-editor reveals it has a pretty simple format we can deduce:
As you can see above, the sought-after aboot image is stored within this file, along with the TrustZone image, and various stages of the boot-chain. Good.
After analysing the structure above, I've written a python script which can be used to unpack all the images from a given Motorola bootloader image, you can find it here.
Much ado aboot nothing
We'll start by inspecting the aboot image. Discouragingly, it is 1MB large, so going over it all would be a waste of time. However, as we've mentioned above, when booting the device into the special "bootloader" mode, the actual interaction with the user is provided by aboot itself. This means that we can start by searching for the strings which are displayed when the unlocking process is performed - and continue from there.
A short search for the "unlock..." string which is printed after starting the unlock process brings us straight to the function (@0xFF4B874) which deals with the unlocking logic:
That was pretty fast!
As you can see, after printing the string to the console, three functions are called consecutively, and if all three of them succeed, the device is considered unlocked.
Going over the last two functions reveals their purpose is to erase the user's data partitions (which is always performed after the bootloader is unlocked, in order to protect the device owner's privacy). In any case, this means they are irrelevant to the unlocking process itself and are simply side-effects.
This leaves us with a single function which, when called, should unlock the bootloader.
So does this mean we're done already? Can we just call this function and unlock the device?
Actually, not yet. Although the TrustZone exploit allows us to achieve code-execution within the TrustZone kernel, this is only done after the operating system is loaded, at which point, executing aboot code directly could cause all sorts of side-effects (since, for example, the code might assume that there is no operating system/the MMU could be disabled, etc.). And even if it were that simple, perhaps there is something interesting to be learned by fully understanding the locking mechanism itself.
Regardless, if we can understand the logic behind the code, we can simply emulate it ourselves, and perform the meaningful parts of it from our TrustZone exploit. Analysing the unlocking function reveals a surprisingly simple high-level logic:
Unfortunately, these two functions wreak havoc within IDA (which fails to even display a meaningful call-graph for them).
Manually analysing the functions reveals that they are in fact quite similar to one another. They both don't contain much logic of their own, but instead they prepare arguments and call the following function:
This is a little surprising - instead of handling the logic itself, this function issues an an SMC (Supervisor Mode Call) in order to invoke a TrustZone system-call from aboot itself! (as we've discussed in previous blog posts). In this case, both functions issue an SMC with the request code 0x3F801. Here is the relevant pseudo-code for each of them:
At this point we've gleaned all the information we need from aboot, now lets switch over to the TrustZone kernel to find out what this SMC call does.
Enter Stage Left, TrustZone
Now that we've established that an SMC call is made with the command-code 0x3F801, we are left with the task of finding this command within the TrustZone kernel.
Going over the TrustZone kernel system calls, we arrive at the following entry:
This is a huge function which performs widely different tasks based on the first argument supplied, which we'll call the "command code" from now on.
It should be noted an additional flag is passed into this system-call indicating whether or not it was called from a "secure" context. This means that if we try invoking it from the Android OS itself, an argument will be passed marking our invocation is insecure, and will prevent us from performing these operations ourselves. Of course, we can get around this limitation using our TrustZone exploit, but we'll go into that later!
As we've seen above, this SMC call is triggered twice, using the command codes #1 and #2 (I've annotated the functions below to improve readability):
In short, we can see both commands are used to read and write (respectively) values from something called a "QFuse".
QFuses
Much like a real-life fuse, a QFuse is a hardware component which facilitates a "one-time-writeable" piece of memory. Each fuse represents a single bit; fuses which are in-tact represent the bit zero, and "blown" fuses represent the bit one. However, as the name suggests, this operation is irreversible - once a fuse is blown it cannot be "un-blown".
Each SoC has it's own arrangement of QFuses, each with it's own unique purpose. Some fuses are already blown when a device is shipped, but others can be blown depending on the user's actions in order to change the way a specific device feature operates.
Unfortunately, the information regarding the role of each fuse is not public, and we are therefore left with the single option of reversing the various software components to try and deduce their role.
In our case, we call a specific function in order to decide which fuse we are going to read and write:
Since we call this function with the second syscall argument, in our case "4", this means we will operate on the fuse at address 0xFC4B86E8.
Putting it all together
Now that we understand the aboot and the TrustZone logic, we can put them together to get the full flow:
First, aboot calls SMC 0x3F801 with command-code #1
This causes the TrustZone kernel to read and return the QFuse at address 0xFC4B86E8
Then, iff the first bit in the QFuse is disabled, aboot calls SMC 0x3F801 once more, this time with command-code #2
This causes the TrustZone kernel to write the value 1 to the LSB of the aforementioned QFuse.
Turns out to be very simple after all - we just need to set a single bit in a single QFuse, and the bootloader will be considered unlocked.
But how can QFuses be written?
DIY QFuses
Luckily the TrustZone kernel exposes a pair of system-call which allow us to read and write a restricted set of QFuses - tzbsp_qfprom_read_row and tzbsp_qfprom_write_row, respectively. If we can lift those restrictions using our TrustZone exploit, we should be able to use this API in order to blow the wanted QFuse.
Lets take a look at these restrictions within the tzbsp_qfprom_write_row system-call:
So first, there's a DWORD at 0xFE823D5C which must be set to zero in order for the function's logic to continue. Normally this flag is in fact set to one, thus preventing the usage of the QFuse calls, but we can easily enough overwrite the flag using the TrustZone exploit.
Then, there's an additional function called, which is used to make sure that the ranges of fuses being written are "allowed":
As we can see, this function goes over a static list of pairs, each denoting the start and end address of the allowed QFuses. This means that in order to pass this check, we can overwrite this static list to include all QFuses (setting the start address to zero and the end address to the maximal QFuse relative address - 0xFFFF).
Trying it out
Now that we have everything figured out, it's time to try it out ourselves! I've written some code which does the following:
Achieves code-execution within TrustZone
Disables the QFuse protections
Writes the LSB QFuse in QFuse 0xFC4B86E8
In this blog post we went over the flow controlled by a single QFuse. But, as you can probably guess, there are many different interesting QFuses out there, waiting to be discovered.
On the one hand, blowing a fuse is really "dangerous" - making one small mistake can permanently brick you device. On the other hand, some fuses might facilitate a special set of features that we would like to enable.
One such example is the "engineering" fuse; this fuse is mentioned throughout the aboot image, and can be used to enable an amazing range of capabilities such as skipping secure boot, loading unsigned peripheral images, having an unsigned GPT, and much more.
However, this fuse is blown in all consumer devices, marking the device as a "non-engineer" device, and disabling these features. But who knows, maybe there are other fuses which are just as important, which have not yet been discovered...
Spent the last day or so looking at this. The surnia bootloader attempts to call the SMC function 0x3000A0A. That's as far as I got without the symbols for the trustzone kernel.
programmargorp said:
Spent the last day or so looking at this. The surnia bootloader attempts to call the SMC function 0x3000A0A. That's as far as I got without the symbols for the trustzone kernel.
Click to expand...
Click to collapse
I don't know if it will helps, but there in comment section of "Full TrustZone exploit for MSM8974" post , one user mentioned about tz_service of msm8916
also on MSM 8916 there is a function called
tz_service <0x3F802, aTzbsp_oem_svc, 0xF, 0x86500ECB, 1> ; "tzbsp_oem_svc"
which doesnt check the ranges and write 3 dwords to an arbitrary address
int __fastcall tzbsp_oem_svc(int a1, int a2)
{
int v2; // [email protected]
v2 = a2;
*(_DWORD *)a2 = 0xC;
*(_DWORD *)(a2 + 4) = get_tzbsp_params(); =>returns 0x0F
*(_DWORD *)(v2 + 8) = sub_865164FE(); =>returns 0x0
return 0;
}
this can be used to neutralise range check.
3max3 said:
I don't know if it will help, but there in comment section of "Full TrustZone exploit for MSM8974" post , one user mentioned about tz_service of msm8916
also on MSM 8916 there is a function called
tz_service <0x3F802, aTzbsp_oem_svc, 0xF, 0x86500ECB, 1> ; "tzbsp_oem_svc"
which doesnt check the ranges and write 3 dwords to an arbitrary address
int __fastcall tzbsp_oem_svc(int a1, int a2)
{
int v2; // [email protected]
v2 = a2;
*(_DWORD *)a2 = 0xC;
*(_DWORD *)(a2 + 4) = get_tzbsp_params(); =>returns 0x0F
*(_DWORD *)(v2 + 8) = sub_865164FE(); =>returns 0x0
return 0;
}
this can be used to neutralise range check.
Click to expand...
Click to collapse
Yes, that gives you a arbritary write gadget but doesn't bring you any closer to blowing the required qfuse.
You may or may not be able to use the other trustzone exploit (CVE-2016-2431) to trigger trustzone call 0x3000A0A with the provided parameters to force an unlock.
Good exploit, but too bad it was not converted into a more general tool ...
There is low single digit number of people who can use the instructions, and unlock other models.
bibikalka said:
Good exploit, but too bad it was not converted into a more general tool ...
There is low single digit number of people who can use the instructions, and unlock other models.
Click to expand...
Click to collapse
You`re absolutely right! Unfortunately , this device is not so popular and chances to practical use any of these exploits are close to zero.
But hope dies last.
P.S. : I really appreciate your attempt to use Dirty COW exploit on this device.
3max3 said:
You`re absolutely right! Unfortunately , this device is not so popular and chances to practical use any of these exploits are close to zero.
But hope dies last.
P.S. : I really appreciate your attempt to use Dirty COW exploit on this device.
Click to expand...
Click to collapse
Actually, I never posted this, but if your phone still has an older ROM where Dirty Cow works (5.0.2 or 5.1.1?), you should try Kingoroot as per these exact extractions:
https://forum.xda-developers.com/hd8-hd10/general/tut-fire-hd-10-7th-gen-2017-root-box-t3726443
This was the most stable temporary root I've seen on XT1528, since I was using XT1528 to debug my Fire HD how-to, and I was very pleased with how well the temporary root stuck around on XT1528, and it also survived soft reboots very well. So then you can do Titanium Backup and any other software that wants root to access /data, and other parts of the system. But of course, it cannot write into /system.
Then, there is a russian dude on the Internet selling Sunshine unlock for (1300 roubles ~$20) vs the usual $25. I'd guess Sunshine does the exact the same unlock, as that blog posted copy-pasted in post #1.
Update: here is the 1300 roubles Sunshine unlock - http://1droid.ru/?page_id=51
programmargorp said:
Spent the last day or so looking at this. The surnia bootloader attempts to call the SMC function 0x3000A0A. That's as far as I got without the symbols for the trustzone kernel.
Click to expand...
Click to collapse
The blog post seems to say that the issue was reported and fixed back in 2014. It's unlikely that XT1528 has this bug since it's a later device, and the Android versions for XT1528 are some flavor of Lollipop, not KitKat.
bibikalka said:
The blog post seems to say that the issue was reported and fixed back in 2014. It's unlikely that XT1528 has this bug since it's a later device, and the Android versions for XT1528 are some flavor of Lollipop, not KitKat.
Click to expand...
Click to collapse
AFAIK Widevine exploit and the QSEE exploit both weren't patched until late 2015/early 2016. If you have a firmware version previous to the last update, then it's very likely at least the Widevine exploit still exists.
In fact, the bootloader unlock exploit did not rely on either the Widevine OR QSEE exploits, but was completely a trustzone exploit.
programmargorp said:
AFAIK Widevine exploit and the QSEE exploit both weren't patched until late 2015/early 2016. If you have a firmware version previous to the last update, then it's very likely at least the Widevine exploit still exists.
In fact, the bootloader unlock exploit did not rely on either the Widevine OR QSEE exploits, but was completely a trustzone exploit.
Click to expand...
Click to collapse
I see what you mean:
https://googleprojectzero.blogspot.com/2017/07/trust-issues-exploiting-trustzone-tees.html
If only we could get that Project Zero dude to start unlocking our phones, LOL
Btw, my ancient xt1028 with 4.4.4 is a great candidate for this original route : http://bits-please.blogspot.com/2016/02/unlocking-motorola-bootloader.html
I plan to try to replicate exactly what the guy in the post was doing (IDA Pro package, his exact phone firmware version, etc), to understand if I am even able to follow his instructions first, and see the same output that he was getting. If I won't even get that far - no reason to proceed with any any other devices
Edit: btw, any suggestions what's the best package to compile his code? https://github.com/laginimaineb/Alohamora It's all provided as source files.
bibikalka said:
I see what you mean:
https://googleprojectzero.blogspot.com/2017/07/trust-issues-exploiting-trustzone-tees.html
If only we could get that Project Zero dude to start unlocking our phones, LOL
Btw, my ancient xt1028 with 4.4.4 is a great candidate for this original route : http://bits-please.blogspot.com/2016/02/unlocking-motorola-bootloader.html
I plan to try to replicate exactly what the guy in the post was doing (IDA Pro package, his exact phone firmware version, etc), to understand if I am even able to follow his instructions first, and see the same output that he was getting. If I won't even get that far - no reason to proceed with any any other devices
Edit: btw, any suggestions what's the best package to compile his code? https://github.com/laginimaineb/Alohamora It's all provided as source files.
Click to expand...
Click to collapse
FYI the base address and probably fuse values will be completely wrong. His code runs directly in Python.
programmargorp said:
FYI the base address and probably fuse values will be completely wrong. His code runs directly in Python.
Click to expand...
Click to collapse
Would that be Python on Android, I guess ? Something like QPython3?
There is a small piece that needs to be compiled, shellcode.S. It wants arm-eabi-gcc to work. Is there an Android version, or would I need to do it in Linux or Windows?
Well, I tried to open the same TZ as in the blog (for XT1095) in IDA, and for now cannot even find the relevant addresses which should be the same as in the blog (SECURE_BOOT_FUSE = 0xfC4B86E8).
Overall, there are about 20 things that need to be filled for a different model/TZ version:
https://github.com/laginimaineb/Alohamora/blob/master/symbols.py
Anyway, it looks like people who could re-trace the unlocking steps fully don't hang out on this thread
programmargorp said:
Spent the last day or so looking at this. The surnia bootloader attempts to call the SMC function 0x3000A0A. That's as far as I got without the symbols for the trustzone kernel.
Click to expand...
Click to collapse
I've spent some time staring at IDA here - with me being a total IDA novice. I gotta say, @laginimaineb instructions on the blog are incomplete, and have quite a few gaps.
I could locate the fuse address for XT1098 (just to repeat his steps), and then for XT1028 that I have. Overall, I decompiled the whole TZ into a big *.c file, and then text searched for similar strings as in the blog. His variable/function names were cleaned up, only operands look the same.
But the challenge is that his exploit for bootloader unlock seems to rely on the kernel module being compiled and loaded (his "/data/local/tmp/fuzz_zone" utility), which seems to imply having root access at the very least. @beaups says exactly the same here : link
Also, I did not pursue searching for the memory addresses that enable fuse writing, since that code seems to sit in aboot which I was not able to decompile yet due to its non-standard format (TZ is a standard ELF). The blog never talks how he loaded aboot into IDA, and addressed this non-ELF format.
Anyway, now I don't understand how he unlocked bootloader in XT1098, he never mentions that he had root access, while his code is using the fuzz_zone program, which relies on the kernel module to talk to TZ via SCMs. It just seems a bit circular.
Then @laginimaineb posted some later exploits which could escalate from a normal user access all the way to SCMs and TZ, but the later exploits never loop back to the bootloader unlock. The code he uploaded is very research-y, inconsistent, and tough to take forward.
Technically, one could be unlocking a bunch of MOTOs and other phones just like peanuts, with a bit of skill/time. How come nobody bothered to collect misc bounties, especially back in 2016? No skill? Or the blog instructions were not so good after all?
bibikalka said:
I've spent some time staring at IDA here - with me being a total IDA novice. I gotta say, @laginimaineb instructions on the blog are incomplete, and have quite a few gaps.
I could locate the fuse address for XT1098 (just to repeat his steps), and then for XT1028 that I have. Overall, I decompiled the whole TZ into a big *.c file, and then text searched for similar strings as in the blog. His variable/function names were cleaned up, only operands look the same.
But the challenge is that his exploit for bootloader unlock seems to rely on the kernel module being compiled and loaded (his "/data/local/tmp/fuzz_zone" utility), which seems to imply having root access at the very least. @beaups says exactly the same here : link
Also, I did not pursue searching for the memory addresses that enable fuse writing, since that code seems to sit in aboot which I was not able to decompile yet due to its non-standard format (TZ is a standard ELF). The blog never talks how he loaded aboot into IDA, and addressed this non-ELF format.
Anyway, now I don't understand how he unlocked bootloader in XT1098, he never mentions that he had root access, while his code is using the fuzz_zone program, which relies on the kernel module to talk to TZ via SCMs. It just seems a bit circular.
Then @laginimaineb posted some later exploits which could escalate from a normal user access all the way to SCMs and TZ, but the later exploits never loop back to the bootloader unlock. The code he uploaded is very research-y, inconsistent, and tough to take forward.
Technically, one could be unlocking a bunch of MOTOs and other phones just like peanuts, with a bit of skill/time. How come nobody bothered to collect misc bounties, especially back in 2016? No skill? Or the blog instructions were not so good after all?
Click to expand...
Click to collapse
That answer is easy. The bugs you've seen publicly disclosed are long patched. In fact the E2 (and E for that matter) were never vulnerable.
beaups said:
That answer is easy. The bugs you've seen publicly disclosed are long patched. In fact the E2 (and E for that matter) were never vulnerable.
Click to expand...
Click to collapse
How about an ancient XT1028 with 4.4.4 ? It's ROM dates to about mid-2014.
bibikalka said:
How about an ancient XT1028 with 4.4.4 ? It's ROM dates to about mid-2014.
Click to expand...
Click to collapse
No publicly disclosed tz bugs impacted that device past 4.4.3
bibikalka said:
I've spent some time staring at IDA here - with me being a total IDA novice. I gotta say, @laginimaineb instructions on the blog are incomplete, and have quite a few gaps.
I could locate the fuse address for XT1098 (just to repeat his steps), and then for XT1028 that I have. Overall, I decompiled the whole TZ into a big *.c file, and then text searched for similar strings as in the blog. His variable/function names were cleaned up, only operands look the same.
But the challenge is that his exploit for bootloader unlock seems to rely on the kernel module being compiled and loaded (his "/data/local/tmp/fuzz_zone" utility), which seems to imply having root access at the very least. @beaups says exactly the same here : link
Also, I did not pursue searching for the memory addresses that enable fuse writing, since that code seems to sit in aboot which I was not able to decompile yet due to its non-standard format (TZ is a standard ELF). The blog never talks how he loaded aboot into IDA, and addressed this non-ELF format.
Anyway, now I don't understand how he unlocked bootloader in XT1098, he never mentions that he had root access, while his code is using the fuzz_zone program, which relies on the kernel module to talk to TZ via SCMs. It just seems a bit circular.
Then @laginimaineb posted some later exploits which could escalate from a normal user access all the way to SCMs and TZ, but the later exploits never loop back to the bootloader unlock. The code he uploaded is very research-y, inconsistent, and tough to take forward.
Technically, one could be unlocking a bunch of MOTOs and other phones just like peanuts, with a bit of skill/time. How come nobody bothered to collect misc bounties, especially back in 2016? No skill? Or the blog instructions were not so good after all?
Click to expand...
Click to collapse
The aboot image should be a standard ELF image.
Getting root on the xt1528 is possible (without system rw) with initrd root.
beaups said:
No publicly disclosed tz bugs impacted that device past 4.4.3
Click to expand...
Click to collapse
Interesting. Not even the Widevine QSEE issue? How can it be? Or is Widevine too difficult to exploit?
programmargorp said:
The aboot image should be a standard ELF image.
Getting root on the xt1528 is possible (without system rw) with initrd root.
Click to expand...
Click to collapse
Yes, root is very easy. But it's getting further, to bootloader unlock that is the big challenge. Can you double check your "aboot", it does not look like ELF ...
bibikalka said:
Interesting. Not even the Widevine QSEE issue? How can it be? Or is Widevine too difficult to exploit?
Yes, root is very easy. But it's getting further, to bootloader unlock that is the big challenge. Can you double check your "aboot", it does not look like ELF ...
Click to expand...
Click to collapse
Correct, the device does not have that widevine vulnerability.