!UUYziobKGGxpovWyAN:nixos.org

Robotnix

243 Members
Build Android (AOSP) using Nix | https://github.com/danielfullmer/robotnix75 Servers

Load older messages


SenderMessageTime
25 Nov 2024
@oak:universumi.fioak 🏳️‍🌈♥️What is cool about Pixel 8 and Pixel 9 is that they support the new ARM Memory Tagging Extensions21:10:25
@oak:universumi.fioak 🏳️‍🌈♥️https://googleprojectzero.blogspot.com/2023/11/first-handset-with-mte-on-market.html21:13:35
26 Nov 2024
@ryantrinkle:matrix.orgryantrinkle samueldr: oak Thanks for the info! 14:45:40
@jack:unredacted.org@jack:unredacted.org joined the room.20:40:37
@jack:unredacted.org@jack:unredacted.orgHi. Anyone here know how to compare pvmfw.img between two AOSP (Android 15) builds? It is an unusual file structure and I don't know how to unpack this before comparing with diffoscope.20:41:42
@atemu12:matrix.orgatemu12 What is the file type? 20:42:25
@jack:unredacted.org@jack:unredacted.org

pvmfw.img is an Android boot image (according to the file command). The unpack_bootimg.py from AOSP can unpack it, but the "kernel" file inside of it has just "data" as its file type.

20:44:55
@jack:unredacted.org@jack:unredacted.org
$ file pvmfw.img.unpacked/* 
kernel:  data
ramdisk: empty
20:49:37
@jack:unredacted.org@jack:unredacted.orgIt's a really weird couple of files and I really don't know how to make diffoscope compare them properly.20:51:38
@jack:unredacted.org@jack:unredacted.org

The pvmfw.img file itself has this type:

$ file pvmfw.img
pvmfw.img: Android bootimg, kernel
20:53:22
@jack:unredacted.org@jack:unredacted.orgMaybe I should try checking if platform/build/+/master/tools/extract_kernel.py can make some sense out of it.20:54:16
@jack:unredacted.org@jack:unredacted.orgYeah, not even AOSP's extract_kernel.py can make some sense out of pvmfw.img.unpacked/kernel.21:06:33
@jack:unredacted.org@jack:unredacted.org *

Yeah, not even AOSP's extract_kernel.py can make some sense out of pvmfw.img.unpacked/kernel.

21:06:41
@jack:unredacted.org@jack:unredacted.org *

Maybe I should try checking if platform/build/+/master/tools/extract_kernel.py can make some sense out of it.

21:06:48
@jack:unredacted.org@jack:unredacted.org

Aside from this, I've also used unpack_bootimg.pyto unpack the boot.img. That one has a perfectly normal ARM64 kernel image, but diffoscope still shows a large diff when comparing two builds of the same boot.img.unpacked/kernel. Do I need to preprocess this kernel image somehow before comparing? Not sure if I need to strip signatures from the image or how to do this, by the way.

21:14:19
@jack:unredacted.org@jack:unredacted.org *

Aside from this, I've also used unpack_bootimg.py to unpack the boot.img. That one has a perfectly normal ARM64 kernel image, but diffoscope still shows a large diff when comparing two builds of the same boot.img.unpacked/kernel. Do I need to preprocess this kernel image somehow before comparing? Not sure if I need to strip signatures from the image or how to do this, by the way.

21:14:29
@samueldr:matrix.orgsamueldr

jack: when I hit a weird filetype, especially when it's likely others have it it, searching its filename in github code search helps... when you know how to use it, which is non-obvious: use a regex to search

21:14:54
@jack:unredacted.org@jack:unredacted.org
In reply to @samueldr:matrix.org

jack: when I hit a weird filetype, especially when it's likely others have it it, searching its filename in github code search helps... when you know how to use it, which is non-obvious: use a regex to search

That's a good idea. Thanks.
21:15:32
@samueldr:matrix.orgsamueldr given I see Protected VM firmware image along it, maybe that's not "just" a linux kernel 21:16:05
@samueldr:matrix.orgsamueldreven though it's packaged within the "container" of an android boot image21:16:24
@jack:unredacted.org@jack:unredacted.org
In reply to @jack:unredacted.org

Aside from this, I've also used unpack_bootimg.py to unpack the boot.img. That one has a perfectly normal ARM64 kernel image, but diffoscope still shows a large diff when comparing two builds of the same boot.img.unpacked/kernel. Do I need to preprocess this kernel image somehow before comparing? Not sure if I need to strip signatures from the image or how to do this, by the way.

Any idea on this one?
21:16:48
@samueldr:matrix.orgsamueldr https://github.com/GrapheneOS/platform_packages_modules_Virtualization/tree/9ec1c44084125875aa9619635e4e6fdf3172b57c/pvmfw#integration 21:17:07
@samueldr:matrix.orgsamueldr
In reply to @jack:unredacted.org
Any idea on this one?
nope
21:17:27
@jack:unredacted.org@jack:unredacted.org
In reply to @jack:unredacted.org

Aside from this, I've also used unpack_bootimg.py to unpack the boot.img. That one has a perfectly normal ARM64 kernel image, but diffoscope still shows a large diff when comparing two builds of the same boot.img.unpacked/kernel. Do I need to preprocess this kernel image somehow before comparing? Not sure if I need to strip signatures from the image or how to do this, by the way.

boot.img unpacked perfectly and has a perfect ARM64 kernel image, but I personally don't know how to compare different kernel images, maybe I have to strip signatures like I successfully did with the kernel modules before.
21:18:06
@jack:unredacted.org@jack:unredacted.org
In reply to @jack:unredacted.org

Aside from this, I've also used unpack_bootimg.pyto unpack the boot.img. That one has a perfectly normal ARM64 kernel image, but diffoscope still shows a large diff when comparing two builds of the same boot.img.unpacked/kernel. Do I need to preprocess this kernel image somehow before comparing? Not sure if I need to strip signatures from the image or how to do this, by the way.

*
21:18:20
@samueldr:matrix.orgsamueldr 21:18:41
@samueldr:matrix.orgsamueldr

The pVM firmware (pvmfw) is the first code executed by a pVM, similar to the boot ROM of a physical device.

21:18:52
@samueldr:matrix.orgsamueldrso it's definitely not a linux kernel21:19:09
@jack:unredacted.org@jack:unredacted.org

Yes, maybe unpack_bootimg.py isn't prepared to handle it, so it just considers whatever is in the supposed kernel position on the file.

21:20:19
27 Nov 2024
@oak:universumi.fioak 🏳️‍🌈♥️I got a little forward by creating the FHS env, but I still can't completely run the mk-vendor-file.py for new Chromium version, now I get same errors I think it tries to pull some Google internal stuff they use for building ChromeOS and Chrome20:40:12

Show newer messages


Back to Room ListRoom Version: 6