Or Continue to Thread: [05.09.2012] Got Brickbug ? v1…
Find Your Device:
6th June 2012, 11:24 PM   |  #69  
Senior Recognized Developer
Flag Owego, NY
Thanks Meter: 24,638
 
13,460 posts
Join Date:Joined: Aug 2007
Donate to Me
More
It would be beneficial to provide more information on the brick bug to avoid some people getting unnecessarily scared (such as most I9100 users).

This bug requires three things for you to be in danger, and ALL of these conditions must be met for danger:
1) A defective eMMC chip/fwrev that is unable to handle eMMC ERASE commands (command 38) properly. (I'll provide a link with more detail on the nature of the bug later) - This condition is the one Chainfire's new app checks for. By the way, M8G2FA fwrev 0x11 (seen on some Kindle Fires) is also suspected of being defective.
2) A recovery binary that attempts to erase partitions when formatting them. Most ICS recovery binaries fit in this category, most Gingerbread recoveries do not attempt to perform an erase operation so are safe. Note that also, an affected update-binary in a ZIP could be a cause of problems too. (e.g. flashing a firmware that has an ICS update-binary and formats the partition could cause a problem even with a "safe" recovery.) So a kernel can be repacked with a "safe" CWM (such as the most recent CF-Root releases) but it will still only be partially safe.
3) A kernel that allows attempts to erase a partition to actually happen. (as opposed to reporting "not supported" and doing nothing.) - A common way of rendering a kernel safe is to remove MMC_CAP_ERASE from the capability flags in drivers/mmc/mshci.c

As of June 6, 2012, this is what I know as far as kernels that meet condition 3:
  • All GT-I9100 ICS leaks and official releases prior to July 2012 are SAFE (MMC_CAP_ERASE not present)
  • New GT-I9100 ICS leaks and official releases (starting in July 2012) are UNSAFE - That's right, Samsung ADDED the trigger conditions for this bug to newer releases such as XXLQ5. So much for "we're working on a fix"...
  • All kernels based on GT-I9100 ICS Update4 sources are SAFE (MMC_CAP_ERASE not present) - This includes all CM9 nightlies for SGH-I777, GT-I9100, and GT-N7000, all GT-I9100 custom kernels I am aware of, and all SGH-I777 custom kernels I am aware of
  • All GT-N7000 ICS leaks are UNSAFE
  • All GT-N7000 ICS official kernels are UNSAFE
  • All kernels built from the GT-N7000 sources are UNSAFE unless the following condition is met:
  • MMC_CAP_ERASE is removed from the capability flags in drivers/mmc/host/mshci.c - check the kernel features for this. Franco.kernel R3 and later and all Speedmod ICS releases are SAFE due to this.
  • All SHW-M250S/K/L ICS kernels are suspected to be UNSAFE
  • All SHW-M250S/K/L ICS source releases as of this date are UNSAFE (SHW-M250L Update4 was the cause of the SiyahKernel 3.1rc6 incident. Other Siyah releases are SAFE)
  • All SPH-D710 ICS releases as of this date are UNSAFE - Rumor has it that the official OTA may have a fixed kernel, but it is recommended to consider this kernel UNSAFE until source code is released and can be reviewed.
  • The SGH-I777 UCLD3 leak is UNSAFE, as is most likely every other leak for that device. Fortunately nearly everyone is using I9100 Update4-based custom kernels.
  • SGH-I727 and SGH-T989 ICS leaks are UNSAFE - However as these two devices use separate recovery and operational kernels, if you have a Gingerbread recovery/kernel, you should be safe regardless of what you are booting for normal operation.

It's hard to get ALL of the cases and evaluate them, but in general in terms of levels of danger (As of June 6, 2012 - this could change with time):
SPH-D710 users are in the most danger - They have no official ICS releases AND the I9100 Update4 source base can't be used to build a usable kernel for their device without major developer work
GT-N7000 users are second on the list - They are the only ones outside of Korea to receive official ICS updates that trigger the eMMC firmware defect. However, I9100 Update4 sources required only minor work to create "safe" kernels, and developers know the proper procedure for rendering the official N7000 Update3 source drop "safe"
SGH-I777 users are next - I777 leaks proved to be dangerous a month or so ago. However, the SGH-I777 required the least amount of work to be able to use the GT-I9100 Update4 source base, and as a result, with the exception of the leaks themselves, nearly all I777 ICS kernels are based off of safe source code bases.
GT-I9100 users are in the least danger - No leak, official binary release, or source code release for this device has been dangerous. Only one I9100 kernel has ever proven dangerous and that was quickly pulled by its developer.

I am not evaluating the SHW-M250S/K/L in the above list, as while I know their source and binaries are dangerous, the language/culture barrier means we have very little information on how this fiasco is panning out for those users.

UPDATE:
We have at least one confirmed report of this bug occurring with KYL00M fwrev 0x12 on a Samsung Skyrocket (SGH-I727) with their ICS leak kernels
In addition, Samsung Hercules (SGH-T989) has the same fwrev and I've been told that they have observed bricks of this type with their ICS leaks

UPDATE 2:
I've received an email from a contact at Samsung who has indicated they are working on some sort of fix to be deployed to devices with an "UNSAFE" configuration listed above. I have requested that I receive an explicit list of which binary builds contain this fix, as without that I cannot know for sure which builds are fixed and which are not. Fixes are not yet deployed to affected devices.

UPDATE 3:
So much for the claims of working on a fix above... Not only have fixes not been deployed to any kernel for any device I am aware of, but Samsung added the trigger conditions to the XXLQ5 build for GT-I9100. Yes, that is correct - a device previously unaffected by this bug is now UNSAFE.
Last edited by Entropy512; 9th July 2012 at 10:28 PM.
The Following 133 Users Say Thank You to Entropy512 For This Useful Post: [ View ]