**Ultimate GS3 sudden death thread**

Search This thread

drakester09

Senior Member
Jan 10, 2012
1,292
876
There is no definite reason of what causes SDS. Some were on charger overnight, some were not on charger overnight, some were on table like mine in the afternoon, some with alarm, some without alarm, some in pocket, some in bed, some in toilet, some on stock, some on custom, some stock with custom kernel, some with custom and stock kernel. Only GOD knows what's causing SDS.

My replacement is 1.5 and still last 2 days it's conked off and i had to press power for 10 secs to make it boot up. I'm on XELLA but old bootloader. I dont think the s**tty bootloader is any cure for samsung's crap quality control on 16gb models. As you can see from the poll i started in the link in my signature about 49% of 132 users who polled have had their 16gb get SDS.

As bad as RROD on xbox360.
The poll is flawed, most of the users that haven't experienced SDS will not vote. ;)
Meanwhile the users that did have the flaw are more likely to vote.

Even then, the number is quite high... it's concerning but I feel this thread is a much more contundent message that something is wrong.
 

RSRivers

New member
Jan 10, 2011
1
0
São Paulo
I had my phone charging overnight, i woke up got to my car and made a call, then I locked the screen and went to work. After 5 minutes I arrived and was going to check the time. The phone didn't respond to home screen or power button, I tried to remove the battery, charge the phone tried to go to download and recovery modes but the phone is just a brick.

The phone was 16GB model manufactured in Brazil and was running stock JB from samsung.
I dont know the HW info

The phone was bought and activated in the 9th of June 2012 and died in January 2nd 2013.
 
Last edited:

iridaki

Retired Forum Moderator
Feb 21, 2007
4,532
5,190
37
Edinburgh, Scotland
The poll is flawed, most of the users that haven't experienced SDS will not vote. ;)
Meanwhile the users that did have the flaw are more likely to vote.

Exactly!

An interesting question would be "Does anyone have a i9300 16GB model that is alive for more than 8 months?" because I'm starting to believe that all phones are affected, it's just that the release date is relatively close.

iR¡D@k!* ? from i9300 via Tapatalk 2
 

Egan

Senior Member
May 29, 2010
596
107
Exactly!

An interesting question would be "Does anyone have a i9300 16GB model that is alive for more than 8 months?" because I'm starting to believe that all phones are affected, it's just that the release date is relatively close.

iR¡D@k!* ? from i9300 via Tapatalk 2
Good question! I believe all 16GB models are flawed and eventually die and Samsung keeps it quiet to make sure customers will continue to buy their flagship. I understand it though, it would be a PR disaster.
 

cyberdroxxx

Senior Member
Apr 24, 2009
181
30
Kuala Lumpur
Good question! I believe all 16GB models are flawed and eventually die and Samsung keeps it quiet to make sure customers will continue to buy their flagship. I understand it though, it would be a PR disaster.

And the Galaxy Note 2 will follow eventually. My SGN2 has the same vulnerable eMMC chip VTU00M.
So yes the clock is ticking until samsung
release their new flagship.

I guess they manufactured their components to last about six months. They honour their warranty by repairing those dead phone to show the world that they have good after sales support. But their repairs will only last for another six months when warranty they have finally expire. They are sure that die hard galaxy fanboys will buy the new flagship. They are smart to announce new flagship models every six month.

Sent from my GT-N7100 using xda premium
 

nckeong88

Senior Member
Oct 15, 2012
98
35
Just joined the sudden death club. Phone was fully charged but dead overnight. :crying:
 
Last edited:

Sad56

Senior Member
Mar 2, 2012
570
191
OnePlus 8 Pro
Google Pixel 6
I've mine 16gb since 29 mai 2012. (pre order)

Always charge over the night.
No symptoms for the moment.

Fingers crossed (I don't have flash the new bootloader, but I use LLA firmware +kernel)



Envoyé depuis mon GT-I9300 avec Tapatalk
 

KeesStolk

Senior Member
Aug 30, 2011
3,695
4,380
58
www.keesstolk.nl
I've mine 16gb since 29 mai 2012. (pre order)

Always charge over the night.
No symptoms for the moment.

Fingers crossed (I don't have flash the new bootloader, but I use LLA firmware +kernel)



Envoyé depuis mon GT-I9300 avec Tapatalk

My 16gb white S3 is from 27 may 2012, 9 of 10 charging over night with the standard wall charger.
I did flashed the new bootloader and running Omega roms since v4 i think :)
Still running smooth and fast.
 
  • Like
Reactions: Chris_84

kintol

Member
Jun 9, 2011
26
0
JB
does anyone here monitoring their cpu temp? mine having the so call "early symtoms" of SDS- restarting once hitting 50deg c.
 

Top Liked Posts

  • There are no posts matching your filters.
  • 85
    **** SAMSUNG HAVE APPLIED A SUDDEN DEATH FIX VIA A SOFTWARE UPDATE****

    STOCK KERNEL: XXELKC AND NEWER

    STOCK RECOVERY: XXELKC AND NEWER

    ALSO ANY CUSTOM KERNEL AND RECOVERY THAT ARE BUILT FROM THE UPDATED SOURCES. THERE HAVE BEEN CASES OF SDS ON DEVICES THAT HAVE UPDATED BUT TO WHAT EXTENT THE DAMAGE WAS AT ALREADY BEFORE FLASHING WE DON'T KNOW.



    WHO'S AFFECTED?
    It seems that at the moment the only devices that have suffered from SDS are the 16GB version with:
    eMMC: VTU00M
    FW Rev: 0xF1


    To check your eMMC version download eMMC Brickbug Check from Google Play.
    What you want is the eMMC version and the FW Rev.



    THE FIX
    Samsung have now released a Sudden Death Fix via a patched Kernel and Recovery in latest firmware. Firmware Version: XXELKC

    Simply update OTA or download XXELKC firmware (or newer) from sammobile.com and flash using Odin.

    You need both the stock recovery and kernel installed for it to work properly, or a custom kernel and recovery built from the updated sources
    ie Latest Perseus Kernel and PhilZ Recovery are examples



    LINKS
    For more info about SDS, who could be effected and how to fix check out this awesome thread by rootSU

    For a detailed look at how the fix works check out this thread here.


    WARRANTY
    For those of you in Europe who have rooted your phone it appears that this doesn't void warranty. Check this thread for more info.
    Also this website could prove very handy for anyone with a European or UK handset that has died.
    40
    Just took a look at the diffs and i have to admit, i don't nearly get what this does. What are those "movi commands"? Where can one find a data-sheet to decode the magics? :(

    BTW, just took the kernel image from the WanamLite v5.3 CWM zip (that's what i am currently running), un-gz-ed it, and actually found the "movi operation is failed" error string in there. Good for me, i guess ;)

    AndreiLux, thanks A LOT for your research.
    You won't. These are HIGHLY proprietary to Samsung's storage people.

    I'd hazard a guess that it does - but I'd certainly like someone like Entropy to weigh in.
    Bah, I wish I could see what you quoted. As far as safety goes:
    90%+ chance that the change in Update7 is the fix.
    75% chance that XXELLA/4/etc have the fix (It's possible, but highly unlikely, that the string VTU00M would appear in the kernel without the fix.

    Is there a way to check if I already have any bad blocks on my eMMC?
    This isn't about bad blocks - this is about a firmware bug where a data structure gets suddenly corrupted. You can really only know "is it working" or "is it dead". The one exception seems to be that some people see odd performance issues just before death, similar to the issues people see when using PIT workarounds for Superbrick.

    Just as I said above, the low-level details of what's going on are HIGHLY proprietary to Samsung.
    The patch additionally checks that the firmware date is 2012/04/13 and only applies the commands then.

    So you need type: VTU00M revision: 0xf1 and internal firmware date of 2012/04/13 for the bug to have an effect. The date which eMMC brickbug checker reads is the production date as it seems.

    So there might be phones with VTU00M/0xf1 out there which are not affected, I don't know if that makes sense in regard that if the revision would even be the same then.
    Yeah. I'm wondering if we should add some printk()s to check what the date is. I'm curious if there are other dates floating around.

    No, the date shown in the eMMC app is the production date, the internal firmware date is something else and not possible to read out through normal methods.
    Correct, although we could add a printk to kernels to print out the info.

    eMMC app gives me: 05/2012
    but checking via the SSID gives me: 2012/06/09.
    So two different dates, but none of them is the internal firmware date, correct?
    Correct.

    Most phones died over night after charging. Since there are many defective chargers, can this be related to a faulty charger? For example, I have my sgs3 for about 6 months and a few days ago charging became very slow (didnt charge fully after whole night). I used HTC's charger and charging is nornal again. Ive seen that many sgs3 owners got problems with charger. Can some faulty chargers start charging very skow and others give too much electricty which burns internals?

    Sent from my GT-I9300 using xda app-developers app
    No. Seriously - READ. It is at this point unambiguously an eMMC firmware failure that has NOTHING to do with the charger.

    The ONLY connection with charging is this: CHARGING HOLDS A WAKELOCK. This means the device will do various tasks in the background that it wouldn't do in deep sleep, some of which perform I/O cycles on the eMMC.

    The patch to the MMC driver discovered in the Update 7 sources released by Samsung performs a procedure that is nearly identical to the fix for another mmc firmware bug in a different samsung device.

    The patch also includes some character strings which can be searched for in the binary kernel of XXELLA, as when code is compiled, strings are left as they are.

    The kernel from the XXELLA firmware DOES include these strings, so it's probably safe to assume that the kernel includes the code that performs the in-RAM fix to the mmc firmware.

    The fact that some people have reported that they've experienced SDS on the XXELLA ROM is interesting - none have confirmed 100% that they had the XXELLA kernel running (to the best of my knowledge). This means that for some reason they may have been running another kernel that doesn't have the patch.
    So far all of them were running other kernels.

    It's just like the people who claimed they Superbricked on stock recovery. Turns out that in their eyes, fakeflashing CWM from stock recovery was still in some twisted way stock recovery... It wasn't.

    I'm still confused. Some posts say lla kernel is safe others say you need Perseus. So which one is it

    Sent from my GT-I9300 using xda premium
    Perseus is 90% guaranteed to be safe (I'm not claiming 100% without a detailed technical explanation from Samsung. Even then I'm not claiming 100%, just like I refuse to guarantee that nonsecure erase is safe on Superbrick-vulnerable devices even though Samsung claims it is... As a result anything I release has eMMC TRIM/ERASE completely disabled for those devices.
    ELLA/4/etc are 75%+ guaranteed to be safe - since we THINK they have the same patch

    If i understood it correctly, we have an assumption? that because similar code is implemented in kernels for similar problems in other samsung phones, so that means we have the same problem in S3.
    If this is true then all or almost all 16GB phones are affected, as i didnt saw a phone with different emmc.(maybe some new phones have newer revision?)
    We are talking then for millions S3's that are going to die?
    Maybe this code then doesnt have to do anything with the "SDS issue" and is more of a precaution or even testing trying to figure out the problem from Samsung?
    Samsung's storage guys have a wide variety of chips/models. VTU00M 0xf1 is primarily seen in I9300 units, and almost all 16GB I9300s except very recent ones have it. Some other devices have it, but it isn't nearly as prevalent in other devices. My Note 10.1 has MAG4FB I think (need to check again...) In addition, there appears to be some additional identifying information beyond VTU00M 0xf1 that we haven't had time to collect data on yet (and developers need to make kernel patches to even allow this data to be collected...)

    I think that it's combination is the solution.

    according to this from 1st post:

    ...Kernels >v31 and beyond stock LLA are now the only truly protected ones.

    Can someone confirm this?
    the key in that post is the word "now". That post was made yesterday - the patch has been making the rounds and is getting integrated

    Have you searched for it in older kernels? Why wouldn't that string appear also in those? If it does, then this means nothing.
    That's something that needs to be checked... However if it appears in older kernels Samsung was violating the GPL with them as I'm fairly certain it is nowhere within the source.

    Nothing can fix an SDS because the phone is already dead. :D
    But to prevent it, yes it seems. One of those at this moment.
    Just like Superbrick.

    Samsung haven't fixed the super brick bug yet :p

    Sent from my GT-I9100 using Tapatalk 2
    On a small subset of affected devices they have - I9100s in HK apparently have Jellybean and that has their official fix. But so far, nearly all affected devices are still on ICS and they only put the fix in JB kernels.

    Just did a emmc check and I found out that my fwrev is oxf7 and the date is 11/2012... But I got the same chip like otherss... :(

    So am I on a safer side?

    Sent from my GT-I9300 using xda app-developers app
    Unknown. Often flaws like this are firmware-dependent.

    For example:
    VYL00M/KYL00M/MAG4FA fwrev 0x19 = Superbrick + 32kb-of-zeroes bug
    VYL00M/KYL00M/MAG4FA fwrev 0x25 = 32kb-of-zeroes bug only (immune to Superbrick)

    However for the above, we had confirmation from a trusted source (A Google engineer) that 0x19 had a bug with the symptoms we were seeing and that HE had seen it in GNex prototypes, and that 0x25 was "fixed" in regards to that bug (Superbrick). A fix for the bug in 0x25 is what led us to him.

    Theoretically if you have the same chip, you are candidate for sds sometime.... :(
    Not necessarily. I would put the status as "unknown". If you have VTU00M 0xf7 you're much less likely to have problems than 0xf1 - but with something like this guarantees cannot be made.

    Also: The fix patch was merged to CM10.1 source last night. So today's nightly should be safe. If it's not - no one is safe.
    30
    Will my i9305 die too, or only the i9300 is afected?
    If you have VTU00M fwrev 0xf1 flash, you are probably at risk. I9305 is too new to tell

    In today's SamMobile article, they say is a firm bug!!?????? In other words, sammy firm only have the bug?? If I use (no because my phone deaths on 12/27) aosp rom and custom kernel, bug not affect my device???
    No. In fact, there's nothing that indicates there is a bug in the bootloader/kernel/system firmware yet. Given the behavior of the problem and Samsung's past history, it's likely a bug in internal eMMC firmware (which can, at best be field-patched if it only involves a few bytes of microcode - major changes are not possible in the field.) This upcoming update likely contains a workaround for that eMMC bug.

    Look at the Superbrick bug - There was no underlying "bug" in any of Samsung's firmwares, except that they didn't block commands that would trigger a known bug in the underlying flash memory. Now, in any hardware without that bug, issuing secure erase commands is fine. The workaround for the bug is simple: Don't send secure erase commands to the damn chip.

    Is it true that all 16gigs phone will die soon one day?

    Sent from a better Galaxy designed for humans!
    Unknown. Right now any device with VTU00M flash is at risk - but how high the risk is we don't know.

    but Samsung says that will fix the issue with fw update.....or not?

    there is no fw going to write to the NAND?
    No one knows yet. If it's done in the kernel, we'll know EXACTLY what/how they fixed it and how to apply the fix to custom firmwares. If it's the bootloader, we won't know unless they explicitly states that they changed the bootloader to fix it. If it's in /system (HIGHLY unlikely) we might see something.

    Most likely place they'll fix this is the kernel with a variant of the Sumrall patch from last spring, OR an alteration to the MMC code in order to avoid doing something (we don't know what) that the chip doesn't like (this would be similar to how Superbrick is worked around). So far, every time Samsung has ever fixed or worked around an eMMC bug/defect, it's been in the kernel and not the bootloader. So everyone flashing this new bootloader is just making it more likely they'll be denied warranty support if their device dies.

    Yes the current btu release (today)apparently has sudden death fix via the bootloader.

    I recommend updating via pc odin as mobile odin won't fix the bootloader, Im already on samsung 4.1.2 release at christmas just downlaoding todays release.
    Bull****. You have ZERO evidence to substantiate this claim.

    so we can only wait for the new bootloader from Samsung .... :crying:
    And why do you think it's the bootloader? There's no evidence to say WHERE the fix will be applied because there isn't even any information about HOW the failures are occurring. Right now, I'd say it's most likely going to be a kernel fix.

    Yes but its like a chain reaction: if one component or sector dies mostly the other ones will follow. Freezes and hookups are those signs of hardware failure.

    As long if those symptoms doesnt apair you dont need to worry too much

    Sent from my GT-I9300 using xda app-developers app
    If it's a wear leveller bug, there's a possibility fixed wear leveller firmware might "repair" damage to the internal data structures.

    Can someone please explain... if it's a wear leveler, and thats a part of eMMC (as opposed to software-only wear leveling), how is it even possible to update it? Can one possibly update the eMMC microcode ?!
    Search before posting. I posted an example of how this has been done to Samsung eMMC chips in the past only 1-2 days ago. (Search this thread for Sumrall...)

    Minor eMMC microcode updates can be done at runtime. It's fairly safe since it apparently patches the firmware after it has loaded into volatile memory (and hence a power cycle removes the patch if it's misapplied). This is what the Galaxy Nexus patch for VYL00M/KYL00M/MAG4FA fwrev 0x25 did.

    Major eMMC microcode updates can't be done so easily, which is why the underlying Superbrick flaw was never fixed.


    [KIES]I9300XXELLA 4.1.2->EXYNOS BUG FIXED!!S3 SUDDEN DEATH FIXED!!Jan.02,2013

    http://xdaforums.com/showthread.php?t=2077844
    Don't make definitive claims you have no evidence to support.
    27
    Hi all,

    Assuming that there is no NAND degradation or similar and that SDS come for something specific, why over 90% of the deaths have come from the fourth month onwards? Why have not failed at 4 days or two weeks, for example? What is the secret component involved over time to fail?

    We will likely never know the specific secret component, but with knowledge of the behavior of eMMC, how it behaves, and how Samsung eMMCs have failed in the past, we can guess.

    The wear leveller keeps track of what memory blocks have been used and what haven't, and relocates blocks periodically to spread wear across the device. For example, if you write to the fourth block of the eMMC repeatedly, internally it'll map the fourth block to the 100th, then maybe the 150th, then 200th, etc...

    At some point, after a long time of operation, the wear leveller might reach a corner case where a bug is triggered - my guess would be an integer overflow or a signed vs. unsigned issue. For example, it's working with the 32767th instance of block 5, and tries to increment a counter to 32768 - but instead, gets -32768 instead because something is treating an unsigned int as signed. The next time it tries to work with that counter, BOOM - it crashes. Again, we don't know the exact nature of what's happening, but it's likely something along these lines.

    It's very similar to what happened with Superbrick on the GS2 - if you issued a secure erase command to erase memory that was in a certain specific state (I can't talk about what the exact state is, sorry... And no, there's no way to tell if the memory is in that state unless you're Samsung or you've Superbricked it.), the wear leveller would crash and leave behind corrupted data structures - any attempt to access these structures again would crash the wear leveller again. The symptom to the user was any attempt to access affected regions of the eMMC would cause the chip to hang.
    22
    Entropy512, thank you.
    How can you explain users that had XXELLA stock rom, and still suffered from SDS? There are more than one report of it.
    I have yet to see any such reports... The one report I've seen of an XXELLA failure was XXELLA system + Siyah.

    You are being unecessarilhy harsh here, especially considering that you are addressing people who are under the fear that their expensive phones will die on them suddenly. AdreiLux seems to be more skeptical regarding the possibility of the fix depending on the new bootloader as well. Calling names surely doesn't make you look smarter than the "idiots" who took a step -granted maybe rushed- towards a probable fix of a dreadful issue. And you may know much more than the average joe here, but you still have ZERO evidence that the new bootloader doesn't do anything at all that contributes to fixing the SDS, so you may have as well been nicer. Just my 2 cents.
    I have all of the evidence I need - I now have kernel source for a complete eMMC firmware patch. The fix is in the kernel, not in the bootloader. It's being patched in the EXACT same way as the GNex 32kb-of-zeros fix patch, which had zero bootloader involvement.

    The fact is that flashing a bootloader is a fundamentally dangerous operation, and flashing a bootloader with known regressions in functionality is 100% reckless and stupid.

    The fact is that Samsung has NEVER fixed a problem like this in the bootloader before. There was ZERO evidence pointing there. There was plenty of evidence (the GNex VYL00M/KYL00M/MAG4FA 0x25 patch) pointing to the fix being in the kernel when it came out.

    How come you knew that SDS is related to eMMC (and specifically version VTU00M) before samsung released their code? What led to this assumption?
    All of the symptoms and behavior pointed this way.
    1) Some devices were exhibiting "Superbrick-ish" behavior where certain eMMC regions were working and others were inaccessible
    2) It was ONLY happening on 16GB devices - this is the most obvious piece of evidence. If it weren't the eMMC, it would have been seen on 32/64GB devices
    3) It would be the third time in one year Samsung has ****ed up their wear leveller, their quality control is clearly crap in this regard.