Dalvik VM init failed (check log file)

[anand@android]

Hi All,

I'm trying to port dalvik VM onto PPC.
I could able to build 'dalvikvm' binary and the respective 'jar' files
that are needed for the execution of 'dalvikvm'('core.jar' ,
'ext.jar','framework.jar',android.policy.jar' and 'services.jar')

To test dalvikvm i followed the following steps:-

1:
export LD_LIBRARY_PATH=/home/cupcake/out/debug/target/product/eee_701/
system/lib
2:
export ANDROID_ROOT=/home/cupcake/out/debug/target/product/eee_701/
system
3:
export BOOTCLASSPATH=/home/cupcake/out/debug/target/product/eee_701/
systemframework/core.jar:/home/cupcake/out/debug/target/product/
eee_701/system/framework/ext.jar:/home/cupcake/out/debug/target/
product/eee_701/system/framework/framework.jar
4:
export ANDROID_DATA=/tmp/dalvik_$USER
mkdir -p $ANDROID_DATA/dalvik-cache
5:
$cat Foo.java
public class Foo {
public static void main(String[] args) {
System.out.println("Hello, world");
}
}

$javac Foo.java
$dx --dex --output=foo.jar Foo.class
6:
$./dalvikvm -cp foo.jar Foo


For the first execution of 'dalvikvm' I got segmentation fault.
On debugging in gdb I found where it was crashing.
To fix it I made modification in the following file
'dalvik/vm/mterp/common/asm-constants.h'
Line No :186

ORIGINAL-> MTERP_OFFSET(offArrayObject_contents, ArrayObject,
contents, 12)
MODIFIED-> MTERP_OFFSET(offArrayObject_contents, ArrayObject,
contents, 16)

Reason:-
'contents' member in ArrayObject structure is a 64-bit entity and
there are three 32-bit entities above it.
For satisfying 64-bit alignment 'contents' member should start at 16
byte offset rather than 12 byte.

After making the above change, I followed the same steps to execute
dalvikvm,
but this time i didnt get the seg fault but one message came up saying
"Dalvik VM init failed (check log file)".

On debugging in gdb I found the error line to be at
system/core/libcutils/ashmem-dev.c:60 ret = ioctl(fd,
ASHMEM_SET_SIZE, size)
; -- with errno set to -ENOTTY.

I really need help on this to go ahead

Thanks in advance
Anand

[fadden]

On Mar 24, 8:09 am, "anand@android" <android.an...@gmail.com> wrote:
> ORIGINAL->  MTERP_OFFSET(offArrayObject_contents,   ArrayObject,
> contents, 12)
> MODIFIED->  MTERP_OFFSET(offArrayObject_contents,   ArrayObject,
> contents, 16)

See MTERP_NO_UNALIGN_64 in the current version:

http://android.git.kernel.org/?p=platform/dalvik.git;a=blob;f=vm/mterp/common/asm-constants.h;h=73292a9f2a5863fe55b3d524695f245388ad1780;hb=HEAD

> but this time i didnt get the seg fault but one message came up saying
> "Dalvik VM init failed (check log file)".
>
> On debugging in gdb I found the error line to be at

The log file will usually provide some useful info. Do you have /dev/
log working?

> system/core/libcutils/ashmem-dev.c:60    ret = ioctl(fd,
> ASHMEM_SET_SIZE, size)
> ; -- with errno set to -ENOTTY.

Do you have the ashmem driver working?

[Anand Android]

Hi fadden , thanks for the reply.

i have fixed 'ashmem' issue and am able to proceed till 'dexopt' tries to optimize the jar file,which is failing with following error.
-------------
DexOpt:  locked cache file
DexOpt: successfully initialized new cache file
DexOpt: --- BEGIN 'core.jar' (bootstrap=1) ---
DexOpt: waiting for verify+opt, pid=2460
Continuing optimization (/home/cupcake/out/debug/target/product/eee_701/system/framework/core.jar, isb=1, vfy=1, opt=1)
+++ swapping bytes
DexOpt: --- END 'core.jar' --- status=0xff00, process failed
Unable to extract+optimize DEX from '/home/cupcake/out/debug/target/product/eee_701/system/framework/core.jar'
--------------

On debugging i found it is failing at '~/cupcake/dalvik/libdex/DexSwapVerify.c:2084' where 'swapMap()' function is failing.


Thanks in advance
Ananda

[fadden]

On Mar 26, 9:31 am, Anand Android <android.an...@gmail.com> wrote:
> DexOpt: --- BEGIN 'core.jar' (bootstrap=1) ---
> DexOpt: waiting for verify+opt, pid=2460
> Continuing optimization
> (/home/cupcake/out/debug/target/product/eee_701/system/framework/core.jar,
> isb=1, vfy=1, opt=1)
> +++ swapping bytes
> DexOpt: --- END 'core.jar' --- status=0xff00, process failed

[...]

> On debugging i found it is failing at
> '~/cupcake/dalvik/libdex/DexSwapVerify.c:2084' where 'swapMap()' function is
> failing.

It should log a specific failure message when something goes wrong.
Line 2084 looks like it's scrounging around in annotations; I think
maybe it should have a LOGE:

while (size--) {
data = verifyEncodedValue(state, data, crossVerify);
if (data == NULL) {
LOGE("Bogus encoded_array value\n");
return NULL;
}
}

If you've got the latest cupcake you should have the checksum
verification enabled (around line 2767), which should ensure that
you're not reading a corrupt file.

[Anand Android]

hi Fadden,
thanks for the reply
I am able to proceed swapMap() function by fixing an endian issue in that function.
Fix: pMap->size has to be converted to big-endian before assigning it to count variable.

Now I have another issue in crossVerifyEverything() with "kDexTypeClassDefItem" switch-case. It is becoming increasingly difficult to debug this as this is tightly coupled to dex file-format. I hope I can get some pointers into fixing this issue.

Thanks
Anand

[fadden]

On Mar 30, 8:57 am, Anand Android <android.an...@gmail.com> wrote:
> I am able to proceed swapMap() function by fixing an endian issue in that
> function.
> Fix: pMap->size has to be converted to big-endian before assigning it to
> count variable.

Aha, something like:

365c365
< u4 count = pMap->size;
---
> u4 count;
370a371,373
>
> SWAP_FIELD4(pMap->size);
> count = pMap->size;

> Now I have another issue in crossVerifyEverything() with
> "kDexTypeClassDefItem" switch-case. It is becoming increasingly difficult to
> debug this as this is tightly coupled to dex file-format. I hope I can get
> some pointers into fixing this issue.

Once upon a time the code ran on a big-endian PPC system, but the swap
& verify code got rewritten after we moved away from PPC hardware.

If it helps any, the DEX format is described in dalvik/docs/dex-
format.html.

[Anand Android]

hi fadden,
yes that is exactly what i have done to fix problem in 'swapMap()' function.

is the version source tree that ran on ppc still available ?
if so,is it advisable to use it now and can you point me to that source tree ?

thanks
Anand

[fadden]

On Mar 30, 11:51 pm, Anand Android <android.an...@gmail.com> wrote:
> is the version source tree that ran on ppc still available ?
> if so,is it advisable to use it now and can you point me to that source tree ?

Looking at the perforce history, the axe was dropped on PPC support in
June 2007.

So, no. :-|

[dj]

hi fadden
i am also working on same environmnt and stuck in running
dalvikvm on my ppc64 linux m/c
Following is the issue that am currently facing

while debugging in gdb i found that
from the file dalvik/vm/oo/Object.h line 797
in the below function clazz->status contains CLASS_VERIFYING
instead of CLASS_VERIFIED

INLINE bool dvmIsClassVerified(const ClassObject* clazz) {
return clazz->status >= CLASS_VERIFIED;
}
thats why returnig false from this function in casuses dvmabort() to
fail at following location

0x0feacd38 in dvmAbort () at dalvik/vm/Init.c:1433
1433 *((char*)0xdeadd00d) = 38;
and here i am getting seg fault .
are you having any idea about this ,how to get rid of this problem?
please reply............
thanks in advance.

regards
DJ

[fadden]

On Apr 9, 12:10 am, dj <dhananjayingr...@gmail.com> wrote:
> thats why returnig false from this function in casuses dvmabort() to
> fail at following location
>
> 0x0feacd38 in dvmAbort () at dalvik/vm/Init.c:1433
> 1433        *((char*)0xdeadd00d) = 38;
>  and here i am getting seg fault .
> are you having any idea about this ,how to get rid of this problem?

The VM calls dvmAbort() if it encounters a problem that it can't deal
with. The store to 0xdeadd00d is expected to crash the process.

The log file will have an indication of what went wrong.

[Dhananjay Joshi]

Hi fadden

               Actually i dont know how to check messages in /dev/log directory?

               For that we need to write socket program ?

               can you please throw a light on this?

 

 

regards,

DJ

[fadden]

On Apr 9, 11:26 pm, Dhananjay Joshi <dhananjayingr...@gmail.com>
wrote:

> > The log file will have an indication of what went wrong.

>                Actually i dont know how to check messages in /dev/log
> directory?
>                For that we need to write socket program ?
>                can you please throw a light on this?

The "logcat" command will read and display messages from the log
devices.

[Dhananjay Joshi]

hi fadden

               The "logcat" command will read and display messages from the log
devices.
 

as i have android kernel running on my ppc64 m/c , on top of that am trying to run dalvikvm executable , how can i run logcat command from shell?

as it is not recognising dalvikvm's internal command.

 

regards,

DJ

[Dhananjay Joshi]

also when i tried running logcat binary from command line i am getting this error

 

Unable to open log device '/dev/log/main'

 

do you know what is this talking about ?

[Avtar Singh]

as i have android kernel running on my ppc64 m/c , on top of that am trying to run dalvikvm executable , how can i run logcat command from shell?

as it is not recognising dalvikvm's internal command.

Do you see the shell prompt? Type in "logcat" at the shell prompt.
 

 

also when i tried running logcat binary from command line i am getting this error

 

Unable to open log device '/dev/log/main'

 

do you know what is this talking about ?

Did you rebuild kernel by switching off the log driver in kernel configuration file? You need to have that driver enabled.

[Dhananjay Joshi]

hi

    thanks for your reply.

    but i dont know this log drivers comes under which kernel module?

    which driver module i need to enable?

 

regards

DJ

[Dhananjay Joshi]

Now i have enable log drivers in android kernel module,

and i can able to see /dev/log socket is generated

 

but after getting seg fault from dalvikvm when i run logcat from command line

getting the same error as

 

Unable to open log device '/dev/log/main'

 

so do you have any solution to this problem?

 

regards,

DJ