|
Rex Lam
|
Errors logged in rmat file
|
|
Rex Lam
09/15/2009 7:36 PM
post38027
|
Errors logged in rmat file
I have been getting a lot of errors in my rmat file. I think this is the reason I cannot use the "get leaks" feature of
the memory analysis tool. Can someone help me figure out what I can do to resolve these errors please?
EV={433369,2,1,0xd1976cba,ERR,249704530, merrno=6,loc={fn=delete_arr,bt={0xbabfa084,0xbabf8cb2,0xbaaa2aa5,0xbaaa08a8,
0xba9e6d1a}},dh={uptr=0x3a129f98,fn=new_arr,sz=24,usz=21,id=0x7b127b7,flag={use}}}
E
EV={433370,2,1,0xd19acbe1,ERR,249704530, merrno=6,loc={fn=free,bt={0xb9c54148,0xb97a4e97,0xb97a2bce,0xb97a32e3,
0xb97a344e}},dh={uptr=0x32276fa0,fn=malloc,sz=16,usz=12,id=0x7b945be,flag={use}}}
E
EV={433371,2,1,0xd19afea7,ERR,249704530, merrno=6,loc={fn=delete,bt={0xb88c7039,0xb88ce5f2,0xb86720ec,0xb867861f,
0xb86786c4}},dh={uptr=0x32276fa0,fn=new,sz=16,usz=16,id=0x7b991cb,flag={use}}}
E
EV={433372,2,1,0xd19b1de6,ERR,249704530, merrno=6,loc={fn=delete_arr,bt={0xb82d08ed,0xb8895d04,0xb8896361,0xb85be45c,
0xb8608774}},dh={uptr=0x32276fa0,fn=new_arr,sz=16,usz=16,id=0x7b991d6,flag={use}}}
E
EV={433373,2,1,0xd19b316d,ERR,249704530, merrno=6,loc={fn=delete,bt={0xb88c7039,0xb88ce5f2,0xb86720ec,0xb867861f,
0xb86786c4}},dh={uptr=0x32276fa0,fn=new,sz=16,usz=16,id=0x7b992f6,flag={use}}}
E
EV={433374,2,1,0xd19b44f4,ERR,249704530, merrno=6,loc={fn=delete_arr,bt={0xb82d08ed,0xb8895d04,0xb8896361,0xb85be45c,
0xb8608774}},dh={uptr=0x32276fa0,fn=new_arr,sz=16,usz=16,id=0x7b99303,flag={use}}}
E
|
|
|
|
|
|
Elena Laskavaia
|
Re: Errors logged in rmat file
|
|
Elena Laskavaia
09/18/2009 1:34 PM
post38350
|
Re: Errors logged in rmat file
You have to load this file in IDE using Import command. And open Errors
tab of loaded session - you will see readable version with full backtrace.
Rex Lam wrote:
> I have been getting a lot of errors in my rmat file. I think this is the reason I cannot use the "get leaks" feature
of the memory analysis tool. Can someone help me figure out what I can do to resolve these errors please?
>
> EV={433369,2,1,0xd1976cba,ERR,249704530, merrno=6,loc={fn=delete_arr,bt={0xbabfa084,0xbabf8cb2,0xbaaa2aa5,0xbaaa08a8,
0xba9e6d1a}},dh={uptr=0x3a129f98,fn=new_arr,sz=24,usz=21,id=0x7b127b7,flag={use}}}
> E
> EV={433370,2,1,0xd19acbe1,ERR,249704530, merrno=6,loc={fn=free,bt={0xb9c54148,0xb97a4e97,0xb97a2bce,0xb97a32e3,
0xb97a344e}},dh={uptr=0x32276fa0,fn=malloc,sz=16,usz=12,id=0x7b945be,flag={use}}}
> E
> EV={433371,2,1,0xd19afea7,ERR,249704530, merrno=6,loc={fn=delete,bt={0xb88c7039,0xb88ce5f2,0xb86720ec,0xb867861f,
0xb86786c4}},dh={uptr=0x32276fa0,fn=new,sz=16,usz=16,id=0x7b991cb,flag={use}}}
> E
> EV={433372,2,1,0xd19b1de6,ERR,249704530, merrno=6,loc={fn=delete_arr,bt={0xb82d08ed,0xb8895d04,0xb8896361,0xb85be45c,
0xb8608774}},dh={uptr=0x32276fa0,fn=new_arr,sz=16,usz=16,id=0x7b991d6,flag={use}}}
> E
> EV={433373,2,1,0xd19b316d,ERR,249704530, merrno=6,loc={fn=delete,bt={0xb88c7039,0xb88ce5f2,0xb86720ec,0xb867861f,
0xb86786c4}},dh={uptr=0x32276fa0,fn=new,sz=16,usz=16,id=0x7b992f6,flag={use}}}
> E
> EV={433374,2,1,0xd19b44f4,ERR,249704530, merrno=6,loc={fn=delete_arr,bt={0xb82d08ed,0xb8895d04,0xb8896361,0xb85be45c,
0xb8608774}},dh={uptr=0x32276fa0,fn=new_arr,sz=16,usz=16,id=0x7b99303,flag={use}}}
> E
>
>
>
>
> _______________________________________________
>
> QNX Momentics Community Support
> http://community.qnx.com/sf/go/post38027
>
>
|
|
|
|
|
|
Rex Lam
|
Re: Errors logged in rmat file
|
|
Rex Lam
09/21/2009 10:50 PM
post38448
|
Re: Errors logged in rmat file
Thank you for the suggestion. It appears most of the errors I am getting are:
"data has been written outside allocated memory block"
How does the memory analysis tool detect this error (what does this error mean)? I could see 5 lines of backtrace with
each error event. Some of the backtraces make sense and some do not. Of the ones that make sense, I could not see any
problems. I would like to know what to look for in code that could cause this error event to be logged. The backtrace
is for "FULL TRACE", which I assume is the backtrace of the deallocation. There is no backtrace for "FULL ALLOC TRACE".
Rex
|
|
|
|
|
|
Elena Laskavaia
|
Re: Errors logged in rmat file
|
|
Elena Laskavaia
09/25/2009 2:43 PM
post38808
|
Re: Errors logged in rmat file
Ok full trace does not actually mean full, poor wording. It is a partial backtrace but bigger that 1 line. Because it
costs a lot to store it, by default it only stores 5 items for error and 5 for allocation itself. There are options to
increase it (Memory Analysis tab options in launch configurations or env variables if you running from command line, I
don't remember exact names, they should be documented).
About memory written out of bounds.
So this error is tricky. This is pretty much buffer overflow or underflow. Depends on memory layout it can be harmless
or really bad. The problem is that backtrace tool prints it at the time error is detected, not when it happens. For
example you do something like this:
x = malloc(1);
x[1]= 3;
...
free(x);
You writing to a second byte, while you only allocated one. However at line 2 where it
happens tool has no meaning to detect it (because MAT is not doing instrumentation of code). However when memory is
freed it can be detected because for example garbage data appears at the end of the block or internal links are
corrupted. In this case trace would show "free" call. In this case you have to manually inspect places where memory can
be overriden. In worst case it is not even same memory which is allocated is being manipulated, for example if you
allocate 2 blocks side by side and "underwrite" to second one, the corrupted one would be first block - which does not
seems to be related at all to the code.
And about leaks (original problem) did you try to insert leaks probs directly in the code?
|
|
|
|
|
|
Rex Lam
|
Re: Errors logged in rmat file
|
|
Rex Lam
10/01/2009 12:33 PM
post39216
|
Re: Errors logged in rmat file
Elena,
Thank you for the explanation. It was very helpful. I have not used the list leaks function in code yet. It has
become a lower priority because the big leaks have been resolved already. There appears to be random memory corruption
so we are looking into memory errors.
There are some backtraces that do not make sense to me. Here is an example:
PresentationManager at startup,9/28/09 5:17 PM,7782,424101103,614826065,3,1,data has been written outside allocated
memory block,0xb943f70,delete_arr,new,2,1,<no source code>,<no source code>,"
[0xb84d2d8f] <no source code>
[0xb84d2f52] <no source code>
[0xb90e1a61] <no source code>
[0xb55fd97] <no source code>
[0xb91206db] <no source code>
",,
The fourth address in the back trace is an address in the heap. I don't understand how the return address could be in
heap memory. I decoded these addresses but the succession of calls do not make sense either. Can the backtrace contain
an address on the heap?
Rex
|
|
|
|
|
|
Elena Laskavaia
|
Re: Errors logged in rmat file
|
|
Elena Laskavaia
10/02/2009 12:48 PM
post39300
|
Re: Errors logged in rmat file
This can happened either is stack is corrupted (classic stack buffer
overflow) or backtrace functionality is fulled some-how and cannot
restore proper stack.
Try to use not-optimized code -O0 and without inline function
-fno-inline-functions (or something like that)
But why do you think it is heap address? They all look shared library
addresses with unresolved libraries
Rex Lam wrote:
> Elena,
>
> Thank you for the explanation. It was very helpful. I have not used the list leaks function in code yet. It has
become a lower priority because the big leaks have been resolved already. There appears to be random memory corruption
so we are looking into memory errors.
>
> There are some backtraces that do not make sense to me. Here is an example:
> PresentationManager at startup,9/28/09 5:17 PM,7782,424101103,614826065,3,1,data has been written outside allocated
memory block,0xb943f70,delete_arr,new,2,1,<no source code>,<no source code>,"
> [0xb84d2d8f] <no source code>
> [0xb84d2f52] <no source code>
> [0xb90e1a61] <no source code>
> [0xb55fd97] <no source code>
> [0xb91206db] <no source code>
> ",,
>
> The fourth address in the back trace is an address in the heap. I don't understand how the return address could be in
heap memory. I decoded these addresses but the succession of calls do not make sense either. Can the backtrace
contain an address on the heap?
>
> Rex
>
>
>
> _______________________________________________
>
> QNX Momentics Community Support
> http://community.qnx.com/sf/go/post39216
>
>
|
|
|
|
|
|
Rex Lam
|
Re: Errors logged in rmat file
|
|
Rex Lam
10/02/2009 5:42 PM
post39322
|
Re: Errors logged in rmat file
Elena,
I don't think the stack is corrupted because the program continued to run after this happened. I will try the
suggestion to not use optimized code or inline functions.
I used the system information tool to view memory address ranges for this process.
> [0xb84d2d8f] <no source code>
> [0xb84d2f52] <no source code>
> [0xb90e1a61] <no source code>
> [0xb55fd97] <no source code>
> [0xb91206db] <no source code>
The addresses are usually from the "library" section. The fourth address from this example is one digit short, so it is
actually 0x0b55fd97, which falls in the "heap" section.
Rex
|
|
|
|
|
|
