Re: [Valgrind-users] failed in UME with error 22

[Ivo R. <iv...@iv...>]

2017-11-29 17:55 GMT+01:00 Silva João <joa...@al...>:
> This is the result of readelf -l:
...
> Program Headers:
>   Type           Offset             VirtAddr           PhysAddr
>                  FileSiz            MemSiz              Flags  Align
>   PHDR           0x0000000000000040 0x0000000000400040 0x0000000000400040
>                  0x00000000000001f8 0x00000000000001f8  R E    8
>   INTERP         0x0000000000000238 0x0000000000400238 0x0000000000400238
>                  0x000000000000001c 0x000000000000001c  R      1
>       [Requesting program interpreter: /lib64/ld-linux-x86-64.so.2]
>   LOAD           0x0000000000000000 0x0000000000400000 0x0000000000400000
>                  0x000000000045e870 0x000000000045e870  R E    200000
>   LOAD           0x000000000045e870 0x0000000000a5e870 0x0000000000a5e870
>                  0x0000000000004c98 0x000000006ae482b0  RW     200000

This program header shows:
FileSize: 0x4c98
MemSize: 0x000000006ae482b0 (~1,7 GB)

So Valgrind is interpreting this file correctly.
I have no idea why this file wants to allocate ~1,7 GB of bss.
If that is even remotely correct than Valgrind built as-is cannot map
this ELF file at an absolute address.

You can work around this problem by redefining Valgrind's load address
to a higher one.
It's currently 0x58000000 (check with the SEGMENT OUTPUT from previous
conversation).
Right before it there is a free address space of ~1344m. You need to
enlarge this to be big
enough to accommodate for your large BSS segment.
Have a look at configure.ac (run ./autogen.sh && ./configure).

if that works for you, you can submit a patch to have Valgrind's load
address configurable with configure, for example.
And also supply a better error message and diagnostics about this
problem next time.

I.