I need some urgent help with the below issue,
tnx,
ovi
[EMAIL PROTECTED]

Ovidiu Nastai wrote:
>
> Rod, Bart,
>
> I noticed a different address space layout for an app (i.e. vi) on
> different platforms (see below). is ELF allowing for what I notice on
> T1000/T2000 platforms (.text segment "split") and 2 extra "anon" pages?
> may I know the rationale behind (i.e. is that machine-specific)? is
> there in OpenSolaris.com, same source code that I can walk to better
> understand?
>
> Regards,
> ovi
>
>
> T1000/T2000 (sun4v)
> ---------------------------------
> 00010000     192K r-x--  /usr/bin/vi
> 00040000       8K r-x--  /usr/bin/vi
> 00052000      40K rwx--  /usr/bin/vi
> 0005C000      16K rwx--    [ anon ]
> 00060000     576K rwx--    [ heap ]
> FF1E0000       8K r-x--  /platform/sun4v/lib/libc_psr.so.1
> FF1F0000       8K rw---    [ anon ]
> FF200000     832K r-x--  /lib/libc.so.1
> FF2D0000      32K r-x--  /lib/libc.so.1
> FF2E8000      32K rwx--  /lib/libc.so.1
> FF2F0000       8K rwx--  /lib/libc.so.1
> Non-Txxx system (sun4u)
> -------------------------------------
> 00010000     200K r-x--  /usr/bin/vi
> 00052000      40K rwx--  /usr/bin/vi
> 0005C000     576K rwx--    [ heap ]
> FF1C0000      16K r-x-- 
> /usr/lib/locale/en_US.ISO8859-1/en_US.ISO8859-1.so.3
> FF1D2000      16K rwx-- 
> /usr/lib/locale/en_US.ISO8859-1/en_US.ISO8859-1.so.3
> FF1E8000      16K r-x--  /platform/sun4u/lib/libc_psr.so.1
> FF1F0000       8K rw---    [ anon ]


Sorry, but I have no idea why the mappings look different.

There are differences between platforms in regards to where a
file is mapped, and in some cases (psr libraries) what files
get loaded.  But I don't know enough about our VM to know
what the "split text" or extra anon are all about.

Perhaps some VM savvy folks on the tools-linking alias will come
up with something.  Although, this isn't really a linking issue.
The link editors simply put together ELF files and ask the kernel
to map them.  Where the files get placed (for shared objects at
least) is the kernels choice.  How many mmappings contribute
to an executable don't matter to ld.so.1(1) either - the runtime
linker simply sees a contiguous virtual address space.

You might try opensolaris-discuss@opensolaris.org to see if you
can fish out some more knowledgeable folks.

--

Rod.
 
 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org

Reply via email to