MIPS: Fix NaClSwitch() to set $t9 to the address of the entry point

MIPS: Fix NaClSwitch() to set $t9 to the address of the entry point

The MIPS calling convention is that a function is called with $t9 set
to the address of the function being called.  This is used rather like
%ebx on x86-32 for locating global variables.

This change is needed to make hello_world.nexe work, otherwise the
nexe crashes the first time it dereferences a pointer derived from
$t9.

This also fixes an information leak: we no longer leak the value of
$t9 from trusted code.

BUG=https://code.google.com/p/nativeclient/issues/detail?id=2275
TEST=hello_world (with other changes applied locally)

Committed: https://src.chromium.org/viewvc/native_client?view=rev&revision=10278

[Mark Seaborn, 2012-11-15 16:05:53]

I got hello_world to run.  I'll send you more details in other posts.

Did you ever have MIPS nexes running successfully under sel_ldr without this
change to fix $t9?  Is there a codegen mode that doesn't rely on $t9, perhaps?

[petarj, 2012-11-15 16:50:56]

On 2012/11/15 16:05:53, Mark Seaborn wrote:
> I got hello_world to run.  I'll send you more details in other posts.
> 
> Did you ever have MIPS nexes running successfully under sel_ldr without this
> change to fix $t9?  Is there a codegen mode that doesn't rely on $t9, perhaps?

We compile NEXE with -relocation-model=static, and that does not require $t9 set
to the start of the function. We previously tried to use -relocation-model=pic
but gave up temporarily since nobody was filling GOT table (the task that ld.so
does in Linux).

[Mark Seaborn, 2012-11-15 16:55:24]

On 15 November 2012 08:50, <petarj@mips.com> wrote:

> On 2012/11/15 16:05:53, Mark Seaborn wrote:
>
>> I got hello_world to run.  I'll send you more details in other posts.
>>
>
>  Did you ever have MIPS nexes running successfully under sel_ldr without
>> this
>> change to fix $t9?  Is there a codegen mode that doesn't rely on $t9,
>> perhaps?
>>
>
> We compile NEXE with -relocation-model=static, and that does not require
> $t9 set
> to the start of the function. We previously tried to use
> -relocation-model=pic
> but gave up temporarily since nobody was filling GOT table (the task that
> ld.so
> does in Linux).
>

OK, I see.  Would you like to LGTM the change anyway?  It is trivial to set
$t9 in case untrusted code is compiled to need it.  This change also fixes
an information leak:  previously the code leaked the $t9 value from trusted
code.

Mark

-- 
You received this message because you are subscribed to the Google Groups
"Native-Client-Reviews" group.
To post to this group, send email to native-client-reviews@googlegroups.com.
To unsubscribe from this group, send email to
native-client-reviews+unsubscribe@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/native-client-reviews?hl=en.

[petarj, 2012-11-15 17:01:42]

lgtm