[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] Re: [PATCH RFC V2 3/5] jump_label: if a key has already been initialized, don't nop it out



On Thu, Oct 06, 2011 at 03:34:13PM -0400, Steven Rostedt wrote:
> On Thu, 2011-10-06 at 14:38 -0400, Jason Baron wrote:
> 
> > Right, its certainly doable, but I'm not sure its so simple, since we'll
> > need a pass to eliminate the jumps - which can be keyed off the
> > '__jump_table' section. 
> 
> Look at the code of scripts/recordmcount.c and friends.
> 
> It does two things.
> 
> 1) find all the callers of mcount and make a section for it.
> 
> 2) For those callers of mcount that is in sections that are not
> whitelisted, and therefor will not be patched, to replace the call to
> mcount with a nop.
> 
> 
> We can use this code, or a copy of it, to do the same with jump_label.
> Have the x86 jump_label be:
> 
> 
> static __always_inline bool arch_static_branch(struct jump_label_key
> *key)
> {
>       asm goto("1:"
>               "jmp l_yes\n"
>               ".pushsection __jump_table,  \"aw\" \n\t"
>               _ASM_ALIGN "\n\t"
>               _ASM_PTR "1b, %l[l_yes], %c0 \n\t"
>               ".popsection \n\t"
>               : :  "i" (key) : : l_yes);
>       return false;
> l_yes:
>       return true;
> }
> 
> Then have the record_jumplabel.c (or whatever it's called) find all the
> jmps at run time, and convert them into the appropriate nop.
> 

I'd prefer to do this at build-time as hpa said. We don't want there to
be some race b/w patching in the no-ops and taking an incorrect branch.

> Then at runtime patching, the jumplabel code could figure out what size
> jump it needs to replace it.
> 
> -- Steve
> 
> 

sounds like a good plan. thanks for the pointers!

-Jason

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.