Re: Mach-O binary format support and Darwin syscall personality [Was: uts banner changes]

[Linus Torvalds <torvalds@osdl.org>]

On Wed, 13 Dec 2006, Benjamin Herrenschmidt wrote:
> 
> > Secondly, is there a preferred existing field into which I should  
> > stick said flag or just stuff it somewhere?
> 
> Yes, thread_info->flags.

Well, it may actually make sense to actually stick the whole "syscall 
table pointer" in there, rather than a flag that says which pointer to 
choose.

We already load the thread_info pointer because we need the flags for 
syscall tracing, and since we have the thread_info pointer, it might be 
easier to load the syscall table pointer right off there, rather than 
loading it as a big constant with "lis + ori" (in fact, on ppc64, we 
currently load it off the TOC, which is really sad, since we already 
brought in the thread_info into the cache, and usign the TOC is not just a 
load, it's a load off a separate cacheline).

So on 64-bit ppc, it could actually speed things up to put the system call 
table pointer into thread_info, and make it more flexible at the same 
time, without any conditional flags.

		Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/