Re: [PATCH 5/7] Use %gs for per-cpu sections in kernel

From: Rusty Russell
Date: Mon Sep 25 2006 - 02:04:21 EST


On Sun, 2006-09-24 at 22:25 -0700, Jeremy Fitzhardinge wrote:
> The %gs:per_cpu__foo addressing mode still calculates
> 0xbcef00+0xc0433800, which is still a subtraction. My essential point
> is that *all* kernel addresses (=kernel symbols) are negative, so using
> them as an offset from a segment base (any segment base) is a
> subtraction, which requires a 4G limit.

I don't think so. There's *never* address subtraction, there's
sometimes 32 bit wrap (glibc uses this to effect subtraction, sure).
But there's no wrap here.

To test, I changed the following:

--- smpboot.c.~8~ 2006-09-25 15:51:50.000000000 +1000
+++ smpboot.c 2006-09-25 16:00:36.000000000 +1000
@@ -926,8 +926,9 @@
unsigned long per_cpu_off)
{
unsigned limit, flags;
+ extern char __per_cpu_end[];

- limit = (1 << 20);
+ limit = PAGE_ALIGN((long)__per_cpu_end) >> PAGE_SHIFT;
flags = 0x8; /* 4k granularity */

/* present read-write data segment */


Works fine...

Hope that clarifies!
Rusty.
--
Help! Save Australia from the worst of the DMCA: http://linux.org.au/law

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