2eca40a8 breaks StrongARM compilation

Commit 2eca40a8 which went into 2.6.32-rc6 breaks compilation for ipaq h3600 and probably other SA1100 machines when CONFIG_CPU_FREQ is unset:

CC arch/arm/mach-sa1100/generic.o
arch/arm/mach-sa1100/generic.c:117: error: redefinition of 'cpufreq_get' include/linux/cpufreq.h:299: error: previous definition of 'cpufreq_get' was here
make[1]: *** [arch/arm/mach-sa1100/generic.o] Error 1
make: *** [arch/arm/mach-sa1100] Error 2

The problem is that before 2eca40a8 StrongArm aready had its own version of cpufreq_get for CONFIG_CPU_FREQ=n case and now it clashed with one introduced in 2eca40a8. Removing strongarm version will cure compilation, but it then will break sa1100_fb driver in runtime, as it blindly uses value which cpufreq_get returns for calculating pixel clock divisor (see line 926 in drivers/video/sa1100fb.c) and 2eca40a8's cpufreq_get returns 0.

One option would be to make sa1100_fb use hardcoded frequency 206.4MHz (default on StrongARM CPUs) for CONFIG_CPU_FREQ=n case - see attached patch. But I'm not sure if this is a proper fix.

Any ideas are welcome.

Best regards,
Dmitry "MAD" Artamonow

-------------- next part --------------

index 23cfdd5..59a21c6 100644
@@ -110,15 +110,6 @@ unsigned int sa11x0_getspeed(unsigned int cpu)
 	return cclk_frequency_100khz[PPCR & 0xf] * 100;
 }
 
-#else
-/*
- * We still need to provide this so building without cpufreq works.
- */
-unsigned int cpufreq_get(unsigned int cpu)
-{
-	return cclk_frequency_100khz[PPCR & 0xf] * 100;
-}
-EXPORT_SYMBOL(cpufreq_get);
 #endif
 
 /*

index cdaa873..4f2e4fe 100644
@@ -923,7 +923,12 @@ static int sa1100fb_activate_var(struct fb_var_screeninfo *var, struct sa1100fb_
 		LCCR2_BegFrmDel(var->upper_margin) +
 		LCCR2_EndFrmDel(var->lower_margin);
 
+#ifdef CONFIG_CPU_FREQ
 	pcd = get_pcd(var->pixclock, cpufreq_get(0));
+#else
+	pcd = get_pcd(var->pixclock, 206400);
+#endif
+
 	new_regs.lccr3 = LCCR3_PixClkDiv(pcd) | fbi->lccr3 |
 		(var->sync & FB_SYNC_HOR_HIGH_ACT ? LCCR3_HorSnchH : LCCR3_HorSnchL) |
 		(var->sync & FB_SYNC_VERT_HIGH_ACT ? LCCR3_VrtSnchH : LCCR3_VrtSnchL);

I'm aware, I have a very large patch which fixes SA1100, but there's other ARM machine classes which are similarly broken.

We've been around this loop before, where a change like this was introduced and reverted. The kernel community seems set to constantly repeat the same old mistakes time and time again.