Skip to contentSkip to conversation

powerpc: Use more accurate limit for first segment memory allocations

PATCH In review Anton Blanchard 2 participants 2 messages May 11, 2010 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc: Use more accurate limit for first segment memory allocations

Author: Milton Miller <[email protected]>

On large machines we are running out of room below 256MB. In some cases we only need to ensure the allocation is in the first segment, which may be 256MB or 1TB.

Add slb0_limit and use it to specify the upper limit for the irqstack and emergency stacks.

On a large ppc64 box, this fixes a panic at boot when the crashkernel= option is specified (previously we would run out of memory below 256MB).

---

Some things we should add:

- Specify a more intelligent limit for BookE
- The allocation of the PACAs should use this limit too

arch/powerpc/kernel/setup_64.c +13 -4
Index: linux-2.6/arch/powerpc/kernel/setup_64.c
===================================================================
@@ -424,9 +424,18 @@ void __init setup_system(void)
 	DBG(" <- setup_system()\n");
 }
 
+static u64 slb0_limit(void)
+{
+	if (cpu_has_feature(CPU_FTR_1T_SEGMENT)) {
+		return 1UL << SID_SHIFT_1T;
+	}
+	return 1UL << SID_SHIFT;
+}
+
 #ifdef CONFIG_IRQSTACKS
 static void __init irqstack_early_init(void)
 {
+	u64 limit = slb0_limit();
 	unsigned int i;
 
 	/*
@@ -436,10 +445,10 @@ static void __init irqstack_early_init(v
 	for_each_possible_cpu(i) {
 		softirq_ctx[i] = (struct thread_info *)
 			__va(lmb_alloc_base(THREAD_SIZE,
-					    THREAD_SIZE, 0x10000000));
+					    THREAD_SIZE, limit));
 		hardirq_ctx[i] = (struct thread_info *)
 			__va(lmb_alloc_base(THREAD_SIZE,
-					    THREAD_SIZE, 0x10000000));
+					    THREAD_SIZE, limit));
 	}
 }
 #else
@@ -470,7 +479,7 @@ static void __init exc_lvl_early_init(vo
  */
 static void __init emergency_stack_init(void)
 {
-	unsigned long limit;
+	u64 limit;
 	unsigned int i;
 
 	/*
@@ -482,7 +491,7 @@ static void __init emergency_stack_init(
 	 * bringup, we need to get at them in real mode. This means they
 	 * must also be within the RMO region.
 	 */
-	limit = min(0x10000000ULL, lmb.rmo_size);
+	limit = min(slb0_limit(), lmb.rmo_size);
 
 	for_each_possible_cpu(i) {
 		unsigned long sp;
#1 lore raw
Quoted 12 lines · Anton Blanchard

Author: Milton Miller <[email protected]>

On large machines we are running out of room below 256MB. In some cases we only need to ensure the allocation is in the first segment, which may be 256MB or 1TB.

Add slb0_limit and use it to specify the upper limit for the irqstack and emergency stacks.

On a large ppc64 box, this fixes a panic at boot when the crashkernel= option is specified (previously we would run out of memory below 256MB).

Quoted 8 lines

---

Some things we should add:

- Specify a more intelligent limit for BookE
- The allocation of the PACAs should use this limit too

We access the PACA from real mode?

Quoted 5 lines

Index: linux-2.6/arch/powerpc/kernel/setup_64.c
===================================================================

arch/powerpc/kernel/setup_64.c void __init setup_system(void) · line 433

Quoted 10 lines
arch/powerpc/kernel/setup_64.c +7 -0
@@ -424,9 +424,18 @@ void __init setup_system(void)
 	DBG(" <- setup_system()\n");
 }
 
+static u64 slb0_limit(void)
+{
+	if (cpu_has_feature(CPU_FTR_1T_SEGMENT)) {
+		return 1UL << SID_SHIFT_1T;
+	}
+	return 1UL << SID_SHIFT;
+}

I take it there's no chance the CPU feature is there but we're using 256 for whatever reason?

cheers