The initial TLB mapping for the kernel boot didn't set the memory coherent
attribute, MAS2[M], in SMP mode.
If this code supported booting a secondary processor, which it doesn't yet,
but suppose it did, then when a secondary processor boots, it would have
probably signaled the primary processor by setting a variable called
something like __secondary_hold_acknowledge. However, due to the lack of
the M bit, the primary processor would not have snooped the transaction
(even if a transaction were broadcast). If primary CPU's L1 D-cache had a
copy, it would not have been flushed and the CPU would have never seen the
ack. Which would have resulted in the primary CPU spinning for a long
time, perhaps a full second before it would have given up, while it would
have waited for the ack from the secondary CPU that it wouldn't have been
able to see because of the stale cache.
The value of MAS2 for the boot page TLB1 entry is a compile time constant,
so there is no need to calculate it in powerpc assembly language.
Also, from the MPC8572 manual section 6.12.5.3, "Bits that represent
offsets within a page are ignored and should be cleared." Existing code
didn't clear them, this code does.
The same when the page of KERNELBASE is found; we don't need to use asm to
mask the lower 12 bits off.
In the code that computes the address to rfi from, don't hard code the
offset to 24 bytes, but have the assembler figure that out for us.
Signed-off-by: Trent Piepho <[email protected]>
Acked-by: Kumar Gala <[email protected]>
---
arch/powerpc/include/asm/mmu-fsl-booke.h | 2 ++
arch/powerpc/kernel/head_fsl_booke.S | 22 +++++++++++++---------
2 files changed, 15 insertions(+), 9 deletions(-)
diff --git a/arch/powerpc/include/asm/mmu-fsl-booke.h b/arch/powerpc/include/asm/mmu-fsl-booke.h
index 925d93c..5588a41 100644
--- a/arch/powerpc/include/asm/mmu-fsl-booke.h
+++ b/arch/powerpc/include/asm/mmu-fsl-booke.h
@@ -40,6 +40,8 @@
#define MAS2_M 0x00000004
#define MAS2_G 0x00000002
#define MAS2_E 0x00000001
+#define MAS2_EPN_MASK(size) (~0 << (2*(size) + 10))
+#define MAS2_VAL(addr, size, flags) ((addr) & MAS2_EPN_MASK(size) | (flags))
#define MAS3_RPN 0xFFFFF000
#define MAS3_U0 0x00000200
diff --git a/arch/powerpc/kernel/head_fsl_booke.S b/arch/powerpc/kernel/head_fsl_booke.S
index 590304c..e621eac 100644
--- a/arch/powerpc/kernel/head_fsl_booke.S
+++ b/arch/powerpc/kernel/head_fsl_booke.S
@@ -235,36 +235,40 @@ skpinv: addi r6,r6,1 /* Increment */
tlbivax 0,r9
TLBSYNC
+/* The mapping only needs to be cache-coherent on SMP */
+#ifdef CONFIG_SMP
+#define M_IF_SMP MAS2_M
+#else
+#define M_IF_SMP 0
+#endif
+
/* 6. Setup KERNELBASE mapping in TLB1[0] */
lis r6,0x1000 /* Set MAS0(TLBSEL) = TLB1(1), ESEL = 0 */
mtspr SPRN_MAS0,r6
lis r6,(MAS1_VALID|MAS1_IPROT)@h
ori r6,r6,(MAS1_TSIZE(BOOKE_PAGESZ_64M))@l
mtspr SPRN_MAS1,r6
- li r7,0
- lis r6,PAGE_OFFSET@h
- ori r6,r6,PAGE_OFFSET@l
- rlwimi r6,r7,0,20,31
+ lis r6,MAS2_VAL(PAGE_OFFSET, BOOKE_PAGESZ_64M, M_IF_SMP)@h
+ ori r6,r6,MAS2_VAL(PAGE_OFFSET, BOOKE_PAGESZ_64M, M_IF_SMP)@l
mtspr SPRN_MAS2,r6
mtspr SPRN_MAS3,r8
tlbwe
/* 7. Jump to KERNELBASE mapping */
- lis r6,KERNELBASE@h
- ori r6,r6,KERNELBASE@l
- rlwimi r6,r7,0,20,31
+ lis r6,(KERNELBASE & ~0xfff)@h
+ ori r6,r6,(KERNELBASE & ~0xfff)@l
lis r7,MSR_KERNEL@h
ori r7,r7,MSR_KERNEL@l
bl 1f /* Find our address */
1: mflr r9
rlwimi r6,r9,0,20,31
- addi r6,r6,24
+ addi r6,r6,(2f - 1b)
mtspr SPRN_SRR0,r6
mtspr SPRN_SRR1,r7
rfi /* start execution out of TLB1[0] entry */
/* 8. Clear out the temp mapping */
- lis r7,0x1000 /* Set MAS0(TLBSEL) = 1 */
+2: lis r7,0x1000 /* Set MAS0(TLBSEL) = 1 */
rlwimi r7,r5,16,4,15 /* Setup MAS0 = TLBSEL | ESEL(r5) */
mtspr SPRN_MAS0,r7
tlbre
--
1.5.4.1
On Nov 19, 2008, at 1:14 PM, Trent Piepho wrote:
> The initial TLB mapping for the kernel boot didn't set the memory
> coherent
> attribute, MAS2[M], in SMP mode.
>
> If this code supported booting a secondary processor, which it
> doesn't yet,
> but suppose it did, then when a secondary processor boots, it would
> have
> probably signaled the primary processor by setting a variable called
> something like __secondary_hold_acknowledge. However, due to the
> lack of
> the M bit, the primary processor would not have snooped the
> transaction
> (even if a transaction were broadcast). If primary CPU's L1 D-cache
> had a
> copy, it would not have been flushed and the CPU would have never
> seen the
> ack. Which would have resulted in the primary CPU spinning for a long
> time, perhaps a full second before it would have given up, while it
> would
> have waited for the ack from the secondary CPU that it wouldn't have
> been
> able to see because of the stale cache.
>
> The value of MAS2 for the boot page TLB1 entry is a compile time
> constant,
> so there is no need to calculate it in powerpc assembly language.
>
> Also, from the MPC8572 manual section 6.12.5.3, "Bits that represent
> offsets within a page are ignored and should be cleared." Existing
> code
> didn't clear them, this code does.
>
> The same when the page of KERNELBASE is found; we don't need to use
> asm to
> mask the lower 12 bits off.
>
> In the code that computes the address to rfi from, don't hard code the
> offset to 24 bytes, but have the assembler figure that out for us.
>
> Signed-off-by: Trent Piepho <[email protected]>
> Acked-by: Kumar Gala <[email protected]>
> ---
> arch/powerpc/include/asm/mmu-fsl-booke.h | 2 ++
> arch/powerpc/kernel/head_fsl_booke.S | 22 +++++++++++++---------
> 2 files changed, 15 insertions(+), 9 deletions(-)
applied to next.
- k