On Wed, May 18, 2016 at 03:45:22PM +0100, Matt Redfearn wrote:Good point - thanks Paul :-)
When starting secondary VPEs which support EVA and the SegCtl registers,Hi Matt,
copy the memory segmentation configuration from the running VPE to ensure
that all VPEs in the core have a consitent virtual memory map.
The EVA configuration of secondary cores is dealt with when starting the
core via the CM.
Signed-off-by: Matt Redfearn <matt.redfearn@xxxxxxxxxx>
---
arch/mips/kernel/cps-vec.S | 16 ++++++++++++++++
1 file changed, 16 insertions(+)
diff --git a/arch/mips/kernel/cps-vec.S b/arch/mips/kernel/cps-vec.S
index ac81edd44563..07b3274c8ae1 100644
--- a/arch/mips/kernel/cps-vec.S
+++ b/arch/mips/kernel/cps-vec.S
@@ -431,6 +431,22 @@ LEAF(mips_cps_boot_vpes)
mfc0 t0, CP0_CONFIG
mttc0 t0, CP0_CONFIG
+ /* Copy the EVA config from this VPE if the CPU supports it */
+ mfc0 t0, CP0_CONFIG, 1
+ bgez t0, 1f
+ mfc0 t0, CP0_CONFIG, 2
+ bgez t0, 1f
+ mfc0 t0, CP0_CONFIG, 3
+ and t0, t0, MIPS_CONF3_SC
+ beqz t0, 1f
+ nop
The checks here aren't *quite* right since they do the mfc0 of the next
register in the delay slot which will happen even if the M bit of the
preceeding register wasn't set. There are other cases in cps-vec.S where
I've made that mistake... Luckily, in this particular case, we know that
we have MT ASE support which means we know that Config3 exists. So I
think you can just remove the checks of Config1.M & Config2.M and just
read Config3 straight away.
Thanks,
Paul
+ mfc0 t0, CP0_SEGCTL0
+ mttc0 t0, CP0_SEGCTL0
+ mfc0 t0, CP0_SEGCTL1
+ mttc0 t0, CP0_SEGCTL1
+ mfc0 t0, CP0_SEGCTL2
+ mttc0 t0, CP0_SEGCTL2
+1:
/* Ensure no software interrupts are pending */
mttc0 zero, CP0_CAUSE
mttc0 zero, CP0_STATUS
--
2.5.0