[PULL 14/35] target/arm/ptw: Don't report GPC faults on stage 1 ptw as stage2 faults

Maintainers: Paolo Bonzini <pbonzini@redhat.com>, Peter Maydell <peter.maydell@linaro.org>, Joel Stanley <joel@jms.id.au>, Huacai Chen <chenhuacai@kernel.org>, Jiaxun Yang <jiaxun.yang@flygoat.com>, "Philippe Mathieu-Daudé" <philmd@linaro.org>, Daniel Henrique Barboza <danielhb413@gmail.com>, "Cédric Le Goater" <clg@kaod.org>, David Gibson <david@gibson.dropbear.id.au>, Greg Kurz <groug@kaod.org>, Harsh Prateek Bora <harshpb@linux.ibm.com>, Thomas Huth <thuth@redhat.com>, Laurent Vivier <lvivier@redhat.com>, Marcelo Tosatti <mtosatti@redhat.com>, Aurelien Jarno <aurelien@aurel32.net>, Aleksandar Rikalo <aleksandar.rikalo@syrmia.com>, Nicholas Piggin <npiggin@gmail.com>, Palmer Dabbelt <palmer@dabbelt.com>, Alistair Francis <alistair.francis@wdc.com>, Bin Meng <bin.meng@windriver.com>, Weiwei Li <liweiwei@iscas.ac.cn>, Liu Zhiwei <zhiwei_liu@linux.alibaba.com>, Halil Pasic <pasic@linux.ibm.com>, Christian Borntraeger <borntraeger@linux.ibm.com>, Richard Henderson <richard.henderson@linaro.org>, David Hildenbrand <david@redhat.com>, Ilya Leoshkevich <iii@linux.ibm.com>
There is a newer version of this series
[PULL 14/35] target/arm/ptw: Don't report GPC faults on stage 1 ptw as stage2 faults
Posted by Peter Maydell 1 year, 7 months ago
In S1_ptw_translate() we set up the ARMMMUFaultInfo if the attempt to
translate the page descriptor address into a physical address fails.
This used to only be possible if we are doing a stage 2 ptw for that
descriptor address, and so the code always sets fi->stage2 and
fi->s1ptw to true.  However, with FEAT_RME it is also possible for
the lookup of the page descriptor address to fail because of a
Granule Protection Check fault.  These should not be reported as
stage 2, otherwise arm_deliver_fault() will incorrectly set
HPFAR_EL2.  Similarly the s1ptw bit should only be set for stage 2
faults on stage 1 translation table walks, i.e.  not for GPC faults.

Add a comment to the the other place where we might detect a
stage2-fault-on-stage-1-ptw, in arm_casq_ptw(), noting why we know in
that case that it must really be a stage 2 fault and not a GPC fault.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20230807141514.19075-3-peter.maydell@linaro.org
---
 target/arm/ptw.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/target/arm/ptw.c b/target/arm/ptw.c
index bafeb876ad7..eb57ebd897b 100644
--- a/target/arm/ptw.c
+++ b/target/arm/ptw.c
@@ -600,8 +600,8 @@ static bool S1_ptw_translate(CPUARMState *env, S1Translate *ptw,
         fi->type = ARMFault_GPCFOnWalk;
     }
     fi->s2addr = addr;
-    fi->stage2 = true;
-    fi->s1ptw = true;
+    fi->stage2 = regime_is_stage2(s2_mmu_idx);
+    fi->s1ptw = fi->stage2;
     fi->s1ns = !is_secure;
     return false;
 }
@@ -719,6 +719,12 @@ static uint64_t arm_casq_ptw(CPUARMState *env, uint64_t old_val,
         env->tlb_fi = NULL;
 
         if (unlikely(flags & TLB_INVALID_MASK)) {
+            /*
+             * We know this must be a stage 2 fault because the granule
+             * protection table does not separately track read and write
+             * permission, so all GPC faults are caught in S1_ptw_translate():
+             * we only get here for "readable but not writeable".
+             */
             assert(fi->type != ARMFault_None);
             fi->s2addr = ptw->out_virt;
             fi->stage2 = true;
-- 
2.34.1