Patchwork [3/3] PPC: Fix for the gdb single step problem on an rfi instruction

login
register
mail settings
Submitter Alexander Graf
Date Nov. 11, 2011, 4:37 p.m.
Message ID <1321029448-11010-4-git-send-email-agraf@suse.de>
Download mbox | patch
Permalink /patch/125213/
State New
Headers show

Comments

Alexander Graf - Nov. 11, 2011, 4:37 p.m.
From: Sebastian Bauer <mail@sebastianbauer.info>

When using gdb to single step a ppc interrupt routine, the execution
flow passes the rfi instruction without actually returning from the
interrupt.

The patch fixes this by avoiding to update the nip when the debug
exception is raised and a previous POWERPC_EXCP_SYNC was set.

The latter is the case only, if code for rfi or a related instruction
was generated.

Signed-off-by: Sebastian Bauer <mail@sebastianbauer.info>
Signed-off-by: Alexander Graf <agraf@suse.de>
---
 target-ppc/translate.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

Patch

diff --git a/target-ppc/translate.c b/target-ppc/translate.c
index 99e995c..66eae30 100644
--- a/target-ppc/translate.c
+++ b/target-ppc/translate.c
@@ -298,8 +298,10 @@  static inline void gen_debug_exception(DisasContext *ctx)
 {
     TCGv_i32 t0;
 
-    if (ctx->exception != POWERPC_EXCP_BRANCH)
+    if ((ctx->exception != POWERPC_EXCP_BRANCH) &&
+        (ctx->exception != POWERPC_EXCP_SYNC)) {
         gen_update_nip(ctx, ctx->nip);
+    }
     t0 = tcg_const_i32(EXCP_DEBUG);
     gen_helper_raise_exception(t0);
     tcg_temp_free_i32(t0);