sparc32 MXCC support

I'm trying to add SuperSparc II MXCC support and need some feedback.

Is there a better way to read and write physical memory in 64bit chunks? I'm not sure what I'm doing is portable between 32/64 and big/little endian.

Index: hw/sun4m.c
===================================================================
RCS file: /sources/qemu/qemu/hw/sun4m.c,v
retrieving revision 1.55

diff -p -u -r1.55 sun4m.c
@@ -327,7 +327,7 @@ static void *sun4m_hw_init(const struct 
 
     for(i = 0; i < smp_cpus; i++) {
         env = cpu_init();
-        cpu_sparc_register(env, def);
+        cpu_sparc_register(env, def, i);
         envs[i] = env;
         if (i == 0) {
             qemu_register_reset(main_cpu_reset, env);

Index: linux-user/main.c
===================================================================
RCS file: /sources/qemu/qemu/linux-user/main.c,v
retrieving revision 1.132

diff -p -u -r1.132 main.c
@@ -2139,7 +2139,7 @@ int main(int argc, char **argv)
             fprintf(stderr, "Unable to find Sparc CPU definition\n");
             exit(1);
         }
-        cpu_sparc_register(env, def);
+        cpu_sparc_register(env, def, 0);
 	env->pc = regs->pc;
 	env->npc = regs->npc;
         env->y = regs->y;

Index: target-sparc/cpu.h
===================================================================
RCS file: /sources/qemu/qemu/target-sparc/cpu.h,v
retrieving revision 1.53

diff -p -u -r1.53 cpu.h
@@ -210,6 +210,8 @@ typedef struct CPUSPARCState {
     uint64_t dtlb_tte[64];
 #else
     uint32_t mmuregs[16];
+    uint64_t mxccdata[4];
+    uint64_t mxccregs[8];
 #endif
     /* temporary float registers */
     float32 ft0, ft1;

@@ -266,7 +268,7 @@ int cpu_sparc_close(CPUSPARCState *s);
 int sparc_find_by_name (const unsigned char *name, const sparc_def_t **def);
 void sparc_cpu_list (FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt,
                                                  ...));
-int cpu_sparc_register (CPUSPARCState *env, const sparc_def_t *def);
+int cpu_sparc_register (CPUSPARCState *env, const sparc_def_t *def, int cpu);
 
 #define GET_PSR(env) (env->version | (env->psr & PSR_ICC) |             \
                       (env->psref? PSR_EF : 0) |                        \

Index: target-sparc/op_helper.c
===================================================================
RCS file: /sources/qemu/qemu/target-sparc/op_helper.c,v retrieving revision 1.41

Index: target-sparc/translate.c
===================================================================
RCS file: /sources/qemu/qemu/target-sparc/translate.c,v retrieving revision 1.74

diff -p -u -r1.74 translate.c
@@ -3618,12 +3618,13 @@ void sparc_cpu_list (FILE *f, int (*cpu_
     }
 }
 
-int cpu_sparc_register (CPUSPARCState *env, const sparc_def_t *def)
+int cpu_sparc_register (CPUSPARCState *env, const sparc_def_t *def, int cpu)
 {
     env->version = def->iu_version;
     env->fsr = def->fpu_version;
 #if !defined(TARGET_SPARC64)
     env->mmuregs[0] |= def->mmu_version;
+    env->mxccregs[7] = ((cpu + 8) & 0xf) << 24;
 #endif
     return 0;
 }

Thank you for your effort. Applying the patch allows NetBSD on SS-10 to boot as far as on SS-5, previously it crashed.

I think the code is portable, but I think changing the register type to uint32_t and using a DPRINTF system like used in for example hw/iommu.c should make the code slightly clearer.

For the memory access, ldl/q_phys/stl/q_phys is used in other block copy routines. For longer blocks or if the address can be unaligned, cpu_physical_memory_read() could be a better choice.

unsigned int cpu?

If it's an error to access the registers with different size, you should use do_unassigned_access() to report this.

I don't have access to chip specs but from reading linux and *bsd code it appears that the registers are 64 bits wide but sometimes accessed as 32 bits to get only the bottom 32 bits. Do you mean using two uint32_t for a 64 bit register? How about a union of 2 uint32_t and a uint64_t. Is that portable? Never mind. I just found out there is an ldq_phys and stq_phys. I'll use those.

I looked at DPRINTF but would need a DPRINTF_MMU and DPRINTF_MXCC ... Would that be acceptable?

OK

These are more for me to make sure I catch all the necessary accesses without adding unnecessary code. linux 2.4 and netbsd only use the ones I have implemented. openboot uses more and also some undocumented ones which are caught by this but I don't know how to handle them yet. I will use do_unassigned_access where necessary now.

Thanks for the review.