diff --git a/cpu-defs.h b/cpu-defs.h
index 5e0f04674e6c01c0a2b1e7fad6e7da9d07825fe9..b581d949f27c6198d26ced0d28bc1558dfd902f9 100644
--- a/cpu-defs.h
+++ b/cpu-defs.h
@@ -102,6 +102,12 @@ typedef unsigned long ram_addr_t;
 #define CPU_TLB_BITS 8
 #define CPU_TLB_SIZE (1 << CPU_TLB_BITS)
 
+#if TARGET_PHYS_ADDR_BITS == 32 && TARGET_LONG_BITS == 32
+#define CPU_TLB_ENTRY_BITS 4
+#else
+#define CPU_TLB_ENTRY_BITS 5
+#endif
+
 typedef struct CPUTLBEntry {
     /* bit 31 to TARGET_PAGE_BITS : virtual address
        bit TARGET_PAGE_BITS-1..IO_MEM_SHIFT : if non zero, memory io
@@ -113,7 +119,17 @@ typedef struct CPUTLBEntry {
     target_ulong addr_write;
     target_ulong addr_code;
     /* addend to virtual address to get physical address */
+#if TARGET_PHYS_ADDR_BITS == 64
+    /* on i386 Linux make sure it is aligned */
+    target_phys_addr_t addend __attribute__((aligned(8)));
+#else
     target_phys_addr_t addend;
+#endif
+    /* padding to get a power of two size */
+    uint8_t dummy[(1 << CPU_TLB_ENTRY_BITS) - 
+                  (sizeof(target_ulong) * 3 + 
+                   ((-sizeof(target_ulong) * 3) & (sizeof(target_phys_addr_t) - 1)) + 
+                   sizeof(target_phys_addr_t))];
 } CPUTLBEntry;
 
 #define CPU_COMMON                                                      \
diff --git a/osdep.h b/osdep.h
index bc513ad3d5279e93b9c095306245390b281df6ca..0666e78b3ffff6a674dc88e584c860d62be52eac 100644
--- a/osdep.h
+++ b/osdep.h
@@ -36,9 +36,9 @@
 #define inline always_inline
 
 #ifdef __i386__
-#define REGPARM(n) __attribute((regparm(n)))
+#define REGPARM __attribute((regparm(3)))
 #else
-#define REGPARM(n)
+#define REGPARM
 #endif
 
 #define qemu_printf printf
diff --git a/softmmu_header.h b/softmmu_header.h
index fbf22b160364c59b0b71c10b0e741e48fb5f7035..51bd22d96f674ce7748dd337a834e7ca9633e803 100644
--- a/softmmu_header.h
+++ b/softmmu_header.h
@@ -70,15 +70,13 @@
 #define ADDR_READ addr_read
 #endif
 
-DATA_TYPE REGPARM(1) glue(glue(__ld, SUFFIX), MMUSUFFIX)(target_ulong addr,
+DATA_TYPE REGPARM glue(glue(__ld, SUFFIX), MMUSUFFIX)(target_ulong addr,
                                                          int mmu_idx);
-void REGPARM(2) glue(glue(__st, SUFFIX), MMUSUFFIX)(target_ulong addr, DATA_TYPE v, int mmu_idx);
+void REGPARM glue(glue(__st, SUFFIX), MMUSUFFIX)(target_ulong addr, DATA_TYPE v, int mmu_idx);
 
 #if (DATA_SIZE <= 4) && (TARGET_LONG_BITS == 32) && defined(__i386__) && \
     (ACCESS_TYPE < NB_MMU_MODES) && defined(ASM_SOFTMMU)
 
-#define CPU_TLB_ENTRY_BITS 4
-
 static inline RES_TYPE glue(glue(ld, USUFFIX), MEMSUFFIX)(target_ulong ptr)
 {
     int res;
@@ -92,9 +90,8 @@ static inline RES_TYPE glue(glue(ld, USUFFIX), MEMSUFFIX)(target_ulong ptr)
                   "cmpl (%%edx), %%eax\n"
                   "movl %1, %%eax\n"
                   "je 1f\n"
-                  "pushl %6\n"
+                  "movl %6, %%edx\n"
                   "call %7\n"
-                  "popl %%edx\n"
                   "movl %%eax, %0\n"
                   "jmp 2f\n"
                   "1:\n"
@@ -135,9 +132,8 @@ static inline int glue(glue(lds, SUFFIX), MEMSUFFIX)(target_ulong ptr)
                   "cmpl (%%edx), %%eax\n"
                   "movl %1, %%eax\n"
                   "je 1f\n"
-                  "pushl %6\n"
+                  "movl %6, %%edx\n"
                   "call %7\n"
-                  "popl %%edx\n"
 #if DATA_SIZE == 1
                   "movsbl %%al, %0\n"
 #elif DATA_SIZE == 2
@@ -189,9 +185,8 @@ static inline void glue(glue(st, SUFFIX), MEMSUFFIX)(target_ulong ptr, RES_TYPE
 #else
 #error unsupported size
 #endif
-                  "pushl %6\n"
+                  "movl %6, %%ecx\n"
                   "call %7\n"
-                  "popl %%eax\n"
                   "jmp 2f\n"
                   "1:\n"
                   "addl 8(%%edx), %%eax\n"
diff --git a/softmmu_template.h b/softmmu_template.h
index 45fcd4e1b6c6082e2a4c8a5b09dcdf4e4e384327..0a4bc7e0cd0d626a2421b00ec96fd3761b80cc30 100644
--- a/softmmu_template.h
+++ b/softmmu_template.h
@@ -75,8 +75,8 @@ static inline DATA_TYPE glue(io_read, SUFFIX)(target_phys_addr_t physaddr,
 }
 
 /* handle all cases except unaligned access which span two pages */
-DATA_TYPE REGPARM(1) glue(glue(__ld, SUFFIX), MMUSUFFIX)(target_ulong addr,
-                                                         int mmu_idx)
+DATA_TYPE REGPARM glue(glue(__ld, SUFFIX), MMUSUFFIX)(target_ulong addr,
+                                                      int mmu_idx)
 {
     DATA_TYPE res;
     int index;
@@ -209,9 +209,9 @@ static inline void glue(io_write, SUFFIX)(target_phys_addr_t physaddr,
 #endif
 }
 
-void REGPARM(2) glue(glue(__st, SUFFIX), MMUSUFFIX)(target_ulong addr,
-                                                    DATA_TYPE val,
-                                                    int mmu_idx)
+void REGPARM glue(glue(__st, SUFFIX), MMUSUFFIX)(target_ulong addr,
+                                                 DATA_TYPE val,
+                                                 int mmu_idx)
 {
     target_phys_addr_t physaddr;
     target_ulong tlb_addr;