openwrt/target/linux/brcm47xx/patches-4.14/159-cpu_fixes.patch

--- a/arch/mips/include/asm/r4kcache.h
+++ b/arch/mips/include/asm/r4kcache.h
@@ -25,6 +25,38 @@
 extern void (*r4k_blast_dcache)(void);
 extern void (*r4k_blast_icache)(void);
 
+#if defined(CONFIG_BCM47XX) && !defined(CONFIG_CPU_MIPS32_R2)
+#include <asm/paccess.h>
+#include <linux/ssb/ssb.h>
+#define BCM4710_DUMMY_RREG() bcm4710_dummy_rreg()
+
+static inline unsigned long bcm4710_dummy_rreg(void)
+{
+      return *(volatile unsigned long *)(KSEG1ADDR(SSB_ENUM_BASE));
+}
+
+#define BCM4710_FILL_TLB(addr) bcm4710_fill_tlb((void *)(addr))
+
+static inline unsigned long bcm4710_fill_tlb(void *addr)
+{
+      return *(unsigned long *)addr;
+}
+
+#define BCM4710_PROTECTED_FILL_TLB(addr) bcm4710_protected_fill_tlb((void *)(addr))
+
+static inline void bcm4710_protected_fill_tlb(void *addr)
+{
+      unsigned long x;
+      get_dbe(x, (unsigned long *)addr);;
+}
+
+#else
+#define BCM4710_DUMMY_RREG()
+
+#define BCM4710_FILL_TLB(addr)
+#define BCM4710_PROTECTED_FILL_TLB(addr)
+#endif
+
 /*
  * This macro return a properly sign-extended address suitable as base address
  * for indexed cache operations.  Two issues here:
@@ -98,6 +130,7 @@ static inline void flush_icache_line_ind
 static inline void flush_dcache_line_indexed(unsigned long addr)
 {
 	__dflush_prologue
+	BCM4710_DUMMY_RREG();
 	cache_op(Index_Writeback_Inv_D, addr);
 	__dflush_epilogue
 }
@@ -125,6 +158,7 @@ static inline void flush_icache_line(uns
 static inline void flush_dcache_line(unsigned long addr)
 {
 	__dflush_prologue
+	BCM4710_DUMMY_RREG();
 	cache_op(Hit_Writeback_Inv_D, addr);
 	__dflush_epilogue
 }
@@ -132,6 +166,7 @@ static inline void flush_dcache_line(uns
 static inline void invalidate_dcache_line(unsigned long addr)
 {
 	__dflush_prologue
+	BCM4710_DUMMY_RREG();
 	cache_op(Hit_Invalidate_D, addr);
 	__dflush_epilogue
 }
@@ -205,6 +240,7 @@ static inline int protected_flush_icache
 #ifdef CONFIG_EVA
 		return protected_cachee_op(Hit_Invalidate_I, addr);
 #else
+		BCM4710_DUMMY_RREG();
 		return protected_cache_op(Hit_Invalidate_I, addr);
 #endif
 	}
@@ -218,6 +254,7 @@ static inline int protected_flush_icache
  */
 static inline int protected_writeback_dcache_line(unsigned long addr)
 {
+	BCM4710_DUMMY_RREG();
 #ifdef CONFIG_EVA
 	return protected_cachee_op(Hit_Writeback_Inv_D, addr);
 #else
@@ -575,8 +612,51 @@ static inline void invalidate_tcache_pag
 		: "r" (base),						\
 		  "i" (op));
 
+static inline void blast_dcache(void)
+{
+	unsigned long start = KSEG0;
+	unsigned long dcache_size = current_cpu_data.dcache.waysize * current_cpu_data.dcache.ways;
+	unsigned long end = (start + dcache_size);
+
+	do {
+		BCM4710_DUMMY_RREG();
+		cache_op(Index_Writeback_Inv_D, start);
+		start += current_cpu_data.dcache.linesz;
+	} while(start < end);
+}
+
+static inline void blast_dcache_page(unsigned long page)
+{
+	unsigned long start = page;
+	unsigned long end = start + PAGE_SIZE;
+
+	BCM4710_FILL_TLB(start);
+	do {
+		BCM4710_DUMMY_RREG();
+		cache_op(Hit_Writeback_Inv_D, start);
+		start += current_cpu_data.dcache.linesz;
+	} while(start < end);
+}
+
+static inline void blast_dcache_page_indexed(unsigned long page)
+{
+	unsigned long start = page;
+	unsigned long end = start + PAGE_SIZE;
+	unsigned long ws_inc = 1UL << current_cpu_data.dcache.waybit;
+	unsigned long ws_end = current_cpu_data.dcache.ways <<
+	                       current_cpu_data.dcache.waybit;
+	unsigned long ws, addr;
+	for (ws = 0; ws < ws_end; ws += ws_inc) {
+		start = page + ws;
+		for (addr = start; addr < end; addr += current_cpu_data.dcache.linesz) {
+			BCM4710_DUMMY_RREG();
+			cache_op(Index_Writeback_Inv_D, addr);
+		}
+	}
+}
+
 /* build blast_xxx, blast_xxx_page, blast_xxx_page_indexed */
-#define __BUILD_BLAST_CACHE(pfx, desc, indexop, hitop, lsize, extra)	\
+#define __BUILD_BLAST_CACHE(pfx, desc, indexop, hitop, lsize, extra, war) \
 static inline void extra##blast_##pfx##cache##lsize(void)		\
 {									\
 	unsigned long start = INDEX_BASE;				\
@@ -588,6 +668,7 @@ static inline void extra##blast_##pfx##c
 									\
 	__##pfx##flush_prologue						\
 									\
+	war								\
 	for (ws = 0; ws < ws_end; ws += ws_inc)				\
 		for (addr = start; addr < end; addr += lsize * 32)	\
 			cache##lsize##_unroll32(addr|ws, indexop);	\
@@ -602,6 +683,7 @@ static inline void extra##blast_##pfx##c
 									\
 	__##pfx##flush_prologue						\
 									\
+	war								\
 	do {								\
 		cache##lsize##_unroll32(start, hitop);			\
 		start += lsize * 32;					\
@@ -620,6 +702,8 @@ static inline void extra##blast_##pfx##c
 			       current_cpu_data.desc.waybit;		\
 	unsigned long ws, addr;						\
 									\
+	war								\
+									\
 	__##pfx##flush_prologue						\
 									\
 	for (ws = 0; ws < ws_end; ws += ws_inc)				\
@@ -629,26 +713,26 @@ static inline void extra##blast_##pfx##c
 	__##pfx##flush_epilogue						\
 }
 
-__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 16, )
-__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 16, )
-__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 16, )
-__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 32, )
-__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 32, )
-__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I_Loongson2, 32, loongson2_)
-__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 32, )
-__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 64, )
-__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64, )
-__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 64, )
-__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 128, )
-__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 128, )
-__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 128, )
-
-__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 16, )
-__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 32, )
-__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 16, )
-__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 32, )
-__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 64, )
-__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 128, )
+__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 16, , )
+__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 16, , BCM4710_FILL_TLB(start);)
+__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 16, , )
+__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 32, , )
+__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 32, , BCM4710_FILL_TLB(start);)
+__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I_Loongson2, 32, loongson2_, BCM4710_FILL_TLB(start);)
+__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 32, , )
+__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 64, , )
+__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64, , BCM4710_FILL_TLB(start);)
+__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 64, , )
+__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 128, , )
+__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 128, , )
+__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 128, , )
+
+__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 16, , )
+__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 32, , )
+__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 16, , )
+__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 32, , )
+__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 64, , )
+__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 128, , )
 
 #define __BUILD_BLAST_USER_CACHE(pfx, desc, indexop, hitop, lsize) \
 static inline void blast_##pfx##cache##lsize##_user_page(unsigned long page) \
@@ -677,53 +761,23 @@ __BUILD_BLAST_USER_CACHE(d, dcache, Inde
 __BUILD_BLAST_USER_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64)
 
 /* build blast_xxx_range, protected_blast_xxx_range */
-#define __BUILD_BLAST_CACHE_RANGE(pfx, desc, hitop, prot, extra)	\
+#define __BUILD_BLAST_CACHE_RANGE(pfx, desc, hitop, prot, extra, war, war2)	\
 static inline void prot##extra##blast_##pfx##cache##_range(unsigned long start, \
 						    unsigned long end)	\
 {									\
 	unsigned long lsize = cpu_##desc##_line_size();			\
-	unsigned long lsize_2 = lsize * 2;				\
-	unsigned long lsize_3 = lsize * 3;				\
-	unsigned long lsize_4 = lsize * 4;				\
-	unsigned long lsize_5 = lsize * 5;				\
-	unsigned long lsize_6 = lsize * 6;				\
-	unsigned long lsize_7 = lsize * 7;				\
-	unsigned long lsize_8 = lsize * 8;				\
 	unsigned long addr = start & ~(lsize - 1);			\
-	unsigned long aend = (end + lsize - 1) & ~(lsize - 1);		\
-	int lines = (aend - addr) / lsize;				\
+	unsigned long aend = (end - 1) & ~(lsize - 1);			\
+	war								\
 									\
 	__##pfx##flush_prologue						\
 									\
-	while (lines >= 8) {						\
-		prot##cache_op(hitop, addr);				\
-		prot##cache_op(hitop, addr + lsize);			\
-		prot##cache_op(hitop, addr + lsize_2);			\
-		prot##cache_op(hitop, addr + lsize_3);			\
-		prot##cache_op(hitop, addr + lsize_4);			\
-		prot##cache_op(hitop, addr + lsize_5);			\
-		prot##cache_op(hitop, addr + lsize_6);			\
-		prot##cache_op(hitop, addr + lsize_7);			\
-		addr += lsize_8;					\
-		lines -= 8;						\
-	}								\
-									\
-	if (lines & 0x4) {						\
-		prot##cache_op(hitop, addr);				\
-		prot##cache_op(hitop, addr + lsize);			\
-		prot##cache_op(hitop, addr + lsize_2);			\
-		prot##cache_op(hitop, addr + lsize_3);			\
-		addr += lsize_4;					\
-	}								\
-									\
-	if (lines & 0x2) {						\
-		prot##cache_op(hitop, addr);				\
-		prot##cache_op(hitop, addr + lsize);			\
-		addr += lsize_2;					\
-	}								\
-									\
-	if (lines & 0x1) {						\
+	while (1) {							\
+		war2							\
 		prot##cache_op(hitop, addr);				\
+		if (addr == aend)					\
+			break;						\
+		addr += lsize;						\
 	}								\
 									\
 	__##pfx##flush_epilogue						\
@@ -731,8 +785,8 @@ static inline void prot##extra##blast_##
 
 #ifndef CONFIG_EVA
 
-__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, protected_, )
-__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, protected_, )
+__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, protected_, , BCM4710_PROTECTED_FILL_TLB(addr); BCM4710_PROTECTED_FILL_TLB(aend);, BCM4710_DUMMY_RREG();)
+__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, protected_, , , )
 
 #else
 
@@ -769,14 +823,14 @@ __BUILD_PROT_BLAST_CACHE_RANGE(d, dcache
 __BUILD_PROT_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I)
 
 #endif
-__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, protected_, )
+__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, protected_, , , )
 __BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I_Loongson2, \
-	protected_, loongson2_)
-__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, , )
-__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, , )
-__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, , )
+	protected_, loongson2_, , )
+__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, , , BCM4710_FILL_TLB(addr); BCM4710_FILL_TLB(aend);, BCM4710_DUMMY_RREG();)
+__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, , , , )
+__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, , , , )
 /* blast_inv_dcache_range */
-__BUILD_BLAST_CACHE_RANGE(inv_d, dcache, Hit_Invalidate_D, , )
-__BUILD_BLAST_CACHE_RANGE(inv_s, scache, Hit_Invalidate_SD, , )
+__BUILD_BLAST_CACHE_RANGE(inv_d, dcache, Hit_Invalidate_D, , , , BCM4710_DUMMY_RREG();)
+__BUILD_BLAST_CACHE_RANGE(inv_s, scache, Hit_Invalidate_SD, , , , )
 
 #endif /* _ASM_R4KCACHE_H */
--- a/arch/mips/include/asm/stackframe.h
+++ b/arch/mips/include/asm/stackframe.h
@@ -428,6 +428,10 @@
 		eretnc
 #else
 		.set	arch=r4000
+#ifdef CONFIG_BCM47XX
+		nop
+		nop
+#endif
 		eret
 		.set	mips0
 #endif
--- a/arch/mips/kernel/genex.S
+++ b/arch/mips/kernel/genex.S
@@ -21,6 +21,19 @@
 #include <asm/war.h>
 #include <asm/thread_info.h>
 
+#ifdef CONFIG_BCM47XX
+# ifdef eret
+#  undef eret
+# endif
+# define eret 					\
+	.set push;				\
+	.set noreorder;				\
+	 nop; 					\
+	 nop;					\
+	 eret;					\
+	.set pop;
+#endif
+
 	__INIT
 
 /*
@@ -32,6 +45,9 @@
 NESTED(except_vec3_generic, 0, sp)
 	.set	push
 	.set	noat
+#ifdef CONFIG_BCM47XX
+	nop
+#endif
 #if R5432_CP0_INTERRUPT_WAR
 	mfc0	k0, CP0_INDEX
 #endif
@@ -55,6 +71,9 @@ NESTED(except_vec3_r4000, 0, sp)
 	.set	push
 	.set	arch=r4000
 	.set	noat
+#ifdef CONFIG_BCM47XX
+	nop
+#endif
 	mfc0	k1, CP0_CAUSE
 	li	k0, 31<<2
 	andi	k1, k1, 0x7c
--- a/arch/mips/mm/c-r4k.c
+++ b/arch/mips/mm/c-r4k.c
@@ -39,6 +39,9 @@
 #include <asm/dma-coherence.h>
 #include <asm/mips-cps.h>
 
+/* For enabling BCM4710 cache workarounds */
+static int bcm4710 = 0;
+
 /*
  * Bits describing what cache ops an SMP callback function may perform.
  *
@@ -190,6 +193,9 @@ static void r4k_blast_dcache_user_page_s
 {
 	unsigned long  dc_lsize = cpu_dcache_line_size();
 
+	if (bcm4710)
+		r4k_blast_dcache_page = blast_dcache_page;
+	else
 	if (dc_lsize == 0)
 		r4k_blast_dcache_user_page = (void *)cache_noop;
 	else if (dc_lsize == 16)
@@ -208,6 +214,9 @@ static void r4k_blast_dcache_page_indexe
 {
 	unsigned long dc_lsize = cpu_dcache_line_size();
 
+	if (bcm4710)
+		r4k_blast_dcache_page_indexed = blast_dcache_page_indexed;
+	else
 	if (dc_lsize == 0)
 		r4k_blast_dcache_page_indexed = (void *)cache_noop;
 	else if (dc_lsize == 16)
@@ -227,6 +236,9 @@ static void r4k_blast_dcache_setup(void)
 {
 	unsigned long dc_lsize = cpu_dcache_line_size();
 
+	if (bcm4710)
+		r4k_blast_dcache = blast_dcache;
+	else
 	if (dc_lsize == 0)
 		r4k_blast_dcache = (void *)cache_noop;
 	else if (dc_lsize == 16)
@@ -952,6 +964,8 @@ static void local_r4k_flush_cache_sigtra
 	}
 
 	R4600_HIT_CACHEOP_WAR_IMPL;
+	BCM4710_PROTECTED_FILL_TLB(addr);
+	BCM4710_PROTECTED_FILL_TLB(addr + 4);
 	if (!cpu_has_ic_fills_f_dc) {
 		if (dc_lsize)
 			vaddr ? flush_dcache_line(addr & ~(dc_lsize - 1))
@@ -1846,6 +1860,17 @@ static void coherency_setup(void)
 	 * silly idea of putting something else there ...
 	 */
 	switch (current_cpu_type()) {
+	case CPU_BMIPS3300:
+		{
+			u32 cm;
+			cm = read_c0_diag();
+			/* Enable icache */
+			cm |= (1 << 31);
+			/* Enable dcache */
+			cm |= (1 << 30);
+			write_c0_diag(cm);
+		}
+		break;
 	case CPU_R4000PC:
 	case CPU_R4000SC:
 	case CPU_R4000MC:
@@ -1892,6 +1917,15 @@ void r4k_cache_init(void)
 	extern void build_copy_page(void);
 	struct cpuinfo_mips *c = &current_cpu_data;
 
+	/* Check if special workarounds are required */
+#if defined(CONFIG_BCM47XX) && !defined(CONFIG_CPU_MIPS32_R2)
+	if (current_cpu_data.cputype == CPU_BMIPS32 && (current_cpu_data.processor_id & 0xff) == 0) {
+		printk("Enabling BCM4710A0 cache workarounds.\n");
+		bcm4710 = 1;
+	} else
+#endif
+		bcm4710 = 0;
+
 	probe_pcache();
 	probe_vcache();
 	setup_scache();
@@ -1969,7 +2003,15 @@ void r4k_cache_init(void)
 	 */
 	local_r4k___flush_cache_all(NULL);
 
+#ifdef CONFIG_BCM47XX
+	{
+		static void (*_coherency_setup)(void);
+		_coherency_setup = (void (*)(void)) KSEG1ADDR(coherency_setup);
+		_coherency_setup();
+	}
+#else
 	coherency_setup();
+#endif
 	board_cache_error_setup = r4k_cache_error_setup;
 
 	/*
--- a/arch/mips/mm/tlbex.c
+++ b/arch/mips/mm/tlbex.c
@@ -971,6 +971,9 @@ void build_get_pgde32(u32 **p, unsigned
 		uasm_i_srl(p, ptr, ptr, SMP_CPUID_PTRSHIFT);
 		uasm_i_addu(p, ptr, tmp, ptr);
 #else
+#ifdef CONFIG_BCM47XX
+		uasm_i_nop(p);
+#endif
 		UASM_i_LA_mostly(p, ptr, pgdc);
 #endif
 		uasm_i_mfc0(p, tmp, C0_BADVADDR); /* get faulting address */
@@ -1331,6 +1334,9 @@ static void build_r4000_tlb_refill_handl
 #ifdef CONFIG_64BIT
 		build_get_pmde64(&p, &l, &r, K0, K1); /* get pmd in K1 */
 #else
+# ifdef CONFIG_BCM47XX
+		uasm_i_nop(&p);
+# endif
 		build_get_pgde32(&p, K0, K1); /* get pgd in K1 */
 #endif
 
@@ -1342,6 +1348,9 @@ static void build_r4000_tlb_refill_handl
 		build_update_entries(&p, K0, K1);
 		build_tlb_write_entry(&p, &l, &r, tlb_random);
 		uasm_l_leave(&l, p);
+#ifdef CONFIG_BCM47XX
+		uasm_i_nop(&p);
+#endif
 		uasm_i_eret(&p); /* return from trap */
 	}
 #ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT
@@ -2046,6 +2055,9 @@ build_r4000_tlbchange_handler_head(u32 *
 #ifdef CONFIG_64BIT
 	build_get_pmde64(p, l, r, wr.r1, wr.r2); /* get pmd in ptr */
 #else
+# ifdef CONFIG_BCM47XX
+	uasm_i_nop(p);
+# endif
 	build_get_pgde32(p, wr.r1, wr.r2); /* get pgd in ptr */
 #endif
 
@@ -2092,6 +2104,9 @@ build_r4000_tlbchange_handler_tail(u32 *
 	build_tlb_write_entry(p, l, r, tlb_indexed);
 	uasm_l_leave(l, *p);
 	build_restore_work_registers(p);
+#ifdef CONFIG_BCM47XX
+	uasm_i_nop(p);
+#endif
 	uasm_i_eret(p); /* return from trap */
 
 #ifdef CONFIG_64BIT