kernel: update kernel 4.4 to 4.4.61
[lede.git] / target / linux / brcm47xx / patches-4.4 / 159-cpu_fixes.patch
1 --- a/arch/mips/include/asm/r4kcache.h
2 +++ b/arch/mips/include/asm/r4kcache.h
3 @@ -25,6 +25,38 @@
4  extern void (*r4k_blast_dcache)(void);
5  extern void (*r4k_blast_icache)(void);
6  
7 +#if defined(CONFIG_BCM47XX) && !defined(CONFIG_CPU_MIPS32_R2)
8 +#include <asm/paccess.h>
9 +#include <linux/ssb/ssb.h>
10 +#define BCM4710_DUMMY_RREG() bcm4710_dummy_rreg()
11 +
12 +static inline unsigned long bcm4710_dummy_rreg(void)
13 +{
14 +      return *(volatile unsigned long *)(KSEG1ADDR(SSB_ENUM_BASE));
15 +}
16 +
17 +#define BCM4710_FILL_TLB(addr) bcm4710_fill_tlb((void *)(addr))
18 +
19 +static inline unsigned long bcm4710_fill_tlb(void *addr)
20 +{
21 +      return *(unsigned long *)addr;
22 +}
23 +
24 +#define BCM4710_PROTECTED_FILL_TLB(addr) bcm4710_protected_fill_tlb((void *)(addr))
25 +
26 +static inline void bcm4710_protected_fill_tlb(void *addr)
27 +{
28 +      unsigned long x;
29 +      get_dbe(x, (unsigned long *)addr);;
30 +}
31 +
32 +#else
33 +#define BCM4710_DUMMY_RREG()
34 +
35 +#define BCM4710_FILL_TLB(addr)
36 +#define BCM4710_PROTECTED_FILL_TLB(addr)
37 +#endif
38 +
39  /*
40   * This macro return a properly sign-extended address suitable as base address
41   * for indexed cache operations.  Two issues here:
42 @@ -98,6 +130,7 @@ static inline void flush_icache_line_ind
43  static inline void flush_dcache_line_indexed(unsigned long addr)
44  {
45         __dflush_prologue
46 +       BCM4710_DUMMY_RREG();
47         cache_op(Index_Writeback_Inv_D, addr);
48         __dflush_epilogue
49  }
50 @@ -125,6 +158,7 @@ static inline void flush_icache_line(uns
51  static inline void flush_dcache_line(unsigned long addr)
52  {
53         __dflush_prologue
54 +       BCM4710_DUMMY_RREG();
55         cache_op(Hit_Writeback_Inv_D, addr);
56         __dflush_epilogue
57  }
58 @@ -132,6 +166,7 @@ static inline void flush_dcache_line(uns
59  static inline void invalidate_dcache_line(unsigned long addr)
60  {
61         __dflush_prologue
62 +       BCM4710_DUMMY_RREG();
63         cache_op(Hit_Invalidate_D, addr);
64         __dflush_epilogue
65  }
66 @@ -187,6 +222,7 @@ static inline void protected_flush_icach
67  #ifdef CONFIG_EVA
68                 protected_cachee_op(Hit_Invalidate_I, addr);
69  #else
70 +               BCM4710_DUMMY_RREG();
71                 protected_cache_op(Hit_Invalidate_I, addr);
72  #endif
73                 break;
74 @@ -201,6 +237,7 @@ static inline void protected_flush_icach
75   */
76  static inline void protected_writeback_dcache_line(unsigned long addr)
77  {
78 +       BCM4710_DUMMY_RREG();
79  #ifdef CONFIG_EVA
80         protected_cachee_op(Hit_Writeback_Inv_D, addr);
81  #else
82 @@ -554,8 +591,51 @@ static inline void invalidate_tcache_pag
83                 : "r" (base),                                           \
84                   "i" (op));
85  
86 +static inline void blast_dcache(void)
87 +{
88 +       unsigned long start = KSEG0;
89 +       unsigned long dcache_size = current_cpu_data.dcache.waysize * current_cpu_data.dcache.ways;
90 +       unsigned long end = (start + dcache_size);
91 +
92 +       do {
93 +               BCM4710_DUMMY_RREG();
94 +               cache_op(Index_Writeback_Inv_D, start);
95 +               start += current_cpu_data.dcache.linesz;
96 +       } while(start < end);
97 +}
98 +
99 +static inline void blast_dcache_page(unsigned long page)
100 +{
101 +       unsigned long start = page;
102 +       unsigned long end = start + PAGE_SIZE;
103 +
104 +       BCM4710_FILL_TLB(start);
105 +       do {
106 +               BCM4710_DUMMY_RREG();
107 +               cache_op(Hit_Writeback_Inv_D, start);
108 +               start += current_cpu_data.dcache.linesz;
109 +       } while(start < end);
110 +}
111 +
112 +static inline void blast_dcache_page_indexed(unsigned long page)
113 +{
114 +       unsigned long start = page;
115 +       unsigned long end = start + PAGE_SIZE;
116 +       unsigned long ws_inc = 1UL << current_cpu_data.dcache.waybit;
117 +       unsigned long ws_end = current_cpu_data.dcache.ways <<
118 +                              current_cpu_data.dcache.waybit;
119 +       unsigned long ws, addr;
120 +       for (ws = 0; ws < ws_end; ws += ws_inc) {
121 +               start = page + ws;
122 +               for (addr = start; addr < end; addr += current_cpu_data.dcache.linesz) {
123 +                       BCM4710_DUMMY_RREG();
124 +                       cache_op(Index_Writeback_Inv_D, addr);
125 +               }
126 +       }
127 +}
128 +
129  /* build blast_xxx, blast_xxx_page, blast_xxx_page_indexed */
130 -#define __BUILD_BLAST_CACHE(pfx, desc, indexop, hitop, lsize, extra)   \
131 +#define __BUILD_BLAST_CACHE(pfx, desc, indexop, hitop, lsize, extra, war) \
132  static inline void extra##blast_##pfx##cache##lsize(void)              \
133  {                                                                      \
134         unsigned long start = INDEX_BASE;                               \
135 @@ -567,6 +647,7 @@ static inline void extra##blast_##pfx##c
136                                                                         \
137         __##pfx##flush_prologue                                         \
138                                                                         \
139 +       war                                                             \
140         for (ws = 0; ws < ws_end; ws += ws_inc)                         \
141                 for (addr = start; addr < end; addr += lsize * 32)      \
142                         cache##lsize##_unroll32(addr|ws, indexop);      \
143 @@ -581,6 +662,7 @@ static inline void extra##blast_##pfx##c
144                                                                         \
145         __##pfx##flush_prologue                                         \
146                                                                         \
147 +       war                                                             \
148         do {                                                            \
149                 cache##lsize##_unroll32(start, hitop);                  \
150                 start += lsize * 32;                                    \
151 @@ -599,6 +681,8 @@ static inline void extra##blast_##pfx##c
152                                current_cpu_data.desc.waybit;            \
153         unsigned long ws, addr;                                         \
154                                                                         \
155 +       war                                                             \
156 +                                                                       \
157         __##pfx##flush_prologue                                         \
158                                                                         \
159         for (ws = 0; ws < ws_end; ws += ws_inc)                         \
160 @@ -608,26 +692,26 @@ static inline void extra##blast_##pfx##c
161         __##pfx##flush_epilogue                                         \
162  }
163  
164 -__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 16, )
165 -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 16, )
166 -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 16, )
167 -__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 32, )
168 -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 32, )
169 -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I_Loongson2, 32, loongson2_)
170 -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 32, )
171 -__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 64, )
172 -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64, )
173 -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 64, )
174 -__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 128, )
175 -__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 128, )
176 -__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 128, )
177 -
178 -__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 16, )
179 -__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 32, )
180 -__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 16, )
181 -__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 32, )
182 -__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 64, )
183 -__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 128, )
184 +__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 16, , )
185 +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 16, , BCM4710_FILL_TLB(start);)
186 +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 16, , )
187 +__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 32, , )
188 +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 32, , BCM4710_FILL_TLB(start);)
189 +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I_Loongson2, 32, loongson2_, BCM4710_FILL_TLB(start);)
190 +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 32, , )
191 +__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 64, , )
192 +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64, , BCM4710_FILL_TLB(start);)
193 +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 64, , )
194 +__BUILD_BLAST_CACHE(d, dcache, Index_Writeback_Inv_D, Hit_Writeback_Inv_D, 128, , )
195 +__BUILD_BLAST_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 128, , )
196 +__BUILD_BLAST_CACHE(s, scache, Index_Writeback_Inv_SD, Hit_Writeback_Inv_SD, 128, , )
197 +
198 +__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 16, , )
199 +__BUILD_BLAST_CACHE(inv_d, dcache, Index_Writeback_Inv_D, Hit_Invalidate_D, 32, , )
200 +__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 16, , )
201 +__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 32, , )
202 +__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 64, , )
203 +__BUILD_BLAST_CACHE(inv_s, scache, Index_Writeback_Inv_SD, Hit_Invalidate_SD, 128, , )
204  
205  #define __BUILD_BLAST_USER_CACHE(pfx, desc, indexop, hitop, lsize) \
206  static inline void blast_##pfx##cache##lsize##_user_page(unsigned long page) \
207 @@ -656,17 +740,19 @@ __BUILD_BLAST_USER_CACHE(d, dcache, Inde
208  __BUILD_BLAST_USER_CACHE(i, icache, Index_Invalidate_I, Hit_Invalidate_I, 64)
209  
210  /* build blast_xxx_range, protected_blast_xxx_range */
211 -#define __BUILD_BLAST_CACHE_RANGE(pfx, desc, hitop, prot, extra)       \
212 +#define __BUILD_BLAST_CACHE_RANGE(pfx, desc, hitop, prot, extra, war, war2)    \
213  static inline void prot##extra##blast_##pfx##cache##_range(unsigned long start, \
214                                                     unsigned long end)  \
215  {                                                                      \
216         unsigned long lsize = cpu_##desc##_line_size();                 \
217         unsigned long addr = start & ~(lsize - 1);                      \
218         unsigned long aend = (end - 1) & ~(lsize - 1);                  \
219 +       war                                                             \
220                                                                         \
221         __##pfx##flush_prologue                                         \
222                                                                         \
223         while (1) {                                                     \
224 +               war2                                                    \
225                 prot##cache_op(hitop, addr);                            \
226                 if (addr == aend)                                       \
227                         break;                                          \
228 @@ -678,8 +764,8 @@ static inline void prot##extra##blast_##
229  
230  #ifndef CONFIG_EVA
231  
232 -__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, protected_, )
233 -__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, protected_, )
234 +__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, protected_, , BCM4710_PROTECTED_FILL_TLB(addr); BCM4710_PROTECTED_FILL_TLB(aend);, BCM4710_DUMMY_RREG();)
235 +__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, protected_, , , )
236  
237  #else
238  
239 @@ -716,14 +802,14 @@ __BUILD_PROT_BLAST_CACHE_RANGE(d, dcache
240  __BUILD_PROT_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I)
241  
242  #endif
243 -__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, protected_, )
244 +__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, protected_, , , )
245  __BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I_Loongson2, \
246 -       protected_, loongson2_)
247 -__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, , )
248 -__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, , )
249 -__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, , )
250 +       protected_, loongson2_, , )
251 +__BUILD_BLAST_CACHE_RANGE(d, dcache, Hit_Writeback_Inv_D, , , BCM4710_FILL_TLB(addr); BCM4710_FILL_TLB(aend);, BCM4710_DUMMY_RREG();)
252 +__BUILD_BLAST_CACHE_RANGE(i, icache, Hit_Invalidate_I, , , , )
253 +__BUILD_BLAST_CACHE_RANGE(s, scache, Hit_Writeback_Inv_SD, , , , )
254  /* blast_inv_dcache_range */
255 -__BUILD_BLAST_CACHE_RANGE(inv_d, dcache, Hit_Invalidate_D, , )
256 -__BUILD_BLAST_CACHE_RANGE(inv_s, scache, Hit_Invalidate_SD, , )
257 +__BUILD_BLAST_CACHE_RANGE(inv_d, dcache, Hit_Invalidate_D, , , , BCM4710_DUMMY_RREG();)
258 +__BUILD_BLAST_CACHE_RANGE(inv_s, scache, Hit_Invalidate_SD, , , , )
259  
260  #endif /* _ASM_R4KCACHE_H */
261 --- a/arch/mips/include/asm/stackframe.h
262 +++ b/arch/mips/include/asm/stackframe.h
263 @@ -365,6 +365,10 @@
264                 .macro  RESTORE_SP_AND_RET
265                 LONG_L  sp, PT_R29(sp)
266                 .set    arch=r4000
267 +#ifdef CONFIG_BCM47XX
268 +               nop
269 +               nop
270 +#endif
271                 eret
272                 .set    mips0
273                 .endm
274 --- a/arch/mips/kernel/genex.S
275 +++ b/arch/mips/kernel/genex.S
276 @@ -21,6 +21,19 @@
277  #include <asm/war.h>
278  #include <asm/thread_info.h>
279  
280 +#ifdef CONFIG_BCM47XX
281 +# ifdef eret
282 +#  undef eret
283 +# endif
284 +# define eret                                  \
285 +       .set push;                              \
286 +       .set noreorder;                         \
287 +        nop;                                   \
288 +        nop;                                   \
289 +        eret;                                  \
290 +       .set pop;
291 +#endif
292 +
293         __INIT
294  
295  /*
296 @@ -32,6 +45,9 @@
297  NESTED(except_vec3_generic, 0, sp)
298         .set    push
299         .set    noat
300 +#ifdef CONFIG_BCM47XX
301 +       nop
302 +#endif
303  #if R5432_CP0_INTERRUPT_WAR
304         mfc0    k0, CP0_INDEX
305  #endif
306 @@ -55,6 +71,9 @@ NESTED(except_vec3_r4000, 0, sp)
307         .set    push
308         .set    arch=r4000
309         .set    noat
310 +#ifdef CONFIG_BCM47XX
311 +       nop
312 +#endif
313         mfc0    k1, CP0_CAUSE
314         li      k0, 31<<2
315         andi    k1, k1, 0x7c
316 --- a/arch/mips/mm/c-r4k.c
317 +++ b/arch/mips/mm/c-r4k.c
318 @@ -39,6 +39,9 @@
319  #include <asm/dma-coherence.h>
320  #include <asm/mips-cm.h>
321  
322 +/* For enabling BCM4710 cache workarounds */
323 +static int bcm4710 = 0;
324 +
325  /*
326   * Bits describing what cache ops an IPI callback function may perform.
327   *
328 @@ -202,6 +205,9 @@ static void r4k_blast_dcache_user_page_s
329  {
330         unsigned long  dc_lsize = cpu_dcache_line_size();
331  
332 +       if (bcm4710)
333 +               r4k_blast_dcache_page = blast_dcache_page;
334 +       else
335         if (dc_lsize == 0)
336                 r4k_blast_dcache_user_page = (void *)cache_noop;
337         else if (dc_lsize == 16)
338 @@ -220,6 +226,9 @@ static void r4k_blast_dcache_page_indexe
339  {
340         unsigned long dc_lsize = cpu_dcache_line_size();
341  
342 +       if (bcm4710)
343 +               r4k_blast_dcache_page_indexed = blast_dcache_page_indexed;
344 +       else
345         if (dc_lsize == 0)
346                 r4k_blast_dcache_page_indexed = (void *)cache_noop;
347         else if (dc_lsize == 16)
348 @@ -239,6 +248,9 @@ static void r4k_blast_dcache_setup(void)
349  {
350         unsigned long dc_lsize = cpu_dcache_line_size();
351  
352 +       if (bcm4710)
353 +               r4k_blast_dcache = blast_dcache;
354 +       else
355         if (dc_lsize == 0)
356                 r4k_blast_dcache = (void *)cache_noop;
357         else if (dc_lsize == 16)
358 @@ -880,6 +892,8 @@ static void local_r4k_flush_cache_sigtra
359         unsigned long addr = (unsigned long) arg;
360  
361         R4600_HIT_CACHEOP_WAR_IMPL;
362 +       BCM4710_PROTECTED_FILL_TLB(addr);
363 +       BCM4710_PROTECTED_FILL_TLB(addr + 4);
364         if (dc_lsize)
365                 protected_writeback_dcache_line(addr & ~(dc_lsize - 1));
366         if (!cpu_icache_snoops_remote_store && scache_size)
367 @@ -1704,6 +1718,17 @@ static void coherency_setup(void)
368          * silly idea of putting something else there ...
369          */
370         switch (current_cpu_type()) {
371 +       case CPU_BMIPS3300:
372 +               {
373 +                       u32 cm;
374 +                       cm = read_c0_diag();
375 +                       /* Enable icache */
376 +                       cm |= (1 << 31);
377 +                       /* Enable dcache */
378 +                       cm |= (1 << 30);
379 +                       write_c0_diag(cm);
380 +               }
381 +               break;
382         case CPU_R4000PC:
383         case CPU_R4000SC:
384         case CPU_R4000MC:
385 @@ -1750,6 +1775,15 @@ void r4k_cache_init(void)
386         extern void build_copy_page(void);
387         struct cpuinfo_mips *c = &current_cpu_data;
388  
389 +       /* Check if special workarounds are required */
390 +#if defined(CONFIG_BCM47XX) && !defined(CONFIG_CPU_MIPS32_R2)
391 +       if (current_cpu_data.cputype == CPU_BMIPS32 && (current_cpu_data.processor_id & 0xff) == 0) {
392 +               printk("Enabling BCM4710A0 cache workarounds.\n");
393 +               bcm4710 = 1;
394 +       } else
395 +#endif
396 +               bcm4710 = 0;
397 +
398         probe_pcache();
399         setup_scache();
400  
401 @@ -1819,7 +1853,15 @@ void r4k_cache_init(void)
402          */
403         local_r4k___flush_cache_all(NULL);
404  
405 +#ifdef CONFIG_BCM47XX
406 +       {
407 +               static void (*_coherency_setup)(void);
408 +               _coherency_setup = (void (*)(void)) KSEG1ADDR(coherency_setup);
409 +               _coherency_setup();
410 +       }
411 +#else
412         coherency_setup();
413 +#endif
414         board_cache_error_setup = r4k_cache_error_setup;
415  
416         /*
417 --- a/arch/mips/mm/tlbex.c
418 +++ b/arch/mips/mm/tlbex.c
419 @@ -960,6 +960,9 @@ build_get_pgde32(u32 **p, unsigned int t
420                 uasm_i_srl(p, ptr, ptr, SMP_CPUID_PTRSHIFT);
421                 uasm_i_addu(p, ptr, tmp, ptr);
422  #else
423 +#ifdef CONFIG_BCM47XX
424 +               uasm_i_nop(p);
425 +#endif
426                 UASM_i_LA_mostly(p, ptr, pgdc);
427  #endif
428                 uasm_i_mfc0(p, tmp, C0_BADVADDR); /* get faulting address */
429 @@ -1301,6 +1304,9 @@ static void build_r4000_tlb_refill_handl
430  #ifdef CONFIG_64BIT
431                 build_get_pmde64(&p, &l, &r, K0, K1); /* get pmd in K1 */
432  #else
433 +# ifdef CONFIG_BCM47XX
434 +               uasm_i_nop(&p);
435 +# endif
436                 build_get_pgde32(&p, K0, K1); /* get pgd in K1 */
437  #endif
438  
439 @@ -1312,6 +1318,9 @@ static void build_r4000_tlb_refill_handl
440                 build_update_entries(&p, K0, K1);
441                 build_tlb_write_entry(&p, &l, &r, tlb_random);
442                 uasm_l_leave(&l, p);
443 +#ifdef CONFIG_BCM47XX
444 +               uasm_i_nop(&p);
445 +#endif
446                 uasm_i_eret(&p); /* return from trap */
447         }
448  #ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT
449 @@ -1888,6 +1897,9 @@ build_r4000_tlbchange_handler_head(u32 *
450  #ifdef CONFIG_64BIT
451         build_get_pmde64(p, l, r, wr.r1, wr.r2); /* get pmd in ptr */
452  #else
453 +# ifdef CONFIG_BCM47XX
454 +       uasm_i_nop(p);
455 +# endif
456         build_get_pgde32(p, wr.r1, wr.r2); /* get pgd in ptr */
457  #endif
458  
459 @@ -1934,6 +1946,9 @@ build_r4000_tlbchange_handler_tail(u32 *
460         build_tlb_write_entry(p, l, r, tlb_indexed);
461         uasm_l_leave(l, *p);
462         build_restore_work_registers(p);
463 +#ifdef CONFIG_BCM47XX
464 +       uasm_i_nop(p);
465 +#endif
466         uasm_i_eret(p); /* return from trap */
467  
468  #ifdef CONFIG_64BIT