10 files changed, 122 insertions, 10 deletions

diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
index 624c31cd8077..c8d954d6f2c4 100644
--- a/arch/mips/Kconfig
+++ b/arch/mips/Kconfig
@@ -894,6 +894,16 @@ choice
         prompt "CPU type"
         default CPU_R4X00
 
+config CPU_LOONGSON2
+        bool "Loongson 2"
+        depends on SYS_HAS_CPU_LOONGSON2
+        select CPU_SUPPORTS_32BIT_KERNEL
+        select CPU_SUPPORTS_64BIT_KERNEL
+        select CPU_SUPPORTS_HIGHMEM
+        help
+          The Loongson 2E processor implements the MIPS III instruction set
+          with many extensions.
+
 config CPU_MIPS32_R1
         bool "MIPS32 Release 1"
         depends on SYS_HAS_CPU_MIPS32_R1
@@ -1104,6 +1114,9 @@ config CPU_SB1
 
 endchoice
 
+config SYS_HAS_CPU_LOONGSON2
+        bool
+
 config SYS_HAS_CPU_MIPS32_R1
         bool
 
@@ -1438,6 +1451,15 @@ config CPU_HAS_SMARTMIPS
 config CPU_HAS_WB
         bool
 
+config 64BIT_CONTEXT
+        bool "Save 64bit integer registers"
+        depends on 32BIT && CPU_LOONGSON2
+        help
+          Loongson2 CPU is 64bit , when used in 32BIT mode, its integer
+          registers can still be accessed as 64bit, mainly for multimedia
+          instructions. We must have all 64bit save/restored to make sure
+          those instructions to get correct result.
+
 #
 # Vectored interrupt mode is an R2 feature
 #
diff --git a/arch/mips/Makefile b/arch/mips/Makefile
index fc09769b3dd4..1e4579cc54ef 100644
--- a/arch/mips/Makefile
+++ b/arch/mips/Makefile
@@ -118,6 +118,7 @@ cflags-$(CONFIG_CPU_R4300)	+= -march=r4300 -Wa,--trap
 cflags-$(CONFIG_CPU_VR41XX)     += -march=r4100 -Wa,--trap
 cflags-$(CONFIG_CPU_R4X00)      += -march=r4600 -Wa,--trap
 cflags-$(CONFIG_CPU_TX49XX)     += -march=r4600 -Wa,--trap
+cflags-$(CONFIG_CPU_LOONGSON2)  += -march=r4600 -Wa,--trap
 cflags-$(CONFIG_CPU_MIPS32_R1)  += $(call cc-option,-march=mips32,-mips32 -U_MIPS_ISA -D_MIPS_ISA=_MIPS_ISA_MIPS32) \
                         -Wa,-mips32 -Wa,--trap
 cflags-$(CONFIG_CPU_MIPS32_R2)  += $(call cc-option,-march=mips32r2,-mips32r2 -U_MIPS_ISA -D_MIPS_ISA=_MIPS_ISA_MIPS32) \
diff --git a/arch/mips/kernel/Makefile b/arch/mips/kernel/Makefile
index bb25dd8ce437..73983eee1431 100644
--- a/arch/mips/kernel/Makefile
+++ b/arch/mips/kernel/Makefile
@@ -14,14 +14,15 @@ binfmt_irix-objs	:= irixelf.o irixinv.o irixioctl.o irixsig.o	\
 obj-$(CONFIG_STACKTRACE)        += stacktrace.o
 obj-$(CONFIG_MODULES)           += mips_ksyms.o module.o
 
+obj-$(CONFIG_CPU_LOONGSON2)     += r4k_fpu.o r4k_switch.o
+obj-$(CONFIG_CPU_MIPS32)        += r4k_fpu.o r4k_switch.o
+obj-$(CONFIG_CPU_MIPS64)        += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R3000)         += r2300_fpu.o r2300_switch.o
-obj-$(CONFIG_CPU_TX39XX)        += r2300_fpu.o r2300_switch.o
-obj-$(CONFIG_CPU_TX49XX)        += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R4000)         += r4k_fpu.o r4k_switch.o
-obj-$(CONFIG_CPU_VR41XX)        += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R4300)         += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R4X00)         += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R5000)         += r4k_fpu.o r4k_switch.o
+obj-$(CONFIG_CPU_R6000)         += r6000_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R5432)         += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R8000)         += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_RM7000)        += r4k_fpu.o r4k_switch.o
@@ -29,9 +30,9 @@ obj-$(CONFIG_CPU_RM9000)	+= r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_NEVADA)        += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_R10000)        += r4k_fpu.o r4k_switch.o
 obj-$(CONFIG_CPU_SB1)           += r4k_fpu.o r4k_switch.o
-obj-$(CONFIG_CPU_MIPS32)        += r4k_fpu.o r4k_switch.o
-obj-$(CONFIG_CPU_MIPS64)        += r4k_fpu.o r4k_switch.o
-obj-$(CONFIG_CPU_R6000)         += r6000_fpu.o r4k_switch.o
+obj-$(CONFIG_CPU_TX39XX)        += r2300_fpu.o r2300_switch.o
+obj-$(CONFIG_CPU_TX49XX)        += r4k_fpu.o r4k_switch.o
+obj-$(CONFIG_CPU_VR41XX)        += r4k_fpu.o r4k_switch.o
 
 obj-$(CONFIG_SMP)               += smp.o
 
diff --git a/arch/mips/kernel/cpu-probe.c b/arch/mips/kernel/cpu-probe.c
index 27fc1843423e..23d8a3b7dd75 100644
--- a/arch/mips/kernel/cpu-probe.c
+++ b/arch/mips/kernel/cpu-probe.c
@@ -485,6 +485,14 @@ static inline void cpu_probe_legacy(struct cpuinfo_mips *c)
                              MIPS_CPU_LLSC;
                 c->tlbsize = 64;
                 break;
+        case PRID_IMP_LOONGSON2:
+                c->cputype = CPU_LOONGSON2;
+                c->isa_level = MIPS_CPU_ISA_III;
+                c->options = R4K_OPTS |
+                             MIPS_CPU_FPU | MIPS_CPU_LLSC |
+                             MIPS_CPU_32FPR;
+                c->tlbsize = 64;
+                break;
         }
 }
 
diff --git a/arch/mips/kernel/proc.c b/arch/mips/kernel/proc.c
index eb7730d67934..ec04f5a1a5ea 100644
--- a/arch/mips/kernel/proc.c
+++ b/arch/mips/kernel/proc.c
@@ -83,6 +83,7 @@ static const char *cpu_name[] = {
         [CPU_VR4181A]   = "NEC VR4181A",
         [CPU_SR71000]   = "Sandcraft SR71000",
         [CPU_PR4450]    = "Philips PR4450",
+        [CPU_LOONGSON2] = "ICT Loongson-2",
 };
 
 
diff --git a/arch/mips/lib/Makefile b/arch/mips/lib/Makefile
index 1a484e430b3c..91ed1eb33102 100644
--- a/arch/mips/lib/Makefile
+++ b/arch/mips/lib/Makefile
@@ -8,6 +8,7 @@ lib-y	+= csum_partial.o memcpy.o memcpy-inatomic.o memset.o strlen_user.o \
 obj-y                   += iomap.o
 obj-$(CONFIG_PCI)       += iomap-pci.o
 
+obj-$(CONFIG_CPU_LOONGSON2)     += dump_tlb.o
 obj-$(CONFIG_CPU_MIPS32)        += dump_tlb.o
 obj-$(CONFIG_CPU_MIPS64)        += dump_tlb.o
 obj-$(CONFIG_CPU_NEVADA)        += dump_tlb.o
diff --git a/arch/mips/mm/Makefile b/arch/mips/mm/Makefile
index 293697b15603..19a0e544c4e9 100644
--- a/arch/mips/mm/Makefile
+++ b/arch/mips/mm/Makefile
@@ -9,6 +9,7 @@ obj-$(CONFIG_32BIT)		+= ioremap.o pgtable-32.o
 obj-$(CONFIG_64BIT)             += pgtable-64.o
 obj-$(CONFIG_HIGHMEM)           += highmem.o
 
+obj-$(CONFIG_CPU_LOONGSON2)     += c-r4k.o cex-gen.o pg-r4k.o tlb-r4k.o
 obj-$(CONFIG_CPU_MIPS32)        += c-r4k.o cex-gen.o pg-r4k.o tlb-r4k.o
 obj-$(CONFIG_CPU_MIPS64)        += c-r4k.o cex-gen.o pg-r4k.o tlb-r4k.o
 obj-$(CONFIG_CPU_NEVADA)        += c-r4k.o cex-gen.o pg-r4k.o tlb-r4k.o
diff --git a/arch/mips/mm/c-r4k.c b/arch/mips/mm/c-r4k.c
index df04a315d830..be96231dccb6 100644
--- a/arch/mips/mm/c-r4k.c
+++ b/arch/mips/mm/c-r4k.c
@@ -335,6 +335,10 @@ static void r4k_flush_cache_all(void)
 
 static inline void local_r4k___flush_cache_all(void * args)
 {
+#if defined(CONFIG_CPU_LOONGSON2)
+        r4k_blast_scache();
+        return;
+#endif
         r4k_blast_dcache();
         r4k_blast_icache();
 
@@ -848,6 +852,24 @@ static void __init probe_pcache(void)
                 c->options |= MIPS_CPU_PREFETCH;
                 break;
 
+        case CPU_LOONGSON2:
+                icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
+                c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
+                if (prid & 0x3)
+                        c->icache.ways = 4;
+                else
+                        c->icache.ways = 2;
+                c->icache.waybit = 0;
+
+                dcache_size = 1 << (12 + ((config & CONF_DC) >> 6));
+                c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
+                if (prid & 0x3)
+                        c->dcache.ways = 4;
+                else
+                        c->dcache.ways = 2;
+                c->dcache.waybit = 0;
+                break;
+
         default:
                 if (!(config & MIPS_CONF_M))
                         panic("Don't know how to probe P-caches on this cpu.");
@@ -963,6 +985,14 @@ static void __init probe_pcache(void)
                 break;
         }
 
+#ifdef  CONFIG_CPU_LOONGSON2
+        /*
+         * LOONGSON2 has 4 way icache, but when using indexed cache op,
+         * one op will act on all 4 ways
+         */
+        c->icache.ways = 1;
+#endif
+
         printk("Primary instruction cache %ldkB, %s, %s, linesize %d bytes.\n",
                icache_size >> 10,
                cpu_has_vtag_icache ? "virtually tagged" : "physically tagged",
@@ -1036,6 +1066,24 @@ static int __init probe_scache(void)
         return 1;
 }
 
+#if defined(CONFIG_CPU_LOONGSON2)
+static void __init loongson2_sc_init(void)
+{
+        struct cpuinfo_mips *c = &current_cpu_data;
+
+        scache_size = 512*1024;
+        c->scache.linesz = 32;
+        c->scache.ways = 4;
+        c->scache.waybit = 0;
+        c->scache.waysize = scache_size / (c->scache.ways);
+        c->scache.sets = scache_size / (c->scache.linesz * c->scache.ways);
+        pr_info("Unified secondary cache %ldkB %s, linesize %d bytes.\n",
+               scache_size >> 10, way_string[c->scache.ways], c->scache.linesz);
+
+        c->options |= MIPS_CPU_INCLUSIVE_CACHES;
+}
+#endif
+
 extern int r5k_sc_init(void);
 extern int rm7k_sc_init(void);
 extern int mips_sc_init(void);
@@ -1085,6 +1133,12 @@ static void __init setup_scache(void)
 #endif
                 return;
 
+#if defined(CONFIG_CPU_LOONGSON2)
+        case CPU_LOONGSON2:
+                loongson2_sc_init();
+                return;
+#endif
+
         default:
                 if (c->isa_level == MIPS_CPU_ISA_M32R1 ||
                     c->isa_level == MIPS_CPU_ISA_M32R2 ||
diff --git a/arch/mips/mm/tlb-r4k.c b/arch/mips/mm/tlb-r4k.c
index 65160d4984d9..dcd6913dc1ff 100644
--- a/arch/mips/mm/tlb-r4k.c
+++ b/arch/mips/mm/tlb-r4k.c
@@ -48,6 +48,22 @@ extern void build_tlb_refill_handler(void);
 
 #endif /* CONFIG_MIPS_MT_SMTC */
 
+#if defined(CONFIG_CPU_LOONGSON2)
+/*
+ * LOONGSON2 has a 4 entry itlb which is a subset of dtlb,
+ * unfortrunately, itlb is not totally transparent to software.
+ */
+#define FLUSH_ITLB write_c0_diag(4);
+
+#define FLUSH_ITLB_VM(vma) { if ((vma)->vm_flags & VM_EXEC)  write_c0_diag(4); }
+
+#else
+
+#define FLUSH_ITLB
+#define FLUSH_ITLB_VM(vma)
+
+#endif
+
 void local_flush_tlb_all(void)
 {
         unsigned long flags;
@@ -73,6 +89,7 @@ void local_flush_tlb_all(void)
         }
         tlbw_use_hazard();
         write_c0_entryhi(old_ctx);
+        FLUSH_ITLB;
         EXIT_CRITICAL(flags);
 }
 
@@ -136,6 +153,7 @@ void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
                 } else {
                         drop_mmu_context(mm, cpu);
                 }
+                FLUSH_ITLB;
                 EXIT_CRITICAL(flags);
         }
 }
@@ -178,6 +196,7 @@ void local_flush_tlb_kernel_range(unsigned long start, unsigned long end)
         } else {
                 local_flush_tlb_all();
         }
+        FLUSH_ITLB;
         EXIT_CRITICAL(flags);
 }
 
@@ -210,6 +229,7 @@ void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
 
         finish:
                 write_c0_entryhi(oldpid);
+                FLUSH_ITLB_VM(vma);
                 EXIT_CRITICAL(flags);
         }
 }
@@ -241,7 +261,7 @@ void local_flush_tlb_one(unsigned long page)
                 tlbw_use_hazard();
         }
         write_c0_entryhi(oldpid);
-
+        FLUSH_ITLB;
         EXIT_CRITICAL(flags);
 }
 
@@ -293,6 +313,7 @@ void __update_tlb(struct vm_area_struct * vma, unsigned long address, pte_t pte)
         else
                 tlb_write_indexed();
         tlbw_use_hazard();
+        FLUSH_ITLB_VM(vma);
         EXIT_CRITICAL(flags);
 }
 
diff --git a/arch/mips/mm/tlbex.c b/arch/mips/mm/tlbex.c
index e7149290d1cb..4ec0964b8394 100644
--- a/arch/mips/mm/tlbex.c
+++ b/arch/mips/mm/tlbex.c
@@ -893,6 +893,7 @@ static __init void build_tlb_write_entry(u32 **p, struct label **l,
         case CPU_4KSC:
         case CPU_20KC:
         case CPU_25KF:
+        case CPU_LOONGSON2:
                 tlbw(p);
                 break;
 
@@ -1276,7 +1277,8 @@ static void __init build_r4000_tlb_refill_handler(void)
          * need three, with the second nop'ed and the third being
          * unused.
          */
-#ifdef CONFIG_32BIT
+        /* Loongson2 ebase is different than r4k, we have more space */
+#if defined(CONFIG_32BIT) || defined(CONFIG_CPU_LOONGSON2)
         if ((p - tlb_handler) > 64)
                 panic("TLB refill handler space exceeded");
 #else
@@ -1289,7 +1291,7 @@ static void __init build_r4000_tlb_refill_handler(void)
         /*
          * Now fold the handler in the TLB refill handler space.
          */
-#ifdef CONFIG_32BIT
+#if defined(CONFIG_32BIT) || defined(CONFIG_CPU_LOONGSON2)
         f = final_handler;
         /* Simplest case, just copy the handler. */
         copy_handler(relocs, labels, tlb_handler, p, f);
@@ -1336,7 +1338,7 @@ static void __init build_r4000_tlb_refill_handler(void)
                 final_len);
 
         f = final_handler;
-#ifdef CONFIG_64BIT
+#if defined(CONFIG_64BIT) && !defined(CONFIG_CPU_LOONGSON2)
         if (final_len > 32)
                 final_len = 64;
         else