summaryrefslogtreecommitdiff
path: root/arch/mips/mm/c-r4k.c
diff options
context:
space:
mode:
Diffstat (limited to 'arch/mips/mm/c-r4k.c')
-rw-r--r--arch/mips/mm/c-r4k.c1093
1 files changed, 725 insertions, 368 deletions
diff --git a/arch/mips/mm/c-r4k.c b/arch/mips/mm/c-r4k.c
index 21813beec7a5..10413b6f6662 100644
--- a/arch/mips/mm/c-r4k.c
+++ b/arch/mips/mm/c-r4k.c
@@ -7,16 +7,19 @@
* Copyright (C) 1997, 1998, 1999, 2000, 2001, 2002 Ralf Baechle (ralf@gnu.org)
* Copyright (C) 1999, 2000 Silicon Graphics, Inc.
*/
+#include <linux/cpu_pm.h>
#include <linux/hardirq.h>
#include <linux/init.h>
#include <linux/highmem.h>
#include <linux/kernel.h>
#include <linux/linkage.h>
+#include <linux/preempt.h>
#include <linux/sched.h>
#include <linux/smp.h>
#include <linux/mm.h>
-#include <linux/module.h>
+#include <linux/export.h>
#include <linux/bitops.h>
+#include <linux/dma-map-ops.h> /* for dma_default_coherent */
#include <asm/bcache.h>
#include <asm/bootinfo.h>
@@ -24,16 +27,60 @@
#include <asm/cacheops.h>
#include <asm/cpu.h>
#include <asm/cpu-features.h>
+#include <asm/cpu-type.h>
#include <asm/io.h>
#include <asm/page.h>
-#include <asm/pgtable.h>
#include <asm/r4kcache.h>
#include <asm/sections.h>
#include <asm/mmu_context.h>
-#include <asm/war.h>
#include <asm/cacheflush.h> /* for run_uncached() */
#include <asm/traps.h>
-#include <asm/dma-coherence.h>
+#include <asm/mips-cps.h>
+
+/*
+ * Bits describing what cache ops an SMP callback function may perform.
+ *
+ * R4K_HIT - Virtual user or kernel address based cache operations. The
+ * active_mm must be checked before using user addresses, falling
+ * back to kmap.
+ * R4K_INDEX - Index based cache operations.
+ */
+
+#define R4K_HIT BIT(0)
+#define R4K_INDEX BIT(1)
+
+/**
+ * r4k_op_needs_ipi() - Decide if a cache op needs to be done on every core.
+ * @type: Type of cache operations (R4K_HIT or R4K_INDEX).
+ *
+ * Decides whether a cache op needs to be performed on every core in the system.
+ * This may change depending on the @type of cache operation, as well as the set
+ * of online CPUs, so preemption should be disabled by the caller to prevent CPU
+ * hotplug from changing the result.
+ *
+ * Returns: 1 if the cache operation @type should be done on every core in
+ * the system.
+ * 0 if the cache operation @type is globalized and only needs to
+ * be performed on a simple CPU.
+ */
+static inline bool r4k_op_needs_ipi(unsigned int type)
+{
+ /* The MIPS Coherence Manager (CM) globalizes address-based cache ops */
+ if (type == R4K_HIT && mips_cm_present())
+ return false;
+
+ /*
+ * Hardware doesn't globalize the required cache ops, so SMP calls may
+ * be needed, but only if there are foreign CPUs (non-siblings with
+ * separate caches).
+ */
+ /* cpu_foreign_map[] undeclared when !CONFIG_SMP */
+#ifdef CONFIG_SMP
+ return !cpumask_empty(&cpu_foreign_map[0]);
+#else
+ return false;
+#endif
+}
/*
* Special Variant of smp_call_function for use by cache functions:
@@ -44,52 +91,34 @@
* primary cache.
* o doesn't disable interrupts on the local CPU
*/
-static inline void r4k_on_each_cpu(void (*func) (void *info), void *info)
+static inline void r4k_on_each_cpu(unsigned int type,
+ void (*func)(void *info), void *info)
{
preempt_disable();
-
-#if !defined(CONFIG_MIPS_MT_SMP) && !defined(CONFIG_MIPS_MT_SMTC)
- smp_call_function(func, info, 1);
-#endif
+ if (r4k_op_needs_ipi(type))
+ smp_call_function_many(&cpu_foreign_map[smp_processor_id()],
+ func, info, 1);
func(info);
preempt_enable();
}
-#if defined(CONFIG_MIPS_CMP)
-#define cpu_has_safe_index_cacheops 0
-#else
-#define cpu_has_safe_index_cacheops 1
-#endif
-
/*
* Must die.
*/
static unsigned long icache_size __read_mostly;
static unsigned long dcache_size __read_mostly;
+static unsigned long vcache_size __read_mostly;
static unsigned long scache_size __read_mostly;
-/*
- * Dummy cache handling routines for machines without boardcaches
- */
-static void cache_noop(void) {}
-
-static struct bcache_ops no_sc_ops = {
- .bc_enable = (void *)cache_noop,
- .bc_disable = (void *)cache_noop,
- .bc_wback_inv = (void *)cache_noop,
- .bc_inv = (void *)cache_noop
-};
-
-struct bcache_ops *bcops = &no_sc_ops;
-
#define cpu_is_r4600_v1_x() ((read_c0_prid() & 0xfffffff0) == 0x00002010)
#define cpu_is_r4600_v2_x() ((read_c0_prid() & 0xfffffff0) == 0x00002020)
#define R4600_HIT_CACHEOP_WAR_IMPL \
do { \
- if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x()) \
+ if (IS_ENABLED(CONFIG_WAR_R4600_V2_HIT_CACHEOP) && \
+ cpu_is_r4600_v2_x()) \
*(volatile unsigned long *)CKSEG1; \
- if (R4600_V1_HIT_CACHEOP_WAR) \
+ if (IS_ENABLED(CONFIG_WAR_R4600_V1_HIT_CACHEOP)) \
__asm__ __volatile__("nop;nop;nop;nop"); \
} while (0)
@@ -103,44 +132,65 @@ static inline void r4k_blast_dcache_page_dc32(unsigned long addr)
static inline void r4k_blast_dcache_page_dc64(unsigned long addr)
{
- R4600_HIT_CACHEOP_WAR_IMPL;
blast_dcache64_page(addr);
}
-static void __cpuinit r4k_blast_dcache_page_setup(void)
+static inline void r4k_blast_dcache_page_dc128(unsigned long addr)
+{
+ blast_dcache128_page(addr);
+}
+
+static void r4k_blast_dcache_page_setup(void)
{
unsigned long dc_lsize = cpu_dcache_line_size();
- if (dc_lsize == 0)
+ switch (dc_lsize) {
+ case 0:
r4k_blast_dcache_page = (void *)cache_noop;
- else if (dc_lsize == 16)
+ break;
+ case 16:
r4k_blast_dcache_page = blast_dcache16_page;
- else if (dc_lsize == 32)
+ break;
+ case 32:
r4k_blast_dcache_page = r4k_blast_dcache_page_dc32;
- else if (dc_lsize == 64)
+ break;
+ case 64:
r4k_blast_dcache_page = r4k_blast_dcache_page_dc64;
+ break;
+ case 128:
+ r4k_blast_dcache_page = r4k_blast_dcache_page_dc128;
+ break;
+ default:
+ break;
+ }
}
-static void (* r4k_blast_dcache_page_indexed)(unsigned long addr);
+#ifndef CONFIG_EVA
+#define r4k_blast_dcache_user_page r4k_blast_dcache_page
+#else
-static void __cpuinit r4k_blast_dcache_page_indexed_setup(void)
+static void (*r4k_blast_dcache_user_page)(unsigned long addr);
+
+static void r4k_blast_dcache_user_page_setup(void)
{
- unsigned long dc_lsize = cpu_dcache_line_size();
+ unsigned long dc_lsize = cpu_dcache_line_size();
if (dc_lsize == 0)
- r4k_blast_dcache_page_indexed = (void *)cache_noop;
+ r4k_blast_dcache_user_page = (void *)cache_noop;
else if (dc_lsize == 16)
- r4k_blast_dcache_page_indexed = blast_dcache16_page_indexed;
+ r4k_blast_dcache_user_page = blast_dcache16_user_page;
else if (dc_lsize == 32)
- r4k_blast_dcache_page_indexed = blast_dcache32_page_indexed;
+ r4k_blast_dcache_user_page = blast_dcache32_user_page;
else if (dc_lsize == 64)
- r4k_blast_dcache_page_indexed = blast_dcache64_page_indexed;
+ r4k_blast_dcache_user_page = blast_dcache64_user_page;
}
+#endif
+
void (* r4k_blast_dcache)(void);
EXPORT_SYMBOL(r4k_blast_dcache);
-static void __cpuinit r4k_blast_dcache_setup(void)
+static void r4k_blast_dcache_setup(void)
{
unsigned long dc_lsize = cpu_dcache_line_size();
@@ -152,9 +202,11 @@ static void __cpuinit r4k_blast_dcache_setup(void)
r4k_blast_dcache = blast_dcache32;
else if (dc_lsize == 64)
r4k_blast_dcache = blast_dcache64;
+ else if (dc_lsize == 128)
+ r4k_blast_dcache = blast_dcache128;
}
-/* force code alignment (used for TX49XX_ICACHE_INDEX_INV_WAR) */
+/* force code alignment (used for CONFIG_WAR_TX49XX_ICACHE_INDEX_INV) */
#define JUMP_TO_ALIGN(order) \
__asm__ __volatile__( \
"b\t1f\n\t" \
@@ -186,48 +238,19 @@ static inline void tx49_blast_icache32(void)
/* I'm in even chunk. blast odd chunks */
for (ws = 0; ws < ws_end; ws += ws_inc)
for (addr = start + 0x400; addr < end; addr += 0x400 * 2)
- cache32_unroll32(addr|ws, Index_Invalidate_I);
+ cache_unroll(32, kernel_cache, Index_Invalidate_I,
+ addr | ws, 32);
CACHE32_UNROLL32_ALIGN;
/* I'm in odd chunk. blast even chunks */
for (ws = 0; ws < ws_end; ws += ws_inc)
for (addr = start; addr < end; addr += 0x400 * 2)
- cache32_unroll32(addr|ws, Index_Invalidate_I);
-}
-
-static inline void blast_icache32_r4600_v1_page_indexed(unsigned long page)
-{
- unsigned long flags;
-
- local_irq_save(flags);
- blast_icache32_page_indexed(page);
- local_irq_restore(flags);
-}
-
-static inline void tx49_blast_icache32_page_indexed(unsigned long page)
-{
- unsigned long indexmask = current_cpu_data.icache.waysize - 1;
- unsigned long start = INDEX_BASE + (page & indexmask);
- unsigned long end = start + PAGE_SIZE;
- unsigned long ws_inc = 1UL << current_cpu_data.icache.waybit;
- unsigned long ws_end = current_cpu_data.icache.ways <<
- current_cpu_data.icache.waybit;
- unsigned long ws, addr;
-
- CACHE32_UNROLL32_ALIGN2;
- /* I'm in even chunk. blast odd chunks */
- for (ws = 0; ws < ws_end; ws += ws_inc)
- for (addr = start + 0x400; addr < end; addr += 0x400 * 2)
- cache32_unroll32(addr|ws, Index_Invalidate_I);
- CACHE32_UNROLL32_ALIGN;
- /* I'm in odd chunk. blast even chunks */
- for (ws = 0; ws < ws_end; ws += ws_inc)
- for (addr = start; addr < end; addr += 0x400 * 2)
- cache32_unroll32(addr|ws, Index_Invalidate_I);
+ cache_unroll(32, kernel_cache, Index_Invalidate_I,
+ addr | ws, 32);
}
static void (* r4k_blast_icache_page)(unsigned long addr);
-static void __cpuinit r4k_blast_icache_page_setup(void)
+static void r4k_blast_icache_page_setup(void)
{
unsigned long ic_lsize = cpu_icache_line_size();
@@ -235,41 +258,42 @@ static void __cpuinit r4k_blast_icache_page_setup(void)
r4k_blast_icache_page = (void *)cache_noop;
else if (ic_lsize == 16)
r4k_blast_icache_page = blast_icache16_page;
+ else if (ic_lsize == 32 && current_cpu_type() == CPU_LOONGSON2EF)
+ r4k_blast_icache_page = loongson2_blast_icache32_page;
else if (ic_lsize == 32)
r4k_blast_icache_page = blast_icache32_page;
else if (ic_lsize == 64)
r4k_blast_icache_page = blast_icache64_page;
+ else if (ic_lsize == 128)
+ r4k_blast_icache_page = blast_icache128_page;
}
+#ifndef CONFIG_EVA
+#define r4k_blast_icache_user_page r4k_blast_icache_page
+#else
-static void (* r4k_blast_icache_page_indexed)(unsigned long addr);
+static void (*r4k_blast_icache_user_page)(unsigned long addr);
-static void __cpuinit r4k_blast_icache_page_indexed_setup(void)
+static void r4k_blast_icache_user_page_setup(void)
{
unsigned long ic_lsize = cpu_icache_line_size();
if (ic_lsize == 0)
- r4k_blast_icache_page_indexed = (void *)cache_noop;
+ r4k_blast_icache_user_page = (void *)cache_noop;
else if (ic_lsize == 16)
- r4k_blast_icache_page_indexed = blast_icache16_page_indexed;
- else if (ic_lsize == 32) {
- if (R4600_V1_INDEX_ICACHEOP_WAR && cpu_is_r4600_v1_x())
- r4k_blast_icache_page_indexed =
- blast_icache32_r4600_v1_page_indexed;
- else if (TX49XX_ICACHE_INDEX_INV_WAR)
- r4k_blast_icache_page_indexed =
- tx49_blast_icache32_page_indexed;
- else
- r4k_blast_icache_page_indexed =
- blast_icache32_page_indexed;
- } else if (ic_lsize == 64)
- r4k_blast_icache_page_indexed = blast_icache64_page_indexed;
+ r4k_blast_icache_user_page = blast_icache16_user_page;
+ else if (ic_lsize == 32)
+ r4k_blast_icache_user_page = blast_icache32_user_page;
+ else if (ic_lsize == 64)
+ r4k_blast_icache_user_page = blast_icache64_user_page;
}
+#endif
+
void (* r4k_blast_icache)(void);
EXPORT_SYMBOL(r4k_blast_icache);
-static void __cpuinit r4k_blast_icache_setup(void)
+static void r4k_blast_icache_setup(void)
{
unsigned long ic_lsize = cpu_icache_line_size();
@@ -278,19 +302,24 @@ static void __cpuinit r4k_blast_icache_setup(void)
else if (ic_lsize == 16)
r4k_blast_icache = blast_icache16;
else if (ic_lsize == 32) {
- if (R4600_V1_INDEX_ICACHEOP_WAR && cpu_is_r4600_v1_x())
+ if (IS_ENABLED(CONFIG_WAR_R4600_V1_INDEX_ICACHEOP) &&
+ cpu_is_r4600_v1_x())
r4k_blast_icache = blast_r4600_v1_icache32;
- else if (TX49XX_ICACHE_INDEX_INV_WAR)
+ else if (IS_ENABLED(CONFIG_WAR_TX49XX_ICACHE_INDEX_INV))
r4k_blast_icache = tx49_blast_icache32;
+ else if (current_cpu_type() == CPU_LOONGSON2EF)
+ r4k_blast_icache = loongson2_blast_icache32;
else
r4k_blast_icache = blast_icache32;
} else if (ic_lsize == 64)
r4k_blast_icache = blast_icache64;
+ else if (ic_lsize == 128)
+ r4k_blast_icache = blast_icache128;
}
static void (* r4k_blast_scache_page)(unsigned long addr);
-static void __cpuinit r4k_blast_scache_page_setup(void)
+static void r4k_blast_scache_page_setup(void)
{
unsigned long sc_lsize = cpu_scache_line_size();
@@ -306,52 +335,46 @@ static void __cpuinit r4k_blast_scache_page_setup(void)
r4k_blast_scache_page = blast_scache128_page;
}
-static void (* r4k_blast_scache_page_indexed)(unsigned long addr);
+static void (* r4k_blast_scache)(void);
-static void __cpuinit r4k_blast_scache_page_indexed_setup(void)
+static void r4k_blast_scache_setup(void)
{
unsigned long sc_lsize = cpu_scache_line_size();
if (scache_size == 0)
- r4k_blast_scache_page_indexed = (void *)cache_noop;
+ r4k_blast_scache = (void *)cache_noop;
else if (sc_lsize == 16)
- r4k_blast_scache_page_indexed = blast_scache16_page_indexed;
+ r4k_blast_scache = blast_scache16;
else if (sc_lsize == 32)
- r4k_blast_scache_page_indexed = blast_scache32_page_indexed;
+ r4k_blast_scache = blast_scache32;
else if (sc_lsize == 64)
- r4k_blast_scache_page_indexed = blast_scache64_page_indexed;
+ r4k_blast_scache = blast_scache64;
else if (sc_lsize == 128)
- r4k_blast_scache_page_indexed = blast_scache128_page_indexed;
+ r4k_blast_scache = blast_scache128;
}
-static void (* r4k_blast_scache)(void);
+static void (*r4k_blast_scache_node)(long node);
-static void __cpuinit r4k_blast_scache_setup(void)
+static void r4k_blast_scache_node_setup(void)
{
unsigned long sc_lsize = cpu_scache_line_size();
- if (scache_size == 0)
- r4k_blast_scache = (void *)cache_noop;
+ if (current_cpu_type() != CPU_LOONGSON64)
+ r4k_blast_scache_node = (void *)cache_noop;
else if (sc_lsize == 16)
- r4k_blast_scache = blast_scache16;
+ r4k_blast_scache_node = blast_scache16_node;
else if (sc_lsize == 32)
- r4k_blast_scache = blast_scache32;
+ r4k_blast_scache_node = blast_scache32_node;
else if (sc_lsize == 64)
- r4k_blast_scache = blast_scache64;
+ r4k_blast_scache_node = blast_scache64_node;
else if (sc_lsize == 128)
- r4k_blast_scache = blast_scache128;
+ r4k_blast_scache_node = blast_scache128_node;
}
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();
-
switch (current_cpu_type()) {
+ case CPU_LOONGSON2EF:
case CPU_R4000SC:
case CPU_R4000MC:
case CPU_R4400SC:
@@ -359,28 +382,75 @@ static inline void local_r4k___flush_cache_all(void * args)
case CPU_R10000:
case CPU_R12000:
case CPU_R14000:
+ case CPU_R16000:
+ /*
+ * These caches are inclusive caches, that is, if something
+ * is not cached in the S-cache, we know it also won't be
+ * in one of the primary caches.
+ */
r4k_blast_scache();
+ break;
+
+ case CPU_LOONGSON64:
+ /* Use get_ebase_cpunum() for both NUMA=y/n */
+ r4k_blast_scache_node(get_ebase_cpunum() >> 2);
+ break;
+
+ case CPU_BMIPS5000:
+ r4k_blast_scache();
+ __sync();
+ break;
+
+ default:
+ r4k_blast_dcache();
+ r4k_blast_icache();
+ break;
}
}
static void r4k___flush_cache_all(void)
{
- r4k_on_each_cpu(local_r4k___flush_cache_all, NULL);
+ r4k_on_each_cpu(R4K_INDEX, local_r4k___flush_cache_all, NULL);
}
-static inline int has_valid_asid(const struct mm_struct *mm)
+/**
+ * has_valid_asid() - Determine if an mm already has an ASID.
+ * @mm: Memory map.
+ * @type: R4K_HIT or R4K_INDEX, type of cache op.
+ *
+ * Determines whether @mm already has an ASID on any of the CPUs which cache ops
+ * of type @type within an r4k_on_each_cpu() call will affect. If
+ * r4k_on_each_cpu() does an SMP call to a single VPE in each core, then the
+ * scope of the operation is confined to sibling CPUs, otherwise all online CPUs
+ * will need to be checked.
+ *
+ * Must be called in non-preemptive context.
+ *
+ * Returns: 1 if the CPUs affected by @type cache ops have an ASID for @mm.
+ * 0 otherwise.
+ */
+static inline int has_valid_asid(const struct mm_struct *mm, unsigned int type)
{
-#if defined(CONFIG_MIPS_MT_SMP) || defined(CONFIG_MIPS_MT_SMTC)
- int i;
+ unsigned int i;
+ const cpumask_t *mask = cpu_present_mask;
+
+ if (cpu_has_mmid)
+ return cpu_context(0, mm) != 0;
- for_each_online_cpu(i)
+ /* cpu_sibling_map[] undeclared when !CONFIG_SMP */
+#ifdef CONFIG_SMP
+ /*
+ * If r4k_on_each_cpu does SMP calls, it does them to a single VPE in
+ * each foreign core, so we only need to worry about siblings.
+ * Otherwise we need to worry about all present CPUs.
+ */
+ if (r4k_op_needs_ipi(type))
+ mask = &cpu_sibling_map[smp_processor_id()];
+#endif
+ for_each_cpu(i, mask)
if (cpu_context(i, mm))
return 1;
-
return 0;
-#else
- return cpu_context(smp_processor_id(), mm);
-#endif
}
static void r4k__flush_cache_vmap(void)
@@ -393,15 +463,26 @@ static void r4k__flush_cache_vunmap(void)
r4k_blast_dcache();
}
+/*
+ * Note: flush_tlb_range() assumes flush_cache_range() sufficiently flushes
+ * whole caches when vma is executable.
+ */
static inline void local_r4k_flush_cache_range(void * args)
{
struct vm_area_struct *vma = args;
int exec = vma->vm_flags & VM_EXEC;
- if (!(has_valid_asid(vma->vm_mm)))
+ if (!has_valid_asid(vma->vm_mm, R4K_INDEX))
return;
- r4k_blast_dcache();
+ /*
+ * If dcache can alias, we must blast it since mapping is changing.
+ * If executable, we must ensure any dirty lines are written back far
+ * enough to be visible to icache.
+ */
+ if (cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc))
+ r4k_blast_dcache();
+ /* If executable, blast stale lines from icache */
if (exec)
r4k_blast_icache();
}
@@ -411,20 +492,20 @@ static void r4k_flush_cache_range(struct vm_area_struct *vma,
{
int exec = vma->vm_flags & VM_EXEC;
- if (cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc))
- r4k_on_each_cpu(local_r4k_flush_cache_range, vma);
+ if (cpu_has_dc_aliases || exec)
+ r4k_on_each_cpu(R4K_INDEX, local_r4k_flush_cache_range, vma);
}
static inline void local_r4k_flush_cache_mm(void * args)
{
struct mm_struct *mm = args;
- if (!has_valid_asid(mm))
+ if (!has_valid_asid(mm, R4K_INDEX))
return;
/*
* Kludge alert. For obscure reasons R4000SC and R4400SC go nuts if we
- * only flush the primary caches but R10000 and R12000 behave sane ...
+ * only flush the primary caches but R1x000 behave sane ...
* R4000SC and R4400SC indexed S-cache ops also invalidate primary
* caches, so we can bail out early.
*/
@@ -444,7 +525,7 @@ static void r4k_flush_cache_mm(struct mm_struct *mm)
if (!cpu_has_dc_aliases)
return;
- r4k_on_each_cpu(local_r4k_flush_cache_mm, mm);
+ r4k_on_each_cpu(R4K_INDEX, local_r4k_flush_cache_mm, mm);
}
struct flush_cache_page_args {
@@ -462,24 +543,20 @@ static inline void local_r4k_flush_cache_page(void *args)
int exec = vma->vm_flags & VM_EXEC;
struct mm_struct *mm = vma->vm_mm;
int map_coherent = 0;
- pgd_t *pgdp;
- pud_t *pudp;
pmd_t *pmdp;
pte_t *ptep;
void *vaddr;
/*
- * If ownes no valid ASID yet, cannot possibly have gotten
+ * If owns no valid ASID yet, cannot possibly have gotten
* this page into the cache.
*/
- if (!has_valid_asid(mm))
+ if (!has_valid_asid(mm, R4K_HIT))
return;
addr &= PAGE_MASK;
- pgdp = pgd_offset(mm, addr);
- pudp = pud_offset(pgdp, addr);
- pmdp = pmd_offset(pudp, addr);
- ptep = pte_offset(pmdp, addr);
+ pmdp = pmd_off(mm, addr);
+ ptep = pte_offset_kernel(pmdp, addr);
/*
* If the page isn't marked valid, the page cannot possibly be
@@ -491,12 +568,14 @@ static inline void local_r4k_flush_cache_page(void *args)
if ((mm == current->active_mm) && (pte_val(*ptep) & _PAGE_VALID))
vaddr = NULL;
else {
+ struct folio *folio = page_folio(page);
/*
* Use kmap_coherent or kmap_atomic to do flushes for
* another ASID than the current one.
*/
map_coherent = (cpu_has_dc_aliases &&
- page_mapped(page) && !Page_dcache_dirty(page));
+ folio_mapped(folio) &&
+ !folio_test_dcache_dirty(folio));
if (map_coherent)
vaddr = kmap_coherent(page, addr);
else
@@ -505,18 +584,17 @@ static inline void local_r4k_flush_cache_page(void *args)
}
if (cpu_has_dc_aliases || (exec && !cpu_has_ic_fills_f_dc)) {
- r4k_blast_dcache_page(addr);
+ vaddr ? r4k_blast_dcache_page(addr) :
+ r4k_blast_dcache_user_page(addr);
if (exec && !cpu_icache_snoops_remote_store)
r4k_blast_scache_page(addr);
}
if (exec) {
if (vaddr && cpu_has_vtag_icache && mm == current->active_mm) {
- int cpu = smp_processor_id();
-
- if (cpu_context(cpu, mm) != 0)
- drop_mmu_context(mm, cpu);
+ drop_mmu_context(mm);
} else
- r4k_blast_icache_page(addr);
+ vaddr ? r4k_blast_icache_page(addr) :
+ r4k_blast_icache_user_page(addr);
}
if (vaddr) {
@@ -536,7 +614,7 @@ static void r4k_flush_cache_page(struct vm_area_struct *vma,
args.addr = addr;
args.pfn = pfn;
- r4k_on_each_cpu(local_r4k_flush_cache_page, &args);
+ r4k_on_each_cpu(R4K_HIT, local_r4k_flush_cache_page, &args);
}
static inline void local_r4k_flush_data_cache_page(void * addr)
@@ -549,29 +627,64 @@ static void r4k_flush_data_cache_page(unsigned long addr)
if (in_atomic())
local_r4k_flush_data_cache_page((void *)addr);
else
- r4k_on_each_cpu(local_r4k_flush_data_cache_page, (void *) addr);
+ r4k_on_each_cpu(R4K_HIT, local_r4k_flush_data_cache_page,
+ (void *) addr);
}
struct flush_icache_range_args {
unsigned long start;
unsigned long end;
+ unsigned int type;
+ bool user;
};
-static inline void local_r4k_flush_icache_range(unsigned long start, unsigned long end)
+static inline void __local_r4k_flush_icache_range(unsigned long start,
+ unsigned long end,
+ unsigned int type,
+ bool user)
{
if (!cpu_has_ic_fills_f_dc) {
- if (end - start >= dcache_size) {
+ if (type == R4K_INDEX ||
+ (type & R4K_INDEX && end - start >= dcache_size)) {
r4k_blast_dcache();
} else {
R4600_HIT_CACHEOP_WAR_IMPL;
- protected_blast_dcache_range(start, end);
+ if (user)
+ protected_blast_dcache_range(start, end);
+ else
+ blast_dcache_range(start, end);
}
}
- if (end - start > icache_size)
+ if (type == R4K_INDEX ||
+ (type & R4K_INDEX && end - start > icache_size))
r4k_blast_icache();
- else
- protected_blast_icache_range(start, end);
+ else {
+ switch (boot_cpu_type()) {
+ case CPU_LOONGSON2EF:
+ protected_loongson2_blast_icache_range(start, end);
+ break;
+
+ default:
+ if (user)
+ protected_blast_icache_range(start, end);
+ else
+ blast_icache_range(start, end);
+ break;
+ }
+ }
+}
+
+static inline void local_r4k_flush_icache_range(unsigned long start,
+ unsigned long end)
+{
+ __local_r4k_flush_icache_range(start, end, R4K_HIT | R4K_INDEX, false);
+}
+
+static inline void local_r4k_flush_icache_user_range(unsigned long start,
+ unsigned long end)
+{
+ __local_r4k_flush_icache_range(start, end, R4K_HIT | R4K_INDEX, true);
}
static inline void local_r4k_flush_icache_range_ipi(void *args)
@@ -579,33 +692,76 @@ static inline void local_r4k_flush_icache_range_ipi(void *args)
struct flush_icache_range_args *fir_args = args;
unsigned long start = fir_args->start;
unsigned long end = fir_args->end;
+ unsigned int type = fir_args->type;
+ bool user = fir_args->user;
- local_r4k_flush_icache_range(start, end);
+ __local_r4k_flush_icache_range(start, end, type, user);
}
-static void r4k_flush_icache_range(unsigned long start, unsigned long end)
+static void __r4k_flush_icache_range(unsigned long start, unsigned long end,
+ bool user)
{
struct flush_icache_range_args args;
+ unsigned long size, cache_size;
args.start = start;
args.end = end;
+ args.type = R4K_HIT | R4K_INDEX;
+ args.user = user;
- r4k_on_each_cpu(local_r4k_flush_icache_range_ipi, &args);
+ /*
+ * Indexed cache ops require an SMP call.
+ * Consider if that can or should be avoided.
+ */
+ preempt_disable();
+ if (r4k_op_needs_ipi(R4K_INDEX) && !r4k_op_needs_ipi(R4K_HIT)) {
+ /*
+ * If address-based cache ops don't require an SMP call, then
+ * use them exclusively for small flushes.
+ */
+ size = end - start;
+ cache_size = icache_size;
+ if (!cpu_has_ic_fills_f_dc) {
+ size *= 2;
+ cache_size += dcache_size;
+ }
+ if (size <= cache_size)
+ args.type &= ~R4K_INDEX;
+ }
+ r4k_on_each_cpu(args.type, local_r4k_flush_icache_range_ipi, &args);
+ preempt_enable();
instruction_hazard();
}
+static void r4k_flush_icache_range(unsigned long start, unsigned long end)
+{
+ return __r4k_flush_icache_range(start, end, false);
+}
+
+static void r4k_flush_icache_user_range(unsigned long start, unsigned long end)
+{
+ return __r4k_flush_icache_range(start, end, true);
+}
+
#ifdef CONFIG_DMA_NONCOHERENT
static void r4k_dma_cache_wback_inv(unsigned long addr, unsigned long size)
{
/* Catch bad driver code */
- BUG_ON(size == 0);
+ if (WARN_ON(size == 0))
+ return;
+ preempt_disable();
if (cpu_has_inclusive_pcaches) {
- if (size >= scache_size)
- r4k_blast_scache();
- else
+ if (size >= scache_size) {
+ if (current_cpu_type() != CPU_LOONGSON64)
+ r4k_blast_scache();
+ else
+ r4k_blast_scache_node(pa_to_nid(addr));
+ } else {
blast_scache_range(addr, addr + size);
+ }
+ preempt_enable();
__sync();
return;
}
@@ -613,28 +769,66 @@ static void r4k_dma_cache_wback_inv(unsigned long addr, unsigned long size)
/*
* Either no secondary cache or the available caches don't have the
* subset property so we have to flush the primary caches
- * explicitly
+ * explicitly.
+ * If we would need IPI to perform an INDEX-type operation, then
+ * we have to use the HIT-type alternative as IPI cannot be used
+ * here due to interrupts possibly being disabled.
*/
- if (cpu_has_safe_index_cacheops && size >= dcache_size) {
+ if (!r4k_op_needs_ipi(R4K_INDEX) && size >= dcache_size) {
r4k_blast_dcache();
} else {
R4600_HIT_CACHEOP_WAR_IMPL;
blast_dcache_range(addr, addr + size);
}
+ preempt_enable();
bc_wback_inv(addr, size);
__sync();
}
+static void prefetch_cache_inv(unsigned long addr, unsigned long size)
+{
+ unsigned int linesz = cpu_scache_line_size();
+ unsigned long addr0 = addr, addr1;
+
+ addr0 &= ~(linesz - 1);
+ addr1 = (addr0 + size - 1) & ~(linesz - 1);
+
+ protected_writeback_scache_line(addr0);
+ if (likely(addr1 != addr0))
+ protected_writeback_scache_line(addr1);
+ else
+ return;
+
+ addr0 += linesz;
+ if (likely(addr1 != addr0))
+ protected_writeback_scache_line(addr0);
+ else
+ return;
+
+ addr1 -= linesz;
+ if (likely(addr1 > addr0))
+ protected_writeback_scache_line(addr0);
+}
+
static void r4k_dma_cache_inv(unsigned long addr, unsigned long size)
{
/* Catch bad driver code */
- BUG_ON(size == 0);
+ if (WARN_ON(size == 0))
+ return;
+
+ preempt_disable();
+
+ if (current_cpu_type() == CPU_BMIPS5000)
+ prefetch_cache_inv(addr, size);
if (cpu_has_inclusive_pcaches) {
- if (size >= scache_size)
- r4k_blast_scache();
- else {
+ if (size >= scache_size) {
+ if (current_cpu_type() != CPU_LOONGSON64)
+ r4k_blast_scache();
+ else
+ r4k_blast_scache_node(pa_to_nid(addr));
+ } else {
/*
* There is no clearly documented alignment requirement
* for the cache instruction on MIPS processors and
@@ -645,68 +839,24 @@ static void r4k_dma_cache_inv(unsigned long addr, unsigned long size)
*/
blast_inv_scache_range(addr, addr + size);
}
+ preempt_enable();
__sync();
return;
}
- if (cpu_has_safe_index_cacheops && size >= dcache_size) {
+ if (!r4k_op_needs_ipi(R4K_INDEX) && size >= dcache_size) {
r4k_blast_dcache();
} else {
R4600_HIT_CACHEOP_WAR_IMPL;
blast_inv_dcache_range(addr, addr + size);
}
+ preempt_enable();
bc_inv(addr, size);
__sync();
}
#endif /* CONFIG_DMA_NONCOHERENT */
-/*
- * While we're protected against bad userland addresses we don't care
- * very much about what happens in that case. Usually a segmentation
- * fault will dump the process later on anyway ...
- */
-static void local_r4k_flush_cache_sigtramp(void * arg)
-{
- unsigned long ic_lsize = cpu_icache_line_size();
- unsigned long dc_lsize = cpu_dcache_line_size();
- unsigned long sc_lsize = cpu_scache_line_size();
- unsigned long addr = (unsigned long) arg;
-
- R4600_HIT_CACHEOP_WAR_IMPL;
- if (dc_lsize)
- protected_writeback_dcache_line(addr & ~(dc_lsize - 1));
- if (!cpu_icache_snoops_remote_store && scache_size)
- protected_writeback_scache_line(addr & ~(sc_lsize - 1));
- if (ic_lsize)
- protected_flush_icache_line(addr & ~(ic_lsize - 1));
- if (MIPS4K_ICACHE_REFILL_WAR) {
- __asm__ __volatile__ (
- ".set push\n\t"
- ".set noat\n\t"
- ".set mips3\n\t"
-#ifdef CONFIG_32BIT
- "la $at,1f\n\t"
-#endif
-#ifdef CONFIG_64BIT
- "dla $at,1f\n\t"
-#endif
- "cache %0,($at)\n\t"
- "nop; nop; nop\n"
- "1:\n\t"
- ".set pop"
- :
- : "i" (Hit_Invalidate_I));
- }
- if (MIPS_CACHE_SYNC_WAR)
- __asm__ __volatile__ ("sync");
-}
-
-static void r4k_flush_cache_sigtramp(unsigned long addr)
-{
- r4k_on_each_cpu(local_r4k_flush_cache_sigtramp, (void *) addr);
-}
-
static void r4k_flush_icache_all(void)
{
if (cpu_has_vtag_icache)
@@ -718,6 +868,15 @@ struct flush_kernel_vmap_range_args {
int size;
};
+static inline void local_r4k_flush_kernel_vmap_range_index(void *args)
+{
+ /*
+ * Aliases only affect the primary caches so don't bother with
+ * S-caches or T-caches.
+ */
+ r4k_blast_dcache();
+}
+
static inline void local_r4k_flush_kernel_vmap_range(void *args)
{
struct flush_kernel_vmap_range_args *vmra = args;
@@ -728,12 +887,8 @@ static inline void local_r4k_flush_kernel_vmap_range(void *args)
* Aliases only affect the primary caches so don't bother with
* S-caches or T-caches.
*/
- if (cpu_has_safe_index_cacheops && size >= dcache_size)
- r4k_blast_dcache();
- else {
- R4600_HIT_CACHEOP_WAR_IMPL;
- blast_dcache_range(vaddr, vaddr + size);
- }
+ R4600_HIT_CACHEOP_WAR_IMPL;
+ blast_dcache_range(vaddr, vaddr + size);
}
static void r4k_flush_kernel_vmap_range(unsigned long vaddr, int size)
@@ -743,7 +898,12 @@ static void r4k_flush_kernel_vmap_range(unsigned long vaddr, int size)
args.vaddr = (unsigned long) vaddr;
args.size = size;
- r4k_on_each_cpu(local_r4k_flush_kernel_vmap_range, &args);
+ if (size >= dcache_size)
+ r4k_on_each_cpu(R4K_INDEX,
+ local_r4k_flush_kernel_vmap_range_index, NULL);
+ else
+ r4k_on_each_cpu(R4K_HIT, local_r4k_flush_kernel_vmap_range,
+ &args);
}
static inline void rm7k_erratum31(void)
@@ -774,42 +934,73 @@ static inline void rm7k_erratum31(void)
"cache\t%1, 0x3000(%0)\n\t"
".set pop\n"
:
- : "r" (addr), "i" (Index_Store_Tag_I), "i" (Fill));
+ : "r" (addr), "i" (Index_Store_Tag_I), "i" (Fill_I));
}
}
-static inline void alias_74k_erratum(struct cpuinfo_mips *c)
+static inline int alias_74k_erratum(struct cpuinfo_mips *c)
{
+ unsigned int imp = c->processor_id & PRID_IMP_MASK;
+ unsigned int rev = c->processor_id & PRID_REV_MASK;
+ int present = 0;
+
/*
* Early versions of the 74K do not update the cache tags on a
* vtag miss/ptag hit which can occur in the case of KSEG0/KUSEG
- * aliases. In this case it is better to treat the cache as always
- * having aliases.
+ * aliases. In this case it is better to treat the cache as always
+ * having aliases. Also disable the synonym tag update feature
+ * where available. In this case no opportunistic tag update will
+ * happen where a load causes a virtual address miss but a physical
+ * address hit during a D-cache look-up.
*/
- if ((c->processor_id & 0xff) <= PRID_REV_ENCODE_332(2, 4, 0))
- c->dcache.flags |= MIPS_CACHE_VTAG;
- if ((c->processor_id & 0xff) == PRID_REV_ENCODE_332(2, 4, 0))
- write_c0_config6(read_c0_config6() | MIPS_CONF6_SYND);
- if (((c->processor_id & 0xff00) == PRID_IMP_1074K) &&
- ((c->processor_id & 0xff) <= PRID_REV_ENCODE_332(1, 1, 0))) {
- c->dcache.flags |= MIPS_CACHE_VTAG;
- write_c0_config6(read_c0_config6() | MIPS_CONF6_SYND);
+ switch (imp) {
+ case PRID_IMP_74K:
+ if (rev <= PRID_REV_ENCODE_332(2, 4, 0))
+ present = 1;
+ if (rev == PRID_REV_ENCODE_332(2, 4, 0))
+ write_c0_config6(read_c0_config6() | MTI_CONF6_SYND);
+ break;
+ case PRID_IMP_1074K:
+ if (rev <= PRID_REV_ENCODE_332(1, 1, 0)) {
+ present = 1;
+ write_c0_config6(read_c0_config6() | MTI_CONF6_SYND);
+ }
+ break;
+ default:
+ BUG();
}
+
+ return present;
+}
+
+static void b5k_instruction_hazard(void)
+{
+ __sync();
+ __sync();
+ __asm__ __volatile__(
+ " nop; nop; nop; nop; nop; nop; nop; nop\n"
+ " nop; nop; nop; nop; nop; nop; nop; nop\n"
+ " nop; nop; nop; nop; nop; nop; nop; nop\n"
+ " nop; nop; nop; nop; nop; nop; nop; nop\n"
+ : : : "memory");
}
-static char *way_string[] __cpuinitdata = { NULL, "direct mapped", "2-way",
- "3-way", "4-way", "5-way", "6-way", "7-way", "8-way"
+static char *way_string[] = { NULL, "direct mapped", "2-way",
+ "3-way", "4-way", "5-way", "6-way", "7-way", "8-way",
+ "9-way", "10-way", "11-way", "12-way",
+ "13-way", "14-way", "15-way", "16-way",
};
-static void __cpuinit probe_pcache(void)
+static void probe_pcache(void)
{
struct cpuinfo_mips *c = &current_cpu_data;
unsigned int config = read_c0_config();
unsigned int prid = read_c0_prid();
+ int has_74k_erratum = 0;
unsigned long config1;
unsigned int lsize;
- switch (c->cputype) {
+ switch (current_cpu_type()) {
case CPU_R4600: /* QED style two way caches? */
case CPU_R4700:
case CPU_R5000:
@@ -827,7 +1018,6 @@ static void __cpuinit probe_pcache(void)
c->options |= MIPS_CPU_CACHE_CDEX_P;
break;
- case CPU_R5432:
case CPU_R5500:
icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
@@ -880,6 +1070,7 @@ static void __cpuinit probe_pcache(void)
case CPU_R10000:
case CPU_R12000:
case CPU_R14000:
+ case CPU_R16000:
icache_size = 1 << (12 + ((config & R10K_CONF_IC) >> 29));
c->icache.linesz = 64;
c->icache.ways = 2;
@@ -893,49 +1084,6 @@ static void __cpuinit probe_pcache(void)
c->options |= MIPS_CPU_PREFETCH;
break;
- case CPU_VR4133:
- write_c0_config(config & ~VR41_CONF_P4K);
- case CPU_VR4131:
- /* Workaround for cache instruction bug of VR4131 */
- if (c->processor_id == 0x0c80U || c->processor_id == 0x0c81U ||
- c->processor_id == 0x0c82U) {
- config |= 0x00400000U;
- if (c->processor_id == 0x0c80U)
- config |= VR41_CONF_BP;
- write_c0_config(config);
- } else
- c->options |= MIPS_CPU_CACHE_CDEX_P;
-
- icache_size = 1 << (10 + ((config & CONF_IC) >> 9));
- c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
- c->icache.ways = 2;
- c->icache.waybit = __ffs(icache_size/2);
-
- dcache_size = 1 << (10 + ((config & CONF_DC) >> 6));
- c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
- c->dcache.ways = 2;
- c->dcache.waybit = __ffs(dcache_size/2);
- break;
-
- case CPU_VR41XX:
- case CPU_VR4111:
- case CPU_VR4121:
- case CPU_VR4122:
- case CPU_VR4181:
- case CPU_VR4181A:
- icache_size = 1 << (10 + ((config & CONF_IC) >> 9));
- c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
- c->icache.ways = 1;
- c->icache.waybit = 0; /* doesn't matter */
-
- dcache_size = 1 << (10 + ((config & CONF_DC) >> 6));
- c->dcache.linesz = 16 << ((config & CONF_DB) >> 4);
- c->dcache.ways = 1;
- c->dcache.waybit = 0; /* does not matter */
-
- c->options |= MIPS_CPU_CACHE_CDEX_P;
- break;
-
case CPU_RM7000:
rm7k_erratum31();
@@ -953,7 +1101,7 @@ static void __cpuinit probe_pcache(void)
c->options |= MIPS_CPU_PREFETCH;
break;
- case CPU_LOONGSON2:
+ case CPU_LOONGSON2EF:
icache_size = 1 << (12 + ((config & CONF_IC) >> 9));
c->icache.linesz = 16 << ((config & CONF_IB) >> 5);
if (prid & 0x3)
@@ -971,6 +1119,52 @@ static void __cpuinit probe_pcache(void)
c->dcache.waybit = 0;
break;
+ case CPU_LOONGSON64:
+ config1 = read_c0_config1();
+ lsize = (config1 >> 19) & 7;
+ if (lsize)
+ c->icache.linesz = 2 << lsize;
+ else
+ c->icache.linesz = 0;
+ c->icache.sets = 64 << ((config1 >> 22) & 7);
+ c->icache.ways = 1 + ((config1 >> 16) & 7);
+ icache_size = c->icache.sets *
+ c->icache.ways *
+ c->icache.linesz;
+ c->icache.waybit = 0;
+
+ lsize = (config1 >> 10) & 7;
+ if (lsize)
+ c->dcache.linesz = 2 << lsize;
+ else
+ c->dcache.linesz = 0;
+ c->dcache.sets = 64 << ((config1 >> 13) & 7);
+ c->dcache.ways = 1 + ((config1 >> 7) & 7);
+ dcache_size = c->dcache.sets *
+ c->dcache.ways *
+ c->dcache.linesz;
+ c->dcache.waybit = 0;
+ if ((c->processor_id & (PRID_IMP_MASK | PRID_REV_MASK)) >=
+ (PRID_IMP_LOONGSON_64C | PRID_REV_LOONGSON3A_R2_0) ||
+ (c->processor_id & PRID_IMP_MASK) == PRID_IMP_LOONGSON_64R)
+ c->options |= MIPS_CPU_PREFETCH;
+ break;
+
+ case CPU_CAVIUM_OCTEON3:
+ /* For now lie about the number of ways. */
+ c->icache.linesz = 128;
+ c->icache.sets = 16;
+ c->icache.ways = 8;
+ c->icache.flags |= MIPS_CACHE_VTAG;
+ icache_size = c->icache.sets * c->icache.ways * c->icache.linesz;
+
+ c->dcache.linesz = 128;
+ c->dcache.ways = 8;
+ c->dcache.sets = 8;
+ dcache_size = c->dcache.sets * c->dcache.ways * c->dcache.linesz;
+ c->options |= MIPS_CPU_PREFETCH;
+ break;
+
default:
if (!(config & MIPS_CONF_M))
panic("Don't know how to probe P-caches on this cpu.");
@@ -981,10 +1175,14 @@ static void __cpuinit probe_pcache(void)
*/
config1 = read_c0_config1();
- if ((lsize = ((config1 >> 19) & 7)))
- c->icache.linesz = 2 << lsize;
- else
- c->icache.linesz = lsize;
+ lsize = (config1 >> 19) & 7;
+
+ /* IL == 7 is reserved */
+ if (lsize == 7)
+ panic("Invalid icache line size");
+
+ c->icache.linesz = lsize ? 2 << lsize : 0;
+
c->icache.sets = 32 << (((config1 >> 22) + 1) & 7);
c->icache.ways = 1 + ((config1 >> 16) & 7);
@@ -993,7 +1191,7 @@ static void __cpuinit probe_pcache(void)
c->icache.linesz;
c->icache.waybit = __ffs(icache_size/c->icache.ways);
- if (config & 0x8) /* VI bit */
+ if (config & MIPS_CONF_VI)
c->icache.flags |= MIPS_CACHE_VTAG;
/*
@@ -1001,10 +1199,14 @@ static void __cpuinit probe_pcache(void)
*/
c->dcache.flags = 0;
- if ((lsize = ((config1 >> 10) & 7)))
- c->dcache.linesz = 2 << lsize;
- else
- c->dcache.linesz= lsize;
+ lsize = (config1 >> 10) & 7;
+
+ /* DL == 7 is reserved */
+ if (lsize == 7)
+ panic("Invalid dcache line size");
+
+ c->dcache.linesz = lsize ? 2 << lsize : 0;
+
c->dcache.sets = 32 << (((config1 >> 13) + 1) & 7);
c->dcache.ways = 1 + ((config1 >> 7) & 7);
@@ -1025,7 +1227,8 @@ static void __cpuinit probe_pcache(void)
* presumably no vendor is shipping his hardware in the "bad"
* configuration.
*/
- if ((prid & 0xff00) == PRID_IMP_R4000 && (prid & 0xff) < 0x40 &&
+ if ((prid & PRID_IMP_MASK) == PRID_IMP_R4000 &&
+ (prid & PRID_REV_MASK) < PRID_REV_R4400 &&
!(config & CONF_SC) && c->icache.linesz != 16 &&
PAGE_SIZE <= 0x8000)
panic("Improper R4000SC processor configuration detected");
@@ -1040,45 +1243,73 @@ static void __cpuinit probe_pcache(void)
dcache_size / (c->dcache.linesz * c->dcache.ways) : 0;
/*
- * R10000 and R12000 P-caches are odd in a positive way. They're 32kB
- * 2-way virtually indexed so normally would suffer from aliases. So
+ * R1x000 P-caches are odd in a positive way. They're 32kB 2-way
+ * virtually indexed so normally would suffer from aliases. So
* normally they'd suffer from aliases but magic in the hardware deals
* with that for us so we don't need to take care ourselves.
*/
- switch (c->cputype) {
+ switch (current_cpu_type()) {
case CPU_20KC:
case CPU_25KF:
+ case CPU_I6400:
+ case CPU_I6500:
case CPU_SB1:
case CPU_SB1A:
- case CPU_XLR:
c->dcache.flags |= MIPS_CACHE_PINDEX;
break;
case CPU_R10000:
case CPU_R12000:
case CPU_R14000:
+ case CPU_R16000:
break;
+ case CPU_74K:
+ case CPU_1074K:
+ has_74k_erratum = alias_74k_erratum(c);
+ fallthrough;
case CPU_M14KC:
case CPU_M14KEC:
case CPU_24K:
case CPU_34K:
- case CPU_74K:
case CPU_1004K:
- if (c->cputype == CPU_74K)
- alias_74k_erratum(c);
- if ((read_c0_config7() & (1 << 16))) {
- /* effectively physically indexed dcache,
- thus no virtual aliases. */
+ case CPU_INTERAPTIV:
+ case CPU_P5600:
+ case CPU_PROAPTIV:
+ case CPU_M5150:
+ case CPU_QEMU_GENERIC:
+ case CPU_P6600:
+ case CPU_M6250:
+ if (!(read_c0_config7() & MIPS_CONF7_IAR) &&
+ (c->icache.waysize > PAGE_SIZE))
+ c->icache.flags |= MIPS_CACHE_ALIASES;
+ if (!has_74k_erratum && (read_c0_config7() & MIPS_CONF7_AR)) {
+ /*
+ * Effectively physically indexed dcache,
+ * thus no virtual aliases.
+ */
c->dcache.flags |= MIPS_CACHE_PINDEX;
break;
}
+ fallthrough;
default:
- if (c->dcache.waysize > PAGE_SIZE)
+ if (has_74k_erratum || c->dcache.waysize > PAGE_SIZE)
c->dcache.flags |= MIPS_CACHE_ALIASES;
}
- switch (c->cputype) {
+ /* Physically indexed caches don't suffer from virtual aliasing */
+ if (c->dcache.flags & MIPS_CACHE_PINDEX)
+ c->dcache.flags &= ~MIPS_CACHE_ALIASES;
+
+ /*
+ * In systems with CM the icache fills from L2 or closer caches, and
+ * thus sees remote stores without needing to write them back any
+ * further than that.
+ */
+ if (mips_cm_present())
+ c->icache.flags |= MIPS_IC_SNOOPS_REMOTE;
+
+ switch (current_cpu_type()) {
case CPU_20KC:
/*
* Some older 20Kc chips doesn't have the 'VI' bit in
@@ -1088,29 +1319,62 @@ static void __cpuinit probe_pcache(void)
break;
case CPU_ALCHEMY:
+ case CPU_I6400:
+ case CPU_I6500:
c->icache.flags |= MIPS_CACHE_IC_F_DC;
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
+ case CPU_BMIPS5000:
+ c->icache.flags |= MIPS_CACHE_IC_F_DC;
+ /* Cache aliases are handled in hardware; allow HIGHMEM */
+ c->dcache.flags &= ~MIPS_CACHE_ALIASES;
+ break;
+
+ case CPU_LOONGSON2EF:
+ /*
+ * LOONGSON2 has 4 way icache, but when using indexed cache op,
+ * one op will act on all 4 ways
+ */
+ c->icache.ways = 1;
+ }
- printk("Primary instruction cache %ldkB, %s, %s, linesize %d bytes.\n",
- icache_size >> 10,
- c->icache.flags & MIPS_CACHE_VTAG ? "VIVT" : "VIPT",
- way_string[c->icache.ways], c->icache.linesz);
+ pr_info("Primary instruction cache %ldkB, %s, %s, linesize %d bytes.\n",
+ icache_size >> 10,
+ c->icache.flags & MIPS_CACHE_VTAG ? "VIVT" : "VIPT",
+ way_string[c->icache.ways], c->icache.linesz);
- printk("Primary data cache %ldkB, %s, %s, %s, linesize %d bytes\n",
- dcache_size >> 10, way_string[c->dcache.ways],
- (c->dcache.flags & MIPS_CACHE_PINDEX) ? "PIPT" : "VIPT",
- (c->dcache.flags & MIPS_CACHE_ALIASES) ?
+ pr_info("Primary data cache %ldkB, %s, %s, %s, linesize %d bytes\n",
+ dcache_size >> 10, way_string[c->dcache.ways],
+ (c->dcache.flags & MIPS_CACHE_PINDEX) ? "PIPT" : "VIPT",
+ (c->dcache.flags & MIPS_CACHE_ALIASES) ?
"cache aliases" : "no aliases",
- c->dcache.linesz);
+ c->dcache.linesz);
+}
+
+static void probe_vcache(void)
+{
+ struct cpuinfo_mips *c = &current_cpu_data;
+ unsigned int config2, lsize;
+
+ if (current_cpu_type() != CPU_LOONGSON64)
+ return;
+
+ config2 = read_c0_config2();
+ if ((lsize = ((config2 >> 20) & 15)))
+ c->vcache.linesz = 2 << lsize;
+ else
+ c->vcache.linesz = lsize;
+
+ c->vcache.sets = 64 << ((config2 >> 24) & 15);
+ c->vcache.ways = 1 + ((config2 >> 16) & 15);
+
+ vcache_size = c->vcache.sets * c->vcache.ways * c->vcache.linesz;
+
+ c->vcache.waybit = 0;
+ c->vcache.waysize = vcache_size / c->vcache.ways;
+
+ pr_info("Unified victim cache %ldkB %s, linesize %d bytes.\n",
+ vcache_size >> 10, way_string[c->vcache.ways], c->vcache.linesz);
}
/*
@@ -1119,7 +1383,7 @@ static void __cpuinit probe_pcache(void)
* executes in KSEG1 space or else you will crash and burn badly. You have
* been warned.
*/
-static int __cpuinit probe_scache(void)
+static int probe_scache(void)
{
unsigned long flags, addr, begin, end, pow2;
unsigned int config = read_c0_config();
@@ -1169,13 +1433,12 @@ static int __cpuinit probe_scache(void)
scache_size = addr;
c->scache.linesz = 16 << ((config & R4K_CONF_SB) >> 22);
c->scache.ways = 1;
- c->dcache.waybit = 0; /* does not matter */
+ c->scache.waybit = 0; /* does not matter */
return 1;
}
-#if defined(CONFIG_CPU_LOONGSON2)
-static void __init loongson2_sc_init(void)
+static void loongson2_sc_init(void)
{
struct cpuinfo_mips *c = &current_cpu_data;
@@ -1190,13 +1453,39 @@ static void __init loongson2_sc_init(void)
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);
+static void loongson3_sc_init(void)
+{
+ struct cpuinfo_mips *c = &current_cpu_data;
+ unsigned int config2, lsize;
+
+ config2 = read_c0_config2();
+ lsize = (config2 >> 4) & 15;
+ if (lsize)
+ c->scache.linesz = 2 << lsize;
+ else
+ c->scache.linesz = 0;
+ c->scache.sets = 64 << ((config2 >> 8) & 15);
+ c->scache.ways = 1 + (config2 & 15);
+
+ /* Loongson-3 has 4-Scache banks, while Loongson-2K have only 2 banks */
+ if ((c->processor_id & PRID_IMP_MASK) == PRID_IMP_LOONGSON_64R)
+ c->scache.sets *= 2;
+ else
+ c->scache.sets *= 4;
+
+ scache_size = c->scache.sets * c->scache.ways * c->scache.linesz;
-static void __cpuinit setup_scache(void)
+ c->scache.waybit = 0;
+ c->scache.waysize = scache_size / 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);
+ if (scache_size)
+ c->options |= MIPS_CPU_INCLUSIVE_CACHES;
+ return;
+}
+
+static void setup_scache(void)
{
struct cpuinfo_mips *c = &current_cpu_data;
unsigned int config = read_c0_config();
@@ -1207,7 +1496,7 @@ static void __cpuinit setup_scache(void)
* processors don't have a S-cache that would be relevant to the
* Linux memory management.
*/
- switch (c->cputype) {
+ switch (current_cpu_type()) {
case CPU_R4000SC:
case CPU_R4000MC:
case CPU_R4400SC:
@@ -1220,6 +1509,7 @@ static void __cpuinit setup_scache(void)
case CPU_R10000:
case CPU_R12000:
case CPU_R14000:
+ case CPU_R16000:
scache_size = 0x80000 << ((config & R10K_CONF_SS) >> 16);
c->scache.linesz = 64 << ((config >> 13) & 1);
c->scache.ways = 2;
@@ -1240,25 +1530,34 @@ static void __cpuinit setup_scache(void)
#endif
return;
-#if defined(CONFIG_CPU_LOONGSON2)
- case CPU_LOONGSON2:
+ case CPU_LOONGSON2EF:
loongson2_sc_init();
return;
-#endif
- case CPU_XLP:
+
+ case CPU_LOONGSON64:
+ loongson3_sc_init();
+ return;
+
+ case CPU_CAVIUM_OCTEON3:
/* don't need to worry about L2, fully coherent */
return;
default:
- if (c->isa_level & (MIPS_CPU_ISA_M32R1 | MIPS_CPU_ISA_M32R2 |
- MIPS_CPU_ISA_M64R1 | MIPS_CPU_ISA_M64R2)) {
+ if (c->isa_level & (MIPS_CPU_ISA_M32R1 | MIPS_CPU_ISA_M64R1 |
+ MIPS_CPU_ISA_M32R2 | MIPS_CPU_ISA_M64R2 |
+ MIPS_CPU_ISA_M32R5 | MIPS_CPU_ISA_M64R5 |
+ MIPS_CPU_ISA_M32R6 | MIPS_CPU_ISA_M64R6)) {
#ifdef CONFIG_MIPS_CPU_SCACHE
if (mips_sc_init ()) {
scache_size = c->scache.ways * c->scache.sets * c->scache.linesz;
printk("MIPS secondary cache %ldkB, %s, linesize %d bytes.\n",
scache_size >> 10,
way_string[c->scache.ways], c->scache.linesz);
+
+ if (current_cpu_type() == CPU_BMIPS5000)
+ c->options |= MIPS_CPU_INCLUSIVE_CACHES;
}
+
#else
if (!(c->scache.flags & MIPS_CACHE_NOT_PRESENT))
panic("Dunno how to handle MIPS32 / MIPS64 second level cache");
@@ -1329,7 +1628,7 @@ static void nxp_pr4450_fixup_config(void)
NXP_BARRIER();
}
-static int __cpuinitdata cca = -1;
+static int cca = -1;
static int __init cca_setup(char *str)
{
@@ -1340,7 +1639,7 @@ static int __init cca_setup(char *str)
early_param("cca", cca_setup);
-static void __cpuinit coherency_setup(void)
+static void coherency_setup(void)
{
if (cca < 0 || cca > 7)
cca = read_c0_config() & CONF_CM_CMASK;
@@ -1351,7 +1650,7 @@ static void __cpuinit coherency_setup(void)
/*
* c0_status.cu=0 specifies that updates by the sc instruction use
- * the coherency mode specified by the TLB; 1 means cachable
+ * the coherency mode specified by the TLB; 1 means cacheable
* coherent update on write will be used. Not all processors have
* this bit and; some wire it to zero, others like Toshiba had the
* silly idea of putting something else there ...
@@ -1380,13 +1679,12 @@ static void __cpuinit coherency_setup(void)
}
}
-static void __cpuinit r4k_cache_error_setup(void)
+static void r4k_cache_error_setup(void)
{
extern char __weak except_vec2_generic;
extern char __weak except_vec2_sb1;
- struct cpuinfo_mips *c = &current_cpu_data;
- switch (c->cputype) {
+ switch (current_cpu_type()) {
case CPU_SB1:
case CPU_SB1A:
set_uncached_handler(0x100, &except_vec2_sb1, 0x80);
@@ -1398,31 +1696,34 @@ static void __cpuinit r4k_cache_error_setup(void)
}
}
-void __cpuinit r4k_cache_init(void)
+void r4k_cache_init(void)
{
extern void build_clear_page(void);
extern void build_copy_page(void);
struct cpuinfo_mips *c = &current_cpu_data;
probe_pcache();
+ probe_vcache();
setup_scache();
r4k_blast_dcache_page_setup();
- r4k_blast_dcache_page_indexed_setup();
r4k_blast_dcache_setup();
r4k_blast_icache_page_setup();
- r4k_blast_icache_page_indexed_setup();
r4k_blast_icache_setup();
r4k_blast_scache_page_setup();
- r4k_blast_scache_page_indexed_setup();
r4k_blast_scache_setup();
+ r4k_blast_scache_node_setup();
+#ifdef CONFIG_EVA
+ r4k_blast_dcache_user_page_setup();
+ r4k_blast_icache_user_page_setup();
+#endif
/*
* Some MIPS32 and MIPS64 processors have physically indexed caches.
* This code supports virtually indexed processors and will be
* unnecessarily inefficient on physically indexed processors.
*/
- if (c->dcache.linesz)
+ if (c->dcache.linesz && cpu_has_dc_aliases)
shm_align_mask = max_t( unsigned long,
c->dcache.sets * c->dcache.linesz - 1,
PAGE_SIZE - 1);
@@ -1440,24 +1741,18 @@ void __cpuinit r4k_cache_init(void)
__flush_kernel_vmap_range = r4k_flush_kernel_vmap_range;
- flush_cache_sigtramp = r4k_flush_cache_sigtramp;
flush_icache_all = r4k_flush_icache_all;
- local_flush_data_cache_page = local_r4k_flush_data_cache_page;
flush_data_cache_page = r4k_flush_data_cache_page;
flush_icache_range = r4k_flush_icache_range;
local_flush_icache_range = local_r4k_flush_icache_range;
+ __flush_icache_user_range = r4k_flush_icache_user_range;
+ __local_flush_icache_user_range = local_r4k_flush_icache_user_range;
-#if defined(CONFIG_DMA_NONCOHERENT)
- if (coherentio) {
- _dma_cache_wback_inv = (void *)cache_noop;
- _dma_cache_wback = (void *)cache_noop;
- _dma_cache_inv = (void *)cache_noop;
- } else {
- _dma_cache_wback_inv = r4k_dma_cache_wback_inv;
- _dma_cache_wback = r4k_dma_cache_wback_inv;
- _dma_cache_inv = r4k_dma_cache_inv;
- }
-#endif
+#ifdef CONFIG_DMA_NONCOHERENT
+ _dma_cache_wback_inv = r4k_dma_cache_wback_inv;
+ _dma_cache_wback = r4k_dma_cache_wback_inv;
+ _dma_cache_inv = r4k_dma_cache_inv;
+#endif /* CONFIG_DMA_NONCOHERENT */
build_clear_page();
build_copy_page();
@@ -1471,4 +1766,66 @@ void __cpuinit r4k_cache_init(void)
coherency_setup();
board_cache_error_setup = r4k_cache_error_setup;
+
+ /*
+ * Per-CPU overrides
+ */
+ switch (current_cpu_type()) {
+ case CPU_BMIPS4350:
+ case CPU_BMIPS4380:
+ /* No IPI is needed because all CPUs share the same D$ */
+ flush_data_cache_page = r4k_blast_dcache_page;
+ break;
+ case CPU_BMIPS5000:
+ /* We lose our superpowers if L2 is disabled */
+ if (c->scache.flags & MIPS_CACHE_NOT_PRESENT)
+ break;
+
+ /* I$ fills from D$ just by emptying the write buffers */
+ flush_cache_page = (void *)b5k_instruction_hazard;
+ flush_cache_range = (void *)b5k_instruction_hazard;
+ flush_data_cache_page = (void *)b5k_instruction_hazard;
+ flush_icache_range = (void *)b5k_instruction_hazard;
+ local_flush_icache_range = (void *)b5k_instruction_hazard;
+
+
+ /* Optimization: an L2 flush implicitly flushes the L1 */
+ current_cpu_data.options |= MIPS_CPU_INCLUSIVE_CACHES;
+ break;
+ case CPU_LOONGSON64:
+ /* Loongson-3 maintains cache coherency by hardware */
+ __flush_cache_all = cache_noop;
+ __flush_cache_vmap = cache_noop;
+ __flush_cache_vunmap = cache_noop;
+ __flush_kernel_vmap_range = (void *)cache_noop;
+ flush_cache_mm = (void *)cache_noop;
+ flush_cache_page = (void *)cache_noop;
+ flush_cache_range = (void *)cache_noop;
+ flush_icache_all = (void *)cache_noop;
+ flush_data_cache_page = (void *)cache_noop;
+ break;
+ }
+}
+
+static int r4k_cache_pm_notifier(struct notifier_block *self, unsigned long cmd,
+ void *v)
+{
+ switch (cmd) {
+ case CPU_PM_ENTER_FAILED:
+ case CPU_PM_EXIT:
+ coherency_setup();
+ break;
+ }
+
+ return NOTIFY_OK;
+}
+
+static struct notifier_block r4k_cache_pm_notifier_block = {
+ .notifier_call = r4k_cache_pm_notifier,
+};
+
+static int __init r4k_cache_init_pm(void)
+{
+ return cpu_pm_register_notifier(&r4k_cache_pm_notifier_block);
}
+arch_initcall(r4k_cache_init_pm);
generated by cgit (git 2.34.1) at 2025-12-25 06:11:45 +0000