mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-15 10:06:52 +07:00
6d3e6d3640
Since commit 3e4d3af501
"mm: stack based kmap_atomic()", it is actively
wrong to rely on fixed kmap type indices (namely KM_L2_CACHE) as
kmap_atomic() totally ignores them and a concurrent instance of it may
happily reuse any slot for any purpose. Because kmap_atomic() is now
able to deal with reentrancy, we can get rid of the ad hoc mapping here.
While the code is made much simpler, there is a needless cache flush
introduced by the usage of __kunmap_atomic(). It is not clear if the
performance difference to remove that is worth the cost in code
maintenance (I don't think there are that many highmem users on that
platform anyway) but that should be reconsidered when/if someone cares
enough to do some measurements.
Signed-off-by: Nicolas Pitre <nicolas.pitre@linaro.org>
351 lines
8.1 KiB
C
351 lines
8.1 KiB
C
/*
|
|
* arch/arm/mm/cache-feroceon-l2.c - Feroceon L2 cache controller support
|
|
*
|
|
* Copyright (C) 2008 Marvell Semiconductor
|
|
*
|
|
* This file is licensed under the terms of the GNU General Public
|
|
* License version 2. This program is licensed "as is" without any
|
|
* warranty of any kind, whether express or implied.
|
|
*
|
|
* References:
|
|
* - Unified Layer 2 Cache for Feroceon CPU Cores,
|
|
* Document ID MV-S104858-00, Rev. A, October 23 2007.
|
|
*/
|
|
|
|
#include <linux/init.h>
|
|
#include <linux/highmem.h>
|
|
#include <asm/cacheflush.h>
|
|
#include <plat/cache-feroceon-l2.h>
|
|
|
|
/*
|
|
* Low-level cache maintenance operations.
|
|
*
|
|
* As well as the regular 'clean/invalidate/flush L2 cache line by
|
|
* MVA' instructions, the Feroceon L2 cache controller also features
|
|
* 'clean/invalidate L2 range by MVA' operations.
|
|
*
|
|
* Cache range operations are initiated by writing the start and
|
|
* end addresses to successive cp15 registers, and process every
|
|
* cache line whose first byte address lies in the inclusive range
|
|
* [start:end].
|
|
*
|
|
* The cache range operations stall the CPU pipeline until completion.
|
|
*
|
|
* The range operations require two successive cp15 writes, in
|
|
* between which we don't want to be preempted.
|
|
*/
|
|
|
|
static inline unsigned long l2_get_va(unsigned long paddr)
|
|
{
|
|
#ifdef CONFIG_HIGHMEM
|
|
/*
|
|
* Because range ops can't be done on physical addresses,
|
|
* we simply install a virtual mapping for it only for the
|
|
* TLB lookup to occur, hence no need to flush the untouched
|
|
* memory mapping afterwards (note: a cache flush may happen
|
|
* in some circumstances depending on the path taken in kunmap_atomic).
|
|
*/
|
|
void *vaddr = kmap_atomic_pfn(paddr >> PAGE_SHIFT);
|
|
return (unsigned long)vaddr + (paddr & ~PAGE_MASK);
|
|
#else
|
|
return __phys_to_virt(paddr);
|
|
#endif
|
|
}
|
|
|
|
static inline void l2_put_va(unsigned long vaddr)
|
|
{
|
|
#ifdef CONFIG_HIGHMEM
|
|
kunmap_atomic((void *)vaddr);
|
|
#endif
|
|
}
|
|
|
|
static inline void l2_clean_pa(unsigned long addr)
|
|
{
|
|
__asm__("mcr p15, 1, %0, c15, c9, 3" : : "r" (addr));
|
|
}
|
|
|
|
static inline void l2_clean_pa_range(unsigned long start, unsigned long end)
|
|
{
|
|
unsigned long va_start, va_end, flags;
|
|
|
|
/*
|
|
* Make sure 'start' and 'end' reference the same page, as
|
|
* L2 is PIPT and range operations only do a TLB lookup on
|
|
* the start address.
|
|
*/
|
|
BUG_ON((start ^ end) >> PAGE_SHIFT);
|
|
|
|
va_start = l2_get_va(start);
|
|
va_end = va_start + (end - start);
|
|
raw_local_irq_save(flags);
|
|
__asm__("mcr p15, 1, %0, c15, c9, 4\n\t"
|
|
"mcr p15, 1, %1, c15, c9, 5"
|
|
: : "r" (va_start), "r" (va_end));
|
|
raw_local_irq_restore(flags);
|
|
l2_put_va(va_start);
|
|
}
|
|
|
|
static inline void l2_clean_inv_pa(unsigned long addr)
|
|
{
|
|
__asm__("mcr p15, 1, %0, c15, c10, 3" : : "r" (addr));
|
|
}
|
|
|
|
static inline void l2_inv_pa(unsigned long addr)
|
|
{
|
|
__asm__("mcr p15, 1, %0, c15, c11, 3" : : "r" (addr));
|
|
}
|
|
|
|
static inline void l2_inv_pa_range(unsigned long start, unsigned long end)
|
|
{
|
|
unsigned long va_start, va_end, flags;
|
|
|
|
/*
|
|
* Make sure 'start' and 'end' reference the same page, as
|
|
* L2 is PIPT and range operations only do a TLB lookup on
|
|
* the start address.
|
|
*/
|
|
BUG_ON((start ^ end) >> PAGE_SHIFT);
|
|
|
|
va_start = l2_get_va(start);
|
|
va_end = va_start + (end - start);
|
|
raw_local_irq_save(flags);
|
|
__asm__("mcr p15, 1, %0, c15, c11, 4\n\t"
|
|
"mcr p15, 1, %1, c15, c11, 5"
|
|
: : "r" (va_start), "r" (va_end));
|
|
raw_local_irq_restore(flags);
|
|
l2_put_va(va_start);
|
|
}
|
|
|
|
static inline void l2_inv_all(void)
|
|
{
|
|
__asm__("mcr p15, 1, %0, c15, c11, 0" : : "r" (0));
|
|
}
|
|
|
|
/*
|
|
* Linux primitives.
|
|
*
|
|
* Note that the end addresses passed to Linux primitives are
|
|
* noninclusive, while the hardware cache range operations use
|
|
* inclusive start and end addresses.
|
|
*/
|
|
#define CACHE_LINE_SIZE 32
|
|
#define MAX_RANGE_SIZE 1024
|
|
|
|
static int l2_wt_override;
|
|
|
|
static unsigned long calc_range_end(unsigned long start, unsigned long end)
|
|
{
|
|
unsigned long range_end;
|
|
|
|
BUG_ON(start & (CACHE_LINE_SIZE - 1));
|
|
BUG_ON(end & (CACHE_LINE_SIZE - 1));
|
|
|
|
/*
|
|
* Try to process all cache lines between 'start' and 'end'.
|
|
*/
|
|
range_end = end;
|
|
|
|
/*
|
|
* Limit the number of cache lines processed at once,
|
|
* since cache range operations stall the CPU pipeline
|
|
* until completion.
|
|
*/
|
|
if (range_end > start + MAX_RANGE_SIZE)
|
|
range_end = start + MAX_RANGE_SIZE;
|
|
|
|
/*
|
|
* Cache range operations can't straddle a page boundary.
|
|
*/
|
|
if (range_end > (start | (PAGE_SIZE - 1)) + 1)
|
|
range_end = (start | (PAGE_SIZE - 1)) + 1;
|
|
|
|
return range_end;
|
|
}
|
|
|
|
static void feroceon_l2_inv_range(unsigned long start, unsigned long end)
|
|
{
|
|
/*
|
|
* Clean and invalidate partial first cache line.
|
|
*/
|
|
if (start & (CACHE_LINE_SIZE - 1)) {
|
|
l2_clean_inv_pa(start & ~(CACHE_LINE_SIZE - 1));
|
|
start = (start | (CACHE_LINE_SIZE - 1)) + 1;
|
|
}
|
|
|
|
/*
|
|
* Clean and invalidate partial last cache line.
|
|
*/
|
|
if (start < end && end & (CACHE_LINE_SIZE - 1)) {
|
|
l2_clean_inv_pa(end & ~(CACHE_LINE_SIZE - 1));
|
|
end &= ~(CACHE_LINE_SIZE - 1);
|
|
}
|
|
|
|
/*
|
|
* Invalidate all full cache lines between 'start' and 'end'.
|
|
*/
|
|
while (start < end) {
|
|
unsigned long range_end = calc_range_end(start, end);
|
|
l2_inv_pa_range(start, range_end - CACHE_LINE_SIZE);
|
|
start = range_end;
|
|
}
|
|
|
|
dsb();
|
|
}
|
|
|
|
static void feroceon_l2_clean_range(unsigned long start, unsigned long end)
|
|
{
|
|
/*
|
|
* If L2 is forced to WT, the L2 will always be clean and we
|
|
* don't need to do anything here.
|
|
*/
|
|
if (!l2_wt_override) {
|
|
start &= ~(CACHE_LINE_SIZE - 1);
|
|
end = (end + CACHE_LINE_SIZE - 1) & ~(CACHE_LINE_SIZE - 1);
|
|
while (start != end) {
|
|
unsigned long range_end = calc_range_end(start, end);
|
|
l2_clean_pa_range(start, range_end - CACHE_LINE_SIZE);
|
|
start = range_end;
|
|
}
|
|
}
|
|
|
|
dsb();
|
|
}
|
|
|
|
static void feroceon_l2_flush_range(unsigned long start, unsigned long end)
|
|
{
|
|
start &= ~(CACHE_LINE_SIZE - 1);
|
|
end = (end + CACHE_LINE_SIZE - 1) & ~(CACHE_LINE_SIZE - 1);
|
|
while (start != end) {
|
|
unsigned long range_end = calc_range_end(start, end);
|
|
if (!l2_wt_override)
|
|
l2_clean_pa_range(start, range_end - CACHE_LINE_SIZE);
|
|
l2_inv_pa_range(start, range_end - CACHE_LINE_SIZE);
|
|
start = range_end;
|
|
}
|
|
|
|
dsb();
|
|
}
|
|
|
|
|
|
/*
|
|
* Routines to disable and re-enable the D-cache and I-cache at run
|
|
* time. These are necessary because the L2 cache can only be enabled
|
|
* or disabled while the L1 Dcache and Icache are both disabled.
|
|
*/
|
|
static int __init flush_and_disable_dcache(void)
|
|
{
|
|
u32 cr;
|
|
|
|
cr = get_cr();
|
|
if (cr & CR_C) {
|
|
unsigned long flags;
|
|
|
|
raw_local_irq_save(flags);
|
|
flush_cache_all();
|
|
set_cr(cr & ~CR_C);
|
|
raw_local_irq_restore(flags);
|
|
return 1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
static void __init enable_dcache(void)
|
|
{
|
|
u32 cr;
|
|
|
|
cr = get_cr();
|
|
set_cr(cr | CR_C);
|
|
}
|
|
|
|
static void __init __invalidate_icache(void)
|
|
{
|
|
__asm__("mcr p15, 0, %0, c7, c5, 0" : : "r" (0));
|
|
}
|
|
|
|
static int __init invalidate_and_disable_icache(void)
|
|
{
|
|
u32 cr;
|
|
|
|
cr = get_cr();
|
|
if (cr & CR_I) {
|
|
set_cr(cr & ~CR_I);
|
|
__invalidate_icache();
|
|
return 1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
static void __init enable_icache(void)
|
|
{
|
|
u32 cr;
|
|
|
|
cr = get_cr();
|
|
set_cr(cr | CR_I);
|
|
}
|
|
|
|
static inline u32 read_extra_features(void)
|
|
{
|
|
u32 u;
|
|
|
|
__asm__("mrc p15, 1, %0, c15, c1, 0" : "=r" (u));
|
|
|
|
return u;
|
|
}
|
|
|
|
static inline void write_extra_features(u32 u)
|
|
{
|
|
__asm__("mcr p15, 1, %0, c15, c1, 0" : : "r" (u));
|
|
}
|
|
|
|
static void __init disable_l2_prefetch(void)
|
|
{
|
|
u32 u;
|
|
|
|
/*
|
|
* Read the CPU Extra Features register and verify that the
|
|
* Disable L2 Prefetch bit is set.
|
|
*/
|
|
u = read_extra_features();
|
|
if (!(u & 0x01000000)) {
|
|
printk(KERN_INFO "Feroceon L2: Disabling L2 prefetch.\n");
|
|
write_extra_features(u | 0x01000000);
|
|
}
|
|
}
|
|
|
|
static void __init enable_l2(void)
|
|
{
|
|
u32 u;
|
|
|
|
u = read_extra_features();
|
|
if (!(u & 0x00400000)) {
|
|
int i, d;
|
|
|
|
printk(KERN_INFO "Feroceon L2: Enabling L2\n");
|
|
|
|
d = flush_and_disable_dcache();
|
|
i = invalidate_and_disable_icache();
|
|
l2_inv_all();
|
|
write_extra_features(u | 0x00400000);
|
|
if (i)
|
|
enable_icache();
|
|
if (d)
|
|
enable_dcache();
|
|
}
|
|
}
|
|
|
|
void __init feroceon_l2_init(int __l2_wt_override)
|
|
{
|
|
l2_wt_override = __l2_wt_override;
|
|
|
|
disable_l2_prefetch();
|
|
|
|
outer_cache.inv_range = feroceon_l2_inv_range;
|
|
outer_cache.clean_range = feroceon_l2_clean_range;
|
|
outer_cache.flush_range = feroceon_l2_flush_range;
|
|
|
|
enable_l2();
|
|
|
|
printk(KERN_INFO "Feroceon L2: Cache support initialised%s.\n",
|
|
l2_wt_override ? ", in WT override mode" : "");
|
|
}
|