linux/arch/mips/mm/dma-noncoherent.c
Paul Cercueil 1660710cf5 MIPS: mm: XBurst CPU requires sync after DMA
I am not sure why this is required, but if this is not enabled, reading
from a buffer in which data has been DMA'd may read incorrect values.

This used to happen for instance in mmc_app_send_scr()
(drivers/mmc/core/sd_ops.c), where data is DMA'd to a buffer then copied
by the CPU to a different location.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
2021-06-01 11:44:46 +02:00

146 lines
3.6 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (C) 2000 Ani Joshi <ajoshi@unixbox.com>
* Copyright (C) 2000, 2001, 06 Ralf Baechle <ralf@linux-mips.org>
* swiped from i386, and cloned for MIPS by Geert, polished by Ralf.
*/
#include <linux/dma-direct.h>
#include <linux/dma-map-ops.h>
#include <linux/highmem.h>
#include <asm/cache.h>
#include <asm/cpu-type.h>
#include <asm/io.h>
/*
* The affected CPUs below in 'cpu_needs_post_dma_flush()' can speculatively
* fill random cachelines with stale data at any time, requiring an extra
* flush post-DMA.
*
* Warning on the terminology - Linux calls an uncached area coherent; MIPS
* terminology calls memory areas with hardware maintained coherency coherent.
*
* Note that the R14000 and R16000 should also be checked for in this condition.
* However this function is only called on non-I/O-coherent systems and only the
* R10000 and R12000 are used in such systems, the SGI IP28 Indigo² rsp.
* SGI IP32 aka O2.
*/
static inline bool cpu_needs_post_dma_flush(void)
{
switch (boot_cpu_type()) {
case CPU_R10000:
case CPU_R12000:
case CPU_BMIPS5000:
case CPU_LOONGSON2EF:
case CPU_XBURST:
return true;
default:
/*
* Presence of MAARs suggests that the CPU supports
* speculatively prefetching data, and therefore requires
* the post-DMA flush/invalidate.
*/
return cpu_has_maar;
}
}
void arch_dma_prep_coherent(struct page *page, size_t size)
{
dma_cache_wback_inv((unsigned long)page_address(page), size);
}
void *arch_dma_set_uncached(void *addr, size_t size)
{
return (void *)(__pa(addr) + UNCAC_BASE);
}
static inline void dma_sync_virt_for_device(void *addr, size_t size,
enum dma_data_direction dir)
{
switch (dir) {
case DMA_TO_DEVICE:
dma_cache_wback((unsigned long)addr, size);
break;
case DMA_FROM_DEVICE:
dma_cache_inv((unsigned long)addr, size);
break;
case DMA_BIDIRECTIONAL:
dma_cache_wback_inv((unsigned long)addr, size);
break;
default:
BUG();
}
}
static inline void dma_sync_virt_for_cpu(void *addr, size_t size,
enum dma_data_direction dir)
{
switch (dir) {
case DMA_TO_DEVICE:
break;
case DMA_FROM_DEVICE:
case DMA_BIDIRECTIONAL:
dma_cache_inv((unsigned long)addr, size);
break;
default:
BUG();
}
}
/*
* A single sg entry may refer to multiple physically contiguous pages. But
* we still need to process highmem pages individually. If highmem is not
* configured then the bulk of this loop gets optimized out.
*/
static inline void dma_sync_phys(phys_addr_t paddr, size_t size,
enum dma_data_direction dir, bool for_device)
{
struct page *page = pfn_to_page(paddr >> PAGE_SHIFT);
unsigned long offset = paddr & ~PAGE_MASK;
size_t left = size;
do {
size_t len = left;
void *addr;
if (PageHighMem(page)) {
if (offset + len > PAGE_SIZE)
len = PAGE_SIZE - offset;
}
addr = kmap_atomic(page);
if (for_device)
dma_sync_virt_for_device(addr + offset, len, dir);
else
dma_sync_virt_for_cpu(addr + offset, len, dir);
kunmap_atomic(addr);
offset = 0;
page++;
left -= len;
} while (left);
}
void arch_sync_dma_for_device(phys_addr_t paddr, size_t size,
enum dma_data_direction dir)
{
dma_sync_phys(paddr, size, dir, true);
}
#ifdef CONFIG_ARCH_HAS_SYNC_DMA_FOR_CPU
void arch_sync_dma_for_cpu(phys_addr_t paddr, size_t size,
enum dma_data_direction dir)
{
if (cpu_needs_post_dma_flush())
dma_sync_phys(paddr, size, dir, false);
}
#endif
#ifdef CONFIG_ARCH_HAS_SETUP_DMA_OPS
void arch_setup_dma_ops(struct device *dev, u64 dma_base, u64 size,
const struct iommu_ops *iommu, bool coherent)
{
dev->dma_coherent = coherent;
}
#endif