From b8fad3d02986222fa162d455eca2ffe807b6a15a Mon Sep 17 00:00:00 2001 From: Julius Werner Date: Tue, 27 Aug 2013 15:48:32 -0700 Subject: arm: libpayload: Add cache coherent DMA memory definition and management This patch adds a mechanism to set aside a region of cache-coherent (i.e. usually uncached) virtual memory, which can be used to communicate with DMA devices without automatic cache snooping (common on ARM) without the need of explicit flush/invalidation instructions in the driver code. This works by setting aside said region in the (board-specific) page table setup, as exemplary done in this patch for the Snow and Pit boards. It uses a new mechanism for adding board-specific Coreboot table entries to describe this region in an entry with the LB_DMA tag. Libpayload's memory allocator is enhanced to be able to operate on distinct types/regions of memory. It provides dma_malloc() and dma_memalign() functions for use in drivers, which by default just operate on the same heap as their traditional counterparts. However, if the Coreboot table parsing code finds a CB_DMA section, further requests through the dma_xxx() functions will return memory from the region described therein instead. Change-Id: Ia9c249249e936bbc3eb76e7b4822af2230ffb186 Signed-off-by: Julius Werner Reviewed-on: https://chromium-review.googlesource.com/167155 (cherry picked from commit d142ccdcd902a9d6ab4d495fbe6cbe85c61a5f01) Signed-off-by: Isaac Christensen Reviewed-on: http://review.coreboot.org/6622 Tested-by: build bot (Jenkins) Reviewed-by: Stefan Reinauer --- payloads/libpayload/libc/malloc.c | 126 +++++++++++++++++++++++++------------- 1 file changed, 85 insertions(+), 41 deletions(-) (limited to 'payloads/libpayload/libc') diff --git a/payloads/libpayload/libc/malloc.c b/payloads/libpayload/libc/malloc.c index f022f7349a..618d8b3b93 100644 --- a/payloads/libpayload/libc/malloc.c +++ b/payloads/libpayload/libc/malloc.c @@ -42,10 +42,17 @@ #define IN_MALLOC_C #include +struct memory_type { + void *start; + void *end; + struct align_region_t* align_regions; +}; + extern char _heap, _eheap; /* Defined in the ldscript. */ -static void *hstart = (void *)&_heap; -static void *hend = (void *)&_eheap; +static struct memory_type default_type = { (void *)&_heap, (void *)&_eheap, NULL }; +static struct memory_type *const heap = &default_type; +static struct memory_type *dma = &default_type; typedef u64 hdrtype_t; #define HDRSIZE (sizeof(hdrtype_t)) @@ -65,7 +72,7 @@ typedef u64 hdrtype_t; #define IS_FREE(_h) (((_h) & (MAGIC | FLAG_FREE)) == (MAGIC | FLAG_FREE)) #define HAS_MAGIC(_h) (((_h) & MAGIC) == MAGIC) -static int free_aligned(void* addr); +static int free_aligned(void* addr, struct memory_type *type); void print_malloc_map(void); #ifdef CONFIG_LP_DEBUG_MALLOC @@ -73,6 +80,23 @@ static int heap_initialized = 0; static int minimal_free = 0; #endif +void init_dma_memory(void *start, u32 size) +{ +#ifdef CONFIG_LP_DEBUG_MALLOC + if (dma != heap) { + printf("WARNING: %s called twice!\n"); + return; + } + + printf("Initializing cache-coherent DMA memory at [%p:%p]\n", start, start + size); +#endif + + dma = malloc(sizeof(*dma)); + dma->start = start; + dma->end = start + size; + dma->align_regions = NULL; +} + static void setup(hdrtype_t volatile *start, int size) { *start = FREE_BLOCK(size); @@ -83,10 +107,10 @@ static void setup(hdrtype_t volatile *start, int size) #endif } -static void *alloc(int len) +static void *alloc(int len, struct memory_type *type) { hdrtype_t header; - hdrtype_t volatile *ptr = (hdrtype_t volatile *) hstart; + hdrtype_t volatile *ptr = (hdrtype_t volatile *)type->start; /* Align the size. */ len = (len + HDRSIZE - 1) & ~(HDRSIZE - 1); @@ -137,17 +161,17 @@ static void *alloc(int len) ptr = (hdrtype_t volatile *)((int)ptr + HDRSIZE + size); - } while (ptr < (hdrtype_t *) hend); + } while (ptr < (hdrtype_t *) type->end); /* Nothing available. */ return (void *)NULL; } -static void _consolidate(void) +static void _consolidate(struct memory_type *type) { - void *ptr = hstart; + void *ptr = type->start; - while (ptr < hend) { + while (ptr < type->end) { void *nptr; hdrtype_t hdr = *((hdrtype_t *) ptr); unsigned int size = 0; @@ -160,7 +184,7 @@ static void _consolidate(void) size = SIZE(hdr); nptr = ptr + HDRSIZE + SIZE(hdr); - while (nptr < hend) { + while (nptr < type->end) { hdrtype_t nhdr = *((hdrtype_t *) nptr); if (!(IS_FREE(nhdr))) @@ -181,15 +205,18 @@ static void _consolidate(void) void free(void *ptr) { hdrtype_t hdr; - - if (free_aligned(ptr)) return; - - ptr -= HDRSIZE; + struct memory_type *type = heap; /* Sanity check. */ - if (ptr < hstart || ptr >= hend) - return; + if (ptr < type->start || ptr >= type->end) { + type = dma; + if (ptr < type->start || ptr >= type->end) + return; + } + if (free_aligned(ptr, type)) return; + + ptr -= HDRSIZE; hdr = *((hdrtype_t *) ptr); /* Not our header (we're probably poisoned). */ @@ -201,18 +228,23 @@ void free(void *ptr) return; *((hdrtype_t *) ptr) = FREE_BLOCK(SIZE(hdr)); - _consolidate(); + _consolidate(type); } void *malloc(size_t size) { - return alloc(size); + return alloc(size, heap); +} + +void *dma_malloc(size_t size) +{ + return alloc(size, dma); } void *calloc(size_t nmemb, size_t size) { size_t total = nmemb * size; - void *ptr = alloc(total); + void *ptr = alloc(total, heap); if (ptr) memset(ptr, 0, total); @@ -224,15 +256,19 @@ void *realloc(void *ptr, size_t size) { void *ret, *pptr; unsigned int osize; + struct memory_type *type = heap; if (ptr == NULL) - return alloc(size); + return alloc(size, type); pptr = ptr - HDRSIZE; if (!HAS_MAGIC(*((hdrtype_t *) pptr))) return NULL; + if (ptr < type->start || ptr >= type->end) + type = dma; + /* Get the original size of the block. */ osize = SIZE(*((hdrtype_t *) pptr)); @@ -242,7 +278,7 @@ void *realloc(void *ptr, size_t size) * reallocated the new space. */ free(ptr); - ret = alloc(size); + ret = alloc(size, type); /* * if ret == NULL, then doh - failure. @@ -277,14 +313,12 @@ struct align_region_t struct align_region_t *next; }; -static struct align_region_t* align_regions = 0; - -static struct align_region_t *allocate_region(int alignment, int num_elements) +static struct align_region_t *allocate_region(int alignment, int num_elements, struct memory_type *type) { struct align_region_t *new_region; #ifdef CONFIG_LP_DEBUG_MALLOC printf("%s(old align_regions=%p, alignment=%u, num_elements=%u)\n", - __func__, align_regions, alignment, num_elements); + __func__, type->align_regions, alignment, num_elements); #endif new_region = malloc(sizeof(struct align_region_t)); @@ -292,7 +326,7 @@ static struct align_region_t *allocate_region(int alignment, int num_elements) if (!new_region) return NULL; new_region->alignment = alignment; - new_region->start = malloc((num_elements+1) * alignment + num_elements); + new_region->start = alloc((num_elements+1) * alignment + num_elements, type); if (!new_region->start) { free(new_region); return NULL; @@ -300,16 +334,16 @@ static struct align_region_t *allocate_region(int alignment, int num_elements) new_region->start_data = (void*)((u32)(new_region->start + num_elements + alignment - 1) & (~(alignment-1))); new_region->size = num_elements * alignment; new_region->free = num_elements; - new_region->next = align_regions; + new_region->next = type->align_regions; memset(new_region->start, 0, num_elements); - align_regions = new_region; + type->align_regions = new_region; return new_region; } -static int free_aligned(void* addr) +static int free_aligned(void* addr, struct memory_type *type) { - struct align_region_t *reg = align_regions; + struct align_region_t *reg = type->align_regions; while (reg != 0) { if ((addr >= reg->start_data) && (addr < reg->start_data + reg->size)) @@ -329,16 +363,16 @@ static int free_aligned(void* addr) return 0; } -void *memalign(size_t align, size_t size) +static void *alloc_aligned(size_t align, size_t size, struct memory_type *type) { if (size == 0) return 0; - if (align_regions == 0) { - align_regions = malloc(sizeof(struct align_region_t)); - if (align_regions == NULL) + if (type->align_regions == 0) { + type->align_regions = malloc(sizeof(struct align_region_t)); + if (type->align_regions == NULL) return NULL; - memset(align_regions, 0, sizeof(struct align_region_t)); + memset(type->align_regions, 0, sizeof(struct align_region_t)); } - struct align_region_t *reg = align_regions; + struct align_region_t *reg = type->align_regions; look_further: while (reg != 0) { @@ -357,9 +391,9 @@ look_further: printf(" need to allocate a new memalign region\n"); #endif /* get align regions */ - reg = allocate_region(align, (size<1024)?(1024/align):(((size-1)/align)+1)); + reg = allocate_region(align, (size<1024)?(1024/align):(((size-1)/align)+1), type); #ifdef CONFIG_LP_DEBUG_MALLOC - printf(" ... returned %p\n", align_regions); + printf(" ... returned %p\n", reg); #endif } if (reg == 0) { @@ -393,14 +427,24 @@ look_further: goto look_further; // end condition is once a new region is allocated - it always has enough space } +void *memalign(size_t align, size_t size) +{ + return alloc_aligned(align, size, heap); +} + +void *dma_memalign(size_t align, size_t size) +{ + return alloc_aligned(align, size, dma); +} + /* This is for debugging purposes. */ #ifdef CONFIG_LP_DEBUG_MALLOC void print_malloc_map(void) { - void *ptr = hstart; + void *ptr = heap->start; int free_memory = 0; - while (ptr < hend) { + while (ptr < heap->end) { hdrtype_t hdr = *((hdrtype_t *) ptr); if (!HAS_MAGIC(hdr)) { @@ -414,7 +458,7 @@ void print_malloc_map(void) /* FIXME: Verify the size of the block. */ printf("%x: %s (%x bytes)\n", - (unsigned int)(ptr - hstart), + (unsigned int)(ptr - heap->start), hdr & FLAG_FREE ? "FREE" : "USED", SIZE(hdr)); if (hdr & FLAG_FREE) -- cgit v1.2.3