From: Janne Grunau j@jannau.net

Low lovel LMB functionality will be used to manage IOVA space in the Apple dart iommu driver. This reordering ensures that those function can not access the global LMB memory map variable.

Reviewed-by: Tom Rini trini@konsulko.com Signed-off-by: Janne Grunau j@jannau.net --- lib/lmb.c | 496 ++++++++++++++++++++++++++++++++------------------------------ 1 file changed, 254 insertions(+), 242 deletions(-)

diff --git a/lib/lmb.c b/lib/lmb.c index 5be0d9dd89a506dc1e6e6951cc54441a8d4bb0df..2ce91f7521cfe58577b092fdaea8004b17f0eeff 100644 --- a/lib/lmb.c +++ b/lib/lmb.c @@ -27,97 +27,11 @@ DECLARE_GLOBAL_DATA_PTR; #define MAP_OP_FREE (u8)0x2 #define MAP_OP_ADD (u8)0x3

-static struct lmb lmb; - -static bool lmb_should_notify(enum lmb_flags flags) -{ - return !lmb.test && !(flags & LMB_NONOTIFY) && - CONFIG_IS_ENABLED(EFI_LOADER); -} - -static int __maybe_unused lmb_map_update_notify(phys_addr_t addr, - phys_size_t size, - u8 op, enum lmb_flags flags) -{ - u64 efi_addr; - u64 pages; - efi_status_t status; - - if (op != MAP_OP_RESERVE && op != MAP_OP_FREE && op != MAP_OP_ADD) { - log_err("Invalid map update op received (%d)\n", op); - return -1; - } - - if (!lmb_should_notify(flags)) - return 0; - - efi_addr = (uintptr_t)map_sysmem(addr, 0); - pages = efi_size_in_pages(size + (efi_addr & EFI_PAGE_MASK)); - efi_addr &= ~EFI_PAGE_MASK; - - status = efi_add_memory_map_pg(efi_addr, pages, - op == MAP_OP_RESERVE ? - EFI_BOOT_SERVICES_DATA : - EFI_CONVENTIONAL_MEMORY, - false); - if (status != EFI_SUCCESS) { - log_err("%s: LMB Map notify failure %lu\n", __func__, - status & ~EFI_ERROR_MASK); - return -1; - } - unmap_sysmem((void *)(uintptr_t)efi_addr); - - return 0; -} - -static void lmb_print_region_flags(enum lmb_flags flags) -{ - u64 bitpos; - const char *flag_str[] = { "none", "no-map", "no-overwrite", "no-notify" }; - - do { - bitpos = flags ? fls(flags) - 1 : 0; - assert_noisy(bitpos < ARRAY_SIZE(flag_str)); - printf("%s", flag_str[bitpos]); - flags &= ~(1ull << bitpos); - puts(flags ? ", " : "\n"); - } while (flags); -} - -static void lmb_dump_region(struct alist *lmb_rgn_lst, char *name) -{ - struct lmb_region *rgn = lmb_rgn_lst->data; - unsigned long long base, size, end; - enum lmb_flags flags; - int i; - - printf(" %s.count = 0x%x\n", name, lmb_rgn_lst->count); - - for (i = 0; i < lmb_rgn_lst->count; i++) { - base = rgn[i].base; - size = rgn[i].size; - end = base + size - 1; - flags = rgn[i].flags; - - printf(" %s[%d]\t[0x%llx-0x%llx], 0x%08llx bytes flags: ", - name, i, base, end, size); - lmb_print_region_flags(flags); - } -} - -void lmb_dump_all_force(void) -{ - printf("lmb_dump_all:\n"); - lmb_dump_region(&lmb.free_mem, "memory"); - lmb_dump_region(&lmb.used_mem, "reserved"); -} - -void lmb_dump_all(void) -{ -#ifdef DEBUG - lmb_dump_all_force(); -#endif -} +/* + * The following low level LMB functions must not access the global LMB memory + * map since they are also used to manage IOVA memory maps in iommu drivers like + * apple_dart. + */

static long lmb_addrs_overlap(phys_addr_t base1, phys_size_t size1, phys_addr_t base2, phys_size_t size2) @@ -206,117 +120,6 @@ static void lmb_fix_over_lap_regions(struct alist *lmb_rgn_lst, lmb_remove_region(lmb_rgn_lst, r2); }

-static void lmb_reserve_uboot_region(void) -{ - int bank; - ulong end, bank_end; - phys_addr_t rsv_start; - - rsv_start = gd->start_addr_sp - CONFIG_STACK_SIZE; - end = gd->ram_top; - - /* - * Reserve memory from aligned address below the bottom of U-Boot stack - * until end of RAM area to prevent LMB from overwriting that memory. - */ - debug("## Current stack ends at 0x%08lx ", (ulong)rsv_start); - - /* adjust sp by 16K to be safe */ - rsv_start -= SZ_16K; - for (bank = 0; bank < CONFIG_NR_DRAM_BANKS; bank++) { - if (!gd->bd->bi_dram[bank].size || - rsv_start < gd->bd->bi_dram[bank].start) - continue; - /* Watch out for RAM at end of address space! */ - bank_end = gd->bd->bi_dram[bank].start + - gd->bd->bi_dram[bank].size - 1; - if (rsv_start > bank_end) - continue; - if (bank_end > end) - bank_end = end - 1; - - lmb_reserve_flags(rsv_start, bank_end - rsv_start + 1, - LMB_NOOVERWRITE); - - if (gd->flags & GD_FLG_SKIP_RELOC) - lmb_reserve_flags((phys_addr_t)(uintptr_t)_start, - gd->mon_len, LMB_NOOVERWRITE); - - break; - } -} - -static void lmb_reserve_common(void *fdt_blob) -{ - lmb_reserve_uboot_region(); - - if (CONFIG_IS_ENABLED(OF_LIBFDT) && fdt_blob) - boot_fdt_add_mem_rsv_regions(fdt_blob); -} - -static __maybe_unused void lmb_reserve_common_spl(void) -{ - phys_addr_t rsv_start; - phys_size_t rsv_size; - - /* - * Assume a SPL stack of 16KB. This must be - * more than enough for the SPL stage. - */ - if (IS_ENABLED(CONFIG_SPL_STACK_R_ADDR)) { - rsv_start = gd->start_addr_sp - 16384; - rsv_size = 16384; - lmb_reserve_flags(rsv_start, rsv_size, LMB_NOOVERWRITE); - } - - if (IS_ENABLED(CONFIG_SPL_SEPARATE_BSS)) { - /* Reserve the bss region */ - rsv_start = (phys_addr_t)(uintptr_t)__bss_start; - rsv_size = (phys_addr_t)(uintptr_t)__bss_end - - (phys_addr_t)(uintptr_t)__bss_start; - lmb_reserve_flags(rsv_start, rsv_size, LMB_NOOVERWRITE); - } -} - -/** - * lmb_add_memory() - Add memory range for LMB allocations - * - * Add the entire available memory range to the pool of memory that - * can be used by the LMB module for allocations. - * - * Return: None - */ -void lmb_add_memory(void) -{ - int i; - phys_size_t size; - u64 ram_top = gd->ram_top; - struct bd_info *bd = gd->bd; - - if (CONFIG_IS_ENABLED(LMB_ARCH_MEM_MAP)) - return lmb_arch_add_memory(); - - /* Assume a 4GB ram_top if not defined */ - if (!ram_top) - ram_top = 0x100000000ULL; - - for (i = 0; i < CONFIG_NR_DRAM_BANKS; i++) { - size = bd->bi_dram[i].size; - if (size) { - lmb_add(bd->bi_dram[i].start, size); - - /* - * Reserve memory above ram_top as - * no-overwrite so that it cannot be - * allocated - */ - if (bd->bi_dram[i].start >= ram_top) - lmb_reserve_flags(bd->bi_dram[i].start, size, - LMB_NOOVERWRITE); - } - } -} - static long lmb_resize_regions(struct alist *lmb_rgn_lst, unsigned long idx_start, phys_addr_t base, phys_size_t size) @@ -476,25 +279,6 @@ static long lmb_add_region_flags(struct alist *lmb_rgn_lst, phys_addr_t base, return 0; }

-static long lmb_add_region(struct alist *lmb_rgn_lst, phys_addr_t base, - phys_size_t size) -{ - return lmb_add_region_flags(lmb_rgn_lst, base, size, LMB_NONE); -} - -/* This routine may be called with relocation disabled. */ -long lmb_add(phys_addr_t base, phys_size_t size) -{ - long ret; - struct alist *lmb_rgn_lst = &lmb.free_mem; - - ret = lmb_add_region(lmb_rgn_lst, base, size); - if (ret) - return ret; - - return lmb_map_update_notify(base, size, MAP_OP_ADD, LMB_NONE); -} - static long _lmb_free(struct alist *lmb_rgn_lst, phys_addr_t base, phys_size_t size) { @@ -546,6 +330,255 @@ static long _lmb_free(struct alist *lmb_rgn_lst, phys_addr_t base, rgn[i].flags); }

+static long lmb_overlaps_region(struct alist *lmb_rgn_lst, phys_addr_t base, + phys_size_t size) +{ + unsigned long i; + struct lmb_region *rgn = lmb_rgn_lst->data; + + for (i = 0; i < lmb_rgn_lst->count; i++) { + phys_addr_t rgnbase = rgn[i].base; + phys_size_t rgnsize = rgn[i].size; + if (lmb_addrs_overlap(base, size, rgnbase, rgnsize)) + break; + } + + return (i < lmb_rgn_lst->count) ? i : -1; +} + +static phys_addr_t lmb_align_down(phys_addr_t addr, phys_size_t size) +{ + return addr & ~(size - 1); +} + +/* + * Low level LMB functions are used to manage IOVA memory maps for the Apple + * dart iommu. They must not access the global LMB memory map. + * So keep the global LMB variable declaration unreachable from them. + */ + +static struct lmb lmb; + +static bool lmb_should_notify(enum lmb_flags flags) +{ + return !lmb.test && !(flags & LMB_NONOTIFY) && + CONFIG_IS_ENABLED(EFI_LOADER); +} + +static int __maybe_unused lmb_map_update_notify(phys_addr_t addr, + phys_size_t size, + u8 op, enum lmb_flags flags) +{ + u64 efi_addr; + u64 pages; + efi_status_t status; + + if (op != MAP_OP_RESERVE && op != MAP_OP_FREE && op != MAP_OP_ADD) { + log_err("Invalid map update op received (%d)\n", op); + return -1; + } + + if (!lmb_should_notify(flags)) + return 0; + + efi_addr = (uintptr_t)map_sysmem(addr, 0); + pages = efi_size_in_pages(size + (efi_addr & EFI_PAGE_MASK)); + efi_addr &= ~EFI_PAGE_MASK; + + status = efi_add_memory_map_pg(efi_addr, pages, + op == MAP_OP_RESERVE ? + EFI_BOOT_SERVICES_DATA : + EFI_CONVENTIONAL_MEMORY, + false); + if (status != EFI_SUCCESS) { + log_err("%s: LMB Map notify failure %lu\n", __func__, + status & ~EFI_ERROR_MASK); + return -1; + } + unmap_sysmem((void *)(uintptr_t)efi_addr); + + return 0; +} + +static void lmb_print_region_flags(enum lmb_flags flags) +{ + u64 bitpos; + const char *flag_str[] = { "none", "no-map", "no-overwrite", "no-notify" }; + + do { + bitpos = flags ? fls(flags) - 1 : 0; + assert_noisy(bitpos < ARRAY_SIZE(flag_str)); + printf("%s", flag_str[bitpos]); + flags &= ~(1ull << bitpos); + puts(flags ? ", " : "\n"); + } while (flags); +} + +static void lmb_dump_region(struct alist *lmb_rgn_lst, char *name) +{ + struct lmb_region *rgn = lmb_rgn_lst->data; + unsigned long long base, size, end; + enum lmb_flags flags; + int i; + + printf(" %s.count = 0x%x\n", name, lmb_rgn_lst->count); + + for (i = 0; i < lmb_rgn_lst->count; i++) { + base = rgn[i].base; + size = rgn[i].size; + end = base + size - 1; + flags = rgn[i].flags; + + printf(" %s[%d]\t[0x%llx-0x%llx], 0x%08llx bytes flags: ", + name, i, base, end, size); + lmb_print_region_flags(flags); + } +} + +void lmb_dump_all_force(void) +{ + printf("lmb_dump_all:\n"); + lmb_dump_region(&lmb.free_mem, "memory"); + lmb_dump_region(&lmb.used_mem, "reserved"); +} + +void lmb_dump_all(void) +{ +#ifdef DEBUG + lmb_dump_all_force(); +#endif +} + +static void lmb_reserve_uboot_region(void) +{ + int bank; + ulong end, bank_end; + phys_addr_t rsv_start; + + rsv_start = gd->start_addr_sp - CONFIG_STACK_SIZE; + end = gd->ram_top; + + /* + * Reserve memory from aligned address below the bottom of U-Boot stack + * until end of RAM area to prevent LMB from overwriting that memory. + */ + debug("## Current stack ends at 0x%08lx ", (ulong)rsv_start); + + /* adjust sp by 16K to be safe */ + rsv_start -= SZ_16K; + for (bank = 0; bank < CONFIG_NR_DRAM_BANKS; bank++) { + if (!gd->bd->bi_dram[bank].size || + rsv_start < gd->bd->bi_dram[bank].start) + continue; + /* Watch out for RAM at end of address space! */ + bank_end = gd->bd->bi_dram[bank].start + + gd->bd->bi_dram[bank].size - 1; + if (rsv_start > bank_end) + continue; + if (bank_end > end) + bank_end = end - 1; + + lmb_reserve_flags(rsv_start, bank_end - rsv_start + 1, + LMB_NOOVERWRITE); + + if (gd->flags & GD_FLG_SKIP_RELOC) + lmb_reserve_flags((phys_addr_t)(uintptr_t)_start, + gd->mon_len, LMB_NOOVERWRITE); + + break; + } +} + +static void lmb_reserve_common(void *fdt_blob) +{ + lmb_reserve_uboot_region(); + + if (CONFIG_IS_ENABLED(OF_LIBFDT) && fdt_blob) + boot_fdt_add_mem_rsv_regions(fdt_blob); +} + +static __maybe_unused void lmb_reserve_common_spl(void) +{ + phys_addr_t rsv_start; + phys_size_t rsv_size; + + /* + * Assume a SPL stack of 16KB. This must be + * more than enough for the SPL stage. + */ + if (IS_ENABLED(CONFIG_SPL_STACK_R_ADDR)) { + rsv_start = gd->start_addr_sp - 16384; + rsv_size = 16384; + lmb_reserve_flags(rsv_start, rsv_size, LMB_NOOVERWRITE); + } + + if (IS_ENABLED(CONFIG_SPL_SEPARATE_BSS)) { + /* Reserve the bss region */ + rsv_start = (phys_addr_t)(uintptr_t)__bss_start; + rsv_size = (phys_addr_t)(uintptr_t)__bss_end - + (phys_addr_t)(uintptr_t)__bss_start; + lmb_reserve_flags(rsv_start, rsv_size, LMB_NOOVERWRITE); + } +} + +/** + * lmb_add_memory() - Add memory range for LMB allocations + * + * Add the entire available memory range to the pool of memory that + * can be used by the LMB module for allocations. + * + * Return: None + */ +void lmb_add_memory(void) +{ + int i; + phys_size_t size; + u64 ram_top = gd->ram_top; + struct bd_info *bd = gd->bd; + + if (CONFIG_IS_ENABLED(LMB_ARCH_MEM_MAP)) + return lmb_arch_add_memory(); + + /* Assume a 4GB ram_top if not defined */ + if (!ram_top) + ram_top = 0x100000000ULL; + + for (i = 0; i < CONFIG_NR_DRAM_BANKS; i++) { + size = bd->bi_dram[i].size; + if (size) { + lmb_add(bd->bi_dram[i].start, size); + + /* + * Reserve memory above ram_top as + * no-overwrite so that it cannot be + * allocated + */ + if (bd->bi_dram[i].start >= ram_top) + lmb_reserve_flags(bd->bi_dram[i].start, size, + LMB_NOOVERWRITE); + } + } +} + +static long lmb_add_region(struct alist *lmb_rgn_lst, phys_addr_t base, + phys_size_t size) +{ + return lmb_add_region_flags(lmb_rgn_lst, base, size, LMB_NONE); +} + +/* This routine may be called with relocation disabled. */ +long lmb_add(phys_addr_t base, phys_size_t size) +{ + long ret; + struct alist *lmb_rgn_lst = &lmb.free_mem; + + ret = lmb_add_region(lmb_rgn_lst, base, size); + if (ret) + return ret; + + return lmb_map_update_notify(base, size, MAP_OP_ADD, LMB_NONE); +} + /** * lmb_free_flags() - Free up a region of memory * @base: Base Address of region to be freed @@ -590,27 +623,6 @@ long lmb_reserve(phys_addr_t base, phys_size_t size) return lmb_reserve_flags(base, size, LMB_NONE); }

-static long lmb_overlaps_region(struct alist *lmb_rgn_lst, phys_addr_t base, - phys_size_t size) -{ - unsigned long i; - struct lmb_region *rgn = lmb_rgn_lst->data; - - for (i = 0; i < lmb_rgn_lst->count; i++) { - phys_addr_t rgnbase = rgn[i].base; - phys_size_t rgnsize = rgn[i].size; - if (lmb_addrs_overlap(base, size, rgnbase, rgnsize)) - break; - } - - return (i < lmb_rgn_lst->count) ? i : -1; -} - -static phys_addr_t lmb_align_down(phys_addr_t addr, phys_size_t size) -{ - return addr & ~(size - 1); -} - static phys_addr_t _lmb_alloc_base(phys_size_t size, ulong align, phys_addr_t max_addr, enum lmb_flags flags) {

From: Janne Grunau j@jannau.net

There is no overlap between the IOVA space managed by the iommu (here the 32-bit address space) and physical RAM on Apple silicon systems. The RAM starts at 0x10_0000_0000 or 0x100_0000_0000 so it's not possible to manage the IOVA with the global memory LMB and use 1:1 translation. In addition each device has its own iommu and does not need to share the address space with all other devices. This should not be problem for u-boot's limited use and hardware support. Restore the private per instance LMB IOVA map.

Fixes: ed17a33fed2 ("lmb: make LMB memory map persistent and global") Signed-off-by: Janne Grunau j@jannau.net --- drivers/iommu/apple_dart.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/drivers/iommu/apple_dart.c b/drivers/iommu/apple_dart.c index 611ac7cd6deb4fccd45ccfe8cbb9b3e36e2af753..3e9e7819e517b8fe62b9e429b7a8ca3eca29741d 100644 --- a/drivers/iommu/apple_dart.c +++ b/drivers/iommu/apple_dart.c @@ -73,6 +73,8 @@ struct apple_dart_priv { u64 *l1, *l2; int bypass, shift;

+ struct lmb io_lmb; + dma_addr_t dvabase; dma_addr_t dvaend;

@@ -123,7 +125,7 @@ static dma_addr_t apple_dart_map(struct udevice *dev, void *addr, size_t size) off = (phys_addr_t)addr - paddr; psize = ALIGN(size + off, DART_PAGE_SIZE);

- dva = lmb_alloc(psize, DART_PAGE_SIZE); + dva = io_lmb_alloc(&priv->io_lmb, psize, DART_PAGE_SIZE);

idx = dva / DART_PAGE_SIZE; for (i = 0; i < psize / DART_PAGE_SIZE; i++) { @@ -159,7 +161,7 @@ static void apple_dart_unmap(struct udevice *dev, dma_addr_t addr, size_t size) (unsigned long)&priv->l2[idx + i]); priv->flush_tlb(priv);

- lmb_free(dva, psize); + io_lmb_free(&priv->io_lmb, dva, psize); }

static struct iommu_ops apple_dart_ops = { @@ -173,7 +175,7 @@ static int apple_dart_probe(struct udevice *dev) dma_addr_t addr; phys_addr_t l2; int ntte, nl1, nl2; - int sid, i; + int ret, sid, i; u32 params2, params4;

priv->base = dev_read_addr_ptr(dev); @@ -212,7 +214,13 @@ static int apple_dart_probe(struct udevice *dev) priv->dvabase = DART_PAGE_SIZE; priv->dvaend = SZ_4G - DART_PAGE_SIZE;

- lmb_add(priv->dvabase, priv->dvaend - priv->dvabase); + ret = io_lmb_setup(&priv->io_lmb); + if (ret) + return ret; + ret = io_lmb_add(&priv->io_lmb, priv->dvabase, + priv->dvaend - priv->dvabase); + if (ret) + return -EINVAL;

/* Disable translations. */ for (sid = 0; sid < priv->nsid; sid++) @@ -294,6 +302,8 @@ static int apple_dart_remove(struct udevice *dev) } priv->flush_tlb(priv);

+ io_lmb_teardown(&priv->io_lmb); + return 0; }