mach-k3: add runtime memory carveouts for MMU table
Subsystems (2)
PATCH v3 1/2 mach-k3: add runtime memory carveouts for MMU table
In u-boot we only provide a single MMU table for all k3 platforms, this does not scale for devices with reserved memory outside the range 0x9e780000 - 0xa0000000 (eg j722s[1]) or for devices with < 2GiB of memory (eg am62-SIP with 512MiB of RAM).
To properly configure the MMU on various k3 platforms, the reserved-memory regions need to be queried at runtime from the device-tree and the MMU table should be updated accordingly.
This patch adds the required fixups to the MMU table (during proper U-boot stage) by marking the reserved regions as non cacheable and keeping the remaining area as cacheable.
For the A-core SPL, the 128MiB region starting from SPL_TEXT_BASE is marked as cacheable i.e 0x80080000 to 0x88080000.
The 128MiB size is chosen to allow for future use cases such as falcon boot from the A-Core SPL which would require loading kernel image from the SPL stage. This change also ensures the reserved memory regions that all exist past 0x88080000 are non cacheable preventing speculative accesses to those addresses.
---
Changes for v3:
- Remove unused memory regions in SPL's map
- Add runtime addition of MMU entry for the framebuffer in SPL
- Refactor k3_mem_map_init to use standard u-boot APIs
- Unmap reserved-memory regions instead of keeping them uncached
v2: https://lore.kernel.org/u-boot/[email protected]/
Changes in v2:
- Removed dependency to:
https://lore.kernel.org/u-boot/[email protected]/
v1: https://lore.kernel.org/u-boot/[email protected]/
--- arch/arm/mach-k3/arm64/arm64-mmu.c | 191 +++++++++++++++++++++++-- arch/arm/mach-k3/include/mach/k3-ddr.h | 9 ++ board/ti/common/k3-ddr.c | 10 ++ 3 files changed, 197 insertions(+), 13 deletions(-)
arch/arm/mach-k3/arm64/arm64-mmu.c +178 -13
@@ -11,9 +11,16 @@#include <asm/system.h> #include <asm/armv8/mmu.h> +#include <mach/k3-ddr.h> +#include <sort.h> -struct mm_region k3_mem_map[] = { +#include "../common_fdt.h" + +DECLARE_GLOBAL_DATA_PTR ; + +struct mm_region k3_mem_map[K3_MMU_REGIONS_COUNT ] = { { +/* Peripherals */ .virt =0x0UL , .phys =0x0UL , .size =0x80000000UL ,
@@ -21,30 +28,27 @@ struct mm_region k3_mem_map[] = {PTE_BLOCK_NON_SHARE |PTE_BLOCK_PXN |PTE_BLOCK_UXN }, { - .virt = 0x80000000UL, - .phys = 0x80000000UL, - .size = 0x1e780000UL, - .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | -PTE_BLOCK_INNER_SHARE - }, { - .virt = 0xa0000000UL, - .phys = 0xa0000000UL, - .size = 0x60000000UL, - .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | -PTE_BLOCK_INNER_SHARE - }, { +/* Higher DDR banks */ .virt = 0x880000000UL, .phys = 0x880000000UL, .size = 0x80000000UL, .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) |PTE_BLOCK_INNER_SHARE }, { +/* Flash peripherals */ .virt = 0x500000000UL, .phys = 0x500000000UL, .size = 0x380000000UL, .attrs =PTE_BLOCK_MEMTYPE (MT_DEVICE_NGNRNE ) |PTE_BLOCK_NON_SHARE |PTE_BLOCK_PXN |PTE_BLOCK_UXN + }, { +/* Map SPL load region and the next 128MiB as cacheable */ + .virt = CONFIG_SPL_TEXT_BASE, + .phys = CONFIG_SPL_TEXT_BASE, + .size = SZ_128M, + .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE }, {/* List terminator */ 0,
@@ -52,3 +56,164 @@ struct mm_region k3_mem_map[] = { };struct mm_region *mem_map = k3_mem_map; + +static int dt_reserved_cmp (const void *a,const void *b) +{ +const struct fdt_resource *val_a = a, *val_b = b; + +return (val_a->start - val_b->start); +} + +int k3_mem_map_init (void ) +{ +fdt_addr_t mem_base; +fdt_size_t mem_size; +struct fdt_resource dt_reserved[K3_MMU_REGIONS_COUNT ], + coalesced[K3_MMU_REGIONS_COUNT ]; +int k3_map_idx = -EINVAL , ret, nodeoffset, subnode; +void *blob = (void *)gd->fdt_blob; +unsigned int carveout_len, i, j; + + ret =fdt_fixup_reserved (blob,"tfa" ,CONFIG_K3_ATF_LOAD_ADDR ,0x80000 ); +if (ret) { +pr_err ("%s: Failed to fixup reserved node for tfa [%d]\n" , + __func__, ret); +return ret; + } + + ret =fdt_fixup_reserved (blob,"optee" ,CONFIG_K3_OPTEE_LOAD_ADDR , +0x1800000 ); +if (ret) { +pr_err ("%s: Failed to fixup reserved node for optee [%d]\n" , + __func__, ret); +return ret; + } + + nodeoffset =fdt_subnode_offset (blob,0 ,"memory" ); +if (nodeoffset <0 ) { +pr_err ("%s: Failed to get memory data: %s\n" , __func__, +fdt_strerror (nodeoffset)); +return nodeoffset; + } + + mem_base =fdtdec_get_addr_size (blob, nodeoffset,"reg" , &mem_size); +if (mem_base !=CFG_SYS_SDRAM_BASE ) +return -EINVAL ; + +for (i =0 ; i <K3_MMU_REGIONS_COUNT ; i++) { +if (k3_mem_map[i].virt ==CONFIG_SPL_TEXT_BASE ) { + k3_map_idx = i; +break ; + } + } + +if (k3_map_idx == -EINVAL ) { +pr_err ("%s: Failed to find DDR region in MMU memory map\n" , + __func__); +return -EINVAL ; + } + + i =0 ; + nodeoffset =fdt_subnode_offset (blob,0 ,"reserved-memory" ); +fdt_for_each_subnode (subnode, blob, nodeoffset) { +const char *name; +fdt_addr_t addr, end_addr; +fdt_size_t size; + +if (i >=K3_MMU_REGIONS_COUNT ) { +/* +* This is a recoverable error if the regions can be +* coalesced, the required logic can be implemented once +* requirement arises. +*/ +pr_err ("%s: Not enough space in MMU map for carveouts\n" , + __func__); +return -ENOMEM ; + } + + name =fdt_get_name (blob, subnode,NULL ); + addr =fdtdec_get_addr_size (blob, subnode,"reg" , &size); + +if (addr ==FDT_ADDR_T_NONE ) +continue ; + +if (!fdtdec_get_bool (blob, subnode,"no-map" )) +continue ; + +if (addr >= mem_base + mem_size) +continue ; + + end_addr = addr + size; + +if (end_addr <= mem_base) +continue ; + +debug ("Added memory carveout at 0x%llx, size: 0x%llx for '%s'\n" , + addr, size, name); + + addr =max (addr, mem_base); + end_addr =min (end_addr, mem_base + mem_size); + size = end_addr - addr; + dt_reserved[i].start = addr; + dt_reserved[i].end = end_addr; + i++; + } + carveout_len = i; + +if (!carveout_len) +return 0 ; + +/* sort carveout regions by address required for creating carveouts */ +qsort (dt_reserved, carveout_len,sizeof (dt_reserved[0 ]), + dt_reserved_cmp); + +/* coalesce regions */ +struct fdt_resource coalescing_temp = dt_reserved[0 ]; + + j =0 ; +for (i =1 ; i < carveout_len; i++) { +struct fdt_resource current = dt_reserved[i]; + +if (coalescing_temp.end >= current.start) { + coalescing_temp.end = current.end; +continue ; + } + coalesced[j] = coalescing_temp; + coalescing_temp = current; + j++; + } + + coalesced[j] = coalescing_temp; + carveout_len = j +1 ; + +if (coalesced[0 ].start != mem_base) { + k3_mem_map[k3_map_idx].virt = mem_base; + k3_mem_map[k3_map_idx].phys = mem_base; + k3_mem_map[k3_map_idx].size = coalesced[0 ].start - mem_base; + k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + } + +for (i =1 ; i < carveout_len; i++) { + k3_mem_map[k3_map_idx].virt = coalesced[i -1 ].end; + k3_mem_map[k3_map_idx].phys = coalesced[i -1 ].end; + k3_mem_map[k3_map_idx].size = + coalesced[i].start - coalesced[i -1 ].end; + k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + } + + k3_mem_map[k3_map_idx].virt = coalesced[carveout_len -1 ].end; + k3_mem_map[k3_map_idx].phys = coalesced[carveout_len -1 ].end; + k3_mem_map[k3_map_idx].size = + mem_base + mem_size - coalesced[carveout_len -1 ].end; + k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + + k3_mem_map[k3_map_idx] = (const struct mm_region ){0 }; + +return 0 ; +}
arch/arm/mach-k3/include/mach/k3-ddr.h +9 -0
@@ -8,10 +8,19 @@#include <spl.h> +/* Number of mappable regions in the MMU page table */ +#define K3_MMU_REGIONS_COUNT 32 +int dram_init (void );int dram_init_banksize (void );void fixup_ddr_driver_for_ecc (struct spl_image_info *spl_image);void fixup_memory_node (struct spl_image_info *spl_image); +/* +* Modifies the MMU memory map based on DDR size and reserved-memory +* nodes in DT +*/ +int k3_mem_map_init (void ); +#endif /* _K3_DDR_H_ */
board/ti/common/k3-ddr.c +10 -0
@@ -7,6 +7,7 @@#include <dm/uclass.h> #include <k3-ddrss.h> #include <spl.h> +#include <mach/k3-ddr.h> #include "k3-ddr.h"
@@ -14,6 +15,15 @@ int dram_init(void) {s32 ret; +if (IS_ENABLED (CONFIG_ARM64 ) &&xpl_phase () !=PHASE_SPL ) { + ret =k3_mem_map_init (); +if (ret) { +printf ("%s: Error fixing up MMU memory map: %d\n" , + __func__, ret); +return ret; + } + } + ret =fdtdec_setup_mem_size_base_lowest ();if (ret)printf ("Error setting up mem size and base. %d\n" , ret);
Signature · 1 line
2.49.0
PATCH v3 2/2 mach-k3: add dynamic mmu fixups for SPL stage
On platforms with spl splash support i.e CONFIG_VIDEO=y, the top of DDR is reserved for the framebuffer.
The size for the framebuffer is computed at runtime by video_reserve. During the MMU configuration an entry corresponding to the framebuffer should be dynamically created to properly map the required space for the framebuffer.
Therefore this patch adds k3_spl_mem_map_init which adds the required MMU entry by querying the gd after the framebuffer size has been computed in spl_reserve_video_from_ram_top.
For non VIDEO=y platforms, the added k3_spl_mem_map_init function gets optimized out of the final binary so overall, the spl size is not impacted[1].
[1]: Tested on clang 19.1.7 and gcc 15.1.1
--- arch/arm/mach-k3/arm64/arm64-mmu.c | 44 ++++++++++++++++++++++++++ arch/arm/mach-k3/common.c | 13 ++++++-- arch/arm/mach-k3/include/mach/k3-ddr.h | 3 ++ 3 files changed, 57 insertions(+), 3 deletions(-)
arch/arm/mach-k3/arm64/arm64-mmu.c +44 -0
@@ -49,6 +49,10 @@ struct mm_region k3_mem_map[K3_MMU_REGIONS_COUNT] = { .size = SZ_128M, .attrs = PTE_BLOCK_MEMTYPE(MT_NORMAL) | PTE_BLOCK_INNER_SHARE + }, { +/* Framebuffer reserved, size set at runtime */ + .attrs = PTE_BLOCK_MEMTYPE(MT_NORMAL) | + PTE_BLOCK_INNER_SHARE }, {/* List terminator */ 0,
@@ -57,6 +61,46 @@ struct mm_region k3_mem_map[K3_MMU_REGIONS_COUNT] = {struct mm_region *mem_map = k3_mem_map; +int k3_spl_mem_map_init (void ) +{ +unsigned int i; +int k3_map_idx = -EINVAL ; + +for (i =0 ; i <K3_MMU_REGIONS_COUNT ; i++) { +if (k3_mem_map[i].virt ==0 && k3_mem_map[i].phys ==0 && + k3_mem_map[i].size ==0 ) { + k3_map_idx = i; +break ; + } + } + +if (k3_map_idx == -EINVAL ) { +pr_err ("%s: Failed to find fixup region in MMU memory map\n" , + __func__); +return -EINVAL ; + } + +if (k3_map_idx >=K3_MMU_REGIONS_COUNT -1 ) { +pr_err ("%s: Not enough space in MMU map for the added entry\n" , + __func__); +return -ENOMEM ; + } + +if (CONFIG_IS_ENABLED (VIDEO )) { + k3_mem_map[k3_map_idx].virt =gd_video_bottom (); + k3_mem_map[k3_map_idx].phys =gd_video_bottom (); + k3_mem_map[k3_map_idx].size = +gd_video_top () -gd_video_bottom (); + k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + } + + k3_mem_map[k3_map_idx] = (const struct mm_region ){0 }; + +return 0 ; +} +static int dt_reserved_cmp (const void *a,const void *b) {const struct fdt_resource *val_a = a, *val_b = b;
arch/arm/mach-k3/common.c +10 -3
@@ -15,6 +15,7 @@#include <asm/global_data.h> #include <linux/printk.h> #include "common.h" +#include "mach/k3-ddr.h" #include <dm.h> #include <remoteproc.h> #include <asm/cache.h>
@@ -224,9 +225,6 @@ void spl_enable_cache(void)dram_init (); -/* reserve TLB table */ - gd->arch.tlb_size =PGTABLE_SIZE ; - gd->ram_top +=get_effective_memsize (); gd->relocaddr = gd->ram_top;
@@ -234,6 +232,15 @@ void spl_enable_cache(void)if (ret)panic ("Failed to reserve framebuffer memory (%d)\n" , ret); +if (IS_ENABLED (CONFIG_ARM64 )) { + ret =k3_spl_mem_map_init (); +if (ret) +panic ("Failed to perform MMU fixups: %d\n" , ret); + } + +/* reserve TLB table */ + gd->arch.tlb_size =PGTABLE_SIZE ; + gd->arch.tlb_addr = gd->relocaddr - gd->arch.tlb_size; gd->arch.tlb_addr &= ~(0x10000 -1 ); debug("TLB table from %08lx to %08lx\n", gd->arch.tlb_addr,
arch/arm/mach-k3/include/mach/k3-ddr.h +3 -0
@@ -17,6 +17,9 @@ int dram_init_banksize(void);void fixup_ddr_driver_for_ecc (struct spl_image_info *spl_image);void fixup_memory_node (struct spl_image_info *spl_image); +/* Performs MMU memory map fixups at SPL stage */ +int k3_spl_mem_map_init (void ); +/* * Modifies the MMU memory map based on DDR size and reserved-memory * nodes in DT
Signature · 1 line
2.49.0
Quoted 4 lines · Anshul Dalal
In u-boot we only provide a single MMU table for all k3 platforms, this does not scale for devices with reserved memory outside the range
0x9e780000-0xa0000000(eg j722s[1]) or for devices with < 2GiB of memory (eg am62-SIP with 512MiB of RAM).
please avoid giving reference to internal HACKs
Quoted 120 lines
To properly configure the MMU on various k3 platforms, the reserved-memory regions need to be queried at runtime from the device-tree and the MMU table should be updated accordingly.
This patch adds the required fixups to the MMU table (during proper U-boot stage) by marking the reserved regions as non cacheable and keeping the remaining area as cacheable.
For the A-core SPL, the 128MiB region starting from
SPL_TEXT_BASEis marked as cacheable i.e0x80080000to0x88080000.The 128MiB size is chosen to allow for future use cases such as falcon boot from the A-Core SPL which would require loading kernel image from the SPL stage. This change also ensures the reserved memory regions that all exist past
0x88080000are non cacheable preventing speculative accesses to those addresses.Signed-off-by Anshul Dalal <[email protected]>---
Changes for v3:
- Remove unused memory regions in SPL's map
- Add runtime addition of MMU entry for the framebuffer in SPL
- Refactork3_mem_map_initto use standard u-boot APIs
- Unmap reserved-memory regions instead of keeping them uncached
v2: https://lore.kernel.org/u-boot/[email protected]/Changes in v2:
- Removed dependency to:https://lore.kernel.org/u-boot/[email protected]/v1: https://lore.kernel.org/u-boot/[email protected]/
--- arch/arm/mach-k3/arm64/arm64-mmu.c | 191 +++++++++++++++++++++++-- arch/arm/mach-k3/include/mach/k3-ddr.h | 9 ++ board/ti/common/k3-ddr.c | 10 ++ 3 files changed, 197 insertions(+), 13 deletions(-)arch/arm/mach-k3/arm64/arm64-mmu.c +31
-13@@ -11,9 +11,16 @@#include <asm/system.h> #include <asm/armv8/mmu.h> +#include <mach/k3-ddr.h> +#include <sort.h> -struct mm_region k3_mem_map[] = { +#include "../common_fdt.h" + +DECLARE_GLOBAL_DATA_PTR ; + +struct mm_region k3_mem_map[K3_MMU_REGIONS_COUNT ] = { { +/* Peripherals */ .virt =0x0UL , .phys =0x0UL , .size =0x80000000UL ,@@ -21,30 +28,27 @@ struct mm_region k3_mem_map[] = {PTE_BLOCK_NON_SHARE |PTE_BLOCK_PXN |PTE_BLOCK_UXN }, { - .virt = 0x80000000UL, - .phys = 0x80000000UL, - .size = 0x1e780000UL, - .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | -PTE_BLOCK_INNER_SHARE - }, { - .virt = 0xa0000000UL, - .phys = 0xa0000000UL, - .size = 0x60000000UL, - .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | -PTE_BLOCK_INNER_SHARE - }, { +/* Higher DDR banks */ .virt = 0x880000000UL, .phys = 0x880000000UL, .size = 0x80000000UL, .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) |PTE_BLOCK_INNER_SHARE }, { +/* Flash peripherals */ .virt = 0x500000000UL, .phys = 0x500000000UL, .size = 0x380000000UL, .attrs =PTE_BLOCK_MEMTYPE (MT_DEVICE_NGNRNE ) |PTE_BLOCK_NON_SHARE |PTE_BLOCK_PXN |PTE_BLOCK_UXN + }, { +/* Map SPL load region and the next 128MiB as cacheable */ + .virt = CONFIG_SPL_TEXT_BASE, + .phys = CONFIG_SPL_TEXT_BASE, + .size = SZ_128M, + .attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE }, {/* List terminator */ 0,@@ -52,3 +56,164 @@ struct mm_region k3_mem_map[] = { };struct mm_region *mem_map = k3_mem_map; + +static int dt_reserved_cmp (const void *a,const void *b) +{ +const struct fdt_resource *val_a = a, *val_b = b; + +return (val_a->start - val_b->start); +} + +int k3_mem_map_init (void ) +{ +fdt_addr_t mem_base; +fdt_size_t mem_size; +struct fdt_resource dt_reserved[K3_MMU_REGIONS_COUNT ], + coalesced[K3_MMU_REGIONS_COUNT ];
arch/arm/mach-k3/arm64/arm64-mmu.c struct mm_region k3_mem_map[] = { · line 82
Quoted 10 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +10
-0+int k3_map_idx = -EINVAL , ret, nodeoffset, subnode; +void *blob = (void *)gd->fdt_blob; +unsigned int carveout_len, i, j; + + ret =fdt_fixup_reserved (blob,"tfa" ,CONFIG_K3_ATF_LOAD_ADDR ,0x80000 ); +if (ret) { +pr_err ("%s: Failed to fixup reserved node for tfa [%d]\n" , + __func__, ret); +return ret; + }
am62l3_fdt.c , needs more size.
fdt_fixup_reserved(blob, "tfa", CONFIG_K3_ATF_LOAD_ADDR, 0x200000)
Quoted 16 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +16
-0+ + ret =fdt_fixup_reserved (blob,"optee" ,CONFIG_K3_OPTEE_LOAD_ADDR , +0x1800000 ); +if (ret) { +pr_err ("%s: Failed to fixup reserved node for optee [%d]\n" , + __func__, ret); +return ret; + } + + nodeoffset =fdt_subnode_offset (blob,0 ,"memory" ); +if (nodeoffset <0 ) { +pr_err ("%s: Failed to get memory data: %s\n" , __func__, +fdt_strerror (nodeoffset)); +return nodeoffset; + } +
arch/arm/mach-k3/arm64/arm64-mmu.c struct mm_region k3_mem_map[] = { · line 108
Quoted 10 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +10
-0+ mem_base =fdtdec_get_addr_size (blob, nodeoffset,"reg" , &mem_size); +if (mem_base !=CFG_SYS_SDRAM_BASE ) +return -EINVAL ; + +for (i =0 ; i <K3_MMU_REGIONS_COUNT ; i++) { +if (k3_mem_map[i].virt ==CONFIG_SPL_TEXT_BASE ) { + k3_map_idx = i; +break ; + } + }
you can fix this entry at offset 0, instead of looping all over
Quoted 98 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +98
-0+ +if (k3_map_idx == -EINVAL ) { +pr_err ("%s: Failed to find DDR region in MMU memory map\n" , + __func__); +return -EINVAL ; + } + + i =0 ; + nodeoffset =fdt_subnode_offset (blob,0 ,"reserved-memory" ); +fdt_for_each_subnode (subnode, blob, nodeoffset) { +const char *name; +fdt_addr_t addr, end_addr; +fdt_size_t size; + +if (i >=K3_MMU_REGIONS_COUNT ) { +/* +* This is a recoverable error if the regions can be +* coalesced, the required logic can be implemented once +* requirement arises. +*/ +pr_err ("%s: Not enough space in MMU map for carveouts\n" , + __func__); +return -ENOMEM ; + } + + name =fdt_get_name (blob, subnode,NULL ); + addr =fdtdec_get_addr_size (blob, subnode,"reg" , &size); + +if (addr ==FDT_ADDR_T_NONE ) +continue ; + +if (!fdtdec_get_bool (blob, subnode,"no-map" )) +continue ; + +if (addr >= mem_base + mem_size) +continue ; + + end_addr = addr + size; + +if (end_addr <= mem_base) +continue ; + +debug ("Added memory carveout at 0x%llx, size: 0x%llx for '%s'\n" , + addr, size, name); + + addr =max (addr, mem_base); + end_addr =min (end_addr, mem_base + mem_size); + size = end_addr - addr; + dt_reserved[i].start = addr; + dt_reserved[i].end = end_addr; + i++; + } + carveout_len = i; + +if (!carveout_len) +return 0 ; + +/* sort carveout regions by address required for creating carveouts */ +qsort (dt_reserved, carveout_len,sizeof (dt_reserved[0 ]), + dt_reserved_cmp); + +/* coalesce regions */ +struct fdt_resource coalescing_temp = dt_reserved[0 ]; + + j =0 ; +for (i =1 ; i < carveout_len; i++) { +struct fdt_resource current = dt_reserved[i]; + +if (coalescing_temp.end >= current.start) { + coalescing_temp.end = current.end; +continue ; + } + coalesced[j] = coalescing_temp; + coalescing_temp = current; + j++; + } + + coalesced[j] = coalescing_temp; + carveout_len = j +1 ; + +if (coalesced[0 ].start != mem_base) { + k3_mem_map[k3_map_idx].virt = mem_base; + k3_mem_map[k3_map_idx].phys = mem_base; + k3_mem_map[k3_map_idx].size = coalesced[0 ].start - mem_base; + k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + } + +for (i =1 ; i < carveout_len; i++) { + k3_mem_map[k3_map_idx].virt = coalesced[i -1 ].end; + k3_mem_map[k3_map_idx].phys = coalesced[i -1 ].end; + k3_mem_map[k3_map_idx].size = + coalesced[i].start - coalesced[i -1 ].end; + k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + }
arch/arm/mach-k3/arm64/arm64-mmu.c struct mm_region k3_mem_map[] = { · line 216
Quoted 10 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +10
-0+ + k3_mem_map[k3_map_idx].virt = coalesced[carveout_len -1 ].end; + k3_mem_map[k3_map_idx].phys = coalesced[carveout_len -1 ].end; + k3_mem_map[k3_map_idx].size = + mem_base + mem_size - coalesced[carveout_len -1 ].end; + k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + + k3_mem_map[k3_map_idx] = (const struct mm_region ){0 };
I assume you are looping over reserved -memory and adding to map..
what if , we need some reserved memory as non-cached.
Quoted 8 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +3
-0+ +return 0 ; +}arch/arm/mach-k3/include/mach/k3-ddr.h +0
-0
arch/arm/mach-k3/include/mach/k3-ddr.h line 13
Quoted 6 lines
arch/arm/mach-k3/include/mach/k3-ddr.h +3
-0@@ -8,10 +8,19 @@#include <spl.h> +/* Number of mappable regions in the MMU page table */ +#define K3_MMU_REGIONS_COUNT 32 +
why 32 ?
Quoted 26 lines
arch/arm/mach-k3/include/mach/k3-ddr.h +6
-0int dram_init (void );int dram_init_banksize (void );void fixup_ddr_driver_for_ecc (struct spl_image_info *spl_image);void fixup_memory_node (struct spl_image_info *spl_image); +/* +* Modifies the MMU memory map based on DDR size and reserved-memory +* nodes in DT +*/ +int k3_mem_map_init (void ); +#endif /* _K3_DDR_H_ */ board/ti/common/k3-ddr.c +1
-0@@ -7,6 +7,7 @@#include <dm/uclass.h> #include <k3-ddrss.h> #include <spl.h> +#include <mach/k3-ddr.h> #include "k3-ddr.h"
board/ti/common/k3-ddr.c int dram_init(void) · line 24
Quoted 10 lines
board/ti/common/k3-ddr.c +6
-0@@ -14,6 +15,15 @@ int dram_init(void) {s32 ret; +if (IS_ENABLED (CONFIG_ARM64 ) &&xpl_phase () !=PHASE_SPL ) { + ret =k3_mem_map_init (); +if (ret) { +printf ("%s: Error fixing up MMU memory map: %d\n" , + __func__, ret); +return ret;
If can not set MMU table, then should be Hang , no ?
Quoted 6 lines
board/ti/common/k3-ddr.c +3
-0+ } + } + ret =fdtdec_setup_mem_size_base_lowest ();if (ret)printf("Error setting up mem size and base. %d\n", ret);
Quoted 9 lines · Udit Kumar
Quoted 4 lines · Anshul Dalal
In u-boot we only provide a single MMU table for all k3 platforms, this does not scale for devices with reserved memory outside the range
0x9e780000-0xa0000000(eg j722s[1]) or for devices with < 2GiB of memory (eg am62-SIP with 512MiB of RAM).please avoid giving reference to internal HACKs
Got it :)
Quoted 40 lines
Quoted 40 lines
To properly configure the MMU on various k3 platforms, the reserved-memory regions need to be queried at runtime from the device-tree and the MMU table should be updated accordingly.
This patch adds the required fixups to the MMU table (during proper U-boot stage) by marking the reserved regions as non cacheable and keeping the remaining area as cacheable.
For the A-core SPL, the 128MiB region starting from
SPL_TEXT_BASEis marked as cacheable i.e0x80080000to0x88080000.The 128MiB size is chosen to allow for future use cases such as falcon boot from the A-Core SPL which would require loading kernel image from the SPL stage. This change also ensures the reserved memory regions that all exist past
0x88080000are non cacheable preventing speculative accesses to those addresses.Signed-off-by Anshul Dalal <[email protected]>---
Changes for v3:
- Remove unused memory regions in SPL's map
- Add runtime addition of MMU entry for the framebuffer in SPL
- Refactork3_mem_map_initto use standard u-boot APIs
- Unmap reserved-memory regions instead of keeping them uncached
v2: https://lore.kernel.org/u-boot/[email protected]/Changes in v2:
- Removed dependency to:https://lore.kernel.org/u-boot/[email protected]/v1: https://lore.kernel.org/u-boot/[email protected]/
--- arch/arm/mach-k3/arm64/arm64-mmu.c | 191 +++++++++++++++++++++++-- arch/arm/mach-k3/include/mach/k3-ddr.h | 9 ++ board/ti/common/k3-ddr.c | 10 ++ 3 files changed, 197 insertions(+), 13 deletions(-)
[snip]
Quoted 19 lines
Quoted 6 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +6
-0+int k3_mem_map_init (void ) +{ +fdt_addr_t mem_base; +fdt_size_t mem_size; +struct fdt_resource dt_reserved[K3_MMU_REGIONS_COUNT ], + coalesced[K3_MMU_REGIONS_COUNT ];arch/arm/mach-k3/arm64/arm64-mmu.c struct mm_region k3_mem_map[] = { · line 82
Quoted 10 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +10
-0+int k3_map_idx = -EINVAL , ret, nodeoffset, subnode; +void *blob = (void *)gd->fdt_blob; +unsigned int carveout_len, i, j; + + ret =fdt_fixup_reserved (blob,"tfa" ,CONFIG_K3_ATF_LOAD_ADDR ,0x80000 ); +if (ret) { +pr_err ("%s: Failed to fixup reserved node for tfa [%d]\n" , + __func__, ret); +return ret; + }
am62l3_fdt.c, needs more size.
Quoted 3 lines
fdt_fixup_reserved(blob, "tfa", CONFIG_K3_ATF_LOAD_ADDR, 0x200000)
The fdt_fixup_reserved actually handles these contingencies, if a matching reserved region already exists it ignores the size we provided and uses the size form the dt instead.
Quoted 20 lines
Quoted 20 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +20
-0+ + ret =fdt_fixup_reserved (blob,"optee" ,CONFIG_K3_OPTEE_LOAD_ADDR , +0x1800000 ); +if (ret) { +pr_err ("%s: Failed to fixup reserved node for optee [%d]\n" , + __func__, ret); +return ret; + } + + nodeoffset =fdt_subnode_offset (blob,0 ,"memory" ); +if (nodeoffset <0 ) { +pr_err ("%s: Failed to get memory data: %s\n" , __func__, +fdt_strerror (nodeoffset)); +return nodeoffset; + } + + mem_base =fdtdec_get_addr_size (blob, nodeoffset,"reg" , &mem_size); +if (mem_base !=CFG_SYS_SDRAM_BASE ) +return -EINVAL ; +
Quoted 10 lines
arch/arm/mach-k3/arm64/arm64-mmu.c struct mm_region k3_mem_map[] = { · line 108
Quoted 6 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +6
-0+for (i =0 ; i <K3_MMU_REGIONS_COUNT ; i++) { +if (k3_mem_map[i].virt ==CONFIG_SPL_TEXT_BASE ) { + k3_map_idx = i; +break ; + } + }you can fix this entry at offset 0, instead of looping all over
We have to have some static nodes in the DT as well for peripherals, flash etc. All the nodes that are added as part of this fixup append to the table so as to not interfere with the already defined MMU entries.
If we move this entry to 0, appending to the table would overwrite the existing entries that we do want.
Quoted 8 lines
Quoted 8 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +8
-0+ +if (k3_map_idx == -EINVAL ) { +pr_err ("%s: Failed to find DDR region in MMU memory map\n" , + __func__); +return -EINVAL ; + } + + i =0 ;
[snip]
Quoted 11 lines
arch/arm/mach-k3/arm64/arm64-mmu.c struct mm_region k3_mem_map[] = { · line 216
Quoted 5 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +5
-0+ k3_mem_map[k3_map_idx].attrs =PTE_BLOCK_MEMTYPE (MT_NORMAL ) | +PTE_BLOCK_INNER_SHARE ; + k3_map_idx++; + + k3_mem_map[k3_map_idx] = (const struct mm_region ){0 };I assume you are looping over reserved -memory and adding to map..
what if , we need some reserved memory as non-cached.
Yes, here I loop over all the reserved regions and add the carveouts to the map after coalescing. So far the reserved regions are kept unmapped but that causes issues with remoteproc, I'll fix that in the next revision by marking them as non-cacheable instead.
Quoted 9 lines
Quoted 9 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +3
-0+ +return 0 ; +}arch/arm/mach-k3/include/mach/k3-ddr.h +0
-0@@ -8,10 +8,19 @@
Quoted 6 lines
arch/arm/mach-k3/include/mach/k3-ddr.h line 13
Quoted 5 lines
arch/arm/mach-k3/include/mach/k3-ddr.h +3
-0#include <spl.h> +/* Number of mappable regions in the MMU page table */ +#define K3_MMU_REGIONS_COUNT 32 +why 32 ?
I just took it as a worst case scenario with assuming > 10 reserved-memory nodes with no coalescing and two entries per node (one for the carveout and another for the non-cached entry as per above). The table size is 1KiB at the moment and I think it's a reasonable compromise.
Quoted 29 lines
Quoted 29 lines
arch/arm/mach-k3/include/mach/k3-ddr.h +6
-0int dram_init (void );int dram_init_banksize (void );void fixup_ddr_driver_for_ecc (struct spl_image_info *spl_image);void fixup_memory_node (struct spl_image_info *spl_image); +/* +* Modifies the MMU memory map based on DDR size and reserved-memory +* nodes in DT +*/ +int k3_mem_map_init (void ); +#endif /* _K3_DDR_H_ */ board/ti/common/k3-ddr.c +1
-0@@ -7,6 +7,7 @@#include <dm/uclass.h> #include <k3-ddrss.h> #include <spl.h> +#include <mach/k3-ddr.h> #include "k3-ddr.h" @@ -14,6 +15,15 @@ int dram_init(void) {s32 ret;
Quoted 8 lines
board/ti/common/k3-ddr.c int dram_init(void) · line 23
Quoted 7 lines
board/ti/common/k3-ddr.c +6
-0+if (IS_ENABLED (CONFIG_ARM64 ) &&xpl_phase () !=PHASE_SPL ) { + ret =k3_mem_map_init (); +if (ret) { +printf ("%s: Error fixing up MMU memory map: %d\n" , + __func__, ret); +return ret;If can not set MMU table, then should be Hang , no ?
That I have deferred to the caller which is initcall in our case which will hang the board if dram_init fails.
Quoted 6 lines
Quoted 6 lines
board/ti/common/k3-ddr.c +3
-0+ } + } + ret =fdtdec_setup_mem_size_base_lowest ();if (ret)printf("Error setting up mem size and base. %d\n", ret);
Quoted 72 lines · Anshul Dalal
Quoted 7 lines · Udit Kumar
Quoted 4 lines · Anshul Dalal
In u-boot we only provide a single MMU table for all k3 platforms, this does not scale for devices with reserved memory outside the range
0x9e780000-0xa0000000(eg j722s[1]) or for devices with < 2GiB of memory (eg am62-SIP with 512MiB of RAM).please avoid giving reference to internal HACKs
Got it :)
Quoted 40 lines
Quoted 40 lines
To properly configure the MMU on various k3 platforms, the reserved-memory regions need to be queried at runtime from the device-tree and the MMU table should be updated accordingly.
This patch adds the required fixups to the MMU table (during proper U-boot stage) by marking the reserved regions as non cacheable and keeping the remaining area as cacheable.
For the A-core SPL, the 128MiB region starting from
SPL_TEXT_BASEis marked as cacheable i.e0x80080000to0x88080000.The 128MiB size is chosen to allow for future use cases such as falcon boot from the A-Core SPL which would require loading kernel image from the SPL stage. This change also ensures the reserved memory regions that all exist past
0x88080000are non cacheable preventing speculative accesses to those addresses.Signed-off-by Anshul Dalal <[email protected]>---
Changes for v3:
- Remove unused memory regions in SPL's map
- Add runtime addition of MMU entry for the framebuffer in SPL
- Refactork3_mem_map_initto use standard u-boot APIs
- Unmap reserved-memory regions instead of keeping them uncached
v2: https://lore.kernel.org/u-boot/[email protected]/Changes in v2:
- Removed dependency to:https://lore.kernel.org/u-boot/[email protected]/v1: https://lore.kernel.org/u-boot/[email protected]/
--- arch/arm/mach-k3/arm64/arm64-mmu.c | 191 +++++++++++++++++++++++-- arch/arm/mach-k3/include/mach/k3-ddr.h | 9 ++ board/ti/common/k3-ddr.c | 10 ++ 3 files changed, 197 insertions(+), 13 deletions(-)[snip]
Quoted 20 lines
Quoted 6 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +6
-0+int k3_mem_map_init (void ) +{ +fdt_addr_t mem_base; +fdt_size_t mem_size; +struct fdt_resource dt_reserved[K3_MMU_REGIONS_COUNT ], + coalesced[K3_MMU_REGIONS_COUNT ];arch/arm/mach-k3/arm64/arm64-mmu.c struct mm_region k3_mem_map[] = { · line 82
Quoted 10 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +10
-0+int k3_map_idx = -EINVAL , ret, nodeoffset, subnode; +void *blob = (void *)gd->fdt_blob; +unsigned int carveout_len, i, j; + + ret =fdt_fixup_reserved (blob,"tfa" ,CONFIG_K3_ATF_LOAD_ADDR ,0x80000 ); +if (ret) { +pr_err ("%s: Failed to fixup reserved node for tfa [%d]\n" , + __func__, ret); +return ret; + }[..]+
arch/arm/mach-k3/arm64/arm64-mmu.c +1
-0+ k3_mem_map[k3_map_idx] = (const struct mm_region ){0 };I assume you are looping over reserved -memory and adding to map..
Quoted 7 lines
Quoted 3 lines
what if , we need some reserved memory as non-cached.
Yes, here I loop over all the reserved regions and add the carveouts to the map after coalescing. So far the reserved regions are kept unmapped but that causes issues with remoteproc, I'll fix that in the next revision by marking them as non-cacheable instead.
you should check wrt device capability, If device is coherent then it can be as is
If device is non-coherent (we observed in J722s, still to confirm this) then node should be
non-cacheable
Quoted 58 lines
Quoted 10 lines
Quoted 10 lines
arch/arm/mach-k3/arm64/arm64-mmu.c +3
-0+ +return 0 ; +}arch/arm/mach-k3/include/mach/k3-ddr.h +0
-0@@ -8,10 +8,19 @@Quoted 6 lines
arch/arm/mach-k3/include/mach/k3-ddr.h line 13
Quoted 5 lines
arch/arm/mach-k3/include/mach/k3-ddr.h +3
-0#include <spl.h> +/* Number of mappable regions in the MMU page table */ +#define K3_MMU_REGIONS_COUNT 32 +why 32 ?
I just took it as a worst case scenario with assuming > 10 reserved-memory nodes with no coalescing and two entries per node (one for the carveout and another for the non-cached entry as per above). The table size is 1KiB at the moment and I think it's a reasonable compromise.
Quoted 29 lines
Quoted 29 lines
arch/arm/mach-k3/include/mach/k3-ddr.h +6
-0int dram_init (void );int dram_init_banksize (void );void fixup_ddr_driver_for_ecc (struct spl_image_info *spl_image);void fixup_memory_node (struct spl_image_info *spl_image); +/* +* Modifies the MMU memory map based on DDR size and reserved-memory +* nodes in DT +*/ +int k3_mem_map_init (void ); +#endif /* _K3_DDR_H_ */ board/ti/common/k3-ddr.c +1
-0@@ -7,6 +7,7 @@#include <dm/uclass.h> #include <k3-ddrss.h> #include <spl.h> +#include <mach/k3-ddr.h> #include "k3-ddr.h" @@ -14,6 +15,15 @@ int dram_init(void) {s32 ret;Quoted 7 lines
board/ti/common/k3-ddr.c int dram_init(void) · line 23
Quoted 6 lines
board/ti/common/k3-ddr.c +6
-0+if (IS_ENABLED (CONFIG_ARM64 ) &&xpl_phase () !=PHASE_SPL ) { + ret =k3_mem_map_init (); +if (ret) { +printf ("%s: Error fixing up MMU memory map: %d\n" , + __func__, ret); +return ret;If can not set MMU table, then should be Hang , no ?
Quoted 2 lines
That I have deferred to the caller which is initcall in our case which will hang the board if
dram_initfails.
ok
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
board/ti/common/k3-ddr.c +3
-0+ } + } + ret =fdtdec_setup_mem_size_base_lowest ();if (ret)printf ("Error setting up mem size and base. %d\n" , ret);
Quoted 34 lines · Anshul Dalal
In u-boot we only provide a single MMU table for all k3 platforms, this does not scale for devices with reserved memory outside the range
0x9e780000-0xa0000000(eg j722s[1]) or for devices with < 2GiB of memory (eg am62-SIP with 512MiB of RAM).To properly configure the MMU on various k3 platforms, the reserved-memory regions need to be queried at runtime from the device-tree and the MMU table should be updated accordingly.
This patch adds the required fixups to the MMU table (during proper U-boot stage) by marking the reserved regions as non cacheable and keeping the remaining area as cacheable.
For the A-core SPL, the 128MiB region starting from
SPL_TEXT_BASEis marked as cacheable i.e0x80080000to0x88080000.The 128MiB size is chosen to allow for future use cases such as falcon boot from the A-Core SPL which would require loading kernel image from the SPL stage. This change also ensures the reserved memory regions that all exist past
0x88080000are non cacheable preventing speculative accesses to those addresses.Signed-off-by Anshul Dalal <[email protected]>---
Changes for v3:
- Remove unused memory regions in SPL's map
- Add runtime addition of MMU entry for the framebuffer in SPL
- Refactork3_mem_map_initto use standard u-boot APIs
- Unmap reserved-memory regions instead of keeping them uncached
v2: https://lore.kernel.org/u-boot/[email protected]/
Quoted 6 lines
Changes in v2:
- Removed dependency to:
https://lore.kernel.org/u-boot/[email protected]/
[snip]
Superseded by v4:
https://lore.kernel.org/u-boot/[email protected]/