nrf51 - Add async loader. Performance on nrf51822QAA/stlink-v2 from ~3.5KiB/s to ~19.5KiB/s.

Change-Id: Ib0bf41a0cec85f0bd5728551f8ad7f6255e4ea04
Signed-off-by: Angus Gratton <gus@projectgus.com>
[spamjunkeater@gmail.com: Cleanup buffer allocation, detect -1 for unknown pages]
Signed-off-by: Erdem U. Altunyurt <spamjunkeater@gmail.com>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Reviewed-on: http://openocd.zylin.com/2204
Tested-by: jenkins
Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
__archive__
Angus Gratton 2014-06-27 16:13:24 +10:00 committed by Andreas Fritiofson
parent b2dc1af59a
commit 233f8859c0
3 changed files with 248 additions and 87 deletions

View File

@ -0,0 +1,72 @@
/***************************************************************************
* Copyright (C) 2014 by Angus Gratton *
* Derived from stm32f1x.S:
* Copyright (C) 2011 by Andreas Fritiofson *
* andreas.fritiofson@gmail.com *
* Copyright (C) 2013 by Roman Dmitrienko *
* me@iamroman.org *
* *
* This program is free software; you can redistribute it and/or modify *
* it under the terms of the GNU General Public License as published by *
* the Free Software Foundation; either version 2 of the License, or *
* (at your option) any later version. *
* *
* This program is distributed in the hope that it will be useful, *
* but WITHOUT ANY WARRANTY; without even the implied warranty of *
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *
* GNU General Public License for more details. *
* *
* You should have received a copy of the GNU General Public License *
* along with this program. If not, see <http://www.gnu.org/licenses/>. *
***************************************************************************/
.text
.syntax unified
.cpu cortex-m0
.thumb
.thumb_func
/* Written for NRF51822 (src/flash/nor/nrf51.c) however the NRF NVMC is
* very generic (CPU blocks during flash writes), so this is actually
* just a generic word-oriented copy routine for cortex-m0 (also
* suitable for cortex m0plus/m3/m4.)
*
* To assemble:
* arm-none-eabi-gcc -c cortex-m0.S
*
* To disassemble:
* arm-none-eabi-objdump -o cortex-m0.o
*
* Thanks to Jens Bauer for providing advice on some of the tweaks.
*/
/* Params:
* r0 - byte count (in)
* r1 - workarea start
* r2 - workarea end
* r3 - target address
* Clobbered:
* r4 - rp
* r5 - wp, tmp
*/
wait_fifo:
ldr r5, [r1, #0] /* read wp */
cmp r5, #0 /* abort if wp == 0 */
beq exit
ldr r4, [r1, #4] /* read rp */
cmp r4, r5 /* wait until rp != wp */
beq wait_fifo
ldmia r4!, {r5} /* "*target_address++ = *rp++" */
stmia r3!, {r5}
cmp r4, r2 /* wrap rp at end of work area buffer */
bcc no_wrap
mov r4, r1
adds r4, #8 /* skip rp,wp at start of work area */
no_wrap:
str r4, [r1, #4] /* write back rp */
subs r0, #4 /* decrement byte count */
bne wait_fifo /* loop if not done */
exit:
bkpt #0

View File

@ -1,6 +1,8 @@
/*************************************************************************** /***************************************************************************
* Copyright (C) 2013 Synapse Product Development * * Copyright (C) 2013 Synapse Product Development *
* Andrey Smirnov <andrew.smironv@gmail.com> * * Andrey Smirnov <andrew.smironv@gmail.com> *
* Angus Gratton <gus@projectgus.com> *
* Erdem U. Altunyurt <spamjunkeater@gmail.com> *
* * * *
* This program is free software; you can redistribute it and/or modify * * This program is free software; you can redistribute it and/or modify *
* it under the terms of the GNU General Public License as published by * * it under the terms of the GNU General Public License as published by *
@ -23,6 +25,9 @@
#endif #endif
#include "imp.h" #include "imp.h"
#include <target/algorithm.h>
#include <target/armv7m.h>
#include <helper/types.h>
enum { enum {
NRF51_FLASH_BASE = 0x00000000, NRF51_FLASH_BASE = 0x00000000,
@ -248,6 +253,7 @@ static int nrf51_wait_for_nvmc(struct nrf51_info *chip)
alive_sleep(1); alive_sleep(1);
} while (timeout--); } while (timeout--);
LOG_DEBUG("Timed out waiting for NVMC_READY");
return ERROR_FLASH_BUSY; return ERROR_FLASH_BUSY;
} }
@ -557,6 +563,7 @@ static struct flash_sector *nrf51_find_sector_by_address(struct flash_bank *bank
static int nrf51_erase_all(struct nrf51_info *chip) static int nrf51_erase_all(struct nrf51_info *chip)
{ {
LOG_DEBUG("Erasing all non-volatile memory");
return nrf51_nvmc_generic_erase(chip, return nrf51_nvmc_generic_erase(chip,
NRF51_NVMC_ERASEALL, NRF51_NVMC_ERASEALL,
0x00000001); 0x00000001);
@ -568,8 +575,11 @@ static int nrf51_erase_page(struct flash_bank *bank,
{ {
int res; int res;
if (sector->is_protected) LOG_DEBUG("Erasing page at 0x%"PRIx32, sector->offset);
if (sector->is_protected) {
LOG_ERROR("Cannot erase protected sector at 0x%x", sector->offset);
return ERROR_FAIL; return ERROR_FAIL;
}
if (bank->base == NRF51_UICR_BASE) { if (bank->base == NRF51_UICR_BASE) {
uint32_t ppfc; uint32_t ppfc;
@ -608,19 +618,59 @@ static int nrf51_erase_page(struct flash_bank *bank,
return res; return res;
} }
static int nrf51_ll_flash_write(struct nrf51_info *chip, uint32_t offset, const uint8_t *buffer, uint32_t buffer_size) static const uint8_t nrf51_flash_write_code[] = {
/* See contrib/loaders/flash/cortex-m0.S */
/* <wait_fifo>: */
0x0d, 0x68, /* ldr r5, [r1, #0] */
0x00, 0x2d, /* cmp r5, #0 */
0x0b, 0xd0, /* beq.n 1e <exit> */
0x4c, 0x68, /* ldr r4, [r1, #4] */
0xac, 0x42, /* cmp r4, r5 */
0xf9, 0xd0, /* beq.n 0 <wait_fifo> */
0x20, 0xcc, /* ldmia r4!, {r5} */
0x20, 0xc3, /* stmia r3!, {r5} */
0x94, 0x42, /* cmp r4, r2 */
0x01, 0xd3, /* bcc.n 18 <no_wrap> */
0x0c, 0x46, /* mov r4, r1 */
0x08, 0x34, /* adds r4, #8 */
/* <no_wrap>: */
0x4c, 0x60, /* str r4, [r1, #4] */
0x04, 0x38, /* subs r0, #4 */
0xf0, 0xd1, /* bne.n 0 <wait_fifo> */
/* <exit>: */
0x00, 0xbe /* bkpt 0x0000 */
};
/* Start a low level flash write for the specified region */
static int nrf51_ll_flash_write(struct nrf51_info *chip, uint32_t offset, const uint8_t *buffer, uint32_t bytes)
{ {
int res; struct target *target = chip->target;
assert(buffer_size % 4 == 0); uint32_t buffer_size = 8192;
struct working_area *write_algorithm;
struct working_area *source;
uint32_t address = NRF51_FLASH_BASE + offset;
struct reg_param reg_params[4];
struct armv7m_algorithm armv7m_info;
int retval = ERROR_OK;
for (; buffer_size > 0; buffer_size -= 4) {
res = target_write_memory(chip->target, offset, 4, 1, buffer);
if (res != ERROR_OK)
return res;
res = nrf51_wait_for_nvmc(chip); LOG_DEBUG("Writing buffer to flash offset=0x%"PRIx32" bytes=0x%"PRIx32, offset, bytes);
if (res != ERROR_OK) assert(bytes % 4 == 0);
return res;
/* allocate working area with flash programming code */
if (target_alloc_working_area(target, sizeof(nrf51_flash_write_code),
&write_algorithm) != ERROR_OK) {
LOG_WARNING("no working area available, falling back to slow memory writes");
for (; bytes > 0; bytes -= 4) {
retval = target_write_memory(chip->target, offset, 4, 1, buffer);
if (retval != ERROR_OK)
return retval;
retval = nrf51_wait_for_nvmc(chip);
if (retval != ERROR_OK)
return retval;
offset += 4; offset += 4;
buffer += 4; buffer += 4;
@ -629,34 +679,101 @@ static int nrf51_ll_flash_write(struct nrf51_info *chip, uint32_t offset, const
return ERROR_OK; return ERROR_OK;
} }
static int nrf51_write_page(struct flash_bank *bank, uint32_t offset, const uint8_t *buffer) LOG_WARNING("using fast async flash loader. This is currently supported");
LOG_WARNING("only with ST-Link and CMSIS-DAP. If you have issues, add");
LOG_WARNING("\"set WORKAREASIZE 0\" before sourcing nrf51.cfg to disable it");
retval = target_write_buffer(target, write_algorithm->address,
sizeof(nrf51_flash_write_code),
nrf51_flash_write_code);
if (retval != ERROR_OK)
return retval;
/* memory buffer */
while (target_alloc_working_area(target, buffer_size, &source) != ERROR_OK) {
buffer_size /= 2;
buffer_size &= ~3UL; /* Make sure it's 4 byte aligned */
if (buffer_size <= 256) {
/* free working area, write algorithm already allocated */
target_free_working_area(target, write_algorithm);
LOG_WARNING("No large enough working area available, can't do block memory writes");
return ERROR_TARGET_RESOURCE_NOT_AVAILABLE;
}
}
armv7m_info.common_magic = ARMV7M_COMMON_MAGIC;
armv7m_info.core_mode = ARM_MODE_THREAD;
init_reg_param(&reg_params[0], "r0", 32, PARAM_IN_OUT); /* byte count */
init_reg_param(&reg_params[1], "r1", 32, PARAM_OUT); /* buffer start */
init_reg_param(&reg_params[2], "r2", 32, PARAM_OUT); /* buffer end */
init_reg_param(&reg_params[3], "r3", 32, PARAM_IN_OUT); /* target address */
buf_set_u32(reg_params[0].value, 0, 32, bytes);
buf_set_u32(reg_params[1].value, 0, 32, source->address);
buf_set_u32(reg_params[2].value, 0, 32, source->address + source->size);
buf_set_u32(reg_params[3].value, 0, 32, address);
retval = target_run_flash_async_algorithm(target, buffer, bytes/4, 4,
0, NULL,
4, reg_params,
source->address, source->size,
write_algorithm->address, 0,
&armv7m_info);
target_free_working_area(target, source);
target_free_working_area(target, write_algorithm);
destroy_reg_param(&reg_params[0]);
destroy_reg_param(&reg_params[1]);
destroy_reg_param(&reg_params[2]);
destroy_reg_param(&reg_params[3]);
return retval;
}
/* Check and erase flash sectors in specified range then start a low level page write.
start/end must be sector aligned.
*/
static int nrf51_write_pages(struct flash_bank *bank, uint32_t start, uint32_t end, const uint8_t *buffer)
{ {
assert(offset % 4 == 0);
int res = ERROR_FAIL; int res = ERROR_FAIL;
struct nrf51_info *chip = bank->driver_priv; struct nrf51_info *chip = bank->driver_priv;
struct flash_sector *sector = nrf51_find_sector_by_address(bank, offset); struct flash_sector *sector;
uint32_t offset;
if (!sector) assert(start % chip->code_page_size == 0);
assert(end % chip->code_page_size == 0);
/* Erase all sectors */
for (offset = start; offset < end; offset += chip->code_page_size) {
sector = nrf51_find_sector_by_address(bank, offset);
if (!sector) {
LOG_ERROR("Invalid sector @ 0x%08"PRIx32, offset);
return ERROR_FLASH_SECTOR_INVALID; return ERROR_FLASH_SECTOR_INVALID;
}
if (sector->is_protected) if (sector->is_protected) {
LOG_ERROR("Can't erase protected sector @ 0x%08"PRIx32, offset);
goto error; goto error;
}
if (sector->is_erased != 1) { if (sector->is_erased != 1) { /* 1 = erased, 0= not erased, -1 = unknown */
res = nrf51_erase_page(bank, chip, sector); res = nrf51_erase_page(bank, chip, sector);
if (res != ERROR_OK) { if (res != ERROR_OK) {
LOG_ERROR("Failed to erase sector @ 0x%08"PRIx32, sector->offset); LOG_ERROR("Failed to erase sector @ 0x%08"PRIx32, sector->offset);
goto error; goto error;
} }
} }
sector->is_erased = 0;
}
res = nrf51_nvmc_write_enable(chip); res = nrf51_nvmc_write_enable(chip);
if (res != ERROR_OK) if (res != ERROR_OK)
goto error; goto error;
sector->is_erased = 0; res = nrf51_ll_flash_write(chip, start, buffer, (end - start));
res = nrf51_ll_flash_write(chip, offset, buffer, chip->code_page_size);
if (res != ERROR_OK) if (res != ERROR_OK)
goto set_read_only; goto set_read_only;
@ -665,7 +782,7 @@ static int nrf51_write_page(struct flash_bank *bank, uint32_t offset, const uint
set_read_only: set_read_only:
nrf51_nvmc_read_only(chip); nrf51_nvmc_read_only(chip);
error: error:
LOG_ERROR("Failed to write sector @ 0x%08"PRIx32, sector->offset); LOG_ERROR("Failed to write to nrf51 flash");
return res; return res;
} }
@ -689,79 +806,51 @@ static int nrf51_code_flash_write(struct flash_bank *bank,
struct nrf51_info *chip, struct nrf51_info *chip,
const uint8_t *buffer, uint32_t offset, uint32_t count) const uint8_t *buffer, uint32_t offset, uint32_t count)
{ {
int res; int res;
struct { /* Need to perform reads to fill any gaps we need to preserve in the first page,
uint32_t start, end; before the start of buffer, or in the last page, after the end of buffer */
} region; uint32_t first_page = offset/chip->code_page_size;
uint32_t last_page = DIV_ROUND_UP(offset+count, chip->code_page_size);
region.start = offset; uint32_t first_page_offset = first_page * chip->code_page_size;
region.end = offset + count; uint32_t last_page_offset = last_page * chip->code_page_size;
struct { LOG_DEBUG("Padding write from 0x%08"PRIx32"-0x%08"PRIx32" as 0x%08"PRIx32"-0x%08"PRIx32,
size_t length; offset, offset+count, first_page_offset, last_page_offset);
const uint8_t *buffer;
} start_extra, end_extra;
start_extra.length = region.start % chip->code_page_size; uint32_t page_cnt = last_page - first_page;
start_extra.buffer = buffer; uint8_t buffer_to_flash[page_cnt*chip->code_page_size];
end_extra.length = region.end % chip->code_page_size;
end_extra.buffer = buffer + count - end_extra.length;
if (start_extra.length) {
uint8_t page[chip->code_page_size];
/* Fill in any space between start of first page and start of buffer */
uint32_t pre = offset - first_page_offset;
if (pre > 0) {
res = target_read_memory(bank->target, res = target_read_memory(bank->target,
region.start - start_extra.length, first_page_offset,
1, start_extra.length, page); 1,
if (res != ERROR_OK) pre,
return res; buffer_to_flash);
memcpy(page + start_extra.length,
start_extra.buffer,
chip->code_page_size - start_extra.length);
res = nrf51_write_page(bank,
region.start - start_extra.length,
page);
if (res != ERROR_OK) if (res != ERROR_OK)
return res; return res;
} }
if (end_extra.length) { /* Fill in main contents of buffer */
uint8_t page[chip->code_page_size]; memcpy(buffer_to_flash+pre, buffer, count);
/* Fill in any space between end of buffer and end of last page */
uint32_t post = last_page_offset - (offset+count);
if (post > 0) {
/* Retrieve the full row contents from Flash */ /* Retrieve the full row contents from Flash */
res = target_read_memory(bank->target, res = target_read_memory(bank->target,
region.end, offset + count,
1, 1,
(chip->code_page_size - end_extra.length), post,
page + end_extra.length); buffer_to_flash+pre+count);
if (res != ERROR_OK)
return res;
memcpy(page, end_extra.buffer, end_extra.length);
res = nrf51_write_page(bank,
region.end - end_extra.length,
page);
if (res != ERROR_OK) if (res != ERROR_OK)
return res; return res;
} }
return nrf51_write_pages(bank, first_page_offset, last_page_offset, buffer_to_flash);
region.start += start_extra.length;
region.end -= end_extra.length;
for (uint32_t address = region.start; address < region.end;
address += chip->code_page_size) {
res = nrf51_write_page(bank, address, &buffer[address - region.start]);
if (res != ERROR_OK)
return res;
}
return ERROR_OK;
} }
static int nrf51_uicr_flash_write(struct flash_bank *bank, static int nrf51_uicr_flash_write(struct flash_bank *bank,

View File

@ -17,11 +17,11 @@ if { [info exists ENDIAN] } {
} }
# Work-area is a space in RAM used for flash programming # Work-area is a space in RAM used for flash programming
# By default use 2kB # By default use 16kB
if { [info exists WORKAREASIZE] } { if { [info exists WORKAREASIZE] } {
set _WORKAREASIZE $WORKAREASIZE set _WORKAREASIZE $WORKAREASIZE
} else { } else {
set _WORKAREASIZE 0x800 set _WORKAREASIZE 0x4000
} }
if { [info exists CPUTAPID] } { if { [info exists CPUTAPID] } {