hi
On Sat, Aug 1, 2009 at 7:19 AM, Scott Woodscottwood@freescale.com wrote:
On Thu, Jun 25, 2009 at 05:21:45PM +0900, HeungJun Kim wrote:
+#ifdef CONFIG_S5PC1XX
- unsigned int *p = (unsigned int *) buf;
- int mem_addr, i;
- mem_addr = MEM_ADDR(block, page, 0);
- pagesize >>= 2;
- for (i = 0; i < pagesize; i++)
- *p++ = *(volatile unsigned int *)(CMD_MAP_01(mem_addr));
+#else /* CONFIG_S5PC1XX */
unsigned long *base;
#ifndef __HAVE_ARCH_MEMCPY32 @@ -85,6 +99,7 @@ static inline int onenand_read_page(ulong block, ulong page, *p++ = value; } #endif +#endif /* CONFIG_S5PC1XX */
It should be S5PC100, not used at S5PC110. it uses generic interface with DMA support.
Please put each controller implementation in a separate file, with an appropriate name. Do not make anything which is SoC-specific look generic.
No problem, One question, how to use use generic & s5pc100 read function simultaneously. Since our test board use two cpu, s5pc100 & s5pc110. but each cpu has different read function. I want to use it both with only one u-boot image. In this case we have to distinguish and use it properly. Now I add the function pointer and assign it based on cpu id.
If onenand_read_block() is not controller-specific (looks like everything except the read of ONENAND_REG_TECHNOLOGY is generic), move it to onenand_boot.c.
@@ -114,6 +129,9 @@ int onenand_read_block(unsigned char *buf)
erasesize = ONENAND_PAGES_PER_BLOCK * pagesize; nblocks = (CONFIG_SYS_MONITOR_LEN + erasesize - 1) >> erase_shift; +#ifdef CONFIG_S5PC1XX
- nblocks = 1;
+#endif
Why?
It's developer's mistake, no reason to use 1 block. actually internal ROM read 16KiB data to internal RAM in case of s5pc100.
Thank you, Kyungmin Park