各位大侠，at91sam9260的SPI调试时出现问题
本帖最后由 kinfirge 于 2013-08-13 15:12 编辑各位大侠，本人在at91sam9260调试系统自带SPI出现问题，内核版本2.6.24
由于要用到SPI，考虑到IO模拟SPI速率慢且占CPU资源，打算用自带的SPI驱动，make menuconfig，已经把SPI相关勾选上了，且将内核目录arch/arm/mach-at91/board-sam9260ek.c的static struct spi_board_info ek_spi_devices[] = 也已经修改。
但在驱动中调用 spi_sync()时，出现问题，具体跟进去，发现出错在
void dma_cache_maint(const void *start, size_t size, int direction)
{
BUG_ON(!virt_addr_valid(start) || !virt_addr_valid(end - 1)); 就出错在就这行代码
}
求做过SPI的大神指点，小弟先谢过。
具体错误打印如下：
kernel BUG at arch/arm/mm/consistent.c:487!
Unable to handle kernel NULL pointer dereference at virtual address 00000000
pgd = c1d70000
*pgd=21c3b031, *pte=00000000, *ppte=00000000
Internal error: Oops: 817 [#1]
Modules linked in: spi_esam wdt
CPU: 0 Not tainted(2.6.24 #251)
PC is at __bug+0x24/0x30
LR is at release_console_sem+0x1cc/0x1e8
pc : [<c0028d74>] lr : [<c0039a48>] psr: 60000013
sp : c1d6fe58ip : c1d6fd98fp : c1d6fe64
r10: 00000008r9 : 00000000r8 : 00011008
r7 : c1d03200r6 : 00000001r5 : 00000008r4 : 00011008
r3 : 00000000r2 : c02bff98r1 : 000073adr0 : 0000002f
Flags: nZCvIRQs onFIQs onMode SVC_32ISA ARMSegment user
Control: 0005317fTable: 21d70000DAC: 00000015
Process esam_main (pid: 350, stack limit = 0xc1d6e268)
Stack: (0xc1d6fe58 to 0xc1d70000)
fe40: c1d6fe84 c1d6fe68
fe60: c002a88c c0028d60 c02bff5c c1d6ff04 c1d6fed8 c1d03518 c1d6feac c1d6fe88
fe80: c01835b4 c002a810 c1d6fed8 c1d6feb4 00000008 c1d6ff20 00000000 00000000
fea0: c1d6fed4 c1d6feb0 c01828e8 c01834f4 c004ff78 00000000 c1d6feb8 c1d6feb8
fec0: bf002e00 c1d6fed8 c1d6ff44 c1d6fed8 bf002178 c01828b0 c1d6ff20 c1d6ff20
fee0: c1d03200 00000000 c0182a6c c1d6feb4 00000000 00000000 00000000 00000000
ff00: 00000000 00011008 00000000 00000008 ffffffff ffffffff 00000000 00000000
ff20: c1d6fed8 c1d6fed8 00000008 c1c29620 00011008 c1d6ff70 c1d6ff6c c1d6ff48
ff40: c007eff0 bf002104 00000000 00000200 c1c29620 00011008 00000008 00000004
ff60: c1d6ffa4 c1d6ff70 c007f134 c007ef44 00000000 00000000 c002b4fc 00000000
ff80: 60000010 000000e4 00000140 401a0000 c0024f68 c1d6e000 00000000 c1d6ffa8
ffa0: c0024dc0 c007f100 000000e4 00000140 00000003 00011008 00000008 00011008
ffc0: 000000e4 00000140 401a0000 00000004 00000000 00000000 00008790 bea4ed4c
ffe0: 401a1b68 bea4ed28 000088f8 4003701c 60000010 00000003 00000000 00000000
Backtrace:
[<c0028d50>] (__bug+0x0/0x30) from [<c002a88c>] (dma_cache_maint+0x8c/0xdc)
[<c002a800>] (dma_cache_maint+0x0/0xdc) from [<c01835b4>] (atmel_spi_transfer+0xd0/0x1ec)
r6:c1d03518 r5:c1d6fed8 r4:c1d6ff04 r3:c02bff5c
[<c01834e4>] (atmel_spi_transfer+0x0/0x1ec) from [<c01828e8>] (spi_sync+0x48/0x70)
[<c01828a0>] (spi_sync+0x0/0x70) from [<bf002178>] (esam_write+0x84/0xe4 )
r5:c1d6fed8 r4:bf002e00
[<bf0020f4>] (esam_write+0x0/0xe4 ) from [<c007eff0>] (vfs_write+0xbc/0x138)
r7:c1d6ff70 r6:00011008 r5:c1c29620 r4:00000008
[<c007ef34>] (vfs_write+0x0/0x138) from [<c007f134>] (sys_write+0x44/0x70)
r7:00000004 r6:00000008 r5:00011008 r4:c1c29620
[<c007f0f0>] (sys_write+0x0/0x70) from [<c0024dc0>] (ret_fast_syscall+0x0/0x2c)
r9:c1d6e000 r8:c0024f68 r6:401a0000 r5:00000140 r4:000000e4
Code: e59f0010 e1a01003 eb09f737 e3a03000 (e5833000)
---[ end trace f40c5ded0f9921f6 ]---
Segmentation fault
LZ是如何跟进去调试的？
LZ是用SPI的DMA方式、中断方式还是轮询方式。
看似像虚拟地址跟物理地址转换的问题。
本帖最后由 kinfirge 于 2013-08-13 16:20 编辑
看了下atmel_spi.c，用的是SPI的DMA方式。
通过在各级函数调用都printk，发现执行到我标的那一行，没有执行下去，打印出一大堆错误信息。
驱动调用spi_sync或spi_write之后，接下去调用的层级函数关系是atmel_spi_transfer()——》atmel_spi_dma_map_xfer()——》dma_map_single()——》dma_cache_maint()
在执行到dma_cache_maint()函数里的BUG_ON(!virt_addr_valid(start) || !virt_addr_valid(end - 1));这行代码，出错！
通过google搜了一下，http://permalink.gmane.org/gmane.linux.ports.arm.general/10903，有人通过给mtdblock.c加几行补丁，如下
+#ifdef CONFIG_BSE_TWEAKS // BSE fix for dataflash/spi DMA mapping
+ mtdblk->cache_data = kmalloc(mtdblk->mtd->erasesize, GFP_KERNEL);
+#else
mtdblk->cache_data = vmalloc(mtdblk->mtd->erasesize);
+#endif
但这个参数CONFIG_BSE_TWEAKS，在我的内核以及2.6.25内核都找不到这个配置，不懂了，这个系统SPI怎么那么难弄的，觉得不应该啊
回复 2# 2879618
我也不太懂
LZ用SPI连接什么设备。 本帖最后由 kinfirge 于 2013-08-14 08:56 编辑
通过SPI和加密芯片通信；
求做过SPI的大神指点？？？
页:
[1]