On Sun, 25 Feb 2007 20:51:12 +0100
DI BACCO ANTONIO - technolabs wrote:

Hi Melinda,

the problem in my case was caused by the following instructions to
allocate the buffers for SPI controller tx and rx:

        rxbuffer = (u8*)m8xx_cpm_hostalloc(SPI_MAX_BUFFER_SIZE);  
        txbuffer = (u8*)m8xx_cpm_hostalloc(SPI_MAX_BUFFER_SIZE); 
        // Set the bd's rx and tx buffer address pointers
        rbdf->cbd_bufaddr = virt_to_bus((void
*)rxbuffer); tbdf->cbd_bufaddr = virt_to_bus((void *)txbuffer);   

I don't know why (I'm asking an opinion to Vitaly Bordug), I replaced
the allocation of rx and tx buffers with:

...->cbd_bufaddr = __pa(__get_free_page(GFP_KERNEL | GFP_DMA))

Well this is just evil... even if it works :)

but it will better to use dma_alloc_coherent.
In my case the use of m8xx_cpm_hostalloc led to an immediate kernel
hang after giving the start command to the spi controller.

Heh, this case you'lll have to use dpram for the buffers I guess iiuc.

dma_* stuff might work too but I'm not sure. Moreover, it didn't work  for 
me in case of powerpc, where m8xx_cpm_hostalloc made as a placeholder for 
dma_alloc_coherent...