[U-Boot] [PATCH] mtd: nand: mxs: fix cache alignment for cache lines >32

[Stefan Agner]

On 2016-08-03 11:44, Fabio Estevam wrote:
> Hi Simon,
> 
> On Wed, Aug 3, 2016 at 3:35 PM, Simon Glass <sjg at chromium.org> wrote:
> 
>> Actually I think these are bugs and should be fixed. In this case,

I am completely with you Simon!

check_cache_range return 0, which actually leads to the cache operation
not being executed at all! Depending on the situation this can be quite
catastrophic...


>> from what I can tell netboot_common() should cache-align the size in
>> the call to:
>>
>> /* flush cache */
>> flush_cache(load_addr, size);
> 
> Do you mean like this?
> 
> --- a/cmd/net.c
> +++ b/cmd/net.c
> @@ -244,6 +244,8 @@ static int netboot_common(enum proto_t proto, cmd_tbl_t *cmd
>         }
> 
>         /* flush cache */
> +       load_addr &= ~(CONFIG_SYS_CACHELINE_SIZE - 1);
> +       size = ALIGN(size, CONFIG_SYS_CACHELINE_SIZE);
>         flush_cache(load_addr, size);
> 
>         bootstage_mark(BOOTSTAGE_ID_NET_LOADED);
> 
> This makes the net warnings go away.

Yes, see my patch too.
https://patchwork.ozlabs.org/patch/654585/

So up until now, that stuff did not get flushed whenever the file size
was not cache line aligned, and nobody noticed... Is that cache flush
necessary at all?


> 
> There is still this one that I am seeing:
> 
> Kernel image @ 0x80800000 [ 0x000000 - 0x6fea70 ]
> ## Flattened Device Tree blob at 83000000
>    Booting using the fdt blob at 0x83000000
>    Using Device Tree in place at 83000000, end 83009c5d
> 
> Starting kernel ...
> 
> CACHE: Misaligned operation at range [00900000, 00900529]
> [    0.000000] Booting Linux on physical CPU 0x0
> 
> Any ideas where it may come from?

I did not had that one on i.MX 7

--
Stefan