[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: FW: Zlib compress



Greg Ungerer wrote:

> Hi David,
>
> David Woodhouse wrote:
>
>> startec@xxxxxxx.net said:
>>
>>> It is a alignment exception in the case and I found
>>> jffs2_zlib_compress will cause word aligment problem. I tried to print
>>> datalen and cdatalen using the following codes and I found the result
>>> before crash. 
>>
>>
>>
>> Linux requires that unaligned loads and stores work. We introduced 
>> get_unaligned() into the flash drivers a year or so ago but after 
>> some debate removed it again. The ARM architecture now should not let 
>> you select MTD devices unless you have CONFIG_ALIGNMENT_TRAP set.
>>
>> Greg, could I ask you either to do the same for the armnommu arch 
>> config.in
>> or maybe re-introduce the get_unaligned() macros in the chip drivers for
>> uClinux if you can really make a case for supporting flash but not 
>> including the 0x500-odd bytes it takes to fix up alignment traps.
>
>
> The current uClinux 2.4.x (which is based on linux 2.4.17)
> does have the CONFIG_ALIGNMENT_TRAP configurable for the
> armnommu targets. (Its implementation is identical to that
> of the arm targets - the code carried strait over).
>
>
>> How many of the CPUs supported by uClinux just can't trap unaligned 
>> accesses at all? 
>
>
> I can't think of any of the top of my head that cannot deal with
> it in some way, either supporting it directly, or through a trap.
>

I enable alignment trap and I got the error message during booting.

POSIX conformance testing by UNIFIX
Internal error: Oops - undefined instruction: 10e00

This is a new porting to Samsung S3C4510 and I an not sure that I need 
to modify alignment trap handler for the MCU. Thanks.

Steve Tsai



To unsubscribe from this list: send the line "unsubscribe jffs-dev" in
the body of a message to majordomo@xxxxxxx.com