ok, looks like i did some dirty work there :-)
i will look into it tonight and update the sdk, i'll let you know as
soon as i have fixed it
Quoting "andrea.maccaferri" <[EMAIL PROTECTED]>:
> I'm sorry for my english, I hope everybody can understand.
>
> I think that there is a mistake in the foxbone_syscalls.c,
> the foxbone bus is 16 bit wide, so a "unsigned short int *" pointer
> is used for bulk data write or bulk data read array exchange from
> user code to kernel code, but inside foxbone_syscalls.c function
> an "unsigned int *buffer" is used, this has the effect that the
> "lenght" should be double of the required data, and the user array is
> filled only to even short integer word.
>
> This is the actual wrong (in my opinion) code from foxbone_syscalls.c:
>
> asmlinkage void sys_foxbonebulkread(unsigned short int reg, unsigned
> short int *value, unsigned int length){
> unsigned int i;
> unsigned int *buffer = kmalloc(length * 2, GFP_KERNEL);
> *buffer = foxbone_read(reg);
> for(i = 1; i < length; i++){
> buffer[i] = foxbone_read_next();
> };
> copy_to_user(value, buffer, length * 2);
> kfree(buffer); };
>
> asmlinkage void sys_foxbonebulkwrite(unsigned short int reg, unsigned
> short int *value, unsigned int length){
> unsigned int i;
> unsigned int *buffer = kmalloc(length * 2, GFP_KERNEL);
> copy_from_user(buffer, value, length * 2);
> foxbone_write(reg, *buffer);
> for(i = 1; i < length; i++){
> foxbone_write_next(buffer[i]);
> }; kfree(buffer); };
>
> This is the suggested patch to kmalloc line:
>
> unsigned short int *buffer = kmalloc(length * 2, GFP_KERNEL);
>
> best regards
> Andrea
>
>
