On Nov 7, 2006, at 3:44 PM, Robert Hardy wrote:
> hverkuil: Please consider commiting this patch (or an adaption of it.)
> While debugging I found that fw->size tends to fluctuate between
> reboots and
> in certain instances up to an extra 700KB of random data was being
> read and
> writen after some firmware files (especially common in
> IVTV_DECODE_INIT_MPEG_FILENAME's case.)
I shouldn't be surprised by the variety of these "sometimes happen to
some people" things, but I still am.
> The following code rewrites the ivtv-firmware code so that it has more
> checks and balances and more closely matches recently discussed proper
> firmware_class usage examples provided by a new firmware_class
> maintainer as
> seen on the kernel mailing list.
>
> The most important modification is the code now no longer blindly
> writes all
> the data fw->size returns if it makes no sense to!
Maybe I misunderstand, but it seems to write at most the requested
size, not fw->size.
> if (fw->size >= size) {
> + // If there are at least "size" bytes of valid firmware
> load
> only "size" worth.
> retval = size;
> + IVTV_INFO("loading %s firmware (size=%ld
> bytes;fw->size=%d bytes)\n", fn, size, fw->size);
> + memcpy(dst, src, retval);
> } else {
> - retval = fw->size;
> + // Otherwise fail loudly
> + IVTV_ERR("reported size of firmware %s is %
> d. It should be at least %ld, loading aborting!\n", fn, fw->size,
> size);
> + retval = -EINVAL;
> }
> - for (i = 0; i < retval; i += 4) {
> - writel(*src, dst);
> - dst++;
> - src++;
I'm pretty sure those writel's were there to fix something.... *checks*
Refer to these:
http://ivtvdriver.org/trac/changeset/3089
http://ivtvdriver.org/trac/changeset/3152
--
Tyler Trafford
_______________________________________________
ivtv-devel mailing list
[email protected]
http://ivtvdriver.org/mailman/listinfo/ivtv-devel
