Re: [Acx100-devel] [PATCH] Remove remaining warnings on x86_64

[Larry F. <Lar...@lw...>]

On 05/11/2012 02:40 PM, Stefan Weil wrote:
> Am 11.05.2012 18:56, schrieb Larry Finger:
>> The previous patch left some warnings on 64-bit systems that are now
>> removed.
>>
>> Signed-off-by: Larry Finger <Lar...@lw...>
>> ---
>> common.c | 3 ++-
>> merge.c | 22 +++++++++++-----------
>> 2 files changed, 13 insertions(+), 12 deletions(-)
>>
>> diff --git a/common.c b/common.c
>> index 2f1b338..25fbc27 100644
>> --- a/common.c
>> +++ b/common.c
>> @@ -1118,7 +1118,8 @@ static int acx100_create_dma_regions(acx_device_t * adev)
>> #ifdef CONFIG_ACX_MAC80211_MEM
>> else if (IS_MEM(adev)) {
>> /* sets the beginning of the rx descriptor queue, after the tx descrs */
>> - adev->acx_queue_indicator = (queueindicator_t *) le32_to_cpu
>> (queueconf.QueueEnd);
>> + adev->acx_queue_indicator = (queueindicator_t *)
>> + (ulong)le32_to_cpu(queueconf.QueueEnd);
>
> How can this and the other conversion (see rest of patch) of a 32 bit
> numeric value to a 64 bit pointer work?
>
> Maybe it works on a 64 bit host which only uses memory addresses in the
> first 4 GiB, but I don't expect that this is really safe for any host.
>
> For those cases where conversions from numeric values to pointers need
> a type cast, (uintptr_t) or (intptr_t) would be better, because those
> data types are specially designed to have the same size as a pointer.

Thanks for the comment. I was not aware of those particular type casts. My test 
box has only 2 GB RAM, and I certainly cannot test for memory > 4 GB.

I will respin the patch and the one that did not go to the list. It will have 
the same problems.

Larry