On 05/12/10 19:43, Ivo Van Doorn wrote:
> On Wed, May 12, 2010 at 11:46 AM, Gertjan van Wingerde
> <gwingerde@xxxxxxxxx> wrote:
>> On Wed, May 12, 2010 at 8:59 AM, Ivo Van Doorn <ivdoorn@xxxxxxxxx> wrote:
>>> On Tue, May 11, 2010 at 11:51 PM, Gertjan van Wingerde
>>> <gwingerde@xxxxxxxxx> wrote:
>>>> The buffer address descriptor word is not part of the TXINFO structure
>>>> needed for beacons. The current writing of that word for beacons is
>>>> therefore an out-of-bounds write.
>>>> Fix this by only writing the buffer address descriptor word for TX
>>>> queues.
>>>>
>>>> Signed-off-by: Gertjan van Wingerde <gwingerde@xxxxxxxxx>
>>>> ---
>>>> drivers/net/wireless/rt2x00/rt61pci.c | 10 +++++-----
>>>> 1 files changed, 5 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/drivers/net/wireless/rt2x00/rt61pci.c b/drivers/net/wireless/rt2x00/rt61pci.c
>>>> index 2436363..99c2981 100644
>>>> --- a/drivers/net/wireless/rt2x00/rt61pci.c
>>>> +++ b/drivers/net/wireless/rt2x00/rt61pci.c
>>>> @@ -1801,12 +1801,12 @@ static void rt61pci_write_tx_desc(struct rt2x00_dev *rt2x00dev,
>>>> rt2x00_set_field32(&word, TXD_W5_WAITING_DMA_DONE_INT, 1);
>>>> rt2x00_desc_write(txd, 5, word);
>>>>
>>>> - rt2x00_desc_read(txd, 6, &word);
>>>> - rt2x00_set_field32(&word, TXD_W6_BUFFER_PHYSICAL_ADDRESS,
>>>> - skbdesc->skb_dma);
>>>> - rt2x00_desc_write(txd, 6, word);
>>>> + if (txdesc->queue != QID_BEACON) {
>>>> + rt2x00_desc_read(txd, 6, &word);
>>>> + rt2x00_set_field32(&word, TXD_W6_BUFFER_PHYSICAL_ADDRESS,
>>>> + skbdesc->skb_dma);
>>>> + rt2x00_desc_write(txd, 6, word);
>>>>
>>>> - if (skbdesc->desc_len > TXINFO_SIZE) {
>>>> rt2x00_desc_read(txd, 11, &word);
>>>> rt2x00_set_field32(&word, TXD_W11_BUFFER_LENGTH0,
>>>> txdesc->length);
>>>
>>> Shouldn't the check for TXINFO_SIZE be used rather than explicitly
>>> checking for the QID?
>>>
>>
>> I agree that this is a change that didn't have to be made in this patch.
>> However, after patch 4 of the series we cannot depend on the
>> skbdesc->desc_len being set anymore, and we would have to depend on
>> checking the QID anyway.
>> Note that in reality these two checks are completely equivalent with
>> respect to the result.
>
> Hmm, is that a good idea? I mean we are using the skbdesc inside the
> function, but we can't be sure that one of the basic values contains the right
> value?
>
To be honest, the chipset drivers don't need the skbdesc->desc and skbdesc->desc_len
at all. Based on the queue ID they already have everything they need to know to write
a descriptor. As far as I can tell the only code that really needs those two fields is
the rt2x00debug code, to dump a frame to userspace via debugfs.
As the dumping through debugfs is done after writing the TX descriptor, with this patch
I changed the desc and desc_len fields of the skbdesc to be output of the write_tx_desc
function (for the generic parts of the rt2x00 code) rather than being input.
Note that originally my plan was to get rid of the desc and desc_len field altogether,
but for now I refrained from doing that.
---
Gertjan.
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
