Re: [RFC PATCH] usb: dwc3: ep0: Fix mem corruption on OUT transfers of more than 512 bytes

[Kishon Vijay Abraham I <kishon@ti.com>]

Hi,

On Tuesday 09 June 2015 08:29 PM, Alan Stern wrote:
> On Tue, 9 Jun 2015, Kishon Vijay Abraham I wrote:
>
>> Hi,
>>
>> On Tuesday 09 June 2015 08:09 PM, Michael Trimarchi wrote:
>>> Hi
>>>
>>> On Jun 9, 2015 4:36 PM, "Kishon Vijay Abraham I" <kishon@ti.com
>>> <mailto:kishon@ti.com>> wrote:
>>>   >
>>>   > DWC3 uses bounce buffer to handle non max packet aligned OUT transfers and
>>>   > the size of bounce buffer is 512 bytes. However if the host initiates OUT
>>>   > transfers of size more than 512 bytes (and non max packet aligned), the
>>>   > driver throws a WARN dump but still programs the TRB to receive more than
>>>   > 512 bytes. This will cause bounce buffer to overflow and corrupt the
>>>   > adjacent memory locations which can be fatal.
>>>   >
>>>   > Fix it by programming the TRB to receive a maximum of DWC3_EP0_BOUNCE_SIZE
>>>   > (512) bytes.
>>>   >
>>>   > Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com <mailto:kishon@ti.com>>
>>>   > ---
>>>   > Steps to see the issue (before this patch)
>>>   > 1) Insert g_zero in DUT
>>>   > 2) run './testusb -t 14 -c 1 -s 520 -v 1' in host (size should be > 512)
>>>   >
>>>   > The test should FAIL since bounce buffer can handle only 512 bytes, but the
>>>   > test PASS. There is a WARN dump in DUT but still there will be memory
>>>   > corruption since the bounce buffer overflows.
>>>   >
>>>   > Tested this patch using USB3 Gen X CV (ch9 tests: usb2 and usb3, link layer
>>>   > testing and MSC tests) and using USB2 X CV (ch9 tests, MSC tests).
>>>   >
>>>   > After the patch, the tests timeout!
>>>   > ./testusb -t 14 -c 1 -s 514 -v 1
>>>   > unknown speed   /dev/bus/usb/001/018    0
>>>   > /dev/bus/usb/001/018 test 14 --> 110 (Connection timed out)
>>>   >
>>>   > IMO a patch to fix this is required for stable releases too. So If this
>>>   > patch is alright, I can post the patch cc'ing stable. While the actual fix
>>>   > would be to have chained TRB, I'm not sure if it can go to stable
>>>   > releases.
>>>   >  drivers/usb/dwc3/ep0.c |   12 ++++++++++--
>>>   >  1 file changed, 10 insertions(+), 2 deletions(-)
>>>   >
>>>   > diff --git a/drivers/usb/dwc3/ep0.c b/drivers/usb/dwc3/ep0.c
>>>   > index 2ef3c8d..8858c60 100644
>>>   > --- a/drivers/usb/dwc3/ep0.c
>>>   > +++ b/drivers/usb/dwc3/ep0.c
>>>   > @@ -816,6 +816,11 @@ static void dwc3_ep0_complete_data(struct dwc3 *dwc,
>>>   >                 unsigned maxp = ep0->endpoint.maxpacket;
>>>   >
>>>   >                 transfer_size += (maxp - (transfer_size % maxp));
>>>   > +
>>>   > +               /* Maximum of DWC3_EP0_BOUNCE_SIZE can only be received */
>>>   > +               if (transfer_size > DWC3_EP0_BOUNCE_SIZE)
>>>   > +                       transfer_size = DWC3_EP0_BOUNCE_SIZE;
>>>   > +
>>>
>>> Can you just use maxp in the correct way?
>>
>> what do you mean by correct way? Using roundup() to calculate transfer_size?
>
> Why not just make the bounce buffer size the same as the maxpacket
> size?  In other words, 1024 bytes instead of 512, for ep0 on a USB-3
> device.

It would still be possible for the host to send data more than 1024 bytes no? 
When working with DFU gadget, I've seen host sends data upto 4KB. Changing the 
bounce buffer size might not be able to fix all the cases IMO. The actual fix 
will be something like [1]

[1] -> http://comments.gmane.org/gmane.linux.kernel/1883688

Thanks
Kishon