Message-ID: <e581b566-65fb-c4d8-74fc-1c1b35b57b9f@kernel.org>
Date:   Thu, 1 Aug 2019 08:03:59 -0600
From:   shuah <shuah@...nel.org>
To:     Suwan Kim <suwan.kim027@...il.com>
Cc:     valentina.manea.m@...il.com, stern@...land.harvard.edu,
        gregkh@...uxfoundation.org, linux-usb@...r.kernel.org,
        linux-kernel@...r.kernel.org, shuah <shuah@...nel.org>
Subject: Re: [PATCH v2 2/2] usbip: Implement SG support to vhci

On 8/1/19 12:38 AM, Suwan Kim wrote:
> On Mon, Jul 29, 2019 at 10:32:31AM -0600, shuah wrote:
>> On 7/29/19 8:52 AM, Suwan Kim wrote:
>>> Hi Shuah,
>>>
>>> On Tue, Jul 23, 2019 at 06:21:53PM -0600, shuah wrote:
>>>> Hi Suwan,
>>>>
>>>> On 7/5/19 10:43 AM, Suwan Kim wrote:
>>>>> There are bugs on vhci with usb 3.0 storage device. Originally, vhci
>>>>> doesn't supported SG, so USB storage driver on vhci breaks SG list
>>>>> into multiple URBs and it causes error that a transfer got terminated
>>>>> too early because the transfer length for one of the URBs was not
>>>>> divisible by the maxpacket size.
>>>>>
>>>>> In this patch, vhci basically support SG and it sends each SG list
>>>>> entry to the stub driver. Then, the stub driver sees the total length
>>>>> of the buffer and allocates SG table and pages according to the total
>>>>> buffer length calling sgl_alloc(). After the stub driver receives
>>>>> completed URB, it again sends each SG list entry to vhci.
>>>>>
>>>>> If HCD of the server doesn't support SG, the stub driver breaks a
>>>>> single SG reqeust into several URBs and submit them to the server's
>>>>> HCD. When all the split URBs are completed, the stub driver
>>>>> reassembles the URBs into a single return command and sends it to
>>>>> vhci.
>>>>>
>>>>> Signed-off-by: Suwan Kim <suwan.kim027@...il.com>
>>>>> ---
>>>>>     drivers/usb/usbip/stub.h         |   7 +-
>>>>>     drivers/usb/usbip/stub_main.c    |  52 +++++---
>>>>>     drivers/usb/usbip/stub_rx.c      | 207 ++++++++++++++++++++++---------
>>>>>     drivers/usb/usbip/stub_tx.c      | 108 +++++++++++-----
>>>>>     drivers/usb/usbip/usbip_common.c |  60 +++++++-- >   drivers/usb/usbip/vhci_hcd.c     |  10 +-
>>>>>     drivers/usb/usbip/vhci_tx.c      |  49 ++++++--
>>>>>     7 files changed, 372 insertions(+), 121 deletions(-)
>>>>
>>>> While you are working on v3 to fix chekpatch and other issues
>>>> I pointed out, I have more for you.
>>>>
>>>> What happens when you have mismatched server and client side?
>>>> i.e stub does and vhci doesn't and vice versa.
>>>>
>>>> Make sure to run checkpatch. Also check spelling errors in
>>>> comments and your commit log.
>>>>
>>>> I am not sure if your eeror paths are correct. Run usbip_test.sh
>>>>
>>>> tools/testing/selftests/drivers/usb/usbip
>>>
>>> I don't know what mismatch you mentioned. Are you saying
>>> "match busid table" at the end of usbip_test.sh?
>>> How does it relate to SG support of this patch?
>>> Could you tell me more about the problem situation?
>>>
>>
>> What happens when usbip_host is running a kernel without the sg
>> support and vhci_hcd does? Just make sure this works with the
>> checks for sg support status as a one of your tests for this
>> sg feature.
> 
> Now I understand. Thanks for the details!
> As a result of testing, in the situation where vhci supports SG,
> but stub does not, or vice versa, usbip works normally. Moreover,
> because there is no protocol modification, there is no problem in
> communication between server and client even if the one has a kernel
> without SG support.
> 
> In the case of vhci supports SG and stub doesn't, because vhci sends
> only the total length of the buffer to stub as it did before the
> patch applied, stub only needs to allocate the required length of
> buffers regardless of whether vhci supports SG or not.
> 
> If stub needs to send data buffer to vhci because of IN pipe, stub
> also sends only total length of buffer as metadata and then send real
> data as vhci does. Then vhci receive data from stub and store it to
> the corresponding buffer of SG list entry.
> 
> In the case of stub that supports SG, if SG buffer is requested by
> vhci, buffer is allocated by sgl_alloc(). However, stub that does
> not support SG allocates buffer using only kmalloc(). Therefore, if
> vhci supports SG and stub doesn't, stub has to allocate buffer with
> kmalloc() as much as the total length of SG buffer which is quite
> huge when vhci sends SG request, so it has big overhead in buffer
> allocation.
> 
> And for the case of stub supports SG and vhci doesn't, since the
> USB storage driver checks that vhci doesn't support SG and sends
> the request to stub by splitting the SG list into multiple URBs,
> stub allocates a buffer with kmalloc() as it did before this patch.
> 
> Therefore, everything works normally in a mismatch situation.

Looking for you add a test case for that. Please include this
in the commit log.

> I will send the v3 soon.
> 

Please send it soon. It would be nice to have this done as soon
as possible.

thanks,
-- Shuah

Powered by blists - more mailing lists