Re: [PATCH] staging: MIPS: add Octeon USB HCD support

[David Daney <ddaney.cavm@gmail.com>]

On 06/03/2013 09:52 AM, Greg Kroah-Hartman wrote:
> On Mon, Jun 03, 2013 at 09:35:45AM -0700, David Daney wrote:
>> On 06/01/2013 11:42 AM, Aaro Koskinen wrote:
>>> Add support for Octeon USB HCD. Tested on EdgeRouter Lite with USB
>>> mass storage.
>>>
>>> The driver has been extracted from GPL sources of EdgeRouter Lite firmware
>>> (based on Linux 2.6.32.13). Some minor fixes and cleanups have been done
>>> to make it work with 3.10-rc3.
>>>
>>> $ uname -a
>>> Linux (none) 3.10.0-rc3-edge-00005-g86cb5bc #41 SMP PREEMPT Sat Jun 1 20:41:46 EEST 2013 mips64 GNU/Linux
>>> $ modprobe octeon-usb
>>> [   37.971683] octeon_usb: module is from the staging directory, the quality is unknown, you have been warned.
>>> [   37.983649] OcteonUSB: Detected 1 ports
>>> [   37.999360] OcteonUSB OcteonUSB.0: Octeon Host Controller
>>> [   38.004847] OcteonUSB OcteonUSB.0: new USB bus registered, assigned bus number 1
>>> [   38.012332] OcteonUSB OcteonUSB.0: irq 122, io mem 0x00000000
>>> [   38.019970] hub 1-0:1.0: USB hub found
>>> [   38.023851] hub 1-0:1.0: 1 port detected
>>> [   38.028101] OcteonUSB: Registered HCD for port 0 on irq 122
>>> [   38.391443] usb 1-1: new high-speed USB device number 2 using OcteonUSB
>>> [   38.586922] usb-storage 1-1:1.0: USB Mass Storage device detected
>>> [   38.597375] scsi0 : usb-storage 1-1:1.0
>>> [   39.604111] scsi 0:0:0:0: Direct-Access              USB DISK 2.0     PMAP PQ: 0 ANSI: 4
>>> [   39.619113] sd 0:0:0:0: [sda] 7579008 512-byte logical blocks: (3.88 GB/3.61 GiB)
>>> [   39.630696] sd 0:0:0:0: [sda] Write Protect is off
>>> [   39.635945] sd 0:0:0:0: [sda] No Caching mode page present
>>> [   39.641464] sd 0:0:0:0: [sda] Assuming drive cache: write through
>>> [   39.651341] sd 0:0:0:0: [sda] No Caching mode page present
>>> [   39.656917] sd 0:0:0:0: [sda] Assuming drive cache: write through
>>> [   39.664296]  sda: sda1 sda2
>>> [   39.675574] sd 0:0:0:0: [sda] No Caching mode page present
>>> [   39.681093] sd 0:0:0:0: [sda] Assuming drive cache: write through
>>> [   39.687223] sd 0:0:0:0: [sda] Attached SCSI removable disk
>>>
>>> Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
>>> ---
>>>   drivers/staging/Kconfig                      |    2 +
>>>   drivers/staging/Makefile                     |    1 +
>>>   drivers/staging/octeon-usb/Kconfig           |   10 +
>>>   drivers/staging/octeon-usb/Makefile          |    3 +
>>>   drivers/staging/octeon-usb/TODO              |   11 +
>>>   drivers/staging/octeon-usb/cvmx-usb.c        | 3344 ++++++++++++++++++++++++++
>>>   drivers/staging/octeon-usb/cvmx-usb.h        | 1085 +++++++++
>>>   drivers/staging/octeon-usb/cvmx-usbcx-defs.h | 3086 ++++++++++++++++++++++++
>>>   drivers/staging/octeon-usb/cvmx-usbnx-defs.h | 1596 ++++++++++++
>>>   drivers/staging/octeon-usb/octeon-hcd.c      |  854 +++++++
>>>   10 files changed, 9992 insertions(+)
>>>   create mode 100644 drivers/staging/octeon-usb/Kconfig
>>>   create mode 100644 drivers/staging/octeon-usb/Makefile
>>>   create mode 100644 drivers/staging/octeon-usb/TODO
>>>   create mode 100644 drivers/staging/octeon-usb/cvmx-usb.c
>>>   create mode 100644 drivers/staging/octeon-usb/cvmx-usb.h
>>>   create mode 100644 drivers/staging/octeon-usb/cvmx-usbcx-defs.h
>>>   create mode 100644 drivers/staging/octeon-usb/cvmx-usbnx-defs.h
>>>   create mode 100644 drivers/staging/octeon-usb/octeon-hcd.c
>>>
>>
>>
>> FYI: This is an alternate, host only, driver for the DesignWare USB2
>> controller as found on some OCTEON SoCs.  drivers/staging/dwc2
>> contains the code supplied by Synopsys for the same controller.
>>
>> One might ask why an alternate driver was ever written.  The answer
>> to this question is:  The octeon-hcd driver is much better than the
>> dwc2 driver when running the hardware in host mode.
>
> So, should we not accept this driver because dwc2 is in the tree now and
> we will have two conflicting drivers for the same hardware?  Or accept
> this, and blacklist dwc2 for this hardware?
>

I don't know the answer to these questions.  I will say this though:

o In USB2 mode, the octeon-hcd driver generates far fewer interrupts 
than dwc2.

o octeon-hcd was written because we couldn't figure out how to fix the 
dwc2 driver so that it wouldn't crash.

o I think Raspberry Pi uses the DWC2 block, and there are said to be 
many complaints about the dwc2 driver crashing there.

I would defer to the judgment of the USB maintainers and other users of 
the DWC2 block.  If you think the code is useful, take it.  If it is 
merged, we will probably supply some bug fixes and cleanups we have 
accumulated.

David Daney