Is it possible for Realtek card reader driver to reside in SCSI subsystem?

Is it possible for Realtek card reader driver to reside in SCSI subsystem?

[edwin_rong]

Dear James and all :

Sorry to disturb you again!

I'm an software engineer of Realtek corporation, responsible for writing 
driver for Realtek Card Reader chips.

Our device supports SD/MMC/MS/MSpro/xD series of cards, etc., which is 
implemented as an SCSI device in our driver,
and now our driver rts_pstor is under staging folder of linux kernel, so 
I want to know whether it is possible to move it out of staging folder,
and reside in SCSI subsystem?

I also know that both "mmc" and "memstick" subsystem exist in kernel 
now, but our device is a composition of these types of cards,
so it seems not suitable for keeping our driver there.

All replies are appreciated.

Thanks & BRs
Edwin.Rong

Re: Is it possible for Realtek card reader driver to reside in SCSI subsystem?

[James Bottomley]

On Mon, 2012-04-23 at 11:49 +0800, edwin_rong wrote:
> Dear James and all :
> 
> Sorry to disturb you again!
> 
> I'm an software engineer of Realtek corporation, responsible for writing 
> driver for Realtek Card Reader chips.
> 
> Our device supports SD/MMC/MS/MSpro/xD series of cards, etc., which is 
> implemented as an SCSI device in our driver,
> and now our driver rts_pstor is under staging folder of linux kernel, so 
> I want to know whether it is possible to move it out of staging folder,
> and reside in SCSI subsystem?
> 
> I also know that both "mmc" and "memstick" subsystem exist in kernel 
> now, but our device is a composition of these types of cards,
> so it seems not suitable for keeping our driver there.
> 
> All replies are appreciated.

The general rule is that if the device itself speaks SCSI ... as in
either the firmware or the underlying disk does, then you should be
using SCSI (This doesn't mean you have to have an actual SCSI device
anywhere ... lots of USB devices are some wierd flash or IDE device
fronted by a chip that does SCSI<->whatever translation [usually
badly]).  Conversely if you would be writing your SCSI command emulation
in the driver, don't ... you should be using another subsystem.

James

Re: Is it possible for Realtek card reader driver to reside in SCSIsubsystem?

[edwin_rong]

On 04/23/2012 03:39 PM, James Bottomley wrote:
> On Mon, 2012-04-23 at 11:49 +0800, edwin_rong wrote:
>> Dear James and all :
>>
>> Sorry to disturb you again!
>>
>> I'm an software engineer of Realtek corporation, responsible for writing
>> driver for Realtek Card Reader chips.
>>
>> Our device supports SD/MMC/MS/MSpro/xD series of cards, etc., which is
>> implemented as an SCSI device in our driver,
>> and now our driver rts_pstor is under staging folder of linux kernel, so
>> I want to know whether it is possible to move it out of staging folder,
>> and reside in SCSI subsystem?
>>
>> I also know that both "mmc" and "memstick" subsystem exist in kernel
>> now, but our device is a composition of these types of cards,
>> so it seems not suitable for keeping our driver there.
>>
>> All replies are appreciated.
> The general rule is that if the device itself speaks SCSI ... as in
> either the firmware or the underlying disk does, then you should be
> using SCSI (This doesn't mean you have to have an actual SCSI device
> anywhere ... lots of USB devices are some wierd flash or IDE device
> fronted by a chip that does SCSI<->whatever translation [usually
> badly]).  Conversely if you would be writing your SCSI command emulation
> in the driver, don't ... you should be using another subsystem.
>
> James
>
Dear James,

Got it, thanks for your response, sincerely.

As to my case, which subsystem do you think is fit for our driver to 
stay, could you give me some suggestions?

Thanks & BRs
Edwin

Re: Is it possible for Realtek card reader driver to reside in SCSIsubsystem?

[James Bottomley]

[updated Greg to non-SUSE address]
On Mon, 2012-04-23 at 16:00 +0800, edwin_rong wrote:
> On 04/23/2012 03:39 PM, James Bottomley wrote:
> > On Mon, 2012-04-23 at 11:49 +0800, edwin_rong wrote:
> >> Dear James and all :
> >>
> >> Sorry to disturb you again!
> >>
> >> I'm an software engineer of Realtek corporation, responsible for writing
> >> driver for Realtek Card Reader chips.
> >>
> >> Our device supports SD/MMC/MS/MSpro/xD series of cards, etc., which is
> >> implemented as an SCSI device in our driver,
> >> and now our driver rts_pstor is under staging folder of linux kernel, so
> >> I want to know whether it is possible to move it out of staging folder,
> >> and reside in SCSI subsystem?
> >>
> >> I also know that both "mmc" and "memstick" subsystem exist in kernel
> >> now, but our device is a composition of these types of cards,
> >> so it seems not suitable for keeping our driver there.
> >>
> >> All replies are appreciated.
> > The general rule is that if the device itself speaks SCSI ... as in
> > either the firmware or the underlying disk does, then you should be
> > using SCSI (This doesn't mean you have to have an actual SCSI device
> > anywhere ... lots of USB devices are some wierd flash or IDE device
> > fronted by a chip that does SCSI<->whatever translation [usually
> > badly]).  Conversely if you would be writing your SCSI command emulation
> > in the driver, don't ... you should be using another subsystem.
> >
> > James
> >
> Dear James,
> 
> Got it, thanks for your response, sincerely.
> 
> As to my case, which subsystem do you think is fit for our driver to 
> stay, could you give me some suggestions?

Well, you said it's a combination of mmc and memstick.  If it can be
bound as two separate drivers, I'd say one in each.  If there's magic
that has to be done in the binding (as in you need a single bind driver
that activates each component), I'd say be guided by the maintainers of
those components.  My instinct would be to put it in one and use the
other via Kconfig, but whatever they find most appropriate.

James

Re: Is it possible for Realtek card reader driver to reside inSCSIsubsystem?

[edwin_rong]

On 04/23/2012 04:11 PM, James Bottomley wrote:
> [updated Greg to non-SUSE address]
> On Mon, 2012-04-23 at 16:00 +0800, edwin_rong wrote:
>> On 04/23/2012 03:39 PM, James Bottomley wrote:
>>> On Mon, 2012-04-23 at 11:49 +0800, edwin_rong wrote:
>>>> Dear James and all :
>>>>
>>>> Sorry to disturb you again!
>>>>
>>>> I'm an software engineer of Realtek corporation, responsible for writing
>>>> driver for Realtek Card Reader chips.
>>>>
>>>> Our device supports SD/MMC/MS/MSpro/xD series of cards, etc., which is
>>>> implemented as an SCSI device in our driver,
>>>> and now our driver rts_pstor is under staging folder of linux kernel, so
>>>> I want to know whether it is possible to move it out of staging folder,
>>>> and reside in SCSI subsystem?
>>>>
>>>> I also know that both "mmc" and "memstick" subsystem exist in kernel
>>>> now, but our device is a composition of these types of cards,
>>>> so it seems not suitable for keeping our driver there.
>>>>
>>>> All replies are appreciated.
>>> The general rule is that if the device itself speaks SCSI ... as in
>>> either the firmware or the underlying disk does, then you should be
>>> using SCSI (This doesn't mean you have to have an actual SCSI device
>>> anywhere ... lots of USB devices are some wierd flash or IDE device
>>> fronted by a chip that does SCSI<->whatever translation [usually
>>> badly]).  Conversely if you would be writing your SCSI command emulation
>>> in the driver, don't ... you should be using another subsystem.
>>>
>>> James
>>>
>> Dear James,
>>
>> Got it, thanks for your response, sincerely.
>>
>> As to my case, which subsystem do you think is fit for our driver to
>> stay, could you give me some suggestions?
> Well, you said it's a combination of mmc and memstick.  If it can be
> bound as two separate drivers, I'd say one in each.  If there's magic
> that has to be done in the binding (as in you need a single bind driver
> that activates each component), I'd say be guided by the maintainers of
> those components.  My instinct would be to put it in one and use the
> other via Kconfig, but whatever they find most appropriate.
>
> James
>
>
Gotcha!

Thanks for your suggestion, James.


BRs
Edwin

Re: Is it possible for Realtek card reader driver to reside inSCSIsubsystem?

[Greg KH]

On Mon, Apr 23, 2012 at 04:24:49PM +0800, edwin_rong wrote:
> On 04/23/2012 04:11 PM, James Bottomley wrote:
> >[updated Greg to non-SUSE address]
> >On Mon, 2012-04-23 at 16:00 +0800, edwin_rong wrote:
> >>On 04/23/2012 03:39 PM, James Bottomley wrote:
> >>>On Mon, 2012-04-23 at 11:49 +0800, edwin_rong wrote:
> >>>>Dear James and all :
> >>>>
> >>>>Sorry to disturb you again!
> >>>>
> >>>>I'm an software engineer of Realtek corporation, responsible for writing
> >>>>driver for Realtek Card Reader chips.
> >>>>
> >>>>Our device supports SD/MMC/MS/MSpro/xD series of cards, etc., which is
> >>>>implemented as an SCSI device in our driver,
> >>>>and now our driver rts_pstor is under staging folder of linux kernel, so
> >>>>I want to know whether it is possible to move it out of staging folder,
> >>>>and reside in SCSI subsystem?
> >>>>
> >>>>I also know that both "mmc" and "memstick" subsystem exist in kernel
> >>>>now, but our device is a composition of these types of cards,
> >>>>so it seems not suitable for keeping our driver there.
> >>>>
> >>>>All replies are appreciated.
> >>>The general rule is that if the device itself speaks SCSI ... as in
> >>>either the firmware or the underlying disk does, then you should be
> >>>using SCSI (This doesn't mean you have to have an actual SCSI device
> >>>anywhere ... lots of USB devices are some wierd flash or IDE device
> >>>fronted by a chip that does SCSI<->whatever translation [usually
> >>>badly]).  Conversely if you would be writing your SCSI command emulation
> >>>in the driver, don't ... you should be using another subsystem.
> >>>
> >>>James
> >>>
> >>Dear James,
> >>
> >>Got it, thanks for your response, sincerely.
> >>
> >>As to my case, which subsystem do you think is fit for our driver to
> >>stay, could you give me some suggestions?
> >Well, you said it's a combination of mmc and memstick.  If it can be
> >bound as two separate drivers, I'd say one in each.  If there's magic
> >that has to be done in the binding (as in you need a single bind driver
> >that activates each component), I'd say be guided by the maintainers of
> >those components.  My instinct would be to put it in one and use the
> >other via Kconfig, but whatever they find most appropriate.
> >
> >James
> >
> >
> Gotcha!
> 
> Thanks for your suggestion, James.

Also, at the least, you have a lot of basic coding style issues to clean
up in the drivers/staging/rts_pstor/ directory before your code can move
out of there.  Please start working on that as soon as possible.

thanks,

greg k-h

Re: Is it possible for Realtek card reader driver to resideinSCSIsubsystem?

[edwin_rong]

On 04/23/2012 11:48 PM, Greg KH wrote:
> On Mon, Apr 23, 2012 at 04:24:49PM +0800, edwin_rong wrote:
>> On 04/23/2012 04:11 PM, James Bottomley wrote:
>>> [updated Greg to non-SUSE address]
>>> On Mon, 2012-04-23 at 16:00 +0800, edwin_rong wrote:
>>>> On 04/23/2012 03:39 PM, James Bottomley wrote:
>>>>> On Mon, 2012-04-23 at 11:49 +0800, edwin_rong wrote:
>>>>>> Dear James and all :
>>>>>>
>>>>>> Sorry to disturb you again!
>>>>>>
>>>>>> I'm an software engineer of Realtek corporation, responsible for writing
>>>>>> driver for Realtek Card Reader chips.
>>>>>>
>>>>>> Our device supports SD/MMC/MS/MSpro/xD series of cards, etc., which is
>>>>>> implemented as an SCSI device in our driver,
>>>>>> and now our driver rts_pstor is under staging folder of linux kernel, so
>>>>>> I want to know whether it is possible to move it out of staging folder,
>>>>>> and reside in SCSI subsystem?
>>>>>>
>>>>>> I also know that both "mmc" and "memstick" subsystem exist in kernel
>>>>>> now, but our device is a composition of these types of cards,
>>>>>> so it seems not suitable for keeping our driver there.
>>>>>>
>>>>>> All replies are appreciated.
>>>>> The general rule is that if the device itself speaks SCSI ... as in
>>>>> either the firmware or the underlying disk does, then you should be
>>>>> using SCSI (This doesn't mean you have to have an actual SCSI device
>>>>> anywhere ... lots of USB devices are some wierd flash or IDE device
>>>>> fronted by a chip that does SCSI<->whatever translation [usually
>>>>> badly]).  Conversely if you would be writing your SCSI command emulation
>>>>> in the driver, don't ... you should be using another subsystem.
>>>>>
>>>>> James
>>>>>
>>>> Dear James,
>>>>
>>>> Got it, thanks for your response, sincerely.
>>>>
>>>> As to my case, which subsystem do you think is fit for our driver to
>>>> stay, could you give me some suggestions?
>>> Well, you said it's a combination of mmc and memstick.  If it can be
>>> bound as two separate drivers, I'd say one in each.  If there's magic
>>> that has to be done in the binding (as in you need a single bind driver
>>> that activates each component), I'd say be guided by the maintainers of
>>> those components.  My instinct would be to put it in one and use the
>>> other via Kconfig, but whatever they find most appropriate.
>>>
>>> James
>>>
>>>
>> Gotcha!
>>
>> Thanks for your suggestion, James.
> Also, at the least, you have a lot of basic coding style issues to clean
> up in the drivers/staging/rts_pstor/ directory before your code can move
> out of there.  Please start working on that as soon as possible.
>
> thanks,
>
> greg k-h
Dear Greg,

Yes, it is.
Now we're working on the refactor of Realtek card reader driver to make
it support
all Realtek driver-based chips with USB or PCIe inferface, that is to
say, making all Realtek card reader driver in one,
and it will substitute currently existing "rts_pstro" and "rts5139"
under driver/staging directory in the future.

Thanks
Edwin