USB mass storage package (slave side).

USB mass storage package (slave side).

[Christophe Coutand]

[-- Attachment #1: Type: text/plain, Size: 352 bytes --]

There has been some request for a USB mass storage package (slave side).
This was tested on AT91 architecture with the AT91 eCos USB device
driver. Some patches are required to the AT91 USB device driver for this
package to work properly. If anyone has the possibility to test it with
other architecture it would be most welcome.

Christophe 


[-- Attachment #2: doclist.diff --]
[-- Type: application/octet-stream, Size: 460 bytes --]

diff -r eb303059b6e5 doc/sgml/doclist
--- a/doc/sgml/doclist	Thu Jun 10 20:46:18 2010 +0000
+++ b/doc/sgml/doclist	Sun Jun 27 16:31:24 2010 +0200
@@ -41,6 +41,7 @@
 io/usb/slave/current/doc/usbs.sgml
 io/usb/eth/slave/current/doc/usbseth.sgml
 io/usb/serial/slave/current/doc/usbs_serial.sgml
+io/usb/msd/slave/current/doc/usbs_msd.sgml
 hal/synth/arch/current/doc/synth.sgml
 hal/m68k/arch/current/doc/m68k.sgml
 hal/m68k/mcf52xx/var/current/doc/mcf52xx.sgml

[-- Attachment #3: ecos.db --]
[-- Type: application/octet-stream, Size: 321 bytes --]

package CYGPKG_IO_USB_SLAVE_MSD {
        alias		{ "USB slave-side mass-storage drivers" usbs_msd }
        directory       io/usb/msd/slave
        script          usbs_msd.cdl
        description     "Support for USB peripherals that present themselves 
                         as mass-storage devices to the host."
}

[-- Attachment #4: io_usb_msd.diff.tgz --]
[-- Type: application/x-compressed, Size: 23722 bytes --]

Re: USB mass storage package (slave side).

[John Dallaway]

Hi Christophe

Christophe Coutand wrote:

> There has been some request for a USB mass storage package (slave side).
> This was tested on AT91 architecture with the AT91 eCos USB device
> driver. Some patches are required to the AT91 USB device driver for this
> package to work properly. If anyone has the possibility to test it with
> other architecture it would be most welcome.

I've been working with Chris Holgate (and Ilija Stanislevik) to get his
STM32 USB slave device driver ready for check-in:

  http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001024

Once this is checked-in, I plan to look at your MSD function driver.

One immediate issue is that the STM32 USB slave device driver uses
dynamic I/O endpoint configuration and does not provide the endpoint
devtab entries which are required by your driver. Looking at your code,
it seems that the endpoint devtab entries are only used to look up the
endpoint structures within usbs_msd_init():

  tab = usbs_msd_lookup( tx_ep_dev_name );
  msd->tx_ep     =  (usbs_tx_endpoint *) tab->priv;

  tab = usbs_msd_lookup( rx_ep_dev_name );
  msd->rx_ep     =  (usbs_rx_endpoint *) tab->priv;

So it seems it should be easy to eliminate the requirement for these
devtab entries. Have I understood correctly?

John Dallaway
eCos maintainer

RE: USB mass storage package (slave side).

[Christophe Coutand]

Hi John,

Calling USB endpoint with their devtab entry name ( .e.g. /dev/usbsX )
was meant to be more user friendly but can be easily changed. I guess
that is the meaning of the new USB API as well.

Note that the patch is missing a file for some reason ( usbs_msd_scsi.h
). I can generate a new patch, shall I remove the dependency towards the
endpoint devtab entry?

Regards,
Christophe

-----Original Message-----
From: John Dallaway [mailto:john@dallaway.org.uk] 
Sent: 18. oktober 2010 14:54
To: Christophe Coutand
Cc: ecos-patches@ecos.sourceware.org; Chris Holgate; Ilija Stanislevik
Subject: Re: USB mass storage package (slave side).

Hi Christophe

Christophe Coutand wrote:

> There has been some request for a USB mass storage package (slave
side).
> This was tested on AT91 architecture with the AT91 eCos USB device
> driver. Some patches are required to the AT91 USB device driver for
this
> package to work properly. If anyone has the possibility to test it
with
> other architecture it would be most welcome.

I've been working with Chris Holgate (and Ilija Stanislevik) to get his
STM32 USB slave device driver ready for check-in:

  http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001024

Once this is checked-in, I plan to look at your MSD function driver.

One immediate issue is that the STM32 USB slave device driver uses
dynamic I/O endpoint configuration and does not provide the endpoint
devtab entries which are required by your driver. Looking at your code,
it seems that the endpoint devtab entries are only used to look up the
endpoint structures within usbs_msd_init():

  tab = usbs_msd_lookup( tx_ep_dev_name );
  msd->tx_ep     =  (usbs_tx_endpoint *) tab->priv;

  tab = usbs_msd_lookup( rx_ep_dev_name );
  msd->rx_ep     =  (usbs_rx_endpoint *) tab->priv;

So it seems it should be easy to eliminate the requirement for these
devtab entries. Have I understood correctly?

John Dallaway
eCos maintainer

Re: USB mass storage package (slave side).

[John Dallaway]

Hi Christophe

Christophe Coutand wrote:

> Calling USB endpoint with their devtab entry name ( .e.g. /dev/usbsX )
> was meant to be more user friendly but can be easily changed. I guess
> that is the meaning of the new USB API as well.

The new API is intended to provide more flexibility for modern USB slave
hardware (dynamic endpoint configuration).

> Note that the patch is missing a file for some reason ( usbs_msd_scsi.h
> ). I can generate a new patch, shall I remove the dependency towards the
> endpoint devtab entry?

If you have time to remove the dependency on endpoint devtab entries,
that would be great. Also, could you raise a new bugzilla report at
http://bugs.ecos.sourceware.org for your contribution? It's much easier
to track patches and the discussion about them in Bugzilla.

To be clear, the new USB API presented at:

  http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001024

is not yet formalised, but any future changes required to use the new
API should be independent of work to remove the dependency on endpoint
devtab entries within your MSD function driver package.

Thank you

John Dallaway
eCos maintainer

RE: USB mass storage package (slave side).

[Christophe Coutand]

Hi John,

Bugzilla report was previously created:

   http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000987

I will upload an updated patch.

Christophe

-----Original Message-----
From: John Dallaway [mailto:john@dallaway.org.uk] 
Sent: 18. oktober 2010 16:08
To: Christophe Coutand
Cc: ecos-patches@ecos.sourceware.org; Chris Holgate; Ilija Stanislevik
Subject: Re: USB mass storage package (slave side).

Hi Christophe

Christophe Coutand wrote:

> Calling USB endpoint with their devtab entry name ( .e.g. /dev/usbsX )
> was meant to be more user friendly but can be easily changed. I guess
> that is the meaning of the new USB API as well.

The new API is intended to provide more flexibility for modern USB slave
hardware (dynamic endpoint configuration).

> Note that the patch is missing a file for some reason (
usbs_msd_scsi.h
> ). I can generate a new patch, shall I remove the dependency towards
the
> endpoint devtab entry?

If you have time to remove the dependency on endpoint devtab entries,
that would be great. Also, could you raise a new bugzilla report at
http://bugs.ecos.sourceware.org for your contribution? It's much easier
to track patches and the discussion about them in Bugzilla.

To be clear, the new USB API presented at:

  http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001024

is not yet formalised, but any future changes required to use the new
API should be independent of work to remove the dependency on endpoint
devtab entries within your MSD function driver package.

Thank you

John Dallaway
eCos maintainer