Workarounds in generic EHCI code

[M. Warner Losh]

In message: <47D1606D.3020506 at semihalf.com>
            Rafal Jaworowski <raj at semihalf.com> writes:
: M. Warner Losh wrote:
: 
: > : I'd like to hear comments on the proper way of handling non-standard behaviour
: > : of a host controller when a workaround needs to be aplied at the shared code
: > : level. In the following I followed an example of existing VIA/ATI chip workaround:
: > : 
: > : http://people.freebsd.org/~raj/patches/misc/usb-workarounds.diff
: > : 
: > : The respective flags are potentially set in platform-specific attachment
: > : driver code, which knows if they apply etc.
: > : 
: > : Does anybody see a better way to handle such cases?
: > 
: > This seems fine, but...
: > 
: > +#define EHCI_SCFLG_USBMODEBUG	0x0003	/* workaround for Marvell 88F5281 chipsets */
: > +#define EHCI_SCFLG_FORCESPEED	0x0004	/* workaround for Marvell chipsets */
: > +#define EHCI_SCFLG_NORESTERM	0x0005	/* don't terminate reset sequence on Marvell chipsets */
: > 
: > Two comments.  '3' isn't a bit, so those values need to change.
: 
: Ah, sure, thanks.
: 
: > Second, I'd make the descriptions a little better.  The first two
: > aren't very helpful...
: > 
: 
: I have more detailed descriptions in the platform code that sets this, e.g.
: 
:  /*
:   * Workaround for Marvell 88F5281 integrated EHCI controller: reset of
:   * the EHCI core clears the USBMODE register, which sets the core in
:   * an undefined state (neither host nor agent), so it needs to be set
:   * again for proper operation.
:   *
:   * Refer to errata document (p. 5.24 GL USB-2) for details.
: 
:   */
: 
: But maybe it really better fits in the ehic.c/ehcivar.h.. I'll try to improve
: this.

The platform code description is good for "why it is needed."  Maybe
have a better "what it is" in the ehci.c/ehcivar.h code.

Warner