cvs commit: src/sys/cam/scsi scsi_da.c

[Nate Lawson]

Marcus Alves Grando wrote:
> mnag        2006-01-20 19:21:15 UTC
> 
>   FreeBSD src repository (ports committer)
> 
>   Modified files:        (Branch: RELENG_6)
>     sys/cam/scsi         scsi_da.c 
>   Log:
>   MFC:    src/sys/cam/scsi/scsi_da.c      1.184
>   
>   Allow to use Time DPA20B 1GB MP3 Player [1]
>   Allow to use Qware BeatZkey! Pro [2]
>   
>   PR:             usb/81846 [1], usb/79164 [2]
>   Submitted by:   Ruben de Groot <rdg at bzerk.org> [1], Radek Kozlowski <radek at raadradd.com> [2]
>   Approved by:    njl (implict)
>   
>   Revision   Changes    Path
>   1.180.2.3  +16 -0     src/sys/cam/scsi/scsi_da.c

It looks like we have another quirk proliferation problem that should be 
solved at the architectural level.  Does anyone have an idea if we can 
do something simpler than just disabling SYNC_CACHE completely?  I have 
a theory that Windows only does SYNC_CACHE in the final close of a USB 
device.  So a very simple implementation in the USB device firmware:

switch (cmd)
     case SYNC_CACHE:
         write_nonvolatile();
	while (1) ;

The problem may be that we are calling SYNC_CACHE at a time before final 
close.  If someone could test that a broken device works if we change 
this behavior, we could come up with an architectural fix.

Perhaps, change umass to filter out SYNC_CACHE but send it at final 
device close.  Although, since we don't have the device_eject() method 
currently, there may be no way to send it at final close.  If this is 
the case, we should just make umass disable SYNC_CACHE for all devices.

-- 
Nate