[SeaBIOS] [RFC PATCH 00/16] make SCSI layer more generic, introduce virtio-scsi driver

Paolo Bonzini pbonzini at redhat.com
Wed Nov 16 09:01:58 CET 2011


On 11/16/2011 02:46 AM, Kevin O'Connor wrote:
> Question on patch 4 - can you elaborate on why waittick is needed?

What happens is that if you do two consecutive OUT sends, the second 
fails ~90% of the time with a timeout in wait_td.  I initially wanted to 
add the wait in usb-msc.c and only for writes, but there is no public 
interface for [uoe]hci_waittick.

The speed wasn't measurably different, but uhci is glacial anyway and I 
haven't tested ehci.

> On patch 5, the READ CAPACITY is made unconditional.  The intent was
> to only run READ CAPACITY on harddrives, because CDROMs need to get
> the info from the CD and it can take some time (eg, 10 seconds) to
> spin the CD up on real hardware.

Ah ok, so also the scsi_is_ready is something that you do not want on 
CDs (unless booting from them, in which case it is done by src/cdrom.c). 
  Will adjust.

> One wouldn't want their computer to
> boot longer just because they had a CD in the drive on boot.  (BTW,
> the check for blksize==CDROM_SECTOR_SIZE to force a CDROM type was
> something I added when testing - to the best of my knowledge no drive
> requires it.)

Good, I'll remove it.

Paolo



More information about the SeaBIOS mailing list