Subject: Re: aic-7895
To: None <>
From: Justin T. Gibbs <>
List: current-users
Date: 02/08/1999 11:14:32
>>Really it's not.  The mechanism used by the NetBSD aic7xxx driver for
>>queuing transactions to the firmware will not work on this or any other
>>aic789X chip since the Q{IN/OUT}FIFO registers don't exist there.  The
>>latest firmware uses a different, more efficient, mechanism on all
>The 7895 isn't a 789x.  At least, the BSD/OS driver, which does *NOT* support
>the 7890, works fine on it, by probing it as if it were two 7880's.  It's
>not an ultra 2, it's a dual-channel.

The BSD/OS driver never used the Q{IN/OUT}FIFO registers.  This is why
their driver, which doesn't bother to take advantage of other nice features
of the aic7895 or even the aic7880, can work in this fashion.  I've had
a few conversations with the author of the BSD/OS driver and am familiar
with the internals of at least their original aic7xxx driver.

>The 7895 is really just a 3940 on a motherboard.

No.  It is a 3940AUW on a motherboard.  The 'A' is important.  The original
3940UW used a DEC PCI-PCI bridge and really was two 7880 chips.

>(And I think Adaptec should be shot for giving it a number that looks like
>a 789x.)

Perhaps.  The 7895 has several features that are also found on the Ultra2
chips which is probably the reason they decided to place it in the 9X 

Why do I know this?  Check out the copyright in the aic7xxx driver. 8-)