kern/44002: 3ware 9690 (ld driver) doesn't respond after transfer big amount of data

[novotny%ics.muni.cz@localhost]

>Number:         44002
>Category:       kern
>Synopsis:       3ware 9690 (ld driver) doesn't respond after transfer big 
>amount of data
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Oct 26 17:25:00 +0000 2010
>Originator:     Jiri Novotny
>Release:        5.1RC4, 5.0.2
>Organization:
Masaryk University
>Environment:
NetBSD bumbrdlik5.1RC4 5.1_RC4 NetBSD 5.1_RC4 (GENERIC) #0: Tue Sep 14 00:04:23 
UTC 2010  
builds%b8.netbsd.org@localhost:/home/builds/ab/netbsd-5-1-RC4/i386/201009140238Z-obj/home/builds/ab/netbsd-5-1-RC4/src/sys/arch/i386/compile/GENERIC
 i386
>Description:
3ware 9690 raid doesn't respond after big data transfer. I use i386, GPT and O2 
filesystem. After transfer of big amount of data (write, read) the raid stop to 
respond. When I attempt to read after writting huge amount of data the raid 
stop respond near every time.
>How-To-Repeat:
# gpt show ld1
        start         size  index  contents
            0  19531141087         
  19531141087           32         Sec GPT table
  19531141119            1         Sec GPT header
# 
dkctl ld1d addwedge dk0 0  19531141087 ffs

newfs -O2 -s -64m dk0

mount /dev/dk0 /mnt

# for i in `seq 1 10` ; do dd if=/dev/zero bs=$((1024*1024)) count=$((1024*$i)) 
of=/mnt/file$i 2>&1 | grep -v records ; done
1073741824 bytes transferred in 4.131 secs (259922978 bytes/sec)
2147483648 bytes transferred in 7.094 secs (302718303 bytes/sec)
3221225472 bytes transferred in 12.966 secs (248436331 bytes/sec)
4294967296 bytes transferred in 15.752 secs (272661712 bytes/sec)
5368709120 bytes transferred in 20.360 secs (263689053 bytes/sec)
6442450944 bytes transferred in 24.519 secs (262753413 bytes/sec)
7516192768 bytes transferred in 29.414 secs (255531133 bytes/sec)
8589934592 bytes transferred in 33.420 secs (257029760 bytes/sec)
9663676416 bytes transferred in 39.032 secs (247583429 bytes/sec)
10737418240 bytes transferred in 43.147 secs (248856658 bytes/sec)
# for i in `seq 1 10` ; do dd if=/dev/zero bs=$((1024*1024)) count=$((1024*$i)) 
of=/mnt/file$i 2>&1 | grep -v records ; done

... and disk stop to respond

There is the same behavior under NetBSD 5.0.2 and 5.1RC4. The raid was tested 
under CENTOS 5 as well and works pretty nice (100x write and read the example 
above).
>Fix:
Don't know.