This is, unfortunately, not a recent regression but it's only been

recently diagnosed.  Apparently the SCSI Parallel transport class domain

validation cable width detection wasn't working, leading to cases where

controllers with damaged cables would end up hanging the system (the

reported one was an aic79xx controller, but the potential is there for

all SPI based systems).  This bug would *only* affect systems whose

cable integrity or connectors were compromised, so it isn't life

threatening to every SCSI Parallel installation, but the consequence of

running into it is a system hang.
The fix is available here:
master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6.git
The short changelog is:
James Bottomley (1):

      scsi_transport_spi: fix domain validation failure from incorrect

width setting
And the diffstat:
 scsi_transport_spi.c |   28 ++++++++++++++++++++++------

 1 file changed, 22 insertions(+), 6 deletions(-)
James
-

We've had another late arriving bug fix:
commit ff4abd6cfacf0bb23a077f615d3a5cd17359db1b

Author: David Miller <davem@davemloft.net>

Date:   Fri Aug 24 22:25:58 2007 -0700
    [SCSI] esp: fix instance numbering.
    Because the ->unique_id is set too late, the ESP scsi host

    instance numbers in the kernel log during probing are

    wrong.
    Bug reported by Meelis Roos.
    Signed-off-by: David S. Miller <davem@davemloft.net>

    Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Which I've added.  The diffstat is now:
 esp_scsi.c           |    3 ++-

 scsi_transport_spi.c |   28 ++++++++++++++++++++++------

 2 files changed, 24 insertions(+), 7 deletions(-)
James
-