Description of Problem: Hardware Environment: Adaptec SCSI Host Bus Adapter Software Environment: Any How Reproducible: Steps to Reproduce: 1. Transfer odd byte (read or write) 2. Attach Magstar 3590 or IBM 3584 Actual Results: 1. Adaptec driver unable to handle odd byte transfer 2. Adaptec driver unable to claim LUN 1 devices Expected Results: allow odd byte transfer and lun 1 device attachment Additional Information: Kernel 2.4.2 had version 5.2.1 for aic7xxx and it worked. Kernel 2.4.5 had version 6.1.13 for aic7xxx and it fixes the problem as well. We need to know when would redhat release another version that would address this problem
Does the 6.1.11 version in the 2.4.3-12 kernel work ?
Putting two bugs into the same report makes it hard to close a report out as one bug is fixed/diagnosed. They should always be kept separate. However, to address the two issues: 1) Fails on odd byte transfers. I need exact error messages and I need to know what the OS did in this case. To my knowledge, the only way Linux will ever request the transfer of an odd byte count is during the INQUIRY command, so it's not something that should be seen under normal use, but instead only once at startup or some such. 2) Can't attach LUN 1 devices. What are you comparing this to? Custom built kernels or other Red Hat kernels? Have you considered that in Red Hat kernels we do not (and *NEVER* will) enable the PROBE_ALL_LUNS option? Have you tried an 'echo "scsi-add-single-device a b c d" > /proc/scsi/scsi' command (replacing a with the host number, b with the channel number, c with the device ID, and d with the LUN number) to see if the device is found then? If it is, then there is no bug here, it is expected behaviour. If not, then I need to know the particulars of the device and I would need a trace of the driver attempting to locate the LUN 1 device before I could do anything more.
Note from the submitter states: Got the following info from the bug submitter: ----------------------------------------------------------------------------- Sorry for not replying sooner. Yes, we have found a work round. It would appeared that Red Hat 7.1 ships with a newer version of the aic7xxx driver (named aic7xxx_mod), and that driver is working much better. We did get hold of Arjan, and that information came from him. -------------------------------------------------------------------------------- OK to close this bug.