file system error

I've run accross a problem with a new driver on IA 64 systems,
wondering if anyone has some insight.

With the Red Hat 7.2 Enigma release (2.4.9-18 or 2.4.9-31),
once installation is completed and rebooted, a new scsi driver is loaded.
The driver load proceeds without error - attached SCSI disks are found.

Attached disks can be partitioned and file systems created. Verified
that the disks are properly formatted. If this is a newly created file
system,
mount proceeds without error.  If files system has been mounted previously,
mount says the file system is dirty are requires a fs check.  
There are no SCSI errors and the driver is returning a status of GOOD to
the SCSI mid-layer for all commands.

If the mounted file system is of type FAT, fs is accessible.

If the mounted file system is of type ext2, any fs access produces a kernel
error.
The typical error is:
EXT2-fs error (device sd(8,35)): ext2_check_page: bad entry in directory #2:
rec_len is smaller than minimal - offset=0, inode=0, rec_len=0, name_len=0.

Occasionally, the error returns as

EXT2-fs error (device sd(8,35)): ext2_check_page: bad entry in directory #2:
unaligned directory entry - offset=0, inode=1702129257, rec_len=28274,
name_len=97.

Interestingly, the "ls" command will result in the first error above but the
SCSI analyzer 
shows the lost+found, . and .. directories ..  in the data stream.

Other Infor:
--------------
** SCSI boot (QLogic) is fine.
** Roswell II release of Red Hat 7.2 works fine.
** This appears to be dependent on the amount of memory in the system - the
machine
    exihibiting this failure contains 4GB mem whereas two other machines
with 2GB or less 
    memory have been running stress tests for weeks without error.
** The machine that is failing, for the most part, works fine with the
2.4.18 IA64 kernel.
   The disks are fully accessible.

If you have any thoughts on how to track down the problem - they would be
greatly
appreciated.

pamela delaney
LSI Logic
[email protected]
316 636 8166