Peter Kirchgessner wrote: > Hello, > > some time ago problems have been reported with some 2.4.x kernels and > SCSI-devices. What was working with 2.2.x, failed there with an "error > during device I/O". The problem seemed to be a broken handling of > SCSI-requests in the kernel (from mailing list: kernels 2.4.6, 2.4.10, > 2.4.12 and 2.4.13). > Does anybody know if the problem has been fixed ? And if so, at which > kernel ? The ChangeLogs don't tell me much.
Peter, If the error from the sg driver is actually EIO then it indicates that the version 2 sg interface (based on the sg_header struct) is being used. One shortcoming of this interface was the difficulty of overcoming the default assumption of a SCSI command size based on its opcode (first byte). The SANE Canon backend was tripped up by that and supplied a patch to the sg driver. That patch probably hasn't worked (i.e. breaks when being applied) since lk 2.2.6 (over 3 years ago). The original sg driver (version 1) yielded EIO for just about every error it encountered! Any recent SANE backend driver should be using the sanei layer which on a linux 2.4 series kernel will be using the sg version 3 interface. In this case EIO should not occur. Doug Gilbert
