On 10/30/2014 12:26 PM, Hannes Reinecke wrote: > > Why would req->cmd.len be wrong though? Are you masking another bug? > > No. I'm _fixing_ a bug.
Sure, I just wasn't sure which/how. :) > scsi_disk.c:scsi_new_request() is just calling scsi_req_alloc(), > which does not set cmd.len. > Yet later on scsi_new_request() uses cmd.len to print out the CDB, > which at the time isn't initialized. Ah, right---req->cmd is set after scsi_new_request() returns. Paolo
