Hi MNC, I know Andy has stepped back from TCMU, but I wanted to get his input (CC'ed) as the original author. On Wed, 2017-03-01 at 23:14 -0600, Mike Christie wrote: > This patch removes the tcmu the cmd timeout code. Like with the core > target_core_tmr.c code, the kernel does not know the state of the command > or implemenation details of the device, so it might not be safe to just > return the command to the initiator for a possible retry. > > Signed-off-by: Mike Christie <mchristi@xxxxxxxxxx> Applied for the moment, but I'm curious wrt the implications for this..? Does this mean all outstanding se_cmd descriptors dispatched into user-space must always complete outstanding I/Os, otherwise target-core will block indefinitely waiting for completion a la other in-kernel backend drivers..? Historically we've always considered a backend device that never completes outstanding se_cmd descriptors to be a buggy driver. But existing user-space code that uses TCMU (I assume) is allowed to restart, and is currently not responsible for completing I/Os after the restart..? Is that the case..? I wonder if this change may end up being problematic because of this assumption by user-space..? -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
