On Mon, 6 Feb 2017 11:37:06 +0100, David Disseldorp wrote: > Hi Bart, > > On Fri, 3 Feb 2017 17:20:06 +0000, Bart Van Assche wrote: > > > It would be easy to modify the code such that a response is sent for all LU > > RESET TMFs. However, in SAM-5 section 6.3.3 LU RESET I found the following: > > > > When responding to a logical unit reset condition, the logical unit shall: > > [ ... ] > > c) terminate all task management functions. > > [ ... ] > > > > Your test submits two concurrent LU RESETs to the same logical unit. So I > > think sending a response for only one of these two LU RESETs is compliant > > with the SCSI specs. > > I don't there are any cases where the LU RESETs are concurrent: > - test_multipathio_reset.c issues an LU RESET on each of the two paths, > but the TMF requests are sent synchronously. > - test_async_lu_reset_simple.c dispatches a bunch of WRITE10s and then a > single async LU RESET while the write I/Os are outstanding. s/bunch of WRITE10s/single WRITE10/ > I'll do some network sniffing today, and upload the mainline and > e06fd3154941 captures for comparison. Please find the traces at: https://www.samba.org/~ddiss/netcap/lio_tmf_sync/ The 4.10.0-rc7 trace shows an immediate TMF response. In the e06fd3154941 trace, the initiator waits six seconds for the LU RESET response before logging out (the TMF response never comes). Cheers, David -- 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
