tomof at acm.org wrote on Tue, 11 Sep 2007 02:31 +0900: > On Tue, 4 Sep 2007 16:03:06 -0400 > Pete Wyckoff <pw at osc.edu> wrote: > > > Repair one corner case in MaxRecvDataSegmentLength processing. Default > > target has always kept data segments at 8k, even if the initiator asks > > for more. Some target administrators may increase this by, e.g.: > > > > tgtadm --lld iscsi --mode target --op update --tid 1 \ > > --name MaxXmitDataSegmentLength --value 262144 > > > > so that target will generate up to min(initiator-specified RDSL, 256k). > > But if such a change is made, and the initiator does not specify RDSL, > > which is odd but specification compliant, the target will incorrectly > > use 256k instead of 8k. > > > > Signed-off-by: Pete Wyckoff <pw at osc.edu> > > --- > > usr/iscsi/iscsid.c | 10 +++++++--- > > usr/iscsi/iscsid.h | 1 + > > usr/iscsi/target.c | 2 +- > > 3 files changed, 9 insertions(+), 4 deletions(-) > > What? > > As I explained before, I can't understand your description. Oh, sorry. I thought our last go-round ended with my explaining why this was necessary. I'll drop the patch. It never happens in practice. Open-iscsi always provides MaxRecvDataSegmentLength during login, even if it is using the spec. default of 8k. Presumably other initiators do so too. -- Pete |