On Sat, Sep 22, 2012 at 5:07 PM, Nicholas A. Bellinger <nab@xxxxxxxxxxxxxxx> wrote: > Hi Ronnie, > > On Wed, 2012-09-19 at 19:17 -0700, ronnie sahlberg wrote: >> Ok, >> >> I have a small iscsi/scsi test tool as part of : >> https://github.com/sahlberg/libiscsi >> >> One of the new tests seems to make Fedora 17 LIO unhappy. >> Unhappy enough to stop responding completely and not allowing any >> further logins. >> >> >> This is how to run the test tool/reproducer : >> >> ./bin/iscsi-test --test="*0401*" >> iscsi://10.9.0.2/iqn.2003-01.org.linux-iscsi.fedora17.x8664:sn.84052e857f74/0 >> >> The 0401 test is a test for different sizes of alloc-len for the >> inquiry command. >> In this case it is the case where an inquiry is set to the target with >> an alloc-len of zero that lio does not like. >> >> The test case will "hang" so you will have to terminate it with CTRL-C. >> After this you can no longer login to the target again. I have to >> reboot the box to recover and allow logins to work again. >> >> >> BTW. the test tool does contain a reasonable set of tests so please >> feel free to try other tests too. >> >> ./bin/iscsi-test --list for a lot of all major tests >> and >> ./bin/iscsi-test --list --info for a list that also includes all subtests >> > > Thanks for reporting this zero-length allocation bug on your F17 setup. > > Paolo Bonzini recently did some work in target-core to gracefully handle > zero-length allocations for pSCSI backends (virtual backends like > IBLOCK+ FILEIO where fixed for v3.3.x IIRC). This series went into > v3.6-rc6 recently: > > The main ones in question from target-pending/master are: > > 6abbdf383 target: go through normal processing for zero-length REQUEST_SENSE > 32a8811ff target: support zero allocation length in REQUEST SENSE > 3717ef0c6 target: support zero-size allocation lengths in transport_kmap_data_sg > 9b16b9edb target: fail REPORT LUNS with less than 16 bytes of payload > 0d7f1299c target: report too-small parameter lists everywhere > 306c11b28 target: go through normal processing for zero-length PSCSI commands > d5829eac5 target: fix use-after-free with PSCSI sense data > 27a270991 target: simplify code around transport_get_sense_data Ah, I see Paolo has worked on this recently. In that case I think one of his patches above does address this issue already. > > Just to verify, which type of target backend driver is this with..? I use fileio. regards ronnie sahlberg -- 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
