Bug 159709
| Summary: | Executing "shut/no shut" on iscsi port on MDS crashed the linux host | ||
|---|---|---|---|
| Product: | Red Hat Enterprise Linux 4 | Reporter: | Nitin Chandna <cnitin> |
| Component: | iscsi-initiator-utils | Assignee: | Mike Christie <mchristi> |
| Status: | CLOSED ERRATA | QA Contact: | Brock Organ <borgan> |
| Severity: | high | Docs Contact: | |
| Priority: | medium | ||
| Version: | 4.0 | CC: | 157070.alewis, berthiaume_wayne, coughlan, davidw, jlaska, kaufman_susan, majianpeng, mchristi, rkenna |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i686 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | RHBA-2007-0810 | Doc Type: | Bug Fix |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2007-11-15 16:12:33 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
|
Description
Nitin Chandna
2005-06-07 11:24:44 UTC
Is this the entire log? From the log it looks like a session was never established. Somehow though a sysfs file is refereneced. At first I thought you may have started the driver and iscsid and there was some refcoutning mishap. Did you just insmod the module, run iscsid, mount a fs, run IO then restart the target? Then kablewey. I am not completely sure what "shutdown/no shutdown" is. 1. About the entire log, my terminal had probably less buffer, where I would have lost some logs before this log. 2. I did not install the iniator as module. I did a make/make install to install this driver (rather than insmod) 3. MDS iSCSI Target System interfaces can run FCIP or iSCSI protocols. In this case, I had done to enable iscsi interface with "no shut" and disabled iscsi on the same interface as "shut". 4. This issue - I am unable to reproduce currently, though I will give it a try again. Moving to kernel componant. The iscsi componant is not used anymore. Please select kernel for iscsi_sfnet problems and iscsi-initiator-utils for userspace tool problems. Thanks. I think I moved this to iscsi-initiator-utils. I think vikasx.aggarwal has found the source of the problem. *** Bug 177645 has been marked as a duplicate of this bug. *** Mike do you have a reproducible case or do you not care about fixing this bug? Another tester just hit this here. (In reply to comment #11) > Mike do you have a reproducible case or do you not care about fixing this > bug? Nice way of putting it :) I have this on my U4 TODO list. I will try to send out a list of all BZs I am working on for U4 to give everyone an idea of what is going on. The problem is that I have only reproduced it once. Have you guys found a way to consistantly hit it? *** Bug 201215 has been marked as a duplicate of this bug. *** Did not make 4.5. Moving to 4.6. This bugzilla had previously been approved for engineering consideration but Red Hat Product Management is currently reevaluating this issue for inclusion in RHEL4.6. This request was previously evaluated by Red Hat Product Management for inclusion in the current Red Hat Enterprise Linux release, but Red Hat was unable to resolve it in time. This request will be reviewed for a future Red Hat Enterprise Linux release. Moving to Assigned . . . FAILS_QA is only used by the RHN team these days. An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2007-0810.html I used iscsi-initiator-utils-4.0.3.0-8.rpm(from centos 4.8),occur this bug.
The centos 4.0.3.0-8.rpm'md5sum did not equal redhat 4.0.3.0-8.But the src.rpm equal,so I think this rpm eqal.
I can find a way to hit it.
env:
iscsi-initiator-utils-4.0.3.0-8.rpm
centos 4.v;
kernel:2.6.9-78.EL
this is my testfile:
int main()
{
int fd = open("/sys/class/iscsi_transport/target2:0:0/first_burst_len");
if(fd < 0){
printf("open error:%s\n",strerror(errno));
return -errno;
}
char buff[4096] = {0};
sleep(40);
if(read(fd, buff,4096) == -1){
printf("read error:%s\n",strerror(errno));
}
printf("first_burst_len=%s\n",buff);
close(fd);
return 0;
}
target2.0.0:is iscsi target.
when testfile in sleep status; service iscsi stop.
This way must hit the bug.
I also read the dirver source the cause:
when I open the sysfs,I remove iscsi drive:
static void
__exit iscsi_cleanup(void)
{
unregister_reboot_notifier(&iscsi_reboot_notifier);
iscsi_unregister_interface();
iscsi_destroy_all_hosts();
iscsi_release_transport(iscsi_transportt);
kmem_cache_destroy(iscsi_task_cache);
}
iscsi_release_transport(iscsi_transportt):may clean the iscsi_transport
but stuct scsi_host live,because the open operation.
I did not know how to modify?
I used iscsi-initiator-utils-4.0.3.0-8.rpm(from centos 4.8),occur this bug.
The centos 4.0.3.0-8.rpm'md5sum did not equal redhat 4.0.3.0-8.But the src.rpm equal,so I think this rpm eqal.
I can find a way to hit it.
env:
iscsi-initiator-utils-4.0.3.0-8.rpm
centos 4.v;
kernel:2.6.9-78.EL
this is my testfile:
int main()
{
int fd = open("/sys/class/iscsi_transport/target2:0:0/first_burst_len");
if(fd < 0){
printf("open error:%s\n",strerror(errno));
return -errno;
}
char buff[4096] = {0};
sleep(40);
if(read(fd, buff,4096) == -1){
printf("read error:%s\n",strerror(errno));
}
printf("first_burst_len=%s\n",buff);
close(fd);
return 0;
}
target2.0.0:is iscsi target.
when testfile in sleep status; service iscsi stop.
This way must hit the bug.
I also read the dirver source the cause:
when I open the sysfs,I remove iscsi drive:
static void
__exit iscsi_cleanup(void)
{
unregister_reboot_notifier(&iscsi_reboot_notifier);
iscsi_unregister_interface();
iscsi_destroy_all_hosts();
iscsi_release_transport(iscsi_transportt);
kmem_cache_destroy(iscsi_task_cache);
}
iscsi_release_transport(iscsi_transportt):may clean the iscsi_transport
but stuct scsi_host live,because the open operation.
I did not know how to modify?
majianpeng, Thanks for the info. Sounds like a refcount bug. The open() is probably not getting a ref to all the necessary structs that the read() ends up accessing. Will look into that code. |