Bug 192800
Summary: | Bizarre autofs mount points with script generated maps | ||||||
---|---|---|---|---|---|---|---|
Product: | [Fedora] Fedora | Reporter: | Daniel Qarras <dqarras> | ||||
Component: | autofs | Assignee: | Ian Kent <ikent> | ||||
Status: | CLOSED NOTABUG | QA Contact: | Brock Organ <borgan> | ||||
Severity: | medium | Docs Contact: | |||||
Priority: | medium | ||||||
Version: | 5 | CC: | ikent, jmoyer | ||||
Target Milestone: | --- | ||||||
Target Release: | --- | ||||||
Hardware: | All | ||||||
OS: | Linux | ||||||
Whiteboard: | |||||||
Fixed In Version: | Doc Type: | Bug Fix | |||||
Doc Text: | Story Points: | --- | |||||
Clone Of: | Environment: | ||||||
Last Closed: | 2006-06-28 11:53:58 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: | |||||||
Attachments: |
|
Description
Daniel Qarras
2006-05-23 07:30:36 UTC
Created attachment 129844 [details]
autofs debug output
(In reply to comment #0) > Description of problem: > With static autofs map I get directories /var/autofs/home and /var/autofs/net > all ok but with a script outputting a corresponding map I get bizarre mounts > like /var/autofs/(null) and /var/autofs/whatever that have contents of /home or > /net or recursively directories /home and /net. /var/autofs/$SERVER_IP has > correctly /home and /net directories but also (null) directory. As I said previously I'm not able to track down why we are seeing requests for the path element "(null)". At this stage I think they are actually coming from the VFS so as far as the kernel module is concerned they are valid. However, they seem always to follow a mount fail whch introduces another problem which I have an experimental patch for. Are you able to test the experimental kernel patch? Ian Yes. Feel free to provide a patch for testing. Please also inform what should change with the patch compared to current situation? (In reply to comment #3) > Yes. Feel free to provide a patch for testing. Please also inform what should > change with the patch compared to current situation? Hang on, what kernel version will you use? I'll do a patch for it. What I would expect to see are immediate failures for mounts that fail and consequently no "(null)" directories for a start. I'm using kernel-2.6.16-1.2122_FC5. Thanks. Daniel, This issue has been traced to libselinux and is fixed in libselinux-1.30.3-3.fc5 which is currently in the FC5 updates testing area. Could you please check that this does indeed fix the problem you are seeing. Ian Hi Ian, I now tested this with libselinux-1.30.3-3.fc5 and kernel-2.6.16-1.2133_FC5. The good news is that the strange (null) directory is totally gone, no signs of it at server logs or at the client. However, other bizarre mount points still exist, ie: /etc/init.d/autofs restart cd /var/autofs/hello cd hallo ls ls net /etc/init.d/autofs stop still gets me to directory /var/autofs/hello/hallo and there are the directories home and net recursively (ie, under net are home and net and under them are home and net, etc). Thanks! (In reply to comment #7) Hi Daniel, > However, other bizarre mount points still exist, ie: Ha, never ending story. > > /etc/init.d/autofs restart > cd /var/autofs/hello > cd hallo > ls > ls net > /etc/init.d/autofs stop > > still gets me to directory /var/autofs/hello/hallo and there are the directories > home and net recursively (ie, under net are home and net and under them are home > and net, etc). OK. I should be able to reproduce this with the info. you've provided. I'll let you know how I go. Ian (In reply to comment #8) > (In reply to comment #7) > Hi Daniel, Finally got to this. > > > However, other bizarre mount points still exist, ie: snip ... > > OK. I should be able to reproduce this with the info. you've > provided. > And I have been able to duplicate the problem. Ian (In reply to comment #9) > > > > > OK. I should be able to reproduce this with the info. you've > > provided. > > > > And I have been able to duplicate the problem. I'm fairly sure that if you remove the "--ghost" option this will work the way you expect it to. Removing the "--ghost" option does not help. I can still do this: /etc/init.d/autofs restart cd /var/autofs/hello cd hallo ls ls net/net /etc/init.d/autofs stop In /var/log/messages I see errors like: Jun 26 11:46:43 ws6 automount[2814]: attempting to mount entry /var/autofs/hello Jun 26 11:46:43 ws6 automount[2821]: mount(nfs): host hello: lookup failure Jun 26 11:46:43 ws6 automount[2821]: >> mount: can't get address for hello Jun 26 11:46:43 ws6 automount[2821]: mount(nfs): nfs: mount failure hello:/net on /var/autofs/hello/net Jun 26 11:46:44 ws6 automount[2821]: mount(nfs): host hello: lookup failure Jun 26 11:46:44 ws6 automount[2821]: >> mount: can't get address for hello Jun 26 11:46:44 ws6 automount[2821]: mount(nfs): nfs: mount failure hello:/home on /var/autofs/hello/home Jun 26 11:46:50 ws6 automount[2814]: attempting to mount entry /var/autofs/hello/hallo Jun 26 11:46:50 ws6 automount[2826]: mount(nfs): host hello/hallo: lookup failure Jun 26 11:46:50 ws6 automount[2826]: >> mount: can't get address for hello/hallo Jun 26 11:46:50 ws6 automount[2826]: mount(nfs): nfs: mount failure hello/hallo:/net on /var/autofs/hello/hallo/net Jun 26 11:46:50 ws6 automount[2826]: mount(nfs): host hello/hallo: lookup failure Jun 26 11:46:50 ws6 automount[2826]: >> mount: can't get address for hello/hallo Jun 26 11:46:50 ws6 automount[2826]: mount(nfs): nfs: mount failure hello/hallo:/home on /var/autofs/hello/hallo/home Jun 26 11:47:08 ws6 automount[2814]: attempting to mount entry /var/autofs/hello/hallo/net Jun 26 11:47:08 ws6 automount[2833]: mount(nfs): host hello/hallo/net: lookup failure Jun 26 11:47:08 ws6 automount[2833]: >> mount: can't get address for hello/hallo/net Jun 26 11:47:08 ws6 automount[2833]: mount(nfs): nfs: mount failure hello/hallo/net:/net on /var/autofs/hello/hallo/net/net Jun 26 11:47:08 ws6 automount[2833]: mount(nfs): host hello/hallo/net: lookup failure Jun 26 11:47:08 ws6 automount[2833]: >> mount: can't get address for hello/hallo/net Jun 26 11:47:08 ws6 automount[2833]: mount(nfs): nfs: mount failure hello/hallo/net:/home on /var/autofs/hello/hallo/net/home Jun 26 11:47:08 ws6 automount[2814]: attempting to mount entry /var/autofs/hello/hallo/net/net Jun 26 11:47:09 ws6 automount[2838]: mount(nfs): host hello/hallo/net/net: lookup failure Jun 26 11:47:09 ws6 automount[2838]: >> mount: can't get address for hello/hallo/net/net Jun 26 11:47:09 ws6 automount[2838]: mount(nfs): nfs: mount failure hello/hallo/net/net:/net on /var/autofs/hello/hallo/net/net/net Jun 26 11:47:09 ws6 automount[2838]: mount(nfs): host hello/hallo/net/net: lookup failure Jun 26 11:47:09 ws6 automount[2838]: >> mount: can't get address for hello/hallo/net/net Jun 26 11:47:09 ws6 automount[2838]: mount(nfs): nfs: mount failure hello/hallo/net/net:/home on /var/autofs/hello/hallo/net/net/home (In reply to comment #11) > Removing the "--ghost" option does not help. I can still do this: The only way to work around this is to add the "-strict" option to the mount in the master map. This has come about because there were compaints about requiring all mounts in a multi-mount to succeed in order for any of the mounts to be left mounted so the default option was changed from "strict" to "nonstrict". Try this option and let me know what happens. Ian Hi! Adding the "-strict" option to PARAMS in the above mentioned auto.local indeed fixes the problem! "-strict" also plays nicely with "--ghost" in auto.master. I guess this is fully solved now. Thanks a lot! (In reply to comment #13) > Hi! Adding the "-strict" option to PARAMS in the above mentioned auto.local > indeed fixes the problem! "-strict" also plays nicely with "--ghost" in > auto.master. I guess this is fully solved now. Thanks a lot! Great, Thanks for letting me know. |