[ ssic-linux-Bugs-983095 ] Can't mount below /dev on multiple nodes
"SourceForge.net" <[email protected]>
| Newsgroups | gmane.linux.cluster.ssic.devel |
|---|---|
| Message-ID | <[email protected]> |
Bugs item #983095, was opened at 2004-06-30 19:43 Message generated for change (Comment added) made by rogertsang You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=405834&aid=983095&group_id=32541 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: Filesystem Group: None >Status: Closed >Resolution: Fixed Priority: 5 Private: No Submitted By: David Zafman (dzafman) Assigned to: Nobody/Anonymous (nobody) Summary: Can't mount below /dev on multiple nodes Initial Comment: This was seen when attempting to do profiling on multiple nodes. The opcontrol command does the following: mount -t oprofilefs nodev /dev/oprofile >/dev/null Since this adds the entry to /etc/mtab, a subsequent node can't do another mount even though /dev is unique for each node. While we could simply add a "-n" to the opcontrol script, this doesn't solve the general problem. We could change the mount command to replace /dev directory specification to /cluster/nodeX/dev. This would allow every node to mount below /dev, because the mount command would fix things to reflect the SSI kernel notion of what /dev is. I believe that this change would improve our application compatibility. ---------------------------------------------------------------------- >Comment By: Roger Tsang (rogertsang) Date: 2008-01-01 20:00 Message: Logged In: YES user_id=1246761 Originator: NO This bug is resolved in SSI-1.9. In SSI-1.9 /dev/X resolves to /cluster/nodeX/dev. mtab /dev/X entries are unique for each node. ---------------------------------------------------------------------- Comment By: David Zafman (dzafman) Date: 2004-06-30 21:30 Message: Logged In: YES user_id=297844 The fix for this is harder than I first thought. First if we get mount to mount on /cluster/nodeX/dev instead of /dev, the mount point won't be seen below /dev and vice versa. This is due to the bind mount nature of /dev. Also, since /dev is controlled by the node context, the mount system call currently only works for local node, even though we might want to function ship to the node context node. It is critical that the translation /dev -> /cluster/nodeX/dev at user level matches the results of the mount() system call in terms of which /dev has the mount actually on. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=405834&aid=983095&group_id=32541 ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/