[Linux-cluster] GFS2/DLM deadlock

[Steven Whitehouse]

Hi,

On Sun, 2012-09-09 at 17:31 -0400, Jason Henderson wrote:
> 
> On Sep 8, 2012 9:44 AM, "Bob Peterson" <rpeterso at redhat.com> wrote:
> >
> > ----- Original Message -----
> > | A question on the inode numbers in the hangalyzer output.
> > |
> > | In the glock dump for node2 you have these lines:
> > | G:  s:SH n:2/81523 f:dq t:SH d:UN/0 l:0 a:0 r:4 m:100
> > |     I: n:126/529699 t:4 f:0x10 d:0x00000001 s:3864/3864
> > |
> > | >From docs I've read I understand that the glock field 'n:2/81523'
> > | tells me that 81523 is the inode number in hex (if then type is 2
> or
> > | 5).
> > | What are the fields in the inode line following the glock mean (at
> > | least the n: field)?
> >
> > The numbers after the n: are the glock identifier. It consists of
> > the glock type (2 for inode, 3 for rgrp, 5 for i_open, and a bunch
> > of special ones) followed by "/" followed by the inode number
> > (disk inode's block address) in hex.
> >
> > After f: are the glock flags. For example, "q" means the glock is
> > queued. There are a bunch of flags with a bunch of meanings.
> >
> > t: is the target glock state; the lock state it's trying to achieve.
> > SH is for a shared lock, EX is exclusive, UN is unlocked, etc.
> > d: is the demote glock state; the lock state it needs to transition
> > to when the lock is demoted. In this case, demote to UNlocked.
> > The number after the slash is the demote time.
> > a: is active items count, or the number of "live" buffers to be
> written.
> > r: is the revoke count, or the number of journal items needing to be
> >    revoked due to delete, etc.
> > m: is the minimum hold time for the glock, in milliseconds.
> >
> > On the next line, I: indicates this glock is for an inode.
> > n:126 is a formal inode number (can be ignored). The number after
> the
> > slash, 529699, is the inode disk address in decimal.
> > t: is the mode, f: are the inode flags, d: are the disk flags, and
> > s: is the inode's size in decimal. Before the slash is the size
> stored
> > in one of our internal structures. After the slash is the size
> > according to the vfs inode. In almost all cases they should be the
> same.
> >
> > Note that the format of these fields, the flags, and everything
> > differs from release to release. For example, newer versions of GFS2
> > don't have two different numbers for inode size.
> >
> 
> I'm not clear on the two different inode numbers in the two lines
> above: Which n: number do I use to locate the file the lock is for?
> The one in the glock line or the one in the I: line? From RedHat docs
> I have read, I should convert the 81523 (from the '2/ 81523') to
> decimal, which is  533795, and then use 'find -inum 533795' to locate
> the file after the filesystem has been unfrozen. 
> 
> 
> I guess my confusion is the definition of a "disk inode's block
> address" versus an "inode disk address". Could you clarify the
> difference for me?
> 
In GFS2 the inode number is the same as the disk block address of the
inode since we have exactly one inode per block, so these things are
both the same,

Steve.