xfs: Fix m_agirotor reset during AG selection
authorCarlos Maiolino <cmaiolino@redhat.com>
Thu, 20 Sep 2012 13:32:36 +0000 (10:32 -0300)
committerBen Myers <bpm@sgi.com>
Wed, 26 Sep 2012 20:42:42 +0000 (15:42 -0500)
commit8aea3ff411b2ce8fe7b46644298ed243a920eb24
tree19a8f40bf1f8b8f48d1d77d29fb87020d8c17f81
parentc3a58fecdd1934a8538ada9073107625f5151687
xfs: Fix m_agirotor reset during AG selection

xfs_ialloc_next_ag() currently resets m_agirotor when it is equal to
m_maxagi:

         if (++mp->m_agirotor == mp->m_maxagi)
         mp->m_agirotor = 0;

But, if for some reason mp->m_maxagi changes to a lower value than
current m_agirotor, this condition will never be true, causing
m_agirotor to exceed the maximum allowed value (m_maxagi).

This implies mainly during lookups for xfs_perag structs in its radix
tree, since the agno value used for the lookup is based on m_agirotor.
An out-of-range m_agirotor may cause a lookup failure which in case will
return NULL.

As an example, the value of m_maxagi is decreased during
inode64->inode32 remount process, case where I've found this problem.

Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
fs/xfs/xfs_ialloc.c
This page took 0.026103 seconds and 5 git commands to generate.