Index of /~fabio/linux/cfq-ioc-data/

NameLast modifiedSize

Parent Directory  -  
README04-Apr-2008 01:50 1.1K 
bug-on-ioc-data-surviving-cic.patch04-Apr-2008 00:43 1.1K 
bug-on-wrong-cic-returned.patch04-Apr-2008 00:43 512  
cfq-iosched-fix-ioc-data-leak.patch04-Apr-2008 01:34 2.3K 
oops-bug-on-ioc-data-surviving-cic04-Apr-2008 00:39 1.6K 
oops-bug-on-wrong-cic-returned04-Apr-2008 00:39 3.0K 
oops-unmodified-cfq04-Apr-2008 00:39 2.7K 
qemu_config04-Apr-2008 01:43 33K 
stress-reload04-Apr-2008 00:27 417  

CFQ ioc->ioc_data LEAK
======================

This directory contains some stuff related to a cfq problem on scheduler
switching and module unloading.  It contains the following files (in logical
rather than alphabetical order):

    - qemu_config: .config used to reproduce all of this on kvm/qemu.

    - stress_reload: simple shell script used to reproduce the bug.
      It definitely has no meaning in the real world, but it had to
      be only a stress test...

    - oops-unmodified-cfq: oops obtained with mainline cfq.

    - bug-on-wrong-cic-returned.patch: scream out loud when
      bfq_cic_lookup() founds a cic in ioc->ioc_data that does not
      belong to cic.

    - oops-bug-on-wrong-cic-returned: an oops showing a wrong lookup
      result being returned.

    - bug-on-ioc-data-surviving-cic.patch: assert that when a cic
      is freed it is no more cached into ioc->ioc_data.  [NOTE: this
      is NOT on top of the previous one.]

    - oops-bug-on-ioc-data-surviving-cic: an oops showing the above
      assertion failing during elv_unregister().

    - cfq-iosched-fix-ioc-data-leak.patch: a patch that should fix
      the problem.