[RFC] Transactional CGroup task attachment

Paul Menage menage at google.com
Mon Jul 14 12:16:08 PDT 2008


On Sun, Jul 13, 2008 at 11:28 PM, Daisuke Nishimura
<nishimura at mxp.nes.nec.co.jp> wrote:
>
> - I think page reclaiming code decreases the memory charge
>  without holding mmap_sem(e.g. try_to_unmap(), __remove_mapping()).
>  Shouldn't we handle these cases?

The prepare_attach_nosleep() call could take the res_counter's
spinlock, which would lock out all other charges and uncharges until
the transaction was completed; that might be enough for what you want.
We'd need to export lock/unlock functions from res_counter.

Paul


More information about the Containers mailing list