Re: [PATCH] track number of mnts writing to superblocks

2008-01-09 Thread Andrew Morton
On Wed, 02 Jan 2008 13:51:10 -0800 Dave Hansen <[EMAIL PROTECTED]> wrote: > > One of the benefits of the r/o bind mount patches is that they > make it explicit when a write to a superblock might occur. > We currently search sb->s_files when remounting rw->ro to look > for writable files. But, th

Re: [PATCH] track number of mnts writing to superblocks

2008-01-03 Thread Dave Hansen
On Wed, 2008-01-02 at 22:02 -0600, Serge E. Hallyn wrote: > Ok I'm blabbing quite a bit here while trying to figure out > the patch, and maybe there are some useful hints for where more > comments would be useful. But other than the fact that > mark_mnt_has_writer() needs to the atomic_inc() even

Re: [PATCH] track number of mnts writing to superblocks

2008-01-02 Thread Serge E. Hallyn
Quoting Dave Hansen ([EMAIL PROTECTED]): > > One of the benefits of the r/o bind mount patches is that they > make it explicit when a write to a superblock might occur. > We currently search sb->s_files when remounting rw->ro to look > for writable files. But, that search is not comprehensive, an

[PATCH] track number of mnts writing to superblocks

2008-01-02 Thread Dave Hansen
One of the benefits of the r/o bind mount patches is that they make it explicit when a write to a superblock might occur. We currently search sb->s_files when remounting rw->ro to look for writable files. But, that search is not comprehensive, and it is racy. This replaces that search. The idea