Re: [PATCH dhowells/mount-context] fs: don't call fs_context->free() from fsmount()

[Andrei Vagin <avagin@virtuozzo.com>]

On Wed, Aug 01, 2018 at 01:42:29AM +0100, David Howells wrote:
> Andrei Vagin <avagin@virtuozzo.com> wrote:
> 
> > > > Does it mean that fc->fs_type->init_fs_context() should not be called
> > > > contexts which are created from fspick()?
> > > 
> > > No.  I've put a flag in the context that is set when ->init_fs_context() is
> > > called and cleared when ->free() is called.  ->free() isn't called in the put
> > > routine if the flag isn't set.
> > 
> > > /* We've done the mount bit - now move the file context into
> > >  * more or less the same state as if we'd done an fspick().
> > 
> > According to this comment, a context after fsmount() should be in
> > the same state as after fspick().
> > 
> > In fsmount(), we call ->free() which is oposite to init_fs_context(). If
> > we want to have "more or less the same state" and want to call
> > fs_context->free() in fsmount(), this means that we should not call
> > fc->fs_type->init_fs_context() in fspick()...  Where am I wrong?
> 
> vfs_fsconfig() calls ->init_fs_context() if fc->phase is
> FS_CONTEXT_AWAITING_RECONF.  This is so that we don't actually fully reinit
> the context unnecessarily if the fd is just going to be closed after fsmount()
> is called.

I understood the idea. Thank you for the explanation.

> 
> fspick() assumes that you're definitely going to reconfigure the superblock
> (presumably that's why you used it) and always reinitialises the context,
> shoving fc->phase round to FS_CONTEXT_RECONF_PARAMS.
> 
> David