Re: [PATCH 105/104] virtiofsd: Unref old/new inodes with the same mutex lock in lo_rename()

[Philippe Mathieu-Daudé]

On 1/19/20 9:35 AM, Xiao Yang wrote:
> On 2020/1/17 21:32, Philippe Mathieu-Daudé wrote:
> > We can unref both old/new inodes with the same mutex lock.
> >
> > Signed-off-by: Philippe Mathieu-Daudé<address@hidden>
> > ---
> > Based-on:<address@hidden>
> > "virtiofs daemon"
> > https://www.mail-archive.com/address@hidden/msg664652.html
> >
> >   tools/virtiofsd/passthrough_ll.c | 6 ++++--
> >   1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/tools/virtiofsd/passthrough_ll.c 
> > b/tools/virtiofsd/passthrough_ll.c
> > index 57f58aef26..5c717cb5a1 100644
> > --- a/tools/virtiofsd/passthrough_ll.c
> > +++ b/tools/virtiofsd/passthrough_ll.c
> > @@ -1461,8 +1461,10 @@ static void lo_rename(fuse_req_t req, 
> > fuse_ino_t parent, const char *name,
> >       }
> >
> >   out:
> > -    unref_inode_lolocked(lo, oldinode, 1);
> > -    unref_inode_lolocked(lo, newinode, 1);
> > +    pthread_mutex_lock(&lo->mutex);
> > +    unref_inode(lo, oldinode, 1);
> > +    unref_inode(lo, newinode, 1);
> > +    pthread_mutex_unlock(&lo->mutex);
> Hi,
>
> It seems to avoid calling pthread_mutex_lock and pthread_mutex_unlock 
> twice.
> Does the change fix some issues or improve the performance?

No issue, simply intend to improve the performance.

> Best Regards,
> Xiao Yang
> >       lo_inode_put(lo,&oldinode);
> >       lo_inode_put(lo,&newinode);
> >       lo_inode_put(lo,&parent_inode);