Re: [PATCH] fuse: add a dev ioctl for recovery

[Hao Peng <flyingpenghao@gmail.com>]

On Tue, Sep 7, 2021 at 5:34 PM Miklos Szeredi <miklos@szeredi.hu> wrote:
>
> On Mon, 6 Sept 2021 at 14:36, Hao Peng <flyingpenghao@gmail.com> wrote:
> >
> > For a simple read-only file system, as long as the connection
> > is not broken, the recovery of the user-mode read-only file
> > system can be realized by putting the request of the processing
> > list back into the pending list.
>
> Thanks for the patch.
>
> Do you have example userspace code for this?
>
Under development. When the fuse user-mode file system process is abnormal,
the process does not terminate (/dev/fuse will not be closed), enter
the reset procedure,
and will not open /dev/fuse again during the reinitialization.
Of course, this can only solve part of the abnormal problem.

> > Signed-off-by: Peng Hao <flyingpeng@tencent.com>
> > ---
> >  fs/fuse/dev.c             | 38 +++++++++++++++++++++++++++++++++++++-
> >  include/uapi/linux/fuse.h |  1 +
> >  2 files changed, 38 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/fuse/dev.c b/fs/fuse/dev.c
> > index 491c092d427b..8e1b69a5b503 100644
> > --- a/fs/fuse/dev.c
> > +++ b/fs/fuse/dev.c
> > @@ -2239,7 +2239,7 @@ static int fuse_device_clone(struct fuse_conn
> > *fc, struct file *new)
> >  static long fuse_dev_ioctl(struct file *file, unsigned int cmd,
> >                            unsigned long arg)
> >  {
> > -       int res;
> > +       int res = 0;
> >         int oldfd;
> >         struct fuse_dev *fud = NULL;
> >
> > @@ -2268,6 +2268,42 @@ static long fuse_dev_ioctl(struct file *file,
> > unsigned int cmd,
> >                         }
> >                 }
> >                 break;
> > +
> > +       case FUSE_DEV_IOC_RECOVERY:
> > +       {
> > +               struct fuse_iqueue *fiq;
> > +               struct fuse_pqueue *fpq;
> > +               struct fuse_req *req, *next;
> > +               LIST_HEAD(recovery);
> > +               unsigned int i;
> > +
> > +               fud = fuse_get_dev(file);
> > +               fiq = &fud->fc->iq;
> > +               fpq = &fud->pq;
> > +
> > +               spin_lock(&fpq->lock);
> > +               list_for_each_entry_safe(req, next, &fpq->io, list) {
> > +                       spin_lock(&req->waitq.lock);
> > +                       clear_bit(FR_LOCKED, &req->flags);
> > +                       list_move(&req->list, &recovery);
> > +                       spin_unlock(&req->waitq.lock);
> > +               }
>
> I don't get it.  Recovery means the previous server process is dead.
> If there are requests on the fpq->io queue, than that's obviously not
> true and having two processes messing with the same requests isn't
> going to do any good.
>
> So I suggest just erroring out if fpq->io is not empty.
I will modify it in the next version.
>
>
> > +               for (i = 0; i < FUSE_PQ_HASH_SIZE; i++)
> > +                       list_splice_tail_init(&fpq->processing[i],
> > +                                             &recovery);
> > +               list_for_each_entry_safe(req, next, &recovery, list) {
> > +                       clear_bit(FR_SENT, &req->flags);
> > +               }
> > +               spin_unlock(&fpq->lock);
> > +
> > +               spin_lock(&fiq->lock);
> > +               list_for_each_entry_safe(req, next, &recovery, list) {
> > +                       set_bit(FR_PENDING, &req->flags);
> > +               }
> > +               list_splice(&recovery, &fiq->pending);
> > +               spin_unlock(&fiq->lock);
> > +               break;
> > +       }
> >         default:
> >                 res = -ENOTTY;
> >                 break;
> > diff --git a/include/uapi/linux/fuse.h b/include/uapi/linux/fuse.h
> > index 36ed092227fa..fc07324efa9d 100644
> > --- a/include/uapi/linux/fuse.h
> > +++ b/include/uapi/linux/fuse.h
> > @@ -923,6 +923,7 @@ struct fuse_notify_retrieve_in {
> >  /* Device ioctls: */
> >  #define FUSE_DEV_IOC_MAGIC             229
> >  #define FUSE_DEV_IOC_CLONE             _IOR(FUSE_DEV_IOC_MAGIC, 0, uint32_t)
> > +#define FUSE_DEV_IOC_RECOVERY          _IOR(FUSE_DEV_IOC_MAGIC+1, 0, uint32_t)
> >
> >  struct fuse_lseek_in {
> >         uint64_t        fh;
> > --
> > 2.27.0