On Wed, 2023-01-18 at 13:16 -0500, Olga Kornievskaia wrote:
> On Wed, Jan 18, 2023 at 12:07 PM Olga Kornievskaia <aglo@xxxxxxxxx> wrote:
> >
> > On Wed, Jan 18, 2023 at 11:57 AM Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> > >
> > > On Wed, 2023-01-18 at 11:29 -0500, Olga Kornievskaia wrote:
> > > > On Wed, Jan 18, 2023 at 10:27 AM Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> > > > >
> > > > > On Wed, 2023-01-18 at 09:42 -0500, Olga Kornievskaia wrote:
> > > > > > On Tue, Jan 17, 2023 at 2:38 PM Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> > > > > > >
> > > > > > > There are two different flavors of the nfsd4_copy struct. One is
> > > > > > > embedded in the compound and is used directly in synchronous copies. The
> > > > > > > other is dynamically allocated, refcounted and tracked in the client
> > > > > > > struture. For the embedded one, the cleanup just involves releasing any
> > > > > > > nfsd_files held on its behalf. For the async one, the cleanup is a bit
> > > > > > > more involved, and we need to dequeue it from lists, unhash it, etc.
> > > > > > >
> > > > > > > There is at least one potential refcount leak in this code now. If the
> > > > > > > kthread_create call fails, then both the src and dst nfsd_files in the
> > > > > > > original nfsd4_copy object are leaked.
> > > > > >
> > > > > > I don't believe that's true. If kthread_create thread fails we call
> > > > > > cleanup_async_copy() that does a put on the file descriptors.
> > > > > >
> > > > >
> > > > > You mean this?
> > > > >
> > > > > out_err:
> > > > > if (async_copy)
> > > > > cleanup_async_copy(async_copy);
> > > > >
> > > > > That puts the references that were taken in dup_copy_fields, but the
> > > > > original (embedded) nfsd4_copy also holds references and those are not
> > > > > being put in this codepath.
> > > >
> > > > Can you please point out where do we take a reference on the original copy?
> > > >
> > >
> > > In the case of an inter-server copy, nf_dst is set in
> > > nfsd4_setup_inter_ssc. For intraserver copy, both pointers are set via
> > > the call to nfsd4_verify_copy. Both functions call
> > > nfs4_preprocess_stateid_op, which returns a reference to the nfsd_file
> > > in the second to last arg.
> >
> > Ah. Thank you. I didn't know that nfs4_preprocess_stateid_op() takes a
> > reference on it's 5th argument. I think I was previously looking at
> > nfsd4_read() function which calls nfs4_preprocess_stateid_op() and
> > gets back read->rd_nf but it never does a put on it when it returns.
> > However, I now looked at other functions that call
> > nfs4_preproess_stateid_op() such as nfsd4_fallocate() and I see that
> > it does a put.
>
> So is there a refcount leak in the nfsd4_read() then since it doesn't
> do a put? Or the internals obscure and that even though it calls the
> same function and passes that parameter no refcount was increased. Is
> it based on the "WR_STATE, RD_STATE" parameter.
>
I don't think so. The put is done just below there, in
nfsd4_read_release:
if (u->read.rd_nf)
nfsd_file_put(u->read.rd_nf);
That said, I am hunting a refcount overput with nfsd_files that I've not
been able to nail down yet (which is why I've been auditing the
nfsd_file refcounting). If you see anything that looks hinky, please do
point it out.
> I see that
> nfsd4_write() does do a put. For copy, we call the src_fd with
> RD_STATE and dst_fd with WR_STATE. If I were to follow the logic of
> nfsd4_read/nfsd4_write, the the copy doesn't need to do a put for src
> but will need it for the dst. The proposed patch does it for both.
>
That'd be wrong. READs have to hold a ref to the open file while the
reply is being marshalled. A WRITE can release it once the data has been
written to the file. Maybe that's worth a comment.
Note that I just sent a patch to the list to add a comment that
(hopefully) makes it clear that nfs4_preprocess_stateid_op returns a
reference in that field.
FWIW, it wouldn't be safe for it to do anything else. Returning a
pointer to a refcounted object without taking a reference would be very
problematic.
> So I'm still confused if this patch is the correct solution.
>
Fair enough. I'm not sure I understand the pushback, as the result seems
clearer to me. If you want to propose an alternate fix, I'm happy to
take a look.
> > >
> > > > > > > The cleanup in this codepath is also sort of weird. In the async copy
> > > > > > > case, we'll have up to four nfsd_file references (src and dst for both
> > > > > > > flavors of copy structure).
> > > > > >
> > > > > > That's not true. There is a careful distinction between intra -- which
> > > > > > had 2 valid file pointers and does a get on both as they both point to
> > > > > > something that's opened on this server--- but inter -- only does a get
> > > > > > on the dst file descriptor, the src doesn't exit. And yes I realize
> > > > > > the code checks for nfs_src being null which it should be but it makes
> > > > > > the code less clear and at some point somebody might want to decide to
> > > > > > really do a put on it.
> > > > > >
> > > > >
> > > > > This is part of the problem here. We have a nfsd4_copy structure, and
> > > > > depending on what has been done to it, you need to call different
> > > > > methods to clean it up. That seems like a real antipattern to me.
> > > >
> > > > But they call different methods because different things need to be
> > > > done there and it makes it clear what needs to be for what type of
> > > > copy.
> > > >
> > >
> > >
> > > I sure as hell had a hard time dissecting how all of that was supposed
> > > to work. There is clear bug here, and I think this patch makes the
> > > result clearer and more robust in the face of changes.
> > >
> > > There are actually 4 different cases here: sync vs. async, alongside
> > > intra vs. interserver copy. These are all overloaded onto a nfsd4_copy
> > > structure, seemingly for no good reason.
> > >
> > > The cleanup, in particular seems quite fragile to me, and there is a
> > > dearth of defensive coding measures. If you subtly call the "wrong"
> > > cleanup function at the wrong point in time, then things may go awry.
> > >
> > > I'll leave it up to Chuck to make the final determination, but I see
> > > this patch as an improvement.
> > >
> > > > > > > They are both put at the end of
> > > > > > > nfsd4_do_async_copy, even though the ones held on behalf of the embedded
> > > > > > > one outlive that structure.
> > > > > > >
> > > > > > > Change it so that we always clean up the nfsd_file refs held by the
> > > > > > > embedded copy structure before nfsd4_copy returns. Rework
> > > > > > > cleanup_async_copy to handle both inter and intra copies. Eliminate
> > > > > > > nfsd4_cleanup_intra_ssc since it now becomes a no-op.
> > > > > >
> > > > > > I feel by combining the cleanup for both it obscures a very important
> > > > > > destication that src filehandle doesn't exist for inter.
> > > > >
> > > > > If the src filehandle doesn't exist, then the pointer to it will be
> > > > > NULL. I don't see what we gain by keeping these two distinct, other than
> > > > > avoiding a NULL pointer check.
> > > >
> > > > My reason would be for code clarity because different things are
> > > > supposed to happen for intra and inter. Difference of opinion it
> > > > seems.
> > > >
> > > > >
> > > > > >
> > > > > > > Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > > > > > > ---
> > > > > > > fs/nfsd/nfs4proc.c | 23 ++++++++++-------------
> > > > > > > 1 file changed, 10 insertions(+), 13 deletions(-)
> > > > > > >
> > > > > > > diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> > > > > > > index 37a9cc8ae7ae..62b9d6c1b18b 100644
> > > > > > > --- a/fs/nfsd/nfs4proc.c
> > > > > > > +++ b/fs/nfsd/nfs4proc.c
> > > > > > > @@ -1512,7 +1512,6 @@ nfsd4_cleanup_inter_ssc(struct nfsd4_ssc_umount_item *nsui, struct file *filp,
> > > > > > > long timeout = msecs_to_jiffies(nfsd4_ssc_umount_timeout);
> > > > > > >
> > > > > > > nfs42_ssc_close(filp);
> > > > > > > - nfsd_file_put(dst);
> > > > > > > fput(filp);
> > > > > > >
> > > > > > > spin_lock(&nn->nfsd_ssc_lock);
> > > > > > > @@ -1562,13 +1561,6 @@ nfsd4_setup_intra_ssc(struct svc_rqst *rqstp,
> > > > > > > ©->nf_dst);
> > > > > > > }
> > > > > > >
> > > > > > > -static void
> > > > > > > -nfsd4_cleanup_intra_ssc(struct nfsd_file *src, struct nfsd_file *dst)
> > > > > > > -{
> > > > > > > - nfsd_file_put(src);
> > > > > > > - nfsd_file_put(dst);
> > > > > > > -}
> > > > > > > -
> > > > > > > static void nfsd4_cb_offload_release(struct nfsd4_callback *cb)
> > > > > > > {
> > > > > > > struct nfsd4_cb_offload *cbo =
> > > > > > > @@ -1683,12 +1675,18 @@ static void dup_copy_fields(struct nfsd4_copy *src, struct nfsd4_copy *dst)
> > > > > > > dst->ss_nsui = src->ss_nsui;
> > > > > > > }
> > > > > > >
> > > > > > > +static void release_copy_files(struct nfsd4_copy *copy)
> > > > > > > +{
> > > > > > > + if (copy->nf_src)
> > > > > > > + nfsd_file_put(copy->nf_src);
> > > > > > > + if (copy->nf_dst)
> > > > > > > + nfsd_file_put(copy->nf_dst);
> > > > > > > +}
> > > > > > > +
> > > > > > > static void cleanup_async_copy(struct nfsd4_copy *copy)
> > > > > > > {
> > > > > > > nfs4_free_copy_state(copy);
> > > > > > > - nfsd_file_put(copy->nf_dst);
> > > > > > > - if (!nfsd4_ssc_is_inter(copy))
> > > > > > > - nfsd_file_put(copy->nf_src);
> > > > > > > + release_copy_files(copy);
> > > > > > > spin_lock(©->cp_clp->async_lock);
> > > > > > > list_del(©->copies);
> > > > > > > spin_unlock(©->cp_clp->async_lock);
> > > > > > > @@ -1748,7 +1746,6 @@ static int nfsd4_do_async_copy(void *data)
> > > > > > > } else {
> > > > > > > nfserr = nfsd4_do_copy(copy, copy->nf_src->nf_file,
> > > > > > > copy->nf_dst->nf_file, false);
> > > > > > > - nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
> > > > > > > }
> > > > > > >
> > > > > > > do_callback:
> > > > > > > @@ -1811,9 +1808,9 @@ nfsd4_copy(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
> > > > > > > } else {
> > > > > > > status = nfsd4_do_copy(copy, copy->nf_src->nf_file,
> > > > > > > copy->nf_dst->nf_file, true);
> > > > > > > - nfsd4_cleanup_intra_ssc(copy->nf_src, copy->nf_dst);
> > > > > > > }
> > > > > > > out:
> > > > > > > + release_copy_files(copy);
> > > > > > > return status;
> > > > > > > out_err:
> > > > > > > if (async_copy)
> > > > > > > --
> > > > > > > 2.39.0
> > > > > > >
> > > > >
> > > > > --
> > > > > Jeff Layton <jlayton@xxxxxxxxxx>
> > >
> > > --
> > > Jeff Layton <jlayton@xxxxxxxxxx>
--
Jeff Layton <jlayton@xxxxxxxxxx>
