On Fri, Sep 08, 2017 at 03:52:53PM -0400, Olga Kornievskaia wrote:
> What if instead the client won't send the 2nd FH in case of the inter
> server to server COPY and it would avoid all this stale handle
> business. So for the "intra" COPY it'll still be PUTFH, SAVEFH, PUTFH,
> COPY but for the "inter" COPY it'll be PUTFH, COPY.
That might be easier to implement, but it'd require a change to the
protocol, right? I may not understand what you're proposing.
--b.
>
> On Fri, Sep 8, 2017 at 3:38 PM, J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote:
> > Some questions below, but really I think this approach can't work. Keep
> > in mind this has to work not just for the one compound the current
> > client uses, but for *any* compound containing a COPY, including those
> > sent by future clients that may do something more complicated, or by
> > malcious clients that may intentionally send weird compounds to make the
> > server crash.
> >
> > I think we want to flag the filehandle itself somehow, not the cstate.
> >
> > And check very carefully to make sure that ops other than COPY can't get
> > tripped up by one of these foreign filehandles.
> >
> > On Tue, Jul 11, 2017 at 12:44:01PM -0400, Olga Kornievskaia wrote:
> >> The inter server to server COPY source server filehandle
> >> is guaranteed to be stale as the COPY is sent to the destination
> >> server.
> >>
> >> Signed-off-by: Andy Adamson <andros@xxxxxxxxxx>
> >> ---
> >> fs/nfsd/nfs4proc.c | 47 ++++++++++++++++++++++++++++++++++++++++++++++-
> >> fs/nfsd/nfs4xdr.c | 26 +++++++++++++++++++++++++-
> >> fs/nfsd/nfsd.h | 2 ++
> >> fs/nfsd/xdr4.h | 4 ++++
> >> 4 files changed, 77 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> >> index b49ff31..ceee852 100644
> >> --- a/fs/nfsd/nfs4proc.c
> >> +++ b/fs/nfsd/nfs4proc.c
> >> @@ -496,11 +496,19 @@ static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_stat
> >> nfsd4_putfh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
> >> struct nfsd4_putfh *putfh)
> >> {
> >> + __be32 ret;
> >> +
> >> fh_put(&cstate->current_fh);
> >> cstate->current_fh.fh_handle.fh_size = putfh->pf_fhlen;
> >> memcpy(&cstate->current_fh.fh_handle.fh_base, putfh->pf_fhval,
> >> putfh->pf_fhlen);
> >> - return fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
> >> + ret = fh_verify(rqstp, &cstate->current_fh, 0, NFSD_MAY_BYPASS_GSS);
> >> + if (ret == nfserr_stale && HAS_CSTATE_FLAG(cstate, NO_VERIFY_FH)) {
> >> + CLEAR_CSTATE_FLAG(cstate, NO_VERIFY_FH);
> >> + SET_CSTATE_FLAG(cstate, IS_STALE_FH);
> >> + ret = 0;
> >> + }
> >> + return ret;
> >> }
> >>
> >> static __be32
> >> @@ -533,6 +541,16 @@ static __be32 nfsd4_open_omfg(struct svc_rqst *rqstp, struct nfsd4_compound_stat
> >> nfsd4_savefh(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
> >> void *arg)
> >> {
> >> + /**
> >> + * This is either an inter COPY (most likely) or an intra COPY with a
> >> + * stale file handle. If the latter, nfsd4_copy will reset the PUTFH to
> >> + * return nfserr_stale. No fh_dentry, just copy the file handle
> >> + * to use with the inter COPY READ.
> >> + */
> >> + if (HAS_CSTATE_FLAG(cstate, IS_STALE_FH)) {
> >> + cstate->save_fh = cstate->current_fh;
> >> + return nfs_ok;
> >> + }
> >> if (!cstate->current_fh.fh_dentry)
> >> return nfserr_nofilehandle;
> >>
> >> @@ -1067,6 +1085,13 @@ static int fill_in_write_vector(struct kvec *vec, struct nfsd4_write *write)
> >> if (status)
> >> goto out;
> >>
> >> + /* Intra copy source fh is stale. PUTFH will fail with ESTALE */
> >> + if (HAS_CSTATE_FLAG(cstate, IS_STALE_FH)) {
> >> + CLEAR_CSTATE_FLAG(cstate, IS_STALE_FH);
> >> + cstate->status = nfserr_copy_stalefh;
> >> + goto out_put;
> >> + }
> >> +
> >> bytes = nfsd_copy_file_range(src, copy->cp_src_pos,
> >> dst, copy->cp_dst_pos, copy->cp_count);
> >>
> >> @@ -1081,6 +1106,7 @@ static int fill_in_write_vector(struct kvec *vec, struct nfsd4_write *write)
> >> status = nfs_ok;
> >> }
> >>
> >> +out_put:
> >> fput(src);
> >> fput(dst);
> >> out:
> >> @@ -1759,6 +1785,7 @@ static void svcxdr_init_encode(struct svc_rqst *rqstp,
> >> struct nfsd4_compound_state *cstate = &resp->cstate;
> >> struct svc_fh *current_fh = &cstate->current_fh;
> >> struct svc_fh *save_fh = &cstate->save_fh;
> >> + int i;
> >> __be32 status;
> >>
> >> svcxdr_init_encode(rqstp, resp);
> >> @@ -1791,6 +1818,12 @@ static void svcxdr_init_encode(struct svc_rqst *rqstp,
> >> goto encode_op;
> >> }
> >>
> >> + /* NFSv4.2 COPY source file handle may be from a different server */
> >> + for (i = 0; i < args->opcnt; i++) {
> >> + op = &args->ops[i];
> >> + if (op->opnum == OP_COPY)
> >> + SET_CSTATE_FLAG(cstate, NO_VERIFY_FH);
> >> + }
> >
> > So, if a compound has a COPY anywhere in it, then a PUTFH with a stale
> > filehandle anywhere in the compound will temporarily succeed and result in
> > IS_STALE_FH being set.
> >
> >> while (!status && resp->opcnt < args->opcnt) {
> >> op = &args->ops[resp->opcnt++];
> >>
> >> @@ -1810,6 +1843,9 @@ static void svcxdr_init_encode(struct svc_rqst *rqstp,
> >>
> >> opdesc = OPDESC(op);
> >>
> >> + if (HAS_CSTATE_FLAG(cstate, IS_STALE_FH))
> >> + goto call_op;
> >> +
> >
> > That means that once IS_STALE_FH is set, we will skip:
> >
> > - nofilehandle checking
> > - moved checking
> > - fh_clear_wcc()
> > - resp_size checking
> > - current stateid capture
> >
> > on every subsequent op. Can that really be right?
> >
> >> if (!current_fh->fh_dentry) {
> >> if (!(opdesc->op_flags & ALLOWED_WITHOUT_FH)) {
> >> op->status = nfserr_nofilehandle;
> >> @@ -1844,6 +1880,7 @@ static void svcxdr_init_encode(struct svc_rqst *rqstp,
> >>
> >> if (opdesc->op_get_currentstateid)
> >> opdesc->op_get_currentstateid(cstate, &op->u);
> >> +call_op:
> >> op->status = opdesc->op_func(rqstp, cstate, &op->u);
> >>
> >> /* Only from SEQUENCE */
> >> @@ -1862,6 +1899,14 @@ static void svcxdr_init_encode(struct svc_rqst *rqstp,
> >> if (need_wrongsec_check(rqstp))
> >> op->status = check_nfsd_access(current_fh->fh_export, rqstp);
> >> }
> >> + /* Only from intra COPY */
> >> + if (cstate->status == nfserr_copy_stalefh) {
> >> + dprintk("%s NFS4.2 intra COPY stale src filehandle\n",
> >> + __func__);
> >> + status = nfserr_stale;
> >> + nfsd4_adjust_encode(resp);
> >
> > Are you sure it's safe just to throw away any operations since that
> > stale PUTFH? What if some of those operations had side effects?
> >
> >> + goto out;
> >> + }
> >> encode_op:
> >> if (op->status == nfserr_replay_me) {
> >> op->replay = &cstate->replay_owner->so_replay;
> >> diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> >> index 3e08c15..2896a11 100644
> >> --- a/fs/nfsd/nfs4xdr.c
> >> +++ b/fs/nfsd/nfs4xdr.c
> >> @@ -4624,15 +4624,28 @@ __be32 nfsd4_check_resp_size(struct nfsd4_compoundres *resp, u32 respsize)
> >> return nfserr_rep_too_big;
> >> }
> >>
> >> +/** Rewind the encoding to return nfserr_stale on the PUTFH
> >> + * in this failed Intra COPY compound
> >> + */
> >> +void
> >> +nfsd4_adjust_encode(struct nfsd4_compoundres *resp)
> >> +{
> >> + __be32 *p;
> >> +
> >> + p = resp->cstate.putfh_errp;
> >> + *p++ = nfserr_stale;
> >> +}
> >> +
> >> void
> >> nfsd4_encode_operation(struct nfsd4_compoundres *resp, struct nfsd4_op *op)
> >> {
> >> struct xdr_stream *xdr = &resp->xdr;
> >> struct nfs4_stateowner *so = resp->cstate.replay_owner;
> >> + struct nfsd4_compound_state *cstate = &resp->cstate;
> >> struct svc_rqst *rqstp = resp->rqstp;
> >> int post_err_offset;
> >> nfsd4_enc encoder;
> >> - __be32 *p;
> >> + __be32 *p, *statp;
> >>
> >> p = xdr_reserve_space(xdr, 8);
> >> if (!p) {
> >> @@ -4641,9 +4654,20 @@ __be32 nfsd4_check_resp_size(struct nfsd4_compoundres *resp, u32 respsize)
> >> }
> >> *p++ = cpu_to_be32(op->opnum);
> >> post_err_offset = xdr->buf->len;
> >> + statp = p;
> >>
> >> if (op->opnum == OP_ILLEGAL)
> >> goto status;
> >> +
> >> + /** This is a COPY compound with a stale source server file handle.
> >> + * If OP_COPY processing determines that this is an intra server to
> >> + * server COPY, then this PUTFH should return nfserr_ stale so the
> >> + * putfh_errp will be set to nfserr_stale. If this is an inter server
> >> + * to server COPY, ignore the nfserr_stale.
> >> + */
> >> + if (op->opnum == OP_PUTFH && HAS_CSTATE_FLAG(cstate, IS_STALE_FH))
> >> + cstate->putfh_errp = statp;
> >> +
> >> BUG_ON(op->opnum < 0 || op->opnum >= ARRAY_SIZE(nfsd4_enc_ops) ||
> >> !nfsd4_enc_ops[op->opnum]);
> >> encoder = nfsd4_enc_ops[op->opnum];
> >> diff --git a/fs/nfsd/nfsd.h b/fs/nfsd/nfsd.h
> >> index d966068..8d6fb0f 100644
> >> --- a/fs/nfsd/nfsd.h
> >> +++ b/fs/nfsd/nfsd.h
> >> @@ -272,6 +272,8 @@ static inline bool nfsd4_spo_must_allow(struct svc_rqst *rqstp)
> >> #define nfserr_replay_me cpu_to_be32(11001)
> >> /* nfs41 replay detected */
> >> #define nfserr_replay_cache cpu_to_be32(11002)
> >> +/* nfs42 intra copy failed with nfserr_stale */
> >> +#define nfserr_copy_stalefh cpu_to_be32(1103)
> >>
> >> /* Check for dir entries '.' and '..' */
> >> #define isdotent(n, l) (l < 3 && n[0] == '.' && (l == 1 || n[1] == '.'))
> >> diff --git a/fs/nfsd/xdr4.h b/fs/nfsd/xdr4.h
> >> index 38fcb4f..aa94295 100644
> >> --- a/fs/nfsd/xdr4.h
> >> +++ b/fs/nfsd/xdr4.h
> >> @@ -45,6 +45,8 @@
> >>
> >> #define CURRENT_STATE_ID_FLAG (1<<0)
> >> #define SAVED_STATE_ID_FLAG (1<<1)
> >> +#define NO_VERIFY_FH (1<<2)
> >> +#define IS_STALE_FH (1<<3)
> >>
> >> #define SET_CSTATE_FLAG(c, f) ((c)->sid_flags |= (f))
> >> #define HAS_CSTATE_FLAG(c, f) ((c)->sid_flags & (f))
> >> @@ -63,6 +65,7 @@ struct nfsd4_compound_state {
> >> size_t iovlen;
> >> u32 minorversion;
> >> __be32 status;
> >> + __be32 *putfh_errp;
> >> stateid_t current_stateid;
> >> stateid_t save_stateid;
> >> /* to indicate current and saved state id presents */
> >> @@ -705,6 +708,7 @@ int nfs4svc_decode_compoundargs(struct svc_rqst *, __be32 *,
> >> int nfs4svc_encode_compoundres(struct svc_rqst *, __be32 *,
> >> struct nfsd4_compoundres *);
> >> __be32 nfsd4_check_resp_size(struct nfsd4_compoundres *, u32);
> >> +void nfsd4_adjust_encode(struct nfsd4_compoundres *);
> >> void nfsd4_encode_operation(struct nfsd4_compoundres *, struct nfsd4_op *);
> >> void nfsd4_encode_replay(struct xdr_stream *xdr, struct nfsd4_op *op);
> >> __be32 nfsd4_encode_fattr_to_buf(__be32 **p, int words,
> >> --
> >> 1.8.3.1
> >>
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> >> the body of a message to majordomo@xxxxxxxxxxxxxxx
> >> More majordomo info at http://vger.kernel.org/majordomo-info.html
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
