Re: [PATCH v3 2/2] sunrpc: add a slot to rqstp->rq_bvec for TCP record marker

[Jeff Layton <jlayton@kernel.org>]

On Thu, 2025-10-09 at 11:03 -0400, Chuck Lever wrote:
> On 10/9/25 10:40 AM, Jeff Layton wrote:
> > We've seen some occurrences of messages like this in dmesg on some knfsd
> > servers:
> > 
> >     xdr_buf_to_bvec: bio_vec array overflow
> > 
> > Usually followed by messages like this that indicate a short send (note
> > that this message is from an older kernel and the amount that it reports
> > attempting to send is short by 4 bytes):
> > 
> >     rpc-srv/tcp: nfsd: sent 1048155 when sending 1048152 bytes - shutting down socket
> > 
> > svc_tcp_sendmsg() steals a slot in the rq_bvec array for the TCP record
> > marker. If the send is an unaligned READ call though, then there may not
> > be enough slots in the rq_bvec array in some cases.
> > 
> > Add a rqstp->rq_bvec_len field and use that to keep track of the length
> > of rq_bvec. Use that in place of rq_maxpages where it's iterating over
> > the bvec.
> 
> Granted that the number of items in rq_pages and in rq_bvec don't have
> to coincide, they just happen to be the same, historically. And, each
> bvec in rq_bvec doesn't necessarily have to be a page.
> 
> 
> > Fixes: e18e157bb5c8 ("SUNRPC: Send RPC message on TCP with a single sock_sendmsg() call")
> > Tested-by: Brandon Adams <brandona@meta.com>
> > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > ---
> >  fs/nfsd/vfs.c              | 6 +++---
> >  include/linux/sunrpc/svc.h | 1 +
> >  net/sunrpc/svc.c           | 4 +++-
> >  net/sunrpc/svcsock.c       | 4 ++--
> >  4 files changed, 9 insertions(+), 6 deletions(-)
> > 
> > diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
> > index 77f6879c2e063fa79865100bbc2d1e64eb332f42..6c7224570d2dadae21876e0069e0b2e0551af0fa 100644
> > --- a/fs/nfsd/vfs.c
> > +++ b/fs/nfsd/vfs.c
> > @@ -1111,7 +1111,7 @@ nfsd_direct_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
> >  
> >  	v = 0;
> >  	total = dio_end - dio_start;
> > -	while (total && v < rqstp->rq_maxpages &&
> > +	while (total && v < rqstp->rq_bvec_len &&
> >  	       rqstp->rq_next_page < rqstp->rq_page_end) {
> >  		len = min_t(size_t, total, PAGE_SIZE);
> >  		bvec_set_page(&rqstp->rq_bvec[v], *rqstp->rq_next_page,
> > @@ -1200,7 +1200,7 @@ __be32 nfsd_iter_read(struct svc_rqst *rqstp, struct svc_fh *fhp,
> >  
> >  	v = 0;
> >  	total = *count;
> > -	while (total && v < rqstp->rq_maxpages &&
> > +	while (total && v < rqstp->rq_bvec_len &&
> >  	       rqstp->rq_next_page < rqstp->rq_page_end) {
> >  		len = min_t(size_t, total, PAGE_SIZE - base);
> >  		bvec_set_page(&rqstp->rq_bvec[v], *rqstp->rq_next_page,
> > @@ -1318,7 +1318,7 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp,
> >  	if (stable && !fhp->fh_use_wgather)
> >  		kiocb.ki_flags |= IOCB_DSYNC;
> >  
> > -	nvecs = xdr_buf_to_bvec(rqstp->rq_bvec, rqstp->rq_maxpages, payload);
> > +	nvecs = xdr_buf_to_bvec(rqstp->rq_bvec, rqstp->rq_bvec_len, payload);
> >  	iov_iter_bvec(&iter, ITER_SOURCE, rqstp->rq_bvec, nvecs, *cnt);
> >  	since = READ_ONCE(file->f_wb_err);
> >  	if (verf)
> > diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
> > index 5506d20857c318774cd223272d4b0022cc19ffb8..0ee1f411860e55d5e0131c29766540f673193d5f 100644
> > --- a/include/linux/sunrpc/svc.h
> > +++ b/include/linux/sunrpc/svc.h
> > @@ -206,6 +206,7 @@ struct svc_rqst {
> >  
> >  	struct folio_batch	rq_fbatch;
> >  	struct bio_vec		*rq_bvec;
> > +	u32			rq_bvec_len;
> >  
> >  	__be32			rq_xid;		/* transmission id */
> >  	u32			rq_prog;	/* program number */
> > diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
> > index 4704dce7284eccc9e2bc64cf22947666facfa86a..a6bdd83fba77b13f973da66a1bac00050ae922fe 100644
> > --- a/net/sunrpc/svc.c
> > +++ b/net/sunrpc/svc.c
> > @@ -706,7 +706,9 @@ svc_prepare_thread(struct svc_serv *serv, struct svc_pool *pool, int node)
> >  	if (!svc_init_buffer(rqstp, serv, node))
> >  		goto out_enomem;
> >  
> > -	rqstp->rq_bvec = kcalloc_node(rqstp->rq_maxpages,
> > +	/* +1 for the TCP record marker */
> > +	rqstp->rq_bvec_len = rqstp->rq_maxpages + 1;
> 
> What bugs me about this is that svc_prepare_thread() shouldn't have
> specific knowledge about the needs of transports. But I don't have a
> better idea...
> 

Yeah, it's a minor layering violation. I guess I could phrase it as:

    /* Some transports need an extra bvec. (e.g. TCP needs it for the record marker) */

...but I'm not sure that's any clearer.

> 
> > +	rqstp->rq_bvec = kcalloc_node(rqstp->rq_bvec_len,
> >  				      sizeof(struct bio_vec),
> >  				      GFP_KERNEL, node);
> >  	if (!rqstp->rq_bvec)
> > diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
> > index 377fcaaaa061463fc5c85fc09c7a8eab5e06af77..2075ddec250b3fdb36becca4a53f1c0536f8634a 100644
> > --- a/net/sunrpc/svcsock.c
> > +++ b/net/sunrpc/svcsock.c
> > @@ -740,7 +740,7 @@ static int svc_udp_sendto(struct svc_rqst *rqstp)
> >  	if (svc_xprt_is_dead(xprt))
> >  		goto out_notconn;
> >  
> > -	count = xdr_buf_to_bvec(rqstp->rq_bvec, rqstp->rq_maxpages, xdr);
> > +	count = xdr_buf_to_bvec(rqstp->rq_bvec, rqstp->rq_bvec_len, xdr);
> >  
> >  	iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, rqstp->rq_bvec,
> >  		      count, rqstp->rq_res.len);
> > @@ -1244,7 +1244,7 @@ static int svc_tcp_sendmsg(struct svc_sock *svsk, struct svc_rqst *rqstp,
> >  	memcpy(buf, &marker, sizeof(marker));
> >  	bvec_set_virt(rqstp->rq_bvec, buf, sizeof(marker));
> >  
> > -	count = xdr_buf_to_bvec(rqstp->rq_bvec + 1, rqstp->rq_maxpages - 1,
> > +	count = xdr_buf_to_bvec(rqstp->rq_bvec + 1, rqstp->rq_bvec_len - 1,
> >  				&rqstp->rq_res);
> >  
> >  	iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, rqstp->rq_bvec,
> > 
> 

-- 
Jeff Layton <jlayton@kernel.org>