[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-id: <172912626140.81717.18213359690044489488@noble.neil.brown.name>
Date: Thu, 17 Oct 2024 11:51:01 +1100
From: "NeilBrown" <neilb@...e.de>
To: "Chuck Lever" <chuck.lever@...cle.com>
Cc: "Gustavo A. R. Silva" <gustavoars@...nel.org>,
"Jeff Layton" <jlayton@...nel.org>, "Olga Kornievskaia" <okorniev@...hat.com>,
"Dai Ngo" <Dai.Ngo@...cle.com>, "Tom Talpey" <tom@...pey.com>,
linux-nfs@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-hardening@...r.kernel.org, "Kees Cook" <kees@...nel.org>
Subject:
Re: [PATCH 2/5][next] nfsd: avoid -Wflex-array-member-not-at-end warnings
On Thu, 17 Oct 2024, Chuck Lever wrote:
> On Tue, Oct 15, 2024 at 06:29:31PM -0600, Gustavo A. R. Silva wrote:
> > -Wflex-array-member-not-at-end was introduced in GCC-14, and we are
> > getting ready to enable it, globally.
> >
> > Address the following warnings by changing the type of the middle struct
> > members in `struct nfsd_genl_rqstp`, which are currently causing trouble,
> > from `struct sockaddr` to `struct sockaddr_legacy`. Note that the latter
> > struct doesn't contain a flexible-array member.
> >
> > fs/nfsd/nfsd.h:74:33: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
> > fs/nfsd/nfsd.h:75:33: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
> >
> > Also, update some related code, accordingly.
> >
> > No binary differences are present after these changes.
> >
> > Signed-off-by: Gustavo A. R. Silva <gustavoars@...nel.org>
> > ---
> > fs/nfsd/nfsctl.c | 4 ++--
> > fs/nfsd/nfsd.h | 4 ++--
> > 2 files changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> > index 3adbc05ebaac..884bfdc7a255 100644
> > --- a/fs/nfsd/nfsctl.c
> > +++ b/fs/nfsd/nfsctl.c
> > @@ -1599,9 +1599,9 @@ int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
> > genl_rqstp.rq_stime = rqstp->rq_stime;
> > genl_rqstp.rq_opcnt = 0;
> > memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
> > - sizeof(struct sockaddr));
> > + sizeof(struct sockaddr_legacy));
> > memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
> > - sizeof(struct sockaddr));
> > + sizeof(struct sockaddr_legacy));
> >
> > #ifdef CONFIG_NFSD_V4
> > if (rqstp->rq_vers == NFS4_VERSION &&
> > diff --git a/fs/nfsd/nfsd.h b/fs/nfsd/nfsd.h
> > index 004415651295..44be32510595 100644
> > --- a/fs/nfsd/nfsd.h
> > +++ b/fs/nfsd/nfsd.h
> > @@ -71,8 +71,8 @@ struct readdir_cd {
> > #define NFSD_MAX_OPS_PER_COMPOUND 50
> >
> > struct nfsd_genl_rqstp {
> > - struct sockaddr rq_daddr;
> > - struct sockaddr rq_saddr;
> > + struct sockaddr_legacy rq_daddr;
> > + struct sockaddr_legacy rq_saddr;
>
> I was hoping to find a struct definition that was simply a union of
> sockaddr_in and sockaddr_in6, which we have in nfs-utils but I guess
> not here in the kernel.
>
> Can we use "struct sockaddr_storage" safely here? Then above,
> replace the raw memcpy() with rpc_copy_addr() ?
sockaddr_storage is conceptually the right solution. It is 128 bytes so
2 of them will waste about 200 bytes on the stack. Does that matter
these days? Probably not.
I do like using rpc_copy_addr().
NeilBrown
>
>
> > unsigned long rq_flags;
> > ktime_t rq_stime;
> > __be32 rq_xid;
> > --
> > 2.34.1
> >
>
> --
> Chuck Lever
>
Powered by blists - more mailing lists