3.16.44-rc1 review patch. If anyone has any objections, please let me know. ------------------ From: "J. Bruce Fields" <bfields@xxxxxxxxxx> commit 13bf9fbff0e5e099e2b6f003a0ab8ae145436309 upstream. The NFSv2/v3 code does not systematically check whether we decode past the end of the buffer. This generally appears to be harmless, but there are a few places where we do arithmetic on the pointers involved and don't account for the possibility that a length could be negative. Add checks to catch these. Reported-by: Tuomas Haanpää <thaan@xxxxxxxxxxxx> Reported-by: Ari Kauppi <ari@xxxxxxxxxxxx> Reviewed-by: NeilBrown <neilb@xxxxxxxx> Signed-off-by: J. Bruce Fields <bfields@xxxxxxxxxx> Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx> --- fs/nfsd/nfs3xdr.c | 4 ++++ fs/nfsd/nfsxdr.c | 2 ++ 2 files changed, 6 insertions(+) --- a/fs/nfsd/nfs3xdr.c +++ b/fs/nfsd/nfs3xdr.c @@ -373,6 +373,8 @@ nfs3svc_decode_writeargs(struct svc_rqst args->count = ntohl(*p++); args->stable = ntohl(*p++); len = args->len = ntohl(*p++); + if ((void *)p > head->iov_base + head->iov_len) + return 0; /* * The count must equal the amount of data passed. */ @@ -476,6 +478,8 @@ nfs3svc_decode_symlinkargs(struct svc_rq /* first copy and check from the first page */ old = (char*)p; vec = &rqstp->rq_arg.head[0]; + if ((void *)old > vec->iov_base + vec->iov_len) + return 0; avail = vec->iov_len - (old - (char*)vec->iov_base); while (len && avail && *old) { *new++ = *old++; --- a/fs/nfsd/nfsxdr.c +++ b/fs/nfsd/nfsxdr.c @@ -303,6 +303,8 @@ nfssvc_decode_writeargs(struct svc_rqst * bytes. */ hdr = (void*)p - head->iov_base; + if (hdr > head->iov_len) + return 0; dlen = head->iov_len + rqstp->rq_arg.page_len - hdr; /*