3.16.50-rc1 review patch. If anyone has any objections, please let me know. ------------------ From: Chuck Lever <chuck.lever@xxxxxxxxxx> commit fc788f64f1f3eb31e87d4f53bcf1ab76590d5838 upstream. When processing an NFSv4 WRITE operation, argp->end should never point past the end of the data in the final page of the page list. Otherwise, nfsd4_decode_compound can walk into uninitialized memory. More critical, nfsd4_decode_write is failing to increment argp->pagelen when it increments argp->pagelist. This can cause later xdr decoders to assume more data is available than really is, which can cause server crashes on malformed requests. Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx> Signed-off-by: J. Bruce Fields <bfields@xxxxxxxxxx> Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx> --- fs/nfsd/nfs4xdr.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) --- a/fs/nfsd/nfs4xdr.c +++ b/fs/nfsd/nfs4xdr.c @@ -134,7 +134,7 @@ static void next_decode_page(struct nfsd argp->p = page_address(argp->pagelist[0]); argp->pagelist++; if (argp->pagelen < PAGE_SIZE) { - argp->end = argp->p + (argp->pagelen>>2); + argp->end = argp->p + XDR_QUADLEN(argp->pagelen); argp->pagelen = 0; } else { argp->end = argp->p + (PAGE_SIZE>>2); @@ -1246,9 +1246,7 @@ nfsd4_decode_write(struct nfsd4_compound argp->pagelen -= pages * PAGE_SIZE; len -= pages * PAGE_SIZE; - argp->p = (__be32 *)page_address(argp->pagelist[0]); - argp->pagelist++; - argp->end = argp->p + XDR_QUADLEN(PAGE_SIZE); + next_decode_page(argp); } argp->p += XDR_QUADLEN(len);