From: Jeff Layton Subject: [PATCH 3/3] NFS: clean up short packet handling for NFSv4 readdir Date: Fri, 22 Feb 2008 14:50:01 -0500 Message-ID: <1203709801-20055-4-git-send-email-jlayton@redhat.com> References: <1203709801-20055-1-git-send-email-jlayton@redhat.com> <1203709801-20055-2-git-send-email-jlayton@redhat.com> <1203709801-20055-3-git-send-email-jlayton@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Cc: linux-nfs@vger.kernel.org, nfsv4@linux-nfs.org To: Trond Myklebust Return-path: In-Reply-To: <1203709801-20055-3-git-send-email-jlayton@redhat.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: nfsv4-bounces@linux-nfs.org Errors-To: nfsv4-bounces@linux-nfs.org List-ID: Currently, the NFS readdir decoders have a workaround for buggy servers that send an empty readdir response with the EOF bit unset. If the server sends a malformed response in some cases, this workaround kicks in and just returns an empty response rather than returning a proper error to the caller. This patch does 3 things: 1) have malformed responses with no entries return error (-EIO) 2) preserve existing workaround for servers that send empty responses with the EOF marker unset. 3) Add some comments to clarify the logic in decode_readdir(). Signed-off-by: Jeff Layton --- fs/nfs/nfs4xdr.c | 37 +++++++++++++++++++++++++++---------- 1 files changed, 27 insertions(+), 10 deletions(-) diff --git a/fs/nfs/nfs4xdr.c b/fs/nfs/nfs4xdr.c index db1ed9c..6a203ba 100644 --- a/fs/nfs/nfs4xdr.c +++ b/fs/nfs/nfs4xdr.c @@ -3481,7 +3481,7 @@ static int decode_readdir(struct xdr_stream *xdr, struct rpc_rqst *req, struct n size_t hdrlen; u32 recvd, pglen = rcvbuf->page_len; __be32 *end, *entry, *p, *kaddr; - unsigned int nr; + unsigned int nr = 0; int status; status = decode_op_hdr(xdr, OP_READDIR); @@ -3505,7 +3505,12 @@ static int decode_readdir(struct xdr_stream *xdr, struct rpc_rqst *req, struct n kaddr = p = kmap_atomic(page, KM_USER0); end = p + ((pglen + readdir->pgbase) >> 2); entry = p; - for (nr = 0; *p++; nr++) { + + /* Make sure the packet actually has a value_follows and EOF entry */ + if ((entry + 1) > end) + goto short_pkt; + + for (; *p++; nr++) { u32 len, attrlen, xlen; if (end - p < 3) goto short_pkt; @@ -3532,20 +3537,32 @@ static int decode_readdir(struct xdr_stream *xdr, struct rpc_rqst *req, struct n p += attrlen; /* attributes */ entry = p; } - if (!nr && (entry[0] != 0 || entry[1] == 0)) - goto short_pkt; + /* + * Apparently some server sends responses that are a valid size, but + * contain no entries, and have value_follows==0 and EOF==0. For + * those, just set the EOF marker. + */ + if (!nr && entry[1] == 0) { + dprintk("NFS: readdir reply truncated!\n"); + entry[1] = 1; + } out: kunmap_atomic(kaddr, KM_USER0); return 0; short_pkt: + /* + * When we get a short packet there are 2 possibilities. We can + * return an error, or fix up the response to look like a valid + * response and return what we have so far. If there are no + * entries and the packet was short, then return -EIO. If there + * are valid entries in the response, return them and pretend that + * the call was successful, but incomplete. The caller can retry the + * readdir starting at the last cookie. + */ dprintk("%s: short packet at entry %d\n", __FUNCTION__, nr); entry[0] = entry[1] = 0; - /* truncate listing ? */ - if (!nr) { - dprintk("NFS: readdir reply truncated!\n"); - entry[1] = 1; - } - goto out; + if (nr) + goto out; err_unmap: kunmap_atomic(kaddr, KM_USER0); return -errno_NFSERR_IO; -- 1.5.4.1