Another thinking for that case: is it neccessory to write pipe with all possible xid to wake up all nfsd waiting for nfsdcld ? If nfsdcld is signaled SIGTERM in processing message, nfsdcld may also crash and nfsd wait for it forever too.
On Tue, 25 Feb 2025, zhangjian (CG) wrote:
> When nfsd inits failed for detecting cld version in nfsd4_client_tracking_init, kernel may assume nfsdcld support version 1 message format and try to upcall with v1 message size to nfsdcld. There exists one error case in the following process, causeing nfsd hunging for nfsdcld replay:
>
> kernel write to pipe->msgs (v1 msg length)
> |--------- first msg --------|-------- second message -------|
>
> nfsdcld read from pipe->msgs (v2 msg length)
> |------------ first msg --------------|---second message-----|
> | valid message | ignore | wrong message |
>
> When two nfsd kernel thread add two upcall messages to cld pipe with v1 version cld_msg (size == 1034) concurrently，but nfsdcld reads with v2 version size(size == 1067), 33 bytes of the second message will be read and merged with first message. The 33 bytes in second message will be ignored. Nfsdcld will then read 1001 bytes in second message, which cause FATAL in cld_messaged_size checking. Nfsd kernel thread will hang for it forever until nfs server restarts.
>
> Signed-off-by: zhangjian <zhangjian496@xxxxxxxxxx>
> ---
> utils/nfsdcld/nfsdcld.c | 63 ++++++++++++++++++++++++++++-------------
> 1 file changed, 43 insertions(+), 20 deletions(-)
>
> diff --git a/utils/nfsdcld/nfsdcld.c b/utils/nfsdcld/nfsdcld.c
> index dbc7a57..76308d1 100644
> --- a/utils/nfsdcld/nfsdcld.c
> +++ b/utils/nfsdcld/nfsdcld.c
> @@ -716,35 +716,58 @@ reply:
> }
> }
>
> +static int cld_pipe_read_msg(struct cld_client *clnt) {
> + ssize_t len, left_len;
> + ssize_t hdr_len = sizeof(struct cld_msg_hdr);
> + struct cld_msg_hdr *hdr = (struct cld_msg_hdr *)&clnt->cl_u;;
> +
> + len = atomicio(read, clnt->cl_fd, hdr, hdr_len);
> +
> + if (len <= 0) {
> + xlog(L_ERROR, "%s: pipe read failed: %m", __func__);
> + goto fail_read;
> + }
> +
> + switch (hdr->cm_vers) {
> + case 1:
> + left_len = sizeof(struct cld_msg) - hdr_len;
> + break;
> + case 2:
> + left_len = sizeof(struct cld_msg_v2) - hdr_len;
> + break;
> + default:
> + xlog(L_ERROR, "%s: unsupported upcall version: %hu",
> + __func__, hdr->cm_vers);
> + goto fail_read;
> + }
> +
> + len = atomicio(read, clnt->cl_fd, hdr, left_len);
> +
> + if (len <= 0) {
> + xlog(L_ERROR, "%s: pipe read failed: %m", __func__);
> + goto fail_read;
> + }
> +
> + return 0;
> +
> +fail_read:
> + cld_pipe_open(clnt);
> + return -1;
> +}
> +
> static void
> cldcb(int UNUSED(fd), short which, void *data)
> {
> - ssize_t len;
> struct cld_client *clnt = data;
> -#if UPCALL_VERSION >= 2
> - struct cld_msg_v2 *cmsg = &clnt->cl_u.cl_msg_v2;
> -#else
> - struct cld_msg *cmsg = &clnt->cl_u.cl_msg;
> -#endif
> + struct cld_msg_hdr *hdr = (struct cld_msg_hdr *)&clnt->cl_u;
>
> if (which != EV_READ)
> goto out;
>
> - len = atomicio(read, clnt->cl_fd, cmsg, sizeof(*cmsg));
> - if (len <= 0) {
> - xlog(L_ERROR, "%s: pipe read failed: %m", __func__);
> - cld_pipe_open(clnt);
> + if (cld_pipe_read_msg(clnt) < 0)
> goto out;
> - }
> -
> - if (cmsg->cm_vers > UPCALL_VERSION) {
> - xlog(L_ERROR, "%s: unsupported upcall version: %hu",
> - __func__, cmsg->cm_vers);
> - cld_pipe_open(clnt);
> - goto out;
> - }
>
> - switch(cmsg->cm_cmd) {
> + switch (hdr->cm_cmd) {
> case Cld_Create:
> cld_create(clnt);
> break;
> @@ -765,7 +788,7 @@ cldcb(int UNUSED(fd), short which, void *data)
> break;
> default:
> xlog(L_WARNING, "%s: command %u is not yet implemented",
> - __func__, cmsg->cm_cmd);
> + __func__, hdr->cm_cmd);
> cld_not_implemented(clnt);
> }
> out:
