linux/net/rxrpc/rxperf.c
David Howells e0416e7d33 rxrpc: Fix potential race in error handling in afs_make_call()
If the rxrpc call set up by afs_make_call() receives an error whilst it is
transmitting the request, there's the possibility that it may get to the
point the rxrpc call is ended (after the error_kill_call label) just as the
call is queued for async processing.

This could manifest itself as call->rxcall being seen as NULL in
afs_deliver_to_call() when it tries to lock the call.

Fix this by splitting rxrpc_kernel_end_call() into a function to shut down
an rxrpc call and a function to release the caller's reference and calling
the latter only when we get to afs_put_call().

Reported-by: Jeffrey Altman <jaltman@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: kafs-testing+fedora36_64checkkafs-build-306@auristor.com
cc: Marc Dionne <marc.dionne@auristor.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: linux-afs@lists.infradead.org
cc: netdev@vger.kernel.org
Signed-off-by: David S. Miller <davem@davemloft.net>
2023-04-22 15:16:39 +01:00

626 lines
16 KiB
C

// SPDX-License-Identifier: GPL-2.0-or-later
/* In-kernel rxperf server for testing purposes.
*
* Copyright (C) 2022 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*/
#define pr_fmt(fmt) "rxperf: " fmt
#include <linux/module.h>
#include <linux/slab.h>
#include <net/sock.h>
#include <net/af_rxrpc.h>
#define RXRPC_TRACE_ONLY_DEFINE_ENUMS
#include <trace/events/rxrpc.h>
MODULE_DESCRIPTION("rxperf test server (afs)");
MODULE_AUTHOR("Red Hat, Inc.");
MODULE_LICENSE("GPL");
#define RXPERF_PORT 7009
#define RX_PERF_SERVICE 147
#define RX_PERF_VERSION 3
#define RX_PERF_SEND 0
#define RX_PERF_RECV 1
#define RX_PERF_RPC 3
#define RX_PERF_FILE 4
#define RX_PERF_MAGIC_COOKIE 0x4711
struct rxperf_proto_params {
__be32 version;
__be32 type;
__be32 rsize;
__be32 wsize;
} __packed;
static const u8 rxperf_magic_cookie[] = { 0x00, 0x00, 0x47, 0x11 };
static const u8 secret[8] = { 0xa7, 0x83, 0x8a, 0xcb, 0xc7, 0x83, 0xec, 0x94 };
enum rxperf_call_state {
RXPERF_CALL_SV_AWAIT_PARAMS, /* Server: Awaiting parameter block */
RXPERF_CALL_SV_AWAIT_REQUEST, /* Server: Awaiting request data */
RXPERF_CALL_SV_REPLYING, /* Server: Replying */
RXPERF_CALL_SV_AWAIT_ACK, /* Server: Awaiting final ACK */
RXPERF_CALL_COMPLETE, /* Completed or failed */
};
struct rxperf_call {
struct rxrpc_call *rxcall;
struct iov_iter iter;
struct kvec kvec[1];
struct work_struct work;
const char *type;
size_t iov_len;
size_t req_len; /* Size of request blob */
size_t reply_len; /* Size of reply blob */
unsigned int debug_id;
unsigned int operation_id;
struct rxperf_proto_params params;
__be32 tmp[2];
s32 abort_code;
enum rxperf_call_state state;
short error;
unsigned short unmarshal;
u16 service_id;
int (*deliver)(struct rxperf_call *call);
void (*processor)(struct work_struct *work);
};
static struct socket *rxperf_socket;
static struct key *rxperf_sec_keyring; /* Ring of security/crypto keys */
static struct workqueue_struct *rxperf_workqueue;
static void rxperf_deliver_to_call(struct work_struct *work);
static int rxperf_deliver_param_block(struct rxperf_call *call);
static int rxperf_deliver_request(struct rxperf_call *call);
static int rxperf_process_call(struct rxperf_call *call);
static void rxperf_charge_preallocation(struct work_struct *work);
static DECLARE_WORK(rxperf_charge_preallocation_work,
rxperf_charge_preallocation);
static inline void rxperf_set_call_state(struct rxperf_call *call,
enum rxperf_call_state to)
{
call->state = to;
}
static inline void rxperf_set_call_complete(struct rxperf_call *call,
int error, s32 remote_abort)
{
if (call->state != RXPERF_CALL_COMPLETE) {
call->abort_code = remote_abort;
call->error = error;
call->state = RXPERF_CALL_COMPLETE;
}
}
static void rxperf_rx_discard_new_call(struct rxrpc_call *rxcall,
unsigned long user_call_ID)
{
kfree((struct rxperf_call *)user_call_ID);
}
static void rxperf_rx_new_call(struct sock *sk, struct rxrpc_call *rxcall,
unsigned long user_call_ID)
{
queue_work(rxperf_workqueue, &rxperf_charge_preallocation_work);
}
static void rxperf_queue_call_work(struct rxperf_call *call)
{
queue_work(rxperf_workqueue, &call->work);
}
static void rxperf_notify_rx(struct sock *sk, struct rxrpc_call *rxcall,
unsigned long call_user_ID)
{
struct rxperf_call *call = (struct rxperf_call *)call_user_ID;
if (call->state != RXPERF_CALL_COMPLETE)
rxperf_queue_call_work(call);
}
static void rxperf_rx_attach(struct rxrpc_call *rxcall, unsigned long user_call_ID)
{
struct rxperf_call *call = (struct rxperf_call *)user_call_ID;
call->rxcall = rxcall;
}
static void rxperf_notify_end_reply_tx(struct sock *sock,
struct rxrpc_call *rxcall,
unsigned long call_user_ID)
{
rxperf_set_call_state((struct rxperf_call *)call_user_ID,
RXPERF_CALL_SV_AWAIT_ACK);
}
/*
* Charge the incoming call preallocation.
*/
static void rxperf_charge_preallocation(struct work_struct *work)
{
struct rxperf_call *call;
for (;;) {
call = kzalloc(sizeof(*call), GFP_KERNEL);
if (!call)
break;
call->type = "unset";
call->debug_id = atomic_inc_return(&rxrpc_debug_id);
call->deliver = rxperf_deliver_param_block;
call->state = RXPERF_CALL_SV_AWAIT_PARAMS;
call->service_id = RX_PERF_SERVICE;
call->iov_len = sizeof(call->params);
call->kvec[0].iov_len = sizeof(call->params);
call->kvec[0].iov_base = &call->params;
iov_iter_kvec(&call->iter, READ, call->kvec, 1, call->iov_len);
INIT_WORK(&call->work, rxperf_deliver_to_call);
if (rxrpc_kernel_charge_accept(rxperf_socket,
rxperf_notify_rx,
rxperf_rx_attach,
(unsigned long)call,
GFP_KERNEL,
call->debug_id) < 0)
break;
call = NULL;
}
kfree(call);
}
/*
* Open an rxrpc socket and bind it to be a server for callback notifications
* - the socket is left in blocking mode and non-blocking ops use MSG_DONTWAIT
*/
static int rxperf_open_socket(void)
{
struct sockaddr_rxrpc srx;
struct socket *socket;
int ret;
ret = sock_create_kern(&init_net, AF_RXRPC, SOCK_DGRAM, PF_INET6,
&socket);
if (ret < 0)
goto error_1;
socket->sk->sk_allocation = GFP_NOFS;
/* bind the callback manager's address to make this a server socket */
memset(&srx, 0, sizeof(srx));
srx.srx_family = AF_RXRPC;
srx.srx_service = RX_PERF_SERVICE;
srx.transport_type = SOCK_DGRAM;
srx.transport_len = sizeof(srx.transport.sin6);
srx.transport.sin6.sin6_family = AF_INET6;
srx.transport.sin6.sin6_port = htons(RXPERF_PORT);
ret = rxrpc_sock_set_min_security_level(socket->sk,
RXRPC_SECURITY_ENCRYPT);
if (ret < 0)
goto error_2;
ret = rxrpc_sock_set_security_keyring(socket->sk, rxperf_sec_keyring);
ret = kernel_bind(socket, (struct sockaddr *)&srx, sizeof(srx));
if (ret < 0)
goto error_2;
rxrpc_kernel_new_call_notification(socket, rxperf_rx_new_call,
rxperf_rx_discard_new_call);
ret = kernel_listen(socket, INT_MAX);
if (ret < 0)
goto error_2;
rxperf_socket = socket;
rxperf_charge_preallocation(&rxperf_charge_preallocation_work);
return 0;
error_2:
sock_release(socket);
error_1:
pr_err("Can't set up rxperf socket: %d\n", ret);
return ret;
}
/*
* close the rxrpc socket rxperf was using
*/
static void rxperf_close_socket(void)
{
kernel_listen(rxperf_socket, 0);
kernel_sock_shutdown(rxperf_socket, SHUT_RDWR);
flush_workqueue(rxperf_workqueue);
sock_release(rxperf_socket);
}
/*
* Log remote abort codes that indicate that we have a protocol disagreement
* with the server.
*/
static void rxperf_log_error(struct rxperf_call *call, s32 remote_abort)
{
static int max = 0;
const char *msg;
int m;
switch (remote_abort) {
case RX_EOF: msg = "unexpected EOF"; break;
case RXGEN_CC_MARSHAL: msg = "client marshalling"; break;
case RXGEN_CC_UNMARSHAL: msg = "client unmarshalling"; break;
case RXGEN_SS_MARSHAL: msg = "server marshalling"; break;
case RXGEN_SS_UNMARSHAL: msg = "server unmarshalling"; break;
case RXGEN_DECODE: msg = "opcode decode"; break;
case RXGEN_SS_XDRFREE: msg = "server XDR cleanup"; break;
case RXGEN_CC_XDRFREE: msg = "client XDR cleanup"; break;
case -32: msg = "insufficient data"; break;
default:
return;
}
m = max;
if (m < 3) {
max = m + 1;
pr_info("Peer reported %s failure on %s\n", msg, call->type);
}
}
/*
* deliver messages to a call
*/
static void rxperf_deliver_to_call(struct work_struct *work)
{
struct rxperf_call *call = container_of(work, struct rxperf_call, work);
enum rxperf_call_state state;
u32 abort_code, remote_abort = 0;
int ret = 0;
if (call->state == RXPERF_CALL_COMPLETE)
return;
while (state = call->state,
state == RXPERF_CALL_SV_AWAIT_PARAMS ||
state == RXPERF_CALL_SV_AWAIT_REQUEST ||
state == RXPERF_CALL_SV_AWAIT_ACK
) {
if (state == RXPERF_CALL_SV_AWAIT_ACK) {
if (!rxrpc_kernel_check_life(rxperf_socket, call->rxcall))
goto call_complete;
return;
}
ret = call->deliver(call);
if (ret == 0)
ret = rxperf_process_call(call);
switch (ret) {
case 0:
continue;
case -EINPROGRESS:
case -EAGAIN:
return;
case -ECONNABORTED:
rxperf_log_error(call, call->abort_code);
goto call_complete;
case -EOPNOTSUPP:
abort_code = RXGEN_OPCODE;
rxrpc_kernel_abort_call(rxperf_socket, call->rxcall,
abort_code, ret,
rxperf_abort_op_not_supported);
goto call_complete;
case -ENOTSUPP:
abort_code = RX_USER_ABORT;
rxrpc_kernel_abort_call(rxperf_socket, call->rxcall,
abort_code, ret,
rxperf_abort_op_not_supported);
goto call_complete;
case -EIO:
pr_err("Call %u in bad state %u\n",
call->debug_id, call->state);
fallthrough;
case -ENODATA:
case -EBADMSG:
case -EMSGSIZE:
case -ENOMEM:
case -EFAULT:
rxrpc_kernel_abort_call(rxperf_socket, call->rxcall,
RXGEN_SS_UNMARSHAL, ret,
rxperf_abort_unmarshal_error);
goto call_complete;
default:
rxrpc_kernel_abort_call(rxperf_socket, call->rxcall,
RX_CALL_DEAD, ret,
rxperf_abort_general_error);
goto call_complete;
}
}
call_complete:
rxperf_set_call_complete(call, ret, remote_abort);
/* The call may have been requeued */
rxrpc_kernel_shutdown_call(rxperf_socket, call->rxcall);
rxrpc_kernel_put_call(rxperf_socket, call->rxcall);
cancel_work(&call->work);
kfree(call);
}
/*
* Extract a piece of data from the received data socket buffers.
*/
static int rxperf_extract_data(struct rxperf_call *call, bool want_more)
{
u32 remote_abort = 0;
int ret;
ret = rxrpc_kernel_recv_data(rxperf_socket, call->rxcall, &call->iter,
&call->iov_len, want_more, &remote_abort,
&call->service_id);
pr_debug("Extract i=%zu l=%zu m=%u ret=%d\n",
iov_iter_count(&call->iter), call->iov_len, want_more, ret);
if (ret == 0 || ret == -EAGAIN)
return ret;
if (ret == 1) {
switch (call->state) {
case RXPERF_CALL_SV_AWAIT_REQUEST:
rxperf_set_call_state(call, RXPERF_CALL_SV_REPLYING);
break;
case RXPERF_CALL_COMPLETE:
pr_debug("premature completion %d", call->error);
return call->error;
default:
break;
}
return 0;
}
rxperf_set_call_complete(call, ret, remote_abort);
return ret;
}
/*
* Grab the operation ID from an incoming manager call.
*/
static int rxperf_deliver_param_block(struct rxperf_call *call)
{
u32 version;
int ret;
/* Extract the parameter block */
ret = rxperf_extract_data(call, true);
if (ret < 0)
return ret;
version = ntohl(call->params.version);
call->operation_id = ntohl(call->params.type);
call->deliver = rxperf_deliver_request;
if (version != RX_PERF_VERSION) {
pr_info("Version mismatch %x\n", version);
return -ENOTSUPP;
}
switch (call->operation_id) {
case RX_PERF_SEND:
call->type = "send";
call->reply_len = 0;
call->iov_len = 4; /* Expect req size */
break;
case RX_PERF_RECV:
call->type = "recv";
call->req_len = 0;
call->iov_len = 4; /* Expect reply size */
break;
case RX_PERF_RPC:
call->type = "rpc";
call->iov_len = 8; /* Expect req size and reply size */
break;
case RX_PERF_FILE:
call->type = "file";
fallthrough;
default:
return -EOPNOTSUPP;
}
rxperf_set_call_state(call, RXPERF_CALL_SV_AWAIT_REQUEST);
return call->deliver(call);
}
/*
* Deliver the request data.
*/
static int rxperf_deliver_request(struct rxperf_call *call)
{
int ret;
switch (call->unmarshal) {
case 0:
call->kvec[0].iov_len = call->iov_len;
call->kvec[0].iov_base = call->tmp;
iov_iter_kvec(&call->iter, READ, call->kvec, 1, call->iov_len);
call->unmarshal++;
fallthrough;
case 1:
ret = rxperf_extract_data(call, true);
if (ret < 0)
return ret;
switch (call->operation_id) {
case RX_PERF_SEND:
call->type = "send";
call->req_len = ntohl(call->tmp[0]);
call->reply_len = 0;
break;
case RX_PERF_RECV:
call->type = "recv";
call->req_len = 0;
call->reply_len = ntohl(call->tmp[0]);
break;
case RX_PERF_RPC:
call->type = "rpc";
call->req_len = ntohl(call->tmp[0]);
call->reply_len = ntohl(call->tmp[1]);
break;
default:
pr_info("Can't parse extra params\n");
return -EIO;
}
pr_debug("CALL op=%s rq=%zx rp=%zx\n",
call->type, call->req_len, call->reply_len);
call->iov_len = call->req_len;
iov_iter_discard(&call->iter, READ, call->req_len);
call->unmarshal++;
fallthrough;
case 2:
ret = rxperf_extract_data(call, false);
if (ret < 0)
return ret;
call->unmarshal++;
fallthrough;
default:
return 0;
}
}
/*
* Process a call for which we've received the request.
*/
static int rxperf_process_call(struct rxperf_call *call)
{
struct msghdr msg = {};
struct bio_vec bv;
struct kvec iov[1];
ssize_t n;
size_t reply_len = call->reply_len, len;
rxrpc_kernel_set_tx_length(rxperf_socket, call->rxcall,
reply_len + sizeof(rxperf_magic_cookie));
while (reply_len > 0) {
len = min_t(size_t, reply_len, PAGE_SIZE);
bvec_set_page(&bv, ZERO_PAGE(0), len, 0);
iov_iter_bvec(&msg.msg_iter, WRITE, &bv, 1, len);
msg.msg_flags = MSG_MORE;
n = rxrpc_kernel_send_data(rxperf_socket, call->rxcall, &msg,
len, rxperf_notify_end_reply_tx);
if (n < 0)
return n;
if (n == 0)
return -EIO;
reply_len -= n;
}
len = sizeof(rxperf_magic_cookie);
iov[0].iov_base = (void *)rxperf_magic_cookie;
iov[0].iov_len = len;
iov_iter_kvec(&msg.msg_iter, WRITE, iov, 1, len);
msg.msg_flags = 0;
n = rxrpc_kernel_send_data(rxperf_socket, call->rxcall, &msg, len,
rxperf_notify_end_reply_tx);
if (n >= 0)
return 0; /* Success */
if (n == -ENOMEM)
rxrpc_kernel_abort_call(rxperf_socket, call->rxcall,
RXGEN_SS_MARSHAL, -ENOMEM,
rxperf_abort_oom);
return n;
}
/*
* Add a key to the security keyring.
*/
static int rxperf_add_key(struct key *keyring)
{
key_ref_t kref;
int ret;
kref = key_create_or_update(make_key_ref(keyring, true),
"rxrpc_s",
__stringify(RX_PERF_SERVICE) ":2",
secret,
sizeof(secret),
KEY_POS_VIEW | KEY_POS_READ | KEY_POS_SEARCH
| KEY_USR_VIEW,
KEY_ALLOC_NOT_IN_QUOTA);
if (IS_ERR(kref)) {
pr_err("Can't allocate rxperf server key: %ld\n", PTR_ERR(kref));
return PTR_ERR(kref);
}
ret = key_link(keyring, key_ref_to_ptr(kref));
if (ret < 0)
pr_err("Can't link rxperf server key: %d\n", ret);
key_ref_put(kref);
return ret;
}
/*
* Initialise the rxperf server.
*/
static int __init rxperf_init(void)
{
struct key *keyring;
int ret = -ENOMEM;
pr_info("Server registering\n");
rxperf_workqueue = alloc_workqueue("rxperf", 0, 0);
if (!rxperf_workqueue)
goto error_workqueue;
keyring = keyring_alloc("rxperf_server",
GLOBAL_ROOT_UID, GLOBAL_ROOT_GID, current_cred(),
KEY_POS_VIEW | KEY_POS_READ | KEY_POS_SEARCH |
KEY_POS_WRITE |
KEY_USR_VIEW | KEY_USR_READ | KEY_USR_SEARCH |
KEY_USR_WRITE |
KEY_OTH_VIEW | KEY_OTH_READ | KEY_OTH_SEARCH,
KEY_ALLOC_NOT_IN_QUOTA,
NULL, NULL);
if (IS_ERR(keyring)) {
pr_err("Can't allocate rxperf server keyring: %ld\n",
PTR_ERR(keyring));
goto error_keyring;
}
rxperf_sec_keyring = keyring;
ret = rxperf_add_key(keyring);
if (ret < 0)
goto error_key;
ret = rxperf_open_socket();
if (ret < 0)
goto error_socket;
return 0;
error_socket:
error_key:
key_put(rxperf_sec_keyring);
error_keyring:
destroy_workqueue(rxperf_workqueue);
rcu_barrier();
error_workqueue:
pr_err("Failed to register: %d\n", ret);
return ret;
}
late_initcall(rxperf_init); /* Must be called after net/ to create socket */
static void __exit rxperf_exit(void)
{
pr_info("Server unregistering.\n");
rxperf_close_socket();
key_put(rxperf_sec_keyring);
destroy_workqueue(rxperf_workqueue);
rcu_barrier();
}
module_exit(rxperf_exit);