[linux.git] / fs / afs / vl_probe.c

// SPDX-License-Identifier: GPL-2.0-or-later
/* AFS vlserver probing
 *
 * Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells ([email protected])
 */

#include <linux/sched.h>
#include <linux/slab.h>
#include "afs_fs.h"
#include "internal.h"
#include "protocol_yfs.h"


/*
 * Handle the completion of a set of probes.
 */
static void afs_finished_vl_probe(struct afs_vlserver *server)
{
	if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)) {
		server->rtt = UINT_MAX;
		clear_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags);
	}

	clear_bit_unlock(AFS_VLSERVER_FL_PROBING, &server->flags);
	wake_up_bit(&server->flags, AFS_VLSERVER_FL_PROBING);
}

/*
 * Handle the completion of a probe RPC call.
 */
static void afs_done_one_vl_probe(struct afs_vlserver *server, bool wake_up)
{
	if (atomic_dec_and_test(&server->probe_outstanding)) {
		afs_finished_vl_probe(server);
		wake_up = true;
	}

	if (wake_up)
		wake_up_all(&server->probe_wq);
}

/*
 * Process the result of probing a vlserver.  This is called after successful
 * or failed delivery of an VL.GetCapabilities operation.
 */
void afs_vlserver_probe_result(struct afs_call *call)
{
	struct afs_addr_list *alist = call->vl_probe;
	struct afs_vlserver *server = call->vlserver;
	struct afs_address *addr = &alist->addrs[call->probe_index];
	unsigned int server_index = call->server_index;
	unsigned int rtt_us = 0;
	unsigned int index = call->probe_index;
	bool have_result = false;
	int ret = call->error;

	_enter("%s,%u,%u,%d,%d", server->name, server_index, index, ret, call->abort_code);

	spin_lock(&server->probe_lock);

	switch (ret) {
	case 0:
		server->probe.error = 0;
		goto responded;
	case -ECONNABORTED:
		if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)) {
			server->probe.abort_code = call->abort_code;
			server->probe.error = ret;
		}
		goto responded;
	case -ENOMEM:
	case -ENONET:
	case -EKEYEXPIRED:
	case -EKEYREVOKED:
	case -EKEYREJECTED:
		server->probe.flags |= AFS_VLSERVER_PROBE_LOCAL_FAILURE;
		if (server->probe.error == 0)
			server->probe.error = ret;
		trace_afs_io_error(call->debug_id, ret, afs_io_error_vl_probe_fail);
		goto out;
	case -ECONNRESET: /* Responded, but call expired. */
	case -ERFKILL:
	case -EADDRNOTAVAIL:
	case -ENETUNREACH:
	case -EHOSTUNREACH:
	case -EHOSTDOWN:
	case -ECONNREFUSED:
	case -ETIMEDOUT:
	case -ETIME:
	default:
		clear_bit(index, &alist->responded);
		set_bit(index, &alist->probe_failed);
		if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED) &&
		    (server->probe.error == 0 ||
		     server->probe.error == -ETIMEDOUT ||
		     server->probe.error == -ETIME))
			server->probe.error = ret;
		trace_afs_io_error(call->debug_id, ret, afs_io_error_vl_probe_fail);
		goto out;
	}

responded:
	set_bit(index, &alist->responded);
	clear_bit(index, &alist->probe_failed);

	if (call->service_id == YFS_VL_SERVICE) {
		server->probe.flags |= AFS_VLSERVER_PROBE_IS_YFS;
		set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
		server->service_id = call->service_id;
	} else {
		server->probe.flags |= AFS_VLSERVER_PROBE_NOT_YFS;
		if (!(server->probe.flags & AFS_VLSERVER_PROBE_IS_YFS)) {
			clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
			server->service_id = call->service_id;
		}
	}

	rtt_us = rxrpc_kernel_get_srtt(addr->peer);
	if (rtt_us < server->probe.rtt) {
		server->probe.rtt = rtt_us;
		server->rtt = rtt_us;
		alist->preferred = index;
	}

	smp_wmb(); /* Set rtt before responded. */
	server->probe.flags |= AFS_VLSERVER_PROBE_RESPONDED;
	set_bit(AFS_VLSERVER_FL_PROBED, &server->flags);
	set_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags);
	have_result = true;
out:
	spin_unlock(&server->probe_lock);

	trace_afs_vl_probe(server, false, alist, index, call->error, call->abort_code, rtt_us);
	_debug("probe [%u][%u] %pISpc rtt=%d ret=%d",
	       server_index, index, rxrpc_kernel_remote_addr(addr->peer),
	       rtt_us, ret);

	afs_done_one_vl_probe(server, have_result);
}

/*
 * Probe all of a vlserver's addresses to find out the best route and to
 * query its capabilities.
 */
static bool afs_do_probe_vlserver(struct afs_net *net,
				  struct afs_vlserver *server,
				  struct key *key,
				  unsigned int server_index,
				  struct afs_error *_e)
{
	struct afs_addr_list *alist;
	struct afs_call *call;
	unsigned long unprobed;
	unsigned int index, i;
	bool in_progress = false;
	int best_prio;

	_enter("%s", server->name);

	read_lock(&server->lock);
	alist = rcu_dereference_protected(server->addresses,
					  lockdep_is_held(&server->lock));
	afs_get_addrlist(alist, afs_alist_trace_get_vlprobe);
	read_unlock(&server->lock);

	atomic_set(&server->probe_outstanding, alist->nr_addrs);
	memset(&server->probe, 0, sizeof(server->probe));
	server->probe.rtt = UINT_MAX;

	unprobed = (1UL << alist->nr_addrs) - 1;
	while (unprobed) {
		best_prio = -1;
		index = 0;
		for (i = 0; i < alist->nr_addrs; i++) {
			if (test_bit(i, &unprobed) &&
			    alist->addrs[i].prio > best_prio) {
				index = i;
				best_prio = alist->addrs[i].prio;
			}
		}
		__clear_bit(index, &unprobed);

		trace_afs_vl_probe(server, true, alist, index, 0, 0, 0);
		call = afs_vl_get_capabilities(net, alist, index, key, server,
					       server_index);
		if (!IS_ERR(call)) {
			afs_prioritise_error(_e, call->error, call->abort_code);
			afs_put_call(call);
			in_progress = true;
		} else {
			afs_prioritise_error(_e, PTR_ERR(call), 0);
			afs_done_one_vl_probe(server, false);
		}
	}

	afs_put_addrlist(alist, afs_alist_trace_put_vlprobe);
	return in_progress;
}

/*
 * Send off probes to all unprobed servers.
 */
int afs_send_vl_probes(struct afs_net *net, struct key *key,
		       struct afs_vlserver_list *vllist)
{
	struct afs_vlserver *server;
	struct afs_error e = {};
	bool in_progress = false;
	int i;

	for (i = 0; i < vllist->nr_servers; i++) {
		server = vllist->servers[i].server;
		if (test_bit(AFS_VLSERVER_FL_PROBED, &server->flags))
			continue;

		if (!test_and_set_bit_lock(AFS_VLSERVER_FL_PROBING, &server->flags) &&
		    afs_do_probe_vlserver(net, server, key, i, &e))
			in_progress = true;
	}

	return in_progress ? 0 : e.error;
}

/*
 * Wait for the first as-yet untried server to respond.
 */
int afs_wait_for_vl_probes(struct afs_vlserver_list *vllist,
			   unsigned long untried)
{
	struct wait_queue_entry *waits;
	struct afs_vlserver *server;
	unsigned int rtt = UINT_MAX, rtt_s;
	bool have_responders = false;
	int pref = -1, i;

	_enter("%u,%lx", vllist->nr_servers, untried);

	/* Only wait for servers that have a probe outstanding. */
	for (i = 0; i < vllist->nr_servers; i++) {
		if (test_bit(i, &untried)) {
			server = vllist->servers[i].server;
			if (!test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
				__clear_bit(i, &untried);
			if (server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)
				have_responders = true;
		}
	}
	if (have_responders || !untried)
		return 0;

	waits = kmalloc(array_size(vllist->nr_servers, sizeof(*waits)), GFP_KERNEL);
	if (!waits)
		return -ENOMEM;

	for (i = 0; i < vllist->nr_servers; i++) {
		if (test_bit(i, &untried)) {
			server = vllist->servers[i].server;
			init_waitqueue_entry(&waits[i], current);
			add_wait_queue(&server->probe_wq, &waits[i]);
		}
	}

	for (;;) {
		bool still_probing = false;

		set_current_state(TASK_INTERRUPTIBLE);
		for (i = 0; i < vllist->nr_servers; i++) {
			if (test_bit(i, &untried)) {
				server = vllist->servers[i].server;
				if (server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)
					goto stop;
				if (test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
					still_probing = true;
			}
		}

		if (!still_probing || signal_pending(current))
			goto stop;
		schedule();
	}

stop:
	set_current_state(TASK_RUNNING);

	for (i = 0; i < vllist->nr_servers; i++) {
		if (test_bit(i, &untried)) {
			server = vllist->servers[i].server;
			rtt_s = READ_ONCE(server->rtt);
			if (test_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags) &&
			    rtt_s < rtt) {
				pref = i;
				rtt = rtt_s;
			}

			remove_wait_queue(&server->probe_wq, &waits[i]);
		}
	}

	kfree(waits);

	if (pref == -1 && signal_pending(current))
		return -ERESTARTSYS;

	if (pref >= 0)
		vllist->preferred = pref;

	_leave(" = 0 [%u]", pref);
	return 0;
}
Commit	Line	Data
b4d0d230	1	// SPDX-License-Identifier: GPL-2.0-or-later
3bf0fb6f DH	2	/* AFS vlserver probing
	3	*
	4	* Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
	5	* Written by David Howells ([email protected])
3bf0fb6f DH	6	*/
	7
	8	#include <linux/sched.h>
	9	#include <linux/slab.h>
	10	#include "afs_fs.h"
	11	#include "internal.h"
	12	#include "protocol_yfs.h"
	13
b95b3094 DH	14
	15	/*
	16	* Handle the completion of a set of probes.
	17	*/
	18	static void afs_finished_vl_probe(struct afs_vlserver *server)
3bf0fb6f	19	{
b95b3094 DH	20	if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)) {
	21	server->rtt = UINT_MAX;
	22	clear_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags);
	23	}
3bf0fb6f	24
3bf0fb6f DH	25	clear_bit_unlock(AFS_VLSERVER_FL_PROBING, &server->flags);
3bf0fb6f DH	26	wake_up_bit(&server->flags, AFS_VLSERVER_FL_PROBING);
b95b3094 DH	27	}
	28
	29	/*
	30	* Handle the completion of a probe RPC call.
	31	*/
	32	static void afs_done_one_vl_probe(struct afs_vlserver *server, bool wake_up)
	33	{
	34	if (atomic_dec_and_test(&server->probe_outstanding)) {
	35	afs_finished_vl_probe(server);
	36	wake_up = true;
	37	}
	38
	39	if (wake_up)
	40	wake_up_all(&server->probe_wq);
3bf0fb6f DH	41	}
	42
	43	/*
	44	* Process the result of probing a vlserver. This is called after successful
	45	* or failed delivery of an VL.GetCapabilities operation.
	46	*/
	47	void afs_vlserver_probe_result(struct afs_call *call)
	48	{
f49b594d	49	struct afs_addr_list *alist = call->vl_probe;
ffba718e	50	struct afs_vlserver *server = call->vlserver;
98f9fda2	51	struct afs_address *addr = &alist->addrs[call->probe_index];
ffba718e	52	unsigned int server_index = call->server_index;
c410bf01	53	unsigned int rtt_us = 0;
98f9fda2	54	unsigned int index = call->probe_index;
3bf0fb6f	55	bool have_result = false;
3bf0fb6f DH	56	int ret = call->error;
	57
	58	_enter("%s,%u,%u,%d,%d", server->name, server_index, index, ret, call->abort_code);
	59
	60	spin_lock(&server->probe_lock);
	61
	62	switch (ret) {
	63	case 0:
	64	server->probe.error = 0;
	65	goto responded;
	66	case -ECONNABORTED:
fb72cd3d	67	if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)) {
3bf0fb6f DH	68	server->probe.abort_code = call->abort_code;
	69	server->probe.error = ret;
	70	}
	71	goto responded;
	72	case -ENOMEM:
	73	case -ENONET:
b95b3094 DH	74	case -EKEYEXPIRED:
	75	case -EKEYREVOKED:
	76	case -EKEYREJECTED:
fb72cd3d	77	server->probe.flags \|= AFS_VLSERVER_PROBE_LOCAL_FAILURE;
b95b3094 DH	78	if (server->probe.error == 0)
	79	server->probe.error = ret;
	80	trace_afs_io_error(call->debug_id, ret, afs_io_error_vl_probe_fail);
3bf0fb6f DH	81	goto out;
3bf0fb6f DH	82	case -ECONNRESET: /* Responded, but call expired. */
4584ae96 DH	83	case -ERFKILL:
4584ae96 DH	84	case -EADDRNOTAVAIL:
3bf0fb6f DH	85	case -ENETUNREACH:
3bf0fb6f DH	86	case -EHOSTUNREACH:
4584ae96	87	case -EHOSTDOWN:
3bf0fb6f DH	88	case -ECONNREFUSED:
	89	case -ETIMEDOUT:
	90	case -ETIME:
	91	default:
	92	clear_bit(index, &alist->responded);
aa4917d6	93	set_bit(index, &alist->probe_failed);
fb72cd3d	94	if (!(server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED) &&
3bf0fb6f DH	95	(server->probe.error == 0 \|\|
	96	server->probe.error == -ETIMEDOUT \|\|
	97	server->probe.error == -ETIME))
	98	server->probe.error = ret;
b95b3094	99	trace_afs_io_error(call->debug_id, ret, afs_io_error_vl_probe_fail);
3bf0fb6f DH	100	goto out;
	101	}
	102
	103	responded:
	104	set_bit(index, &alist->responded);
aa4917d6	105	clear_bit(index, &alist->probe_failed);
3bf0fb6f DH	106
3bf0fb6f DH	107	if (call->service_id == YFS_VL_SERVICE) {
fb72cd3d	108	server->probe.flags \|= AFS_VLSERVER_PROBE_IS_YFS;
3bf0fb6f	109	set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
e38f299e	110	server->service_id = call->service_id;
3bf0fb6f	111	} else {
fb72cd3d DH	112	server->probe.flags \|= AFS_VLSERVER_PROBE_NOT_YFS;
fb72cd3d DH	113	if (!(server->probe.flags & AFS_VLSERVER_PROBE_IS_YFS)) {
3bf0fb6f	114	clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
e38f299e	115	server->service_id = call->service_id;
3bf0fb6f DH	116	}
	117	}
	118
72904d7b	119	rtt_us = rxrpc_kernel_get_srtt(addr->peer);
ba00b190	120	if (rtt_us < server->probe.rtt) {
c410bf01	121	server->probe.rtt = rtt_us;
b95b3094	122	server->rtt = rtt_us;
3bf0fb6f	123	alist->preferred = index;
3bf0fb6f DH	124	}
	125
	126	smp_wmb(); /* Set rtt before responded. */
fb72cd3d	127	server->probe.flags \|= AFS_VLSERVER_PROBE_RESPONDED;
3bf0fb6f	128	set_bit(AFS_VLSERVER_FL_PROBED, &server->flags);
b95b3094 DH	129	set_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags);
b95b3094 DH	130	have_result = true;
3bf0fb6f DH	131	out:
	132	spin_unlock(&server->probe_lock);
	133
e6a7d7f7	134	trace_afs_vl_probe(server, false, alist, index, call->error, call->abort_code, rtt_us);
72904d7b DH	135	_debug("probe [%u][%u] %pISpc rtt=%d ret=%d",
	136	server_index, index, rxrpc_kernel_remote_addr(addr->peer),
	137	rtt_us, ret);
3bf0fb6f	138
b95b3094	139	afs_done_one_vl_probe(server, have_result);
3bf0fb6f DH	140	}
	141
	142	/*
	143	* Probe all of a vlserver's addresses to find out the best route and to
	144	* query its capabilities.
	145	*/
4584ae96 DH	146	static bool afs_do_probe_vlserver(struct afs_net *net,
	147	struct afs_vlserver *server,
	148	struct key *key,
	149	unsigned int server_index,
	150	struct afs_error *_e)
3bf0fb6f	151	{
98f9fda2	152	struct afs_addr_list *alist;
0b9bf381	153	struct afs_call *call;
e6a7d7f7 DH	154	unsigned long unprobed;
e6a7d7f7 DH	155	unsigned int index, i;
4584ae96	156	bool in_progress = false;
e6a7d7f7	157	int best_prio;
3bf0fb6f DH	158
	159	_enter("%s", server->name);
	160
	161	read_lock(&server->lock);
98f9fda2 DH	162	alist = rcu_dereference_protected(server->addresses,
	163	lockdep_is_held(&server->lock));
	164	afs_get_addrlist(alist, afs_alist_trace_get_vlprobe);
3bf0fb6f DH	165	read_unlock(&server->lock);
3bf0fb6f DH	166
98f9fda2	167	atomic_set(&server->probe_outstanding, alist->nr_addrs);
3bf0fb6f DH	168	memset(&server->probe, 0, sizeof(server->probe));
	169	server->probe.rtt = UINT_MAX;
	170
e6a7d7f7 DH	171	unprobed = (1UL << alist->nr_addrs) - 1;
	172	while (unprobed) {
	173	best_prio = -1;
	174	index = 0;
	175	for (i = 0; i < alist->nr_addrs; i++) {
	176	if (test_bit(i, &unprobed) &&
	177	alist->addrs[i].prio > best_prio) {
	178	index = i;
	179	best_prio = alist->addrs[i].prio;
	180	}
	181	}
	182	__clear_bit(index, &unprobed);
	183
	184	trace_afs_vl_probe(server, true, alist, index, 0, 0, 0);
98f9fda2	185	call = afs_vl_get_capabilities(net, alist, index, key, server,
0b9bf381 DH	186	server_index);
0b9bf381 DH	187	if (!IS_ERR(call)) {
aa453bec	188	afs_prioritise_error(_e, call->error, call->abort_code);
0b9bf381	189	afs_put_call(call);
4584ae96	190	in_progress = true;
0b9bf381	191	} else {
aa453bec	192	afs_prioritise_error(_e, PTR_ERR(call), 0);
b95b3094	193	afs_done_one_vl_probe(server, false);
0b9bf381	194	}
3bf0fb6f DH	195	}
3bf0fb6f DH	196
98f9fda2	197	afs_put_addrlist(alist, afs_alist_trace_put_vlprobe);
4584ae96	198	return in_progress;
3bf0fb6f DH	199	}
	200
	201	/*
	202	* Send off probes to all unprobed servers.
	203	*/
	204	int afs_send_vl_probes(struct afs_net net, struct key key,
	205	struct afs_vlserver_list *vllist)
	206	{
	207	struct afs_vlserver *server;
aa453bec	208	struct afs_error e = {};
4584ae96 DH	209	bool in_progress = false;
4584ae96 DH	210	int i;
3bf0fb6f DH	211
	212	for (i = 0; i < vllist->nr_servers; i++) {
	213	server = vllist->servers[i].server;
	214	if (test_bit(AFS_VLSERVER_FL_PROBED, &server->flags))
	215	continue;
	216
4584ae96 DH	217	if (!test_and_set_bit_lock(AFS_VLSERVER_FL_PROBING, &server->flags) &&
	218	afs_do_probe_vlserver(net, server, key, i, &e))
	219	in_progress = true;
3bf0fb6f DH	220	}
3bf0fb6f DH	221
4584ae96	222	return in_progress ? 0 : e.error;
3bf0fb6f DH	223	}
	224
	225	/*
	226	* Wait for the first as-yet untried server to respond.
	227	*/
	228	int afs_wait_for_vl_probes(struct afs_vlserver_list *vllist,
	229	unsigned long untried)
	230	{
	231	struct wait_queue_entry *waits;
	232	struct afs_vlserver *server;
b95b3094	233	unsigned int rtt = UINT_MAX, rtt_s;
3bf0fb6f DH	234	bool have_responders = false;
	235	int pref = -1, i;
	236
	237	_enter("%u,%lx", vllist->nr_servers, untried);
	238
	239	/* Only wait for servers that have a probe outstanding. */
	240	for (i = 0; i < vllist->nr_servers; i++) {
	241	if (test_bit(i, &untried)) {
	242	server = vllist->servers[i].server;
	243	if (!test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
	244	__clear_bit(i, &untried);
fb72cd3d	245	if (server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)
3bf0fb6f DH	246	have_responders = true;
	247	}
	248	}
	249	if (have_responders \|\| !untried)
	250	return 0;
	251
	252	waits = kmalloc(array_size(vllist->nr_servers, sizeof(*waits)), GFP_KERNEL);
	253	if (!waits)
	254	return -ENOMEM;
	255
	256	for (i = 0; i < vllist->nr_servers; i++) {
	257	if (test_bit(i, &untried)) {
	258	server = vllist->servers[i].server;
	259	init_waitqueue_entry(&waits[i], current);
	260	add_wait_queue(&server->probe_wq, &waits[i]);
	261	}
	262	}
	263
	264	for (;;) {
	265	bool still_probing = false;
	266
	267	set_current_state(TASK_INTERRUPTIBLE);
	268	for (i = 0; i < vllist->nr_servers; i++) {
	269	if (test_bit(i, &untried)) {
	270	server = vllist->servers[i].server;
fb72cd3d	271	if (server->probe.flags & AFS_VLSERVER_PROBE_RESPONDED)
3bf0fb6f DH	272	goto stop;
	273	if (test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
	274	still_probing = true;
	275	}
	276	}
	277
08d405c8	278	if (!still_probing \|\| signal_pending(current))
3bf0fb6f DH	279	goto stop;
	280	schedule();
	281	}
	282
	283	stop:
	284	set_current_state(TASK_RUNNING);
	285
	286	for (i = 0; i < vllist->nr_servers; i++) {
	287	if (test_bit(i, &untried)) {
	288	server = vllist->servers[i].server;
b95b3094 DH	289	rtt_s = READ_ONCE(server->rtt);
	290	if (test_bit(AFS_VLSERVER_FL_RESPONDING, &server->flags) &&
	291	rtt_s < rtt) {
3bf0fb6f	292	pref = i;
b95b3094	293	rtt = rtt_s;
3bf0fb6f DH	294	}
	295
	296	remove_wait_queue(&server->probe_wq, &waits[i]);
	297	}
	298	}
	299
	300	kfree(waits);
	301
	302	if (pref == -1 && signal_pending(current))
	303	return -ERESTARTSYS;
	304
	305	if (pref >= 0)
	306	vllist->preferred = pref;
	307
	308	_leave(" = 0 [%u]", pref);
	309	return 0;
	310	}