[thirdparty/kernel/stable.git] / fs / afs / vl_probe.c

/* AFS vlserver probing
 *
 * Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells (dhowells@redhat.com)
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public Licence
 * as published by the Free Software Foundation; either version
 * 2 of the Licence, or (at your option) any later version.
 */

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

static bool afs_vl_probe_done(struct afs_vlserver *server)
{
	if (!atomic_dec_and_test(&server->probe_outstanding))
		return false;

	wake_up_var(&server->probe_outstanding);
	clear_bit_unlock(AFS_VLSERVER_FL_PROBING, &server->flags);
	wake_up_bit(&server->flags, AFS_VLSERVER_FL_PROBING);
	return true;
}

/*
 * 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->alist;
	struct afs_vlserver *server = call->reply[0];
	unsigned int server_index = (long)call->reply[1];
	unsigned int index = call->addr_ix;
	unsigned int rtt = UINT_MAX;
	bool have_result = false;
	u64 _rtt;
	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.responded) {
			server->probe.abort_code = call->abort_code;
			server->probe.error = ret;
		}
		goto responded;
	case -ENOMEM:
	case -ENONET:
		server->probe.local_failure = true;
		afs_io_error(call, 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->failed);
		if (!server->probe.responded &&
		    (server->probe.error == 0 ||
		     server->probe.error == -ETIMEDOUT ||
		     server->probe.error == -ETIME))
			server->probe.error = ret;
		afs_io_error(call, afs_io_error_vl_probe_fail);
		goto out;
	}

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

	if (call->service_id == YFS_VL_SERVICE) {
		server->probe.is_yfs = true;
		set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
		alist->addrs[index].srx_service = call->service_id;
	} else {
		server->probe.not_yfs = true;
		if (!server->probe.is_yfs) {
			clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
			alist->addrs[index].srx_service = call->service_id;
		}
	}

	/* Get the RTT and scale it to fit into a 32-bit value that represents
	 * over a minute of time so that we can access it with one instruction
	 * on a 32-bit system.
	 */
	_rtt = rxrpc_kernel_get_rtt(call->net->socket, call->rxcall);
	_rtt /= 64;
	rtt = (_rtt > UINT_MAX) ? UINT_MAX : _rtt;
	if (rtt < server->probe.rtt) {
		server->probe.rtt = rtt;
		alist->preferred = index;
		have_result = true;
	}

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

	_debug("probe [%u][%u] %pISpc rtt=%u ret=%d",
	       server_index, index, &alist->addrs[index].transport,
	       (unsigned int)rtt, ret);

	have_result |= afs_vl_probe_done(server);
	if (have_result) {
		server->probe.have_result = true;
		wake_up_var(&server->probe.have_result);
		wake_up_all(&server->probe_wq);
	}
}

/*
 * 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_cursor ac = {
		.index = 0,
	};
	bool in_progress = false;
	int err;

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

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

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

	for (ac.index = 0; ac.index < ac.alist->nr_addrs; ac.index++) {
		err = afs_vl_get_capabilities(net, &ac, key, server,
					      server_index, true);
		if (err == -EINPROGRESS)
			in_progress = true;
		else
			afs_prioritise_error(_e, err, ac.abort_code);
	}

	if (!in_progress)
		afs_vl_probe_done(server);
	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;

	e.error = 0;
	e.responded = false;
	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;
	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.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.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;
			if (server->probe.responded &&
			    server->probe.rtt < rtt) {
				pref = i;
				rtt = server->probe.rtt;
			}

			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
3bf0fb6f DH	1	/* AFS vlserver probing
	2	*
	3	* Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
	4	* Written by David Howells (dhowells@redhat.com)
	5	*
	6	* This program is free software; you can redistribute it and/or
	7	* modify it under the terms of the GNU General Public Licence
	8	* as published by the Free Software Foundation; either version
	9	* 2 of the Licence, or (at your option) any later version.
	10	*/
	11
	12	#include <linux/sched.h>
	13	#include <linux/slab.h>
	14	#include "afs_fs.h"
	15	#include "internal.h"
	16	#include "protocol_yfs.h"
	17
	18	static bool afs_vl_probe_done(struct afs_vlserver *server)
	19	{
	20	if (!atomic_dec_and_test(&server->probe_outstanding))
	21	return false;
	22
	23	wake_up_var(&server->probe_outstanding);
	24	clear_bit_unlock(AFS_VLSERVER_FL_PROBING, &server->flags);
	25	wake_up_bit(&server->flags, AFS_VLSERVER_FL_PROBING);
	26	return true;
	27	}
	28
	29	/*
	30	* Process the result of probing a vlserver. This is called after successful
	31	* or failed delivery of an VL.GetCapabilities operation.
	32	*/
	33	void afs_vlserver_probe_result(struct afs_call *call)
	34	{
	35	struct afs_addr_list *alist = call->alist;
	36	struct afs_vlserver *server = call->reply[0];
	37	unsigned int server_index = (long)call->reply[1];
	38	unsigned int index = call->addr_ix;
	39	unsigned int rtt = UINT_MAX;
	40	bool have_result = false;
	41	u64 _rtt;
	42	int ret = call->error;
	43
	44	_enter("%s,%u,%u,%d,%d", server->name, server_index, index, ret, call->abort_code);
	45
	46	spin_lock(&server->probe_lock);
	47
	48	switch (ret) {
	49	case 0:
	50	server->probe.error = 0;
	51	goto responded;
	52	case -ECONNABORTED:
	53	if (!server->probe.responded) {
	54	server->probe.abort_code = call->abort_code;
	55	server->probe.error = ret;
	56	}
	57	goto responded;
	58	case -ENOMEM:
	59	case -ENONET:
	60	server->probe.local_failure = true;
	61	afs_io_error(call, afs_io_error_vl_probe_fail);
	62	goto out;
	63	case -ECONNRESET: /* Responded, but call expired. */
4584ae96 DH	64	case -ERFKILL:
4584ae96 DH	65	case -EADDRNOTAVAIL:
3bf0fb6f DH	66	case -ENETUNREACH:
3bf0fb6f DH	67	case -EHOSTUNREACH:
4584ae96	68	case -EHOSTDOWN:
3bf0fb6f DH	69	case -ECONNREFUSED:
	70	case -ETIMEDOUT:
	71	case -ETIME:
	72	default:
	73	clear_bit(index, &alist->responded);
	74	set_bit(index, &alist->failed);
	75	if (!server->probe.responded &&
	76	(server->probe.error == 0 \|\|
	77	server->probe.error == -ETIMEDOUT \|\|
	78	server->probe.error == -ETIME))
	79	server->probe.error = ret;
	80	afs_io_error(call, afs_io_error_vl_probe_fail);
	81	goto out;
	82	}
	83
	84	responded:
	85	set_bit(index, &alist->responded);
	86	clear_bit(index, &alist->failed);
	87
	88	if (call->service_id == YFS_VL_SERVICE) {
	89	server->probe.is_yfs = true;
	90	set_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
	91	alist->addrs[index].srx_service = call->service_id;
	92	} else {
	93	server->probe.not_yfs = true;
	94	if (!server->probe.is_yfs) {
	95	clear_bit(AFS_VLSERVER_FL_IS_YFS, &server->flags);
	96	alist->addrs[index].srx_service = call->service_id;
	97	}
	98	}
	99
	100	/* Get the RTT and scale it to fit into a 32-bit value that represents
	101	* over a minute of time so that we can access it with one instruction
	102	* on a 32-bit system.
	103	*/
	104	_rtt = rxrpc_kernel_get_rtt(call->net->socket, call->rxcall);
	105	_rtt /= 64;
	106	rtt = (_rtt > UINT_MAX) ? UINT_MAX : _rtt;
	107	if (rtt < server->probe.rtt) {
	108	server->probe.rtt = rtt;
	109	alist->preferred = index;
	110	have_result = true;
	111	}
	112
	113	smp_wmb(); /* Set rtt before responded. */
	114	server->probe.responded = true;
	115	set_bit(AFS_VLSERVER_FL_PROBED, &server->flags);
	116	out:
	117	spin_unlock(&server->probe_lock);
	118
	119	_debug("probe [%u][%u] %pISpc rtt=%u ret=%d",
	120	server_index, index, &alist->addrs[index].transport,
	121	(unsigned int)rtt, ret);
	122
	123	have_result \|= afs_vl_probe_done(server);
	124	if (have_result) {
	125	server->probe.have_result = true;
	126	wake_up_var(&server->probe.have_result);
	127	wake_up_all(&server->probe_wq);
	128	}
	129	}
	130
	131	/*
	132	* Probe all of a vlserver's addresses to find out the best route and to
133	* query its capabilities.
134	*/
4584ae96 DH	135	static bool afs_do_probe_vlserver(struct afs_net *net,
	136	struct afs_vlserver *server,
	137	struct key *key,
	138	unsigned int server_index,
	139	struct afs_error *_e)
3bf0fb6f DH	140	{
	141	struct afs_addr_cursor ac = {
	142	.index = 0,
	143	};
4584ae96 DH	144	bool in_progress = false;
4584ae96 DH	145	int err;
3bf0fb6f DH	146
	147	_enter("%s", server->name);
	148
	149	read_lock(&server->lock);
	150	ac.alist = rcu_dereference_protected(server->addresses,
	151	lockdep_is_held(&server->lock));
	152	read_unlock(&server->lock);
	153
	154	atomic_set(&server->probe_outstanding, ac.alist->nr_addrs);
	155	memset(&server->probe, 0, sizeof(server->probe));
	156	server->probe.rtt = UINT_MAX;
	157
	158	for (ac.index = 0; ac.index < ac.alist->nr_addrs; ac.index++) {
4584ae96	159	err = afs_vl_get_capabilities(net, &ac, key, server,
3bf0fb6f	160	server_index, true);
4584ae96 DH	161	if (err == -EINPROGRESS)
	162	in_progress = true;
	163	else
	164	afs_prioritise_error(_e, err, ac.abort_code);
3bf0fb6f DH	165	}
3bf0fb6f DH	166
4584ae96 DH	167	if (!in_progress)
	168	afs_vl_probe_done(server);
	169	return in_progress;
3bf0fb6f DH	170	}
	171
	172	/*
	173	* Send off probes to all unprobed servers.
	174	*/
	175	int afs_send_vl_probes(struct afs_net net, struct key key,
	176	struct afs_vlserver_list *vllist)
	177	{
	178	struct afs_vlserver *server;
4584ae96 DH	179	struct afs_error e;
	180	bool in_progress = false;
	181	int i;
3bf0fb6f	182
4584ae96 DH	183	e.error = 0;
4584ae96 DH	184	e.responded = false;
3bf0fb6f DH	185	for (i = 0; i < vllist->nr_servers; i++) {
	186	server = vllist->servers[i].server;
	187	if (test_bit(AFS_VLSERVER_FL_PROBED, &server->flags))
	188	continue;
	189
4584ae96 DH	190	if (!test_and_set_bit_lock(AFS_VLSERVER_FL_PROBING, &server->flags) &&
	191	afs_do_probe_vlserver(net, server, key, i, &e))
	192	in_progress = true;
3bf0fb6f DH	193	}
3bf0fb6f DH	194
4584ae96	195	return in_progress ? 0 : e.error;
3bf0fb6f DH	196	}
	197
	198	/*
	199	* Wait for the first as-yet untried server to respond.
	200	*/
	201	int afs_wait_for_vl_probes(struct afs_vlserver_list *vllist,
	202	unsigned long untried)
	203	{
	204	struct wait_queue_entry *waits;
	205	struct afs_vlserver *server;
	206	unsigned int rtt = UINT_MAX;
	207	bool have_responders = false;
	208	int pref = -1, i;
	209
	210	_enter("%u,%lx", vllist->nr_servers, untried);
	211
	212	/* Only wait for servers that have a probe outstanding. */
	213	for (i = 0; i < vllist->nr_servers; i++) {
	214	if (test_bit(i, &untried)) {
	215	server = vllist->servers[i].server;
	216	if (!test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
	217	__clear_bit(i, &untried);
	218	if (server->probe.responded)
	219	have_responders = true;
	220	}
	221	}
	222	if (have_responders \|\| !untried)
	223	return 0;
	224
	225	waits = kmalloc(array_size(vllist->nr_servers, sizeof(*waits)), GFP_KERNEL);
	226	if (!waits)
	227	return -ENOMEM;
	228
	229	for (i = 0; i < vllist->nr_servers; i++) {
	230	if (test_bit(i, &untried)) {
	231	server = vllist->servers[i].server;
	232	init_waitqueue_entry(&waits[i], current);
	233	add_wait_queue(&server->probe_wq, &waits[i]);
	234	}
	235	}
	236
	237	for (;;) {
	238	bool still_probing = false;
	239
	240	set_current_state(TASK_INTERRUPTIBLE);
	241	for (i = 0; i < vllist->nr_servers; i++) {
	242	if (test_bit(i, &untried)) {
	243	server = vllist->servers[i].server;
	244	if (server->probe.responded)
	245	goto stop;
	246	if (test_bit(AFS_VLSERVER_FL_PROBING, &server->flags))
	247	still_probing = true;
	248	}
	249	}
	250
08d405c8	251	if (!still_probing \|\| signal_pending(current))
3bf0fb6f DH	252	goto stop;
	253	schedule();
	254	}
	255
	256	stop:
	257	set_current_state(TASK_RUNNING);
	258
	259	for (i = 0; i < vllist->nr_servers; i++) {
	260	if (test_bit(i, &untried)) {
	261	server = vllist->servers[i].server;
	262	if (server->probe.responded &&
	263	server->probe.rtt < rtt) {
	264	pref = i;
	265	rtt = server->probe.rtt;
	266	}
	267
	268	remove_wait_queue(&server->probe_wq, &waits[i]);
	269	}
	270	}
	271
	272	kfree(waits);
	273
	274	if (pref == -1 && signal_pending(current))
	275	return -ERESTARTSYS;
	276
	277	if (pref >= 0)
	278	vllist->preferred = pref;
	279
	280	_leave(" = 0 [%u]", pref);
	281	return 0;
	282	}