1
0
Fork 0
alistair23-linux/fs/afs/fs_probe.c

271 lines
6.7 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-or-later
/* AFS fileserver probing
*
* Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*/
#include <linux/sched.h>
#include <linux/slab.h>
#include "afs_fs.h"
#include "internal.h"
#include "protocol_yfs.h"
static bool afs_fs_probe_done(struct afs_server *server)
{
if (!atomic_dec_and_test(&server->probe_outstanding))
return false;
wake_up_var(&server->probe_outstanding);
clear_bit_unlock(AFS_SERVER_FL_PROBING, &server->flags);
wake_up_bit(&server->flags, AFS_SERVER_FL_PROBING);
return true;
}
/*
* Process the result of probing a fileserver. This is called after successful
* or failed delivery of an FS.GetCapabilities operation.
*/
void afs_fileserver_probe_result(struct afs_call *call)
{
struct afs_addr_list *alist = call->alist;
struct afs_server *server = call->server;
unsigned int server_index = call->server_index;
unsigned int index = call->addr_ix;
rxrpc: Fix the excessive initial retransmission timeout commit c410bf01933e5e09d142c66c3df9ad470a7eec13 upstream. rxrpc currently uses a fixed 4s retransmission timeout until the RTT is sufficiently sampled. This can cause problems with some fileservers with calls to the cache manager in the afs filesystem being dropped from the fileserver because a packet goes missing and the retransmission timeout is greater than the call expiry timeout. Fix this by: (1) Copying the RTT/RTO calculation code from Linux's TCP implementation and altering it to fit rxrpc. (2) Altering the various users of the RTT to make use of the new SRTT value. (3) Replacing the use of rxrpc_resend_timeout to use the calculated RTO value instead (which is needed in jiffies), along with a backoff. Notes: (1) rxrpc provides RTT samples by matching the serial numbers on outgoing DATA packets that have the RXRPC_REQUEST_ACK set and PING ACK packets against the reference serial number in incoming REQUESTED ACK and PING-RESPONSE ACK packets. (2) Each packet that is transmitted on an rxrpc connection gets a new per-connection serial number, even for retransmissions, so an ACK can be cross-referenced to a specific trigger packet. This allows RTT information to be drawn from retransmitted DATA packets also. (3) rxrpc maintains the RTT/RTO state on the rxrpc_peer record rather than on an rxrpc_call because many RPC calls won't live long enough to generate more than one sample. (4) The calculated SRTT value is in units of 8ths of a microsecond rather than nanoseconds. The (S)RTT and RTO values are displayed in /proc/net/rxrpc/peers. Fixes: 17926a79320a ([AF_RXRPC]: Provide secure RxRPC sockets for use by userspace and kernel both"") Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2020-05-11 07:54:34 -06:00
unsigned int rtt_us;
bool have_result = false;
int ret = call->error;
_enter("%pU,%u", &server->uuid, index);
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_fs_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_fs_probe_fail);
goto out;
}
responded:
set_bit(index, &alist->responded);
clear_bit(index, &alist->failed);
if (call->service_id == YFS_FS_SERVICE) {
server->probe.is_yfs = true;
set_bit(AFS_SERVER_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_SERVER_FL_IS_YFS, &server->flags);
alist->addrs[index].srx_service = call->service_id;
}
}
if (rxrpc_kernel_get_srtt(call->net->socket, call->rxcall, &rtt_us) &&
rtt_us < server->probe.rtt) {
rxrpc: Fix the excessive initial retransmission timeout commit c410bf01933e5e09d142c66c3df9ad470a7eec13 upstream. rxrpc currently uses a fixed 4s retransmission timeout until the RTT is sufficiently sampled. This can cause problems with some fileservers with calls to the cache manager in the afs filesystem being dropped from the fileserver because a packet goes missing and the retransmission timeout is greater than the call expiry timeout. Fix this by: (1) Copying the RTT/RTO calculation code from Linux's TCP implementation and altering it to fit rxrpc. (2) Altering the various users of the RTT to make use of the new SRTT value. (3) Replacing the use of rxrpc_resend_timeout to use the calculated RTO value instead (which is needed in jiffies), along with a backoff. Notes: (1) rxrpc provides RTT samples by matching the serial numbers on outgoing DATA packets that have the RXRPC_REQUEST_ACK set and PING ACK packets against the reference serial number in incoming REQUESTED ACK and PING-RESPONSE ACK packets. (2) Each packet that is transmitted on an rxrpc connection gets a new per-connection serial number, even for retransmissions, so an ACK can be cross-referenced to a specific trigger packet. This allows RTT information to be drawn from retransmitted DATA packets also. (3) rxrpc maintains the RTT/RTO state on the rxrpc_peer record rather than on an rxrpc_call because many RPC calls won't live long enough to generate more than one sample. (4) The calculated SRTT value is in units of 8ths of a microsecond rather than nanoseconds. The (S)RTT and RTO values are displayed in /proc/net/rxrpc/peers. Fixes: 17926a79320a ([AF_RXRPC]: Provide secure RxRPC sockets for use by userspace and kernel both"") Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2020-05-11 07:54:34 -06:00
server->probe.rtt = rtt_us;
alist->preferred = index;
have_result = true;
}
smp_wmb(); /* Set rtt before responded. */
server->probe.responded = true;
set_bit(AFS_SERVER_FL_PROBED, &server->flags);
out:
spin_unlock(&server->probe_lock);
_debug("probe [%u][%u] %pISpc rtt=%u ret=%d",
rxrpc: Fix the excessive initial retransmission timeout commit c410bf01933e5e09d142c66c3df9ad470a7eec13 upstream. rxrpc currently uses a fixed 4s retransmission timeout until the RTT is sufficiently sampled. This can cause problems with some fileservers with calls to the cache manager in the afs filesystem being dropped from the fileserver because a packet goes missing and the retransmission timeout is greater than the call expiry timeout. Fix this by: (1) Copying the RTT/RTO calculation code from Linux's TCP implementation and altering it to fit rxrpc. (2) Altering the various users of the RTT to make use of the new SRTT value. (3) Replacing the use of rxrpc_resend_timeout to use the calculated RTO value instead (which is needed in jiffies), along with a backoff. Notes: (1) rxrpc provides RTT samples by matching the serial numbers on outgoing DATA packets that have the RXRPC_REQUEST_ACK set and PING ACK packets against the reference serial number in incoming REQUESTED ACK and PING-RESPONSE ACK packets. (2) Each packet that is transmitted on an rxrpc connection gets a new per-connection serial number, even for retransmissions, so an ACK can be cross-referenced to a specific trigger packet. This allows RTT information to be drawn from retransmitted DATA packets also. (3) rxrpc maintains the RTT/RTO state on the rxrpc_peer record rather than on an rxrpc_call because many RPC calls won't live long enough to generate more than one sample. (4) The calculated SRTT value is in units of 8ths of a microsecond rather than nanoseconds. The (S)RTT and RTO values are displayed in /proc/net/rxrpc/peers. Fixes: 17926a79320a ([AF_RXRPC]: Provide secure RxRPC sockets for use by userspace and kernel both"") Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2020-05-11 07:54:34 -06:00
server_index, index, &alist->addrs[index].transport, rtt_us, ret);
have_result |= afs_fs_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 fileserver's addresses to find out the best route and to
* query its capabilities.
*/
static int afs_do_probe_fileserver(struct afs_net *net,
struct afs_server *server,
struct key *key,
unsigned int server_index,
struct afs_error *_e)
{
struct afs_addr_cursor ac = {
.index = 0,
};
struct afs_call *call;
bool in_progress = false;
_enter("%pU", &server->uuid);
read_lock(&server->fs_lock);
ac.alist = rcu_dereference_protected(server->addresses,
lockdep_is_held(&server->fs_lock));
afs_get_addrlist(ac.alist);
read_unlock(&server->fs_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++) {
call = afs_fs_get_capabilities(net, server, &ac, key, server_index);
if (!IS_ERR(call)) {
afs_put_call(call);
in_progress = true;
} else {
afs_prioritise_error(_e, PTR_ERR(call), ac.abort_code);
}
}
if (!in_progress)
afs_fs_probe_done(server);
afs_put_addrlist(ac.alist);
return in_progress;
}
/*
* Send off probes to all unprobed servers.
*/
int afs_probe_fileservers(struct afs_net *net, struct key *key,
struct afs_server_list *list)
{
struct afs_server *server;
struct afs_error e;
bool in_progress = false;
int i;
e.error = 0;
e.responded = false;
for (i = 0; i < list->nr_servers; i++) {
server = list->servers[i].server;
if (test_bit(AFS_SERVER_FL_PROBED, &server->flags))
continue;
if (!test_and_set_bit_lock(AFS_SERVER_FL_PROBING, &server->flags) &&
afs_do_probe_fileserver(net, server, key, i, &e))
in_progress = true;
}
return in_progress ? 0 : e.error;
}
/*
* Wait for the first as-yet untried fileserver to respond.
*/
int afs_wait_for_fs_probes(struct afs_server_list *slist, unsigned long untried)
{
struct wait_queue_entry *waits;
struct afs_server *server;
unsigned int rtt = UINT_MAX;
bool have_responders = false;
int pref = -1, i;
_enter("%u,%lx", slist->nr_servers, untried);
/* Only wait for servers that have a probe outstanding. */
for (i = 0; i < slist->nr_servers; i++) {
if (test_bit(i, &untried)) {
server = slist->servers[i].server;
if (!test_bit(AFS_SERVER_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(slist->nr_servers, sizeof(*waits)), GFP_KERNEL);
if (!waits)
return -ENOMEM;
for (i = 0; i < slist->nr_servers; i++) {
if (test_bit(i, &untried)) {
server = slist->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 < slist->nr_servers; i++) {
if (test_bit(i, &untried)) {
server = slist->servers[i].server;
if (server->probe.responded)
goto stop;
if (test_bit(AFS_SERVER_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 < slist->nr_servers; i++) {
if (test_bit(i, &untried)) {
server = slist->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)
slist->preferred = pref;
return 0;
}