qemu/blockdev-nbd.c

353 lines
9.4 KiB
C
Raw Normal View History

/*
* Serving QEMU block devices via NBD
*
* Copyright (c) 2012 Red Hat, Inc.
*
* Author: Paolo Bonzini <pbonzini@redhat.com>
*
* This work is licensed under the terms of the GNU GPL, version 2 or
* later. See the COPYING file in the top-level directory.
*/
#include "qemu/osdep.h"
#include "sysemu/blockdev.h"
#include "sysemu/block-backend.h"
#include "hw/block/block.h"
#include "qapi/error.h"
#include "qapi/qapi-commands-block-export.h"
#include "block/nbd.h"
#include "io/channel-socket.h"
#include "io/net-listener.h"
typedef struct NBDServerData {
QIONetListener *listener;
QCryptoTLSCreds *tlscreds;
char *tlsauthz;
uint32_t max_connections;
uint32_t connections;
} NBDServerData;
static NBDServerData *nbd_server;
static bool is_qemu_nbd;
static void nbd_update_server_watch(NBDServerData *s);
void nbd_server_is_qemu_nbd(bool value)
{
is_qemu_nbd = value;
}
nbd: Fix regression on resiliency to port scan Back in qemu 2.5, qemu-nbd was immune to port probes (a transient server would not quit, regardless of how many probe connections came and went, until a connection actually negotiated). But we broke that in commit ee7d7aa when removing the return value to nbd_client_new(), although that patch also introduced a bug causing an assertion failure on a client that fails negotiation. We then made it worse during refactoring in commit 1a6245a (a segfault before we could even assert); the (masked) assertion was cleaned up in d3780c2 (still in 2.6), and just recently we finally fixed the segfault ("nbd: Fully intialize client in case of failed negotiation"). But that still means that ever since we added TLS support to qemu-nbd, we have been vulnerable to an ill-timed port-scan being able to cause a denial of service by taking down qemu-nbd before a real client has a chance to connect. Since negotiation is now handled asynchronously via coroutines, we no longer have a synchronous point of return by re-adding a return value to nbd_client_new(). So this patch instead wires things up to pass the negotiation status through the close_fn callback function. Simple test across two terminals: $ qemu-nbd -f raw -p 30001 file $ nmap 127.0.0.1 -p 30001 && \ qemu-io -c 'r 0 512' -f raw nbd://localhost:30001 Note that this patch does not change what constitutes successful negotiation (thus, a client must enter transmission phase before that client can be considered as a reason to terminate the server when the connection ends). Perhaps we may want to tweak things in a later patch to also treat a client that uses NBD_OPT_ABORT as being a 'successful' negotiation (the client correctly talked the NBD protocol, and informed us it was not going to use our export after all), but that's a discussion for another day. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1451614 Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20170608222617.20376-1-eblake@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-06-09 06:26:17 +08:00
static void nbd_blockdev_client_closed(NBDClient *client, bool ignored)
{
nbd_client_put(client);
assert(nbd_server->connections > 0);
nbd_server->connections--;
nbd_update_server_watch(nbd_server);
nbd: Fix regression on resiliency to port scan Back in qemu 2.5, qemu-nbd was immune to port probes (a transient server would not quit, regardless of how many probe connections came and went, until a connection actually negotiated). But we broke that in commit ee7d7aa when removing the return value to nbd_client_new(), although that patch also introduced a bug causing an assertion failure on a client that fails negotiation. We then made it worse during refactoring in commit 1a6245a (a segfault before we could even assert); the (masked) assertion was cleaned up in d3780c2 (still in 2.6), and just recently we finally fixed the segfault ("nbd: Fully intialize client in case of failed negotiation"). But that still means that ever since we added TLS support to qemu-nbd, we have been vulnerable to an ill-timed port-scan being able to cause a denial of service by taking down qemu-nbd before a real client has a chance to connect. Since negotiation is now handled asynchronously via coroutines, we no longer have a synchronous point of return by re-adding a return value to nbd_client_new(). So this patch instead wires things up to pass the negotiation status through the close_fn callback function. Simple test across two terminals: $ qemu-nbd -f raw -p 30001 file $ nmap 127.0.0.1 -p 30001 && \ qemu-io -c 'r 0 512' -f raw nbd://localhost:30001 Note that this patch does not change what constitutes successful negotiation (thus, a client must enter transmission phase before that client can be considered as a reason to terminate the server when the connection ends). Perhaps we may want to tweak things in a later patch to also treat a client that uses NBD_OPT_ABORT as being a 'successful' negotiation (the client correctly talked the NBD protocol, and informed us it was not going to use our export after all), but that's a discussion for another day. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1451614 Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20170608222617.20376-1-eblake@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-06-09 06:26:17 +08:00
}
static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc,
gpointer opaque)
{
nbd_server->connections++;
nbd_update_server_watch(nbd_server);
qio_channel_set_name(QIO_CHANNEL(cioc), "nbd-server");
nbd_client_new(cioc, nbd_server->tlscreds, nbd_server->tlsauthz,
nbd: Fix regression on resiliency to port scan Back in qemu 2.5, qemu-nbd was immune to port probes (a transient server would not quit, regardless of how many probe connections came and went, until a connection actually negotiated). But we broke that in commit ee7d7aa when removing the return value to nbd_client_new(), although that patch also introduced a bug causing an assertion failure on a client that fails negotiation. We then made it worse during refactoring in commit 1a6245a (a segfault before we could even assert); the (masked) assertion was cleaned up in d3780c2 (still in 2.6), and just recently we finally fixed the segfault ("nbd: Fully intialize client in case of failed negotiation"). But that still means that ever since we added TLS support to qemu-nbd, we have been vulnerable to an ill-timed port-scan being able to cause a denial of service by taking down qemu-nbd before a real client has a chance to connect. Since negotiation is now handled asynchronously via coroutines, we no longer have a synchronous point of return by re-adding a return value to nbd_client_new(). So this patch instead wires things up to pass the negotiation status through the close_fn callback function. Simple test across two terminals: $ qemu-nbd -f raw -p 30001 file $ nmap 127.0.0.1 -p 30001 && \ qemu-io -c 'r 0 512' -f raw nbd://localhost:30001 Note that this patch does not change what constitutes successful negotiation (thus, a client must enter transmission phase before that client can be considered as a reason to terminate the server when the connection ends). Perhaps we may want to tweak things in a later patch to also treat a client that uses NBD_OPT_ABORT as being a 'successful' negotiation (the client correctly talked the NBD protocol, and informed us it was not going to use our export after all), but that's a discussion for another day. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1451614 Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20170608222617.20376-1-eblake@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-06-09 06:26:17 +08:00
nbd_blockdev_client_closed);
}
static void nbd_update_server_watch(NBDServerData *s)
{
if (!s->max_connections || s->connections < s->max_connections) {
qio_net_listener_set_client_func(s->listener, nbd_accept, NULL, NULL);
} else {
qio_net_listener_set_client_func(s->listener, NULL, NULL, NULL);
}
}
static void nbd_server_free(NBDServerData *server)
{
if (!server) {
return;
}
qio_net_listener_disconnect(server->listener);
object_unref(OBJECT(server->listener));
if (server->tlscreds) {
object_unref(OBJECT(server->tlscreds));
}
g_free(server->tlsauthz);
g_free(server);
}
static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, Error **errp)
{
Object *obj;
QCryptoTLSCreds *creds;
obj = object_resolve_path_component(
object_get_objects_root(), id);
if (!obj) {
error_setg(errp, "No TLS credentials with id '%s'",
id);
return NULL;
}
creds = (QCryptoTLSCreds *)
object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS);
if (!creds) {
error_setg(errp, "Object with id '%s' is not TLS credentials",
id);
return NULL;
}
if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) {
error_setg(errp,
"Expecting TLS credentials with a server endpoint");
return NULL;
}
object_ref(obj);
return creds;
}
void nbd_server_start(SocketAddress *addr, const char *tls_creds,
const char *tls_authz, uint32_t max_connections,
Error **errp)
{
if (nbd_server) {
error_setg(errp, "NBD server already running");
return;
}
nbd_server = g_new0(NBDServerData, 1);
nbd_server->max_connections = max_connections;
nbd_server->listener = qio_net_listener_new();
qio_net_listener_set_name(nbd_server->listener,
"nbd-listener");
if (qio_net_listener_open_sync(nbd_server->listener, addr, 1, errp) < 0) {
goto error;
}
if (tls_creds) {
nbd_server->tlscreds = nbd_get_tls_creds(tls_creds, errp);
if (!nbd_server->tlscreds) {
goto error;
}
/* TODO SOCKET_ADDRESS_TYPE_FD where fd has AF_INET or AF_INET6 */
if (addr->type != SOCKET_ADDRESS_TYPE_INET) {
error_setg(errp, "TLS is only supported with IPv4/IPv6");
goto error;
}
}
nbd_server->tlsauthz = g_strdup(tls_authz);
nbd_update_server_watch(nbd_server);
return;
error:
nbd_server_free(nbd_server);
nbd_server = NULL;
}
void nbd_server_start_options(NbdServerOptions *arg, Error **errp)
{
nbd_server_start(arg->addr, arg->tls_creds, arg->tls_authz,
arg->max_connections, errp);
}
void qmp_nbd_server_start(SocketAddressLegacy *addr,
bool has_tls_creds, const char *tls_creds,
bool has_tls_authz, const char *tls_authz,
bool has_max_connections, uint32_t max_connections,
Error **errp)
{
SocketAddress *addr_flat = socket_address_flatten(addr);
nbd_server_start(addr_flat, tls_creds, tls_authz, max_connections, errp);
qapi_free_SocketAddress(addr_flat);
}
int nbd_export_create(BlockExport *exp, BlockExportOptions *exp_args,
Error **errp)
{
BlockExportOptionsNbd *arg = &exp_args->u.nbd;
BlockDriverState *bs = NULL;
AioContext *aio_context;
int ret;
assert(exp_args->type == BLOCK_EXPORT_TYPE_NBD);
if (!nbd_server && !is_qemu_nbd) {
error_setg(errp, "NBD server not running");
return -EINVAL;
}
if (!arg->has_name) {
arg->name = exp_args->node_name;
}
if (strlen(arg->name) > NBD_MAX_STRING_SIZE) {
error_setg(errp, "export name '%s' too long", arg->name);
return -EINVAL;
}
if (arg->description && strlen(arg->description) > NBD_MAX_STRING_SIZE) {
error_setg(errp, "description '%s' too long", arg->description);
return -EINVAL;
}
if (nbd_export_find(arg->name)) {
error_setg(errp, "NBD server already has export named '%s'", arg->name);
return -EEXIST;
}
bs = bdrv_lookup_bs(NULL, exp_args->node_name, errp);
if (!bs) {
return -ENOENT;
}
aio_context = bdrv_get_aio_context(bs);
aio_context_acquire(aio_context);
if (!arg->has_writable) {
arg->writable = false;
}
if (bdrv_is_read_only(bs) && arg->writable) {
ret = -EINVAL;
error_setg(errp, "Cannot export read-only node as writable");
goto out;
}
if (!exp_args->has_writethrough) {
exp_args->writethrough = false;
}
ret = nbd_export_new(exp, bs, arg->name, arg->description, arg->bitmap,
!arg->writable, !arg->writable,
exp_args->writethrough, errp);
if (ret < 0) {
goto out;
}
/* The list of named exports has a strong reference to this export now and
* our only way of accessing it is through nbd_export_find(), so we can drop
* the strong reference that is @exp. */
blk_exp_unref(exp);
ret = 0;
out:
aio_context_release(aio_context);
return ret;
}
void qmp_nbd_server_add(NbdServerAddOptions *arg, Error **errp)
{
BlockExport *export;
BlockDriverState *bs;
BlockBackend *on_eject_blk;
BlockExportOptions *export_opts;
bs = bdrv_lookup_bs(arg->device, arg->device, errp);
if (!bs) {
return;
}
/*
* block-export-add would default to the node-name, but we may have to use
* the device name as a default here for compatibility.
*/
if (!arg->has_name) {
arg->name = arg->device;
}
export_opts = g_new(BlockExportOptions, 1);
*export_opts = (BlockExportOptions) {
.type = BLOCK_EXPORT_TYPE_NBD,
.node_name = g_strdup(bdrv_get_node_name(bs)),
.u.nbd = {
.has_name = true,
.name = g_strdup(arg->name),
.has_description = arg->has_description,
.description = g_strdup(arg->description),
.has_writable = arg->has_writable,
.writable = arg->writable,
.has_bitmap = arg->has_bitmap,
.bitmap = g_strdup(arg->bitmap),
},
};
/*
* nbd-server-add doesn't complain when a read-only device should be
* exported as writable, but simply downgrades it. This is an error with
* block-export-add.
*/
if (bdrv_is_read_only(bs)) {
export_opts->u.nbd.has_writable = true;
export_opts->u.nbd.writable = false;
}
export = blk_exp_add(export_opts, errp);
if (!export) {
goto fail;
}
/*
* nbd-server-add removes the export when the named BlockBackend used for
* @device goes away.
*/
on_eject_blk = blk_by_name(arg->device);
if (on_eject_blk) {
nbd_export_set_on_eject_blk(export, on_eject_blk);
}
fail:
qapi_free_BlockExportOptions(export_opts);
}
void qmp_nbd_server_remove(const char *name,
bool has_mode, NbdServerRemoveMode mode,
Error **errp)
{
NBDExport *exp;
AioContext *aio_context;
if (!nbd_server) {
error_setg(errp, "NBD server not running");
return;
}
exp = nbd_export_find(name);
if (exp == NULL) {
error_setg(errp, "Export '%s' is not found", name);
return;
}
if (!has_mode) {
mode = NBD_SERVER_REMOVE_MODE_SAFE;
}
aio_context = nbd_export_aio_context(exp);
aio_context_acquire(aio_context);
nbd_export_remove(exp, mode, errp);
aio_context_release(aio_context);
}
void qmp_nbd_server_stop(Error **errp)
{
if (!nbd_server) {
error_setg(errp, "NBD server not running");
return;
}
nbd_export_close_all();
nbd_server_free(nbd_server);
nbd_server = NULL;
}