xemu/nbd/server.c

1308 lines
36 KiB
C
Raw Normal View History

/*
* Copyright (C) 2016 Red Hat, Inc.
* Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
*
* Network Block Device Server Side
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; under version 2 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include "qemu/osdep.h"
2016-03-14 08:01:28 +00:00
#include "qapi/error.h"
#include "nbd-internal.h"
qemu-nbd: only send a limited number of errno codes on the wire Right now, NBD includes potentially platform-specific error values in the wire protocol. Luckily, most common error values are more or less universal: in particular, of all errno values <= 34 (up to ERANGE), they are all the same on supported platforms except for 11 (which is EAGAIN on Windows and Linux, but EDEADLK on Darwin and the *BSDs). So, in order to guarantee some portability, only keep a handful of possible error codes and squash everything else to EINVAL. This patch defines a limited set of errno values that are valid for the NBD protocol, and specifies recommendations for what error to return in specific corner cases. The set of errno values is roughly based on the errors listed in the read(2) and write(2) man pages, with some exceptions: - ENOMEM is added for servers that implement copy-on-write or other formats that require dynamic allocation. - EDQUOT is not part of the universal set of errors; it can be changed to ENOSPC on the wire format. - EFBIG is part of the universal set of errors, but it is also changed to ENOSPC because it is pretty similar to ENOSPC or EDQUOT. Incoming values will in general match system errno values, but not on the Hurd which has different errno values (they have a "subsystem code" equal to 0x10 in bits 24-31). The Hurd is probably not something to which QEMU has been ported, but still do the right thing and reverse-map the NBD errno values to the system errno values. The corresponding patch to the NBD protocol description can be found at http://article.gmane.org/gmane.linux.drivers.nbd.general/3154. Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2015-05-07 15:25:10 +00:00
static int system_errno_to_nbd_errno(int err)
{
switch (err) {
case 0:
return NBD_SUCCESS;
case EPERM:
case EROFS:
qemu-nbd: only send a limited number of errno codes on the wire Right now, NBD includes potentially platform-specific error values in the wire protocol. Luckily, most common error values are more or less universal: in particular, of all errno values <= 34 (up to ERANGE), they are all the same on supported platforms except for 11 (which is EAGAIN on Windows and Linux, but EDEADLK on Darwin and the *BSDs). So, in order to guarantee some portability, only keep a handful of possible error codes and squash everything else to EINVAL. This patch defines a limited set of errno values that are valid for the NBD protocol, and specifies recommendations for what error to return in specific corner cases. The set of errno values is roughly based on the errors listed in the read(2) and write(2) man pages, with some exceptions: - ENOMEM is added for servers that implement copy-on-write or other formats that require dynamic allocation. - EDQUOT is not part of the universal set of errors; it can be changed to ENOSPC on the wire format. - EFBIG is part of the universal set of errors, but it is also changed to ENOSPC because it is pretty similar to ENOSPC or EDQUOT. Incoming values will in general match system errno values, but not on the Hurd which has different errno values (they have a "subsystem code" equal to 0x10 in bits 24-31). The Hurd is probably not something to which QEMU has been ported, but still do the right thing and reverse-map the NBD errno values to the system errno values. The corresponding patch to the NBD protocol description can be found at http://article.gmane.org/gmane.linux.drivers.nbd.general/3154. Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2015-05-07 15:25:10 +00:00
return NBD_EPERM;
case EIO:
return NBD_EIO;
case ENOMEM:
return NBD_ENOMEM;
#ifdef EDQUOT
case EDQUOT:
#endif
case EFBIG:
case ENOSPC:
return NBD_ENOSPC;
case ESHUTDOWN:
return NBD_ESHUTDOWN;
qemu-nbd: only send a limited number of errno codes on the wire Right now, NBD includes potentially platform-specific error values in the wire protocol. Luckily, most common error values are more or less universal: in particular, of all errno values <= 34 (up to ERANGE), they are all the same on supported platforms except for 11 (which is EAGAIN on Windows and Linux, but EDEADLK on Darwin and the *BSDs). So, in order to guarantee some portability, only keep a handful of possible error codes and squash everything else to EINVAL. This patch defines a limited set of errno values that are valid for the NBD protocol, and specifies recommendations for what error to return in specific corner cases. The set of errno values is roughly based on the errors listed in the read(2) and write(2) man pages, with some exceptions: - ENOMEM is added for servers that implement copy-on-write or other formats that require dynamic allocation. - EDQUOT is not part of the universal set of errors; it can be changed to ENOSPC on the wire format. - EFBIG is part of the universal set of errors, but it is also changed to ENOSPC because it is pretty similar to ENOSPC or EDQUOT. Incoming values will in general match system errno values, but not on the Hurd which has different errno values (they have a "subsystem code" equal to 0x10 in bits 24-31). The Hurd is probably not something to which QEMU has been ported, but still do the right thing and reverse-map the NBD errno values to the system errno values. The corresponding patch to the NBD protocol description can be found at http://article.gmane.org/gmane.linux.drivers.nbd.general/3154. Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2015-05-07 15:25:10 +00:00
case EINVAL:
default:
return NBD_EINVAL;
}
}
/* Definitions for opaque data types */
typedef struct NBDRequestData NBDRequestData;
struct NBDRequestData {
QSIMPLEQ_ENTRY(NBDRequestData) entry;
NBDClient *client;
uint8_t *data;
bool complete;
};
struct NBDExport {
int refcount;
void (*close)(NBDExport *exp);
BlockBackend *blk;
char *name;
char *description;
off_t dev_offset;
off_t size;
uint16_t nbdflags;
QTAILQ_HEAD(, NBDClient) clients;
QTAILQ_ENTRY(NBDExport) next;
AioContext *ctx;
BlockBackend *eject_notifier_blk;
Notifier eject_notifier;
};
static QTAILQ_HEAD(, NBDExport) exports = QTAILQ_HEAD_INITIALIZER(exports);
struct NBDClient {
int refcount;
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>
2017-06-08 22:26:17 +00:00
void (*close_fn)(NBDClient *client, bool negotiated);
bool no_zeroes;
NBDExport *exp;
QCryptoTLSCreds *tlscreds;
char *tlsaclname;
QIOChannelSocket *sioc; /* The underlying data channel */
QIOChannel *ioc; /* The current I/O channel which may differ (eg TLS) */
Coroutine *recv_coroutine;
CoMutex send_lock;
Coroutine *send_coroutine;
QTAILQ_ENTRY(NBDClient) next;
int nb_requests;
bool closing;
};
/* That's all folks */
static void nbd_client_receive_next_request(NBDClient *client);
/* Basic flow for negotiation
Server Client
Negotiate
or
Server Client
Negotiate #1
Option
Negotiate #2
----
followed by
Server Client
Request
Response
Request
Response
...
...
Request (type == 2)
*/
/* Send a reply header, including length, but no payload.
* Return -errno on error, 0 on success. */
static int nbd_negotiate_send_rep_len(QIOChannel *ioc, uint32_t type,
uint32_t opt, uint32_t len)
{
uint64_t magic;
TRACE("Reply opt=%" PRIx32 " type=%" PRIx32 " len=%" PRIu32,
type, opt, len);
magic = cpu_to_be64(NBD_REP_MAGIC);
if (nbd_write(ioc, &magic, sizeof(magic), NULL) < 0) {
LOG("write failed (rep magic)");
return -EINVAL;
}
opt = cpu_to_be32(opt);
if (nbd_write(ioc, &opt, sizeof(opt), NULL) < 0) {
LOG("write failed (rep opt)");
return -EINVAL;
}
type = cpu_to_be32(type);
if (nbd_write(ioc, &type, sizeof(type), NULL) < 0) {
LOG("write failed (rep type)");
return -EINVAL;
}
len = cpu_to_be32(len);
if (nbd_write(ioc, &len, sizeof(len), NULL) < 0) {
LOG("write failed (rep data length)");
return -EINVAL;
}
return 0;
}
/* Send a reply header with default 0 length.
* Return -errno on error, 0 on success. */
static int nbd_negotiate_send_rep(QIOChannel *ioc, uint32_t type, uint32_t opt)
{
return nbd_negotiate_send_rep_len(ioc, type, opt, 0);
}
/* Send an error reply.
* Return -errno on error, 0 on success. */
static int GCC_FMT_ATTR(4, 5)
nbd_negotiate_send_rep_err(QIOChannel *ioc, uint32_t type,
uint32_t opt, const char *fmt, ...)
{
va_list va;
char *msg;
int ret;
size_t len;
va_start(va, fmt);
msg = g_strdup_vprintf(fmt, va);
va_end(va);
len = strlen(msg);
assert(len < 4096);
TRACE("sending error message \"%s\"", msg);
ret = nbd_negotiate_send_rep_len(ioc, type, opt, len);
if (ret < 0) {
goto out;
}
if (nbd_write(ioc, msg, len, NULL) < 0) {
LOG("write failed (error message)");
ret = -EIO;
} else {
ret = 0;
}
out:
g_free(msg);
return ret;
}
/* Send a single NBD_REP_SERVER reply to NBD_OPT_LIST, including payload.
* Return -errno on error, 0 on success. */
static int nbd_negotiate_send_rep_list(QIOChannel *ioc, NBDExport *exp)
{
size_t name_len, desc_len;
uint32_t len;
const char *name = exp->name ? exp->name : "";
const char *desc = exp->description ? exp->description : "";
int ret;
TRACE("Advertising export name '%s' description '%s'", name, desc);
name_len = strlen(name);
desc_len = strlen(desc);
len = name_len + desc_len + sizeof(len);
ret = nbd_negotiate_send_rep_len(ioc, NBD_REP_SERVER, NBD_OPT_LIST, len);
if (ret < 0) {
return ret;
}
len = cpu_to_be32(name_len);
if (nbd_write(ioc, &len, sizeof(len), NULL) < 0) {
LOG("write failed (name length)");
return -EINVAL;
}
if (nbd_write(ioc, name, name_len, NULL) < 0) {
LOG("write failed (name buffer)");
return -EINVAL;
}
if (nbd_write(ioc, desc, desc_len, NULL) < 0) {
LOG("write failed (description buffer)");
return -EINVAL;
}
return 0;
}
/* Process the NBD_OPT_LIST command, with a potential series of replies.
* Return -errno on error, 0 on success. */
static int nbd_negotiate_handle_list(NBDClient *client, uint32_t length)
{
NBDExport *exp;
if (length) {
if (nbd_drop(client->ioc, length, NULL) < 0) {
return -EIO;
}
return nbd_negotiate_send_rep_err(client->ioc,
NBD_REP_ERR_INVALID, NBD_OPT_LIST,
"OPT_LIST should not have length");
}
/* For each export, send a NBD_REP_SERVER reply. */
QTAILQ_FOREACH(exp, &exports, next) {
if (nbd_negotiate_send_rep_list(client->ioc, exp)) {
return -EINVAL;
}
}
/* Finish with a NBD_REP_ACK. */
return nbd_negotiate_send_rep(client->ioc, NBD_REP_ACK, NBD_OPT_LIST);
}
static int nbd_negotiate_handle_export_name(NBDClient *client, uint32_t length)
{
char name[NBD_MAX_NAME_SIZE + 1];
/* Client sends:
[20 .. xx] export name (length bytes)
*/
TRACE("Checking length");
if (length >= sizeof(name)) {
LOG("Bad length received");
return -EINVAL;
}
if (nbd_read(client->ioc, name, length, NULL) < 0) {
LOG("read failed");
return -EINVAL;
}
name[length] = '\0';
TRACE("Client requested export '%s'", name);
client->exp = nbd_export_find(name);
if (!client->exp) {
LOG("export not found");
return -EINVAL;
}
QTAILQ_INSERT_TAIL(&client->exp->clients, client, next);
nbd_export_get(client->exp);
return 0;
}
/* Handle NBD_OPT_STARTTLS. Return NULL to drop connection, or else the
* new channel for all further (now-encrypted) communication. */
static QIOChannel *nbd_negotiate_handle_starttls(NBDClient *client,
uint32_t length)
{
QIOChannel *ioc;
QIOChannelTLS *tioc;
struct NBDTLSHandshakeData data = { 0 };
TRACE("Setting up TLS");
ioc = client->ioc;
if (length) {
if (nbd_drop(ioc, length, NULL) < 0) {
return NULL;
}
nbd_negotiate_send_rep_err(ioc, NBD_REP_ERR_INVALID, NBD_OPT_STARTTLS,
"OPT_STARTTLS should not have length");
return NULL;
}
if (nbd_negotiate_send_rep(client->ioc, NBD_REP_ACK,
NBD_OPT_STARTTLS) < 0) {
return NULL;
}
tioc = qio_channel_tls_new_server(ioc,
client->tlscreds,
client->tlsaclname,
NULL);
if (!tioc) {
return NULL;
}
qio_channel_set_name(QIO_CHANNEL(tioc), "nbd-server-tls");
TRACE("Starting TLS handshake");
data.loop = g_main_loop_new(g_main_context_default(), FALSE);
qio_channel_tls_handshake(tioc,
nbd_tls_handshake,
&data,
NULL);
if (!data.complete) {
g_main_loop_run(data.loop);
}
g_main_loop_unref(data.loop);
if (data.error) {
object_unref(OBJECT(tioc));
error_free(data.error);
return NULL;
}
return QIO_CHANNEL(tioc);
}
/* nbd_negotiate_options
* Process all NBD_OPT_* client option commands.
* Return:
* -errno on error
* 0 on successful negotiation
* 1 if client sent NBD_OPT_ABORT, i.e. on valid disconnect
*/
static int nbd_negotiate_options(NBDClient *client)
{
uint32_t flags;
bool fixedNewstyle = false;
/* Client sends:
[ 0 .. 3] client flags
[ 0 .. 7] NBD_OPTS_MAGIC
[ 8 .. 11] NBD option
[12 .. 15] Data length
... Rest of request
[ 0 .. 7] NBD_OPTS_MAGIC
[ 8 .. 11] Second NBD option
[12 .. 15] Data length
... Rest of request
*/
if (nbd_read(client->ioc, &flags, sizeof(flags), NULL) < 0) {
LOG("read failed");
return -EIO;
}
TRACE("Checking client flags");
be32_to_cpus(&flags);
if (flags & NBD_FLAG_C_FIXED_NEWSTYLE) {
TRACE("Client supports fixed newstyle handshake");
fixedNewstyle = true;
flags &= ~NBD_FLAG_C_FIXED_NEWSTYLE;
}
if (flags & NBD_FLAG_C_NO_ZEROES) {
TRACE("Client supports no zeroes at handshake end");
client->no_zeroes = true;
flags &= ~NBD_FLAG_C_NO_ZEROES;
}
if (flags != 0) {
TRACE("Unknown client flags 0x%" PRIx32 " received", flags);
return -EIO;
}
while (1) {
int ret;
uint32_t clientflags, length;
uint64_t magic;
if (nbd_read(client->ioc, &magic, sizeof(magic), NULL) < 0) {
LOG("read failed");
return -EINVAL;
}
TRACE("Checking opts magic");
if (magic != be64_to_cpu(NBD_OPTS_MAGIC)) {
LOG("Bad magic received");
return -EINVAL;
}
if (nbd_read(client->ioc, &clientflags,
sizeof(clientflags), NULL) < 0)
nbd: read_sync and friends: return 0 on success functions read_sync, drop_sync, write_sync, and also nbd_negotiate_write, nbd_negotiate_read, nbd_negotiate_drop_sync returns number of processed bytes. But what this number can be, except requested number of bytes? Actually, underlying nbd_wr_syncv function returns a value >= 0 and != requested_bytes only on eof on read operation. So, firstly, it is impossible on write (let's add an assert) and on read it actually means, that communication is broken (except nbd_receive_reply, see below). Most of callers operate like this: if (func(..., size) != size) { /* error path */ } , i.e.: 1. They are not interested in partial success 2. Extra duplications in code (especially bad are duplications of magic numbers) 3. User doesn't see actual error message, as return code is lost. (this patch doesn't fix this point, but it makes fixing easier) Several callers handles ret >= 0 and != requested-size separately, by just returning EINVAL in this case. This patch makes read_sync and friends return EINVAL in this case, so final behavior is the same. And only one caller - nbd_receive_reply() does something not so obvious. It returns EINVAL for ret > 0 and != requested-size, like previous group, but for ret == 0 it returns 0. The only caller of nbd_receive_reply() - nbd_read_reply_entry() handles ret == 0 in the same way as ret < 0, so for now it doesn't matter. However, in following commits error path handling will be improved and we'll need to distinguish success from fail in this case too. So, this patch adds separate helper for this case - read_sync_eof. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Message-Id: <20170516094533.6160-3-vsementsov@virtuozzo.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-05-16 09:45:30 +00:00
{
LOG("read failed");
return -EINVAL;
}
clientflags = be32_to_cpu(clientflags);
if (nbd_read(client->ioc, &length, sizeof(length), NULL) < 0) {
LOG("read failed");
return -EINVAL;
}
length = be32_to_cpu(length);
TRACE("Checking option 0x%" PRIx32, clientflags);
if (client->tlscreds &&
client->ioc == (QIOChannel *)client->sioc) {
QIOChannel *tioc;
if (!fixedNewstyle) {
TRACE("Unsupported option 0x%" PRIx32, clientflags);
return -EINVAL;
}
switch (clientflags) {
case NBD_OPT_STARTTLS:
tioc = nbd_negotiate_handle_starttls(client, length);
if (!tioc) {
return -EIO;
}
object_unref(OBJECT(client->ioc));
client->ioc = QIO_CHANNEL(tioc);
break;
case NBD_OPT_EXPORT_NAME:
/* No way to return an error to client, so drop connection */
TRACE("Option 0x%x not permitted before TLS", clientflags);
return -EINVAL;
default:
if (nbd_drop(client->ioc, length, NULL) < 0) {
return -EIO;
}
ret = nbd_negotiate_send_rep_err(client->ioc,
NBD_REP_ERR_TLS_REQD,
clientflags,
"Option 0x%" PRIx32
"not permitted before TLS",
clientflags);
if (ret < 0) {
return ret;
}
/* Let the client keep trying, unless they asked to quit */
if (clientflags == NBD_OPT_ABORT) {
return 1;
}
break;
}
} else if (fixedNewstyle) {
switch (clientflags) {
case NBD_OPT_LIST:
ret = nbd_negotiate_handle_list(client, length);
if (ret < 0) {
return ret;
}
break;
case NBD_OPT_ABORT:
/* NBD spec says we must try to reply before
* disconnecting, but that we must also tolerate
* guests that don't wait for our reply. */
nbd_negotiate_send_rep(client->ioc, NBD_REP_ACK, clientflags);
return 1;
case NBD_OPT_EXPORT_NAME:
return nbd_negotiate_handle_export_name(client, length);
case NBD_OPT_STARTTLS:
if (nbd_drop(client->ioc, length, NULL) < 0) {
return -EIO;
}
if (client->tlscreds) {
ret = nbd_negotiate_send_rep_err(client->ioc,
NBD_REP_ERR_INVALID,
clientflags,
"TLS already enabled");
} else {
ret = nbd_negotiate_send_rep_err(client->ioc,
NBD_REP_ERR_POLICY,
clientflags,
"TLS not configured");
}
if (ret < 0) {
return ret;
}
break;
default:
if (nbd_drop(client->ioc, length, NULL) < 0) {
return -EIO;
}
ret = nbd_negotiate_send_rep_err(client->ioc,
NBD_REP_ERR_UNSUP,
clientflags,
"Unsupported option 0x%"
PRIx32,
clientflags);
if (ret < 0) {
return ret;
}
break;
}
} else {
/*
* If broken new-style we should drop the connection
* for anything except NBD_OPT_EXPORT_NAME
*/
switch (clientflags) {
case NBD_OPT_EXPORT_NAME:
return nbd_negotiate_handle_export_name(client, length);
default:
TRACE("Unsupported option 0x%" PRIx32, clientflags);
return -EINVAL;
}
}
}
}
/* nbd_negotiate
* Return:
* -errno on error
* 0 on successful negotiation
* 1 if client sent NBD_OPT_ABORT, i.e. on valid disconnect
*/
static coroutine_fn int nbd_negotiate(NBDClient *client)
{
char buf[8 + 8 + 8 + 128];
int ret;
const uint16_t myflags = (NBD_FLAG_HAS_FLAGS | NBD_FLAG_SEND_TRIM |
NBD_FLAG_SEND_FLUSH | NBD_FLAG_SEND_FUA |
NBD_FLAG_SEND_WRITE_ZEROES);
bool oldStyle;
size_t len;
/* Old style negotiation header without options
[ 0 .. 7] passwd ("NBDMAGIC")
[ 8 .. 15] magic (NBD_CLIENT_MAGIC)
[16 .. 23] size
[24 .. 25] server flags (0)
[26 .. 27] export flags
[28 .. 151] reserved (0)
New style negotiation header with options
[ 0 .. 7] passwd ("NBDMAGIC")
[ 8 .. 15] magic (NBD_OPTS_MAGIC)
[16 .. 17] server flags (0)
....options sent....
[18 .. 25] size
[26 .. 27] export flags
[28 .. 151] reserved (0, omit if no_zeroes)
*/
qio_channel_set_blocking(client->ioc, false, NULL);
TRACE("Beginning negotiation.");
memset(buf, 0, sizeof(buf));
memcpy(buf, "NBDMAGIC", 8);
oldStyle = client->exp != NULL && !client->tlscreds;
if (oldStyle) {
TRACE("advertising size %" PRIu64 " and flags %x",
client->exp->size, client->exp->nbdflags | myflags);
stq_be_p(buf + 8, NBD_CLIENT_MAGIC);
stq_be_p(buf + 16, client->exp->size);
stw_be_p(buf + 26, client->exp->nbdflags | myflags);
} else {
stq_be_p(buf + 8, NBD_OPTS_MAGIC);
stw_be_p(buf + 16, NBD_FLAG_FIXED_NEWSTYLE | NBD_FLAG_NO_ZEROES);
}
if (oldStyle) {
if (client->tlscreds) {
TRACE("TLS cannot be enabled with oldstyle protocol");
return -EINVAL;
}
if (nbd_write(client->ioc, buf, sizeof(buf), NULL) < 0) {
LOG("write failed");
return -EINVAL;
}
} else {
if (nbd_write(client->ioc, buf, 18, NULL) < 0) {
LOG("write failed");
return -EINVAL;
}
ret = nbd_negotiate_options(client);
if (ret != 0) {
LOG("option negotiation failed");
return ret;
}
TRACE("advertising size %" PRIu64 " and flags %x",
client->exp->size, client->exp->nbdflags | myflags);
stq_be_p(buf + 18, client->exp->size);
stw_be_p(buf + 26, client->exp->nbdflags | myflags);
len = client->no_zeroes ? 10 : sizeof(buf) - 18;
ret = nbd_write(client->ioc, buf + 18, len, NULL);
if (ret < 0) {
LOG("write failed");
return ret;
}
}
TRACE("Negotiation succeeded.");
return 0;
}
static int nbd_receive_request(QIOChannel *ioc, NBDRequest *request)
{
uint8_t buf[NBD_REQUEST_SIZE];
uint32_t magic;
int ret;
ret = nbd_read(ioc, buf, sizeof(buf), NULL);
if (ret < 0) {
return ret;
}
/* Request
[ 0 .. 3] magic (NBD_REQUEST_MAGIC)
[ 4 .. 5] flags (NBD_CMD_FLAG_FUA, ...)
[ 6 .. 7] type (NBD_CMD_READ, ...)
[ 8 .. 15] handle
[16 .. 23] from
[24 .. 27] len
*/
magic = ldl_be_p(buf);
request->flags = lduw_be_p(buf + 4);
request->type = lduw_be_p(buf + 6);
request->handle = ldq_be_p(buf + 8);
request->from = ldq_be_p(buf + 16);
request->len = ldl_be_p(buf + 24);
TRACE("Got request: { magic = 0x%" PRIx32 ", .flags = %" PRIx16
", .type = %" PRIx16 ", from = %" PRIu64 ", len = %" PRIu32 " }",
magic, request->flags, request->type, request->from, request->len);
if (magic != NBD_REQUEST_MAGIC) {
LOG("invalid magic (got 0x%" PRIx32 ")", magic);
return -EINVAL;
}
return 0;
}
static int nbd_send_reply(QIOChannel *ioc, NBDReply *reply)
{
uint8_t buf[NBD_REPLY_SIZE];
qemu-nbd: only send a limited number of errno codes on the wire Right now, NBD includes potentially platform-specific error values in the wire protocol. Luckily, most common error values are more or less universal: in particular, of all errno values <= 34 (up to ERANGE), they are all the same on supported platforms except for 11 (which is EAGAIN on Windows and Linux, but EDEADLK on Darwin and the *BSDs). So, in order to guarantee some portability, only keep a handful of possible error codes and squash everything else to EINVAL. This patch defines a limited set of errno values that are valid for the NBD protocol, and specifies recommendations for what error to return in specific corner cases. The set of errno values is roughly based on the errors listed in the read(2) and write(2) man pages, with some exceptions: - ENOMEM is added for servers that implement copy-on-write or other formats that require dynamic allocation. - EDQUOT is not part of the universal set of errors; it can be changed to ENOSPC on the wire format. - EFBIG is part of the universal set of errors, but it is also changed to ENOSPC because it is pretty similar to ENOSPC or EDQUOT. Incoming values will in general match system errno values, but not on the Hurd which has different errno values (they have a "subsystem code" equal to 0x10 in bits 24-31). The Hurd is probably not something to which QEMU has been ported, but still do the right thing and reverse-map the NBD errno values to the system errno values. The corresponding patch to the NBD protocol description can be found at http://article.gmane.org/gmane.linux.drivers.nbd.general/3154. Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2015-05-07 15:25:10 +00:00
reply->error = system_errno_to_nbd_errno(reply->error);
TRACE("Sending response to client: { .error = %" PRId32
", handle = %" PRIu64 " }",
reply->error, reply->handle);
/* Reply
[ 0 .. 3] magic (NBD_REPLY_MAGIC)
[ 4 .. 7] error (0 == no error)
[ 7 .. 15] handle
*/
stl_be_p(buf, NBD_REPLY_MAGIC);
stl_be_p(buf + 4, reply->error);
stq_be_p(buf + 8, reply->handle);
return nbd_write(ioc, buf, sizeof(buf), NULL);
}
#define MAX_NBD_REQUESTS 16
void nbd_client_get(NBDClient *client)
{
client->refcount++;
}
void nbd_client_put(NBDClient *client)
{
if (--client->refcount == 0) {
/* The last reference should be dropped by client->close,
* which is called by client_close.
*/
assert(client->closing);
qio_channel_detach_aio_context(client->ioc);
object_unref(OBJECT(client->sioc));
object_unref(OBJECT(client->ioc));
if (client->tlscreds) {
object_unref(OBJECT(client->tlscreds));
}
g_free(client->tlsaclname);
if (client->exp) {
QTAILQ_REMOVE(&client->exp->clients, client, next);
nbd_export_put(client->exp);
}
g_free(client);
}
}
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>
2017-06-08 22:26:17 +00:00
static void client_close(NBDClient *client, bool negotiated)
{
if (client->closing) {
return;
}
client->closing = true;
/* Force requests to finish. They will drop their own references,
* then we'll close the socket and free the NBDClient.
*/
qio_channel_shutdown(client->ioc, QIO_CHANNEL_SHUTDOWN_BOTH,
NULL);
/* Also tell the client, so that they release their reference. */
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>
2017-06-08 22:26:17 +00:00
if (client->close_fn) {
client->close_fn(client, negotiated);
}
}
static NBDRequestData *nbd_request_get(NBDClient *client)
{
NBDRequestData *req;
assert(client->nb_requests <= MAX_NBD_REQUESTS - 1);
client->nb_requests++;
req = g_new0(NBDRequestData, 1);
nbd_client_get(client);
req->client = client;
return req;
}
static void nbd_request_put(NBDRequestData *req)
{
NBDClient *client = req->client;
if (req->data) {
qemu_vfree(req->data);
}
g_free(req);
client->nb_requests--;
nbd_client_receive_next_request(client);
nbd_client_put(client);
}
static void blk_aio_attached(AioContext *ctx, void *opaque)
{
NBDExport *exp = opaque;
NBDClient *client;
TRACE("Export %s: Attaching clients to AIO context %p\n", exp->name, ctx);
exp->ctx = ctx;
QTAILQ_FOREACH(client, &exp->clients, next) {
qio_channel_attach_aio_context(client->ioc, ctx);
if (client->recv_coroutine) {
aio_co_schedule(ctx, client->recv_coroutine);
}
if (client->send_coroutine) {
aio_co_schedule(ctx, client->send_coroutine);
}
}
}
static void blk_aio_detach(void *opaque)
{
NBDExport *exp = opaque;
NBDClient *client;
TRACE("Export %s: Detaching clients from AIO context %p\n", exp->name, exp->ctx);
QTAILQ_FOREACH(client, &exp->clients, next) {
qio_channel_detach_aio_context(client->ioc);
}
exp->ctx = NULL;
}
static void nbd_eject_notifier(Notifier *n, void *data)
{
NBDExport *exp = container_of(n, NBDExport, eject_notifier);
nbd_export_close(exp);
}
NBDExport *nbd_export_new(BlockDriverState *bs, off_t dev_offset, off_t size,
uint16_t nbdflags, void (*close)(NBDExport *),
bool writethrough, BlockBackend *on_eject_blk,
Error **errp)
{
BlockBackend *blk;
NBDExport *exp = g_malloc0(sizeof(NBDExport));
uint64_t perm;
int ret;
/* Don't allow resize while the NBD server is running, otherwise we don't
* care what happens with the node. */
perm = BLK_PERM_CONSISTENT_READ;
if ((nbdflags & NBD_FLAG_READ_ONLY) == 0) {
perm |= BLK_PERM_WRITE;
}
blk = blk_new(perm, BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED |
BLK_PERM_WRITE | BLK_PERM_GRAPH_MOD);
ret = blk_insert_bs(blk, bs, errp);
if (ret < 0) {
goto fail;
}
blk_set_enable_write_cache(blk, !writethrough);
exp->refcount = 1;
QTAILQ_INIT(&exp->clients);
exp->blk = blk;
exp->dev_offset = dev_offset;
exp->nbdflags = nbdflags;
exp->size = size < 0 ? blk_getlength(blk) : size;
if (exp->size < 0) {
error_setg_errno(errp, -exp->size,
"Failed to determine the NBD export's length");
goto fail;
}
exp->size -= exp->size % BDRV_SECTOR_SIZE;
exp->close = close;
exp->ctx = blk_get_aio_context(blk);
blk_add_aio_context_notifier(blk, blk_aio_attached, blk_aio_detach, exp);
if (on_eject_blk) {
blk_ref(on_eject_blk);
exp->eject_notifier_blk = on_eject_blk;
exp->eject_notifier.notify = nbd_eject_notifier;
blk_add_remove_bs_notifier(on_eject_blk, &exp->eject_notifier);
}
/*
* NBD exports are used for non-shared storage migration. Make sure
* that BDRV_O_INACTIVE is cleared and the image is ready for write
* access since the export could be available before migration handover.
*/
aio_context_acquire(exp->ctx);
blk_invalidate_cache(blk, NULL);
aio_context_release(exp->ctx);
return exp;
fail:
blk_unref(blk);
g_free(exp);
return NULL;
}
NBDExport *nbd_export_find(const char *name)
{
NBDExport *exp;
QTAILQ_FOREACH(exp, &exports, next) {
if (strcmp(name, exp->name) == 0) {
return exp;
}
}
return NULL;
}
void nbd_export_set_name(NBDExport *exp, const char *name)
{
if (exp->name == name) {
return;
}
nbd_export_get(exp);
if (exp->name != NULL) {
g_free(exp->name);
exp->name = NULL;
QTAILQ_REMOVE(&exports, exp, next);
nbd_export_put(exp);
}
if (name != NULL) {
nbd_export_get(exp);
exp->name = g_strdup(name);
QTAILQ_INSERT_TAIL(&exports, exp, next);
}
nbd_export_put(exp);
}
void nbd_export_set_description(NBDExport *exp, const char *description)
{
g_free(exp->description);
exp->description = g_strdup(description);
}
void nbd_export_close(NBDExport *exp)
{
NBDClient *client, *next;
nbd_export_get(exp);
QTAILQ_FOREACH_SAFE(client, &exp->clients, next, next) {
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>
2017-06-08 22:26:17 +00:00
client_close(client, true);
}
nbd_export_set_name(exp, NULL);
nbd_export_set_description(exp, NULL);
nbd_export_put(exp);
}
void nbd_export_get(NBDExport *exp)
{
assert(exp->refcount > 0);
exp->refcount++;
}
void nbd_export_put(NBDExport *exp)
{
assert(exp->refcount > 0);
if (exp->refcount == 1) {
nbd_export_close(exp);
}
if (--exp->refcount == 0) {
assert(exp->name == NULL);
assert(exp->description == NULL);
if (exp->close) {
exp->close(exp);
}
if (exp->blk) {
if (exp->eject_notifier_blk) {
notifier_remove(&exp->eject_notifier);
blk_unref(exp->eject_notifier_blk);
}
blk_remove_aio_context_notifier(exp->blk, blk_aio_attached,
blk_aio_detach, exp);
blk_unref(exp->blk);
exp->blk = NULL;
}
g_free(exp);
}
}
BlockBackend *nbd_export_get_blockdev(NBDExport *exp)
{
return exp->blk;
}
void nbd_export_close_all(void)
{
NBDExport *exp, *next;
QTAILQ_FOREACH_SAFE(exp, &exports, next, next) {
nbd_export_close(exp);
}
}
static int nbd_co_send_reply(NBDRequestData *req, NBDReply *reply, int len)
{
NBDClient *client = req->client;
int ret;
g_assert(qemu_in_coroutine());
qemu_co_mutex_lock(&client->send_lock);
client->send_coroutine = qemu_coroutine_self();
if (!len) {
ret = nbd_send_reply(client->ioc, reply);
} else {
qio_channel_set_cork(client->ioc, true);
ret = nbd_send_reply(client->ioc, reply);
if (ret == 0) {
ret = nbd_write(client->ioc, req->data, len, NULL);
if (ret < 0) {
ret = -EIO;
}
}
qio_channel_set_cork(client->ioc, false);
}
client->send_coroutine = NULL;
qemu_co_mutex_unlock(&client->send_lock);
return ret;
}
/* nbd_co_receive_request
* Collect a client request. Return 0 if request looks valid, -EIO to drop
* connection right away, and any other negative value to report an error to
* the client (although the caller may still need to disconnect after reporting
* the error).
*/
static int nbd_co_receive_request(NBDRequestData *req, NBDRequest *request)
{
NBDClient *client = req->client;
g_assert(qemu_in_coroutine());
assert(client->recv_coroutine == qemu_coroutine_self());
if (nbd_receive_request(client->ioc, request) < 0) {
return -EIO;
}
TRACE("Decoding type");
if (request->type != NBD_CMD_WRITE) {
/* No payload, we are ready to read the next request. */
req->complete = true;
}
if (request->type == NBD_CMD_DISC) {
/* Special case: we're going to disconnect without a reply,
* whether or not flags, from, or len are bogus */
TRACE("Request type is DISCONNECT");
return -EIO;
}
/* Check for sanity in the parameters, part 1. Defer as many
* checks as possible until after reading any NBD_CMD_WRITE
* payload, so we can try and keep the connection alive. */
if ((request->from + request->len) < request->from) {
LOG("integer overflow detected, you're probably being attacked");
return -EINVAL;
}
if (request->type == NBD_CMD_READ || request->type == NBD_CMD_WRITE) {
if (request->len > NBD_MAX_BUFFER_SIZE) {
LOG("len (%" PRIu32" ) is larger than max len (%u)",
request->len, NBD_MAX_BUFFER_SIZE);
return -EINVAL;
}
req->data = blk_try_blockalign(client->exp->blk, request->len);
if (req->data == NULL) {
return -ENOMEM;
}
}
if (request->type == NBD_CMD_WRITE) {
TRACE("Reading %" PRIu32 " byte(s)", request->len);
if (nbd_read(client->ioc, req->data, request->len, NULL) < 0) {
LOG("reading from socket failed");
return -EIO;
}
req->complete = true;
}
/* Sanity checks, part 2. */
if (request->from + request->len > client->exp->size) {
LOG("operation past EOF; From: %" PRIu64 ", Len: %" PRIu32
", Size: %" PRIu64, request->from, request->len,
(uint64_t)client->exp->size);
return request->type == NBD_CMD_WRITE ? -ENOSPC : -EINVAL;
}
if (request->flags & ~(NBD_CMD_FLAG_FUA | NBD_CMD_FLAG_NO_HOLE)) {
LOG("unsupported flags (got 0x%x)", request->flags);
return -EINVAL;
}
if (request->type != NBD_CMD_WRITE_ZEROES &&
(request->flags & NBD_CMD_FLAG_NO_HOLE)) {
LOG("unexpected flags (got 0x%x)", request->flags);
return -EINVAL;
}
return 0;
}
/* Owns a reference to the NBDClient passed as opaque. */
static coroutine_fn void nbd_trip(void *opaque)
{
NBDClient *client = opaque;
NBDExport *exp = client->exp;
NBDRequestData *req;
NBDRequest request = { 0 }; /* GCC thinks it can be used uninitialized */
NBDReply reply;
int ret;
int flags;
int reply_data_len = 0;
TRACE("Reading request.");
if (client->closing) {
nbd_client_put(client);
return;
}
req = nbd_request_get(client);
ret = nbd_co_receive_request(req, &request);
client->recv_coroutine = NULL;
nbd_client_receive_next_request(client);
if (ret == -EIO) {
goto disconnect;
}
reply.handle = request.handle;
reply.error = 0;
if (ret < 0) {
reply.error = -ret;
goto reply;
}
if (client->closing) {
/*
* The client may be closed when we are blocked in
* nbd_co_receive_request()
*/
goto done;
}
switch (request.type) {
case NBD_CMD_READ:
TRACE("Request type is READ");
/* XXX: NBD Protocol only documents use of FUA with WRITE */
if (request.flags & NBD_CMD_FLAG_FUA) {
ret = blk_co_flush(exp->blk);
if (ret < 0) {
LOG("flush failed");
reply.error = -ret;
break;
}
}
ret = blk_pread(exp->blk, request.from + exp->dev_offset,
req->data, request.len);
if (ret < 0) {
LOG("reading from file failed");
reply.error = -ret;
break;
}
reply_data_len = request.len;
TRACE("Read %" PRIu32" byte(s)", request.len);
break;
case NBD_CMD_WRITE:
TRACE("Request type is WRITE");
if (exp->nbdflags & NBD_FLAG_READ_ONLY) {
TRACE("Server is read-only, return error");
reply.error = EROFS;
break;
}
TRACE("Writing to device");
flags = 0;
if (request.flags & NBD_CMD_FLAG_FUA) {
flags |= BDRV_REQ_FUA;
}
ret = blk_pwrite(exp->blk, request.from + exp->dev_offset,
req->data, request.len, flags);
if (ret < 0) {
LOG("writing to file failed");
reply.error = -ret;
}
break;
case NBD_CMD_WRITE_ZEROES:
TRACE("Request type is WRITE_ZEROES");
if (exp->nbdflags & NBD_FLAG_READ_ONLY) {
TRACE("Server is read-only, return error");
reply.error = EROFS;
break;
}
TRACE("Writing to device");
flags = 0;
if (request.flags & NBD_CMD_FLAG_FUA) {
flags |= BDRV_REQ_FUA;
}
if (!(request.flags & NBD_CMD_FLAG_NO_HOLE)) {
flags |= BDRV_REQ_MAY_UNMAP;
}
ret = blk_pwrite_zeroes(exp->blk, request.from + exp->dev_offset,
request.len, flags);
if (ret < 0) {
LOG("writing to file failed");
reply.error = -ret;
}
break;
case NBD_CMD_DISC:
/* unreachable, thanks to special case in nbd_co_receive_request() */
abort();
case NBD_CMD_FLUSH:
TRACE("Request type is FLUSH");
ret = blk_co_flush(exp->blk);
if (ret < 0) {
LOG("flush failed");
reply.error = -ret;
}
break;
case NBD_CMD_TRIM:
TRACE("Request type is TRIM");
ret = blk_co_pdiscard(exp->blk, request.from + exp->dev_offset,
request.len);
if (ret < 0) {
LOG("discard failed");
reply.error = -ret;
}
break;
default:
LOG("invalid request type (%" PRIu32 ") received", request.type);
reply.error = EINVAL;
}
reply:
/* We must disconnect after NBD_CMD_WRITE if we did not
* read the payload.
*/
if (nbd_co_send_reply(req, &reply, reply_data_len) < 0 || !req->complete) {
goto disconnect;
}
TRACE("Request/Reply complete");
done:
nbd_request_put(req);
nbd_client_put(client);
return;
disconnect:
nbd_request_put(req);
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>
2017-06-08 22:26:17 +00:00
client_close(client, true);
nbd_client_put(client);
}
static void nbd_client_receive_next_request(NBDClient *client)
{
if (!client->recv_coroutine && client->nb_requests < MAX_NBD_REQUESTS) {
nbd_client_get(client);
client->recv_coroutine = qemu_coroutine_create(nbd_trip, client);
aio_co_schedule(client->exp->ctx, client->recv_coroutine);
}
}
static coroutine_fn void nbd_co_client_start(void *opaque)
{
NBDClient *client = opaque;
NBDExport *exp = client->exp;
if (exp) {
nbd_export_get(exp);
QTAILQ_INSERT_TAIL(&exp->clients, client, next);
}
qemu_co_mutex_init(&client->send_lock);
if (nbd_negotiate(client)) {
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>
2017-06-08 22:26:17 +00:00
client_close(client, false);
return;
}
nbd_client_receive_next_request(client);
}
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>
2017-06-08 22:26:17 +00:00
/*
* Create a new client listener on the given export @exp, using the
* given channel @sioc. Begin servicing it in a coroutine. When the
* connection closes, call @close_fn with an indication of whether the
* client completed negotiation.
*/
void nbd_client_new(NBDExport *exp,
QIOChannelSocket *sioc,
QCryptoTLSCreds *tlscreds,
const char *tlsaclname,
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>
2017-06-08 22:26:17 +00:00
void (*close_fn)(NBDClient *, bool))
{
NBDClient *client;
Coroutine *co;
client = g_malloc0(sizeof(NBDClient));
client->refcount = 1;
client->exp = exp;
client->tlscreds = tlscreds;
if (tlscreds) {
object_ref(OBJECT(client->tlscreds));
}
client->tlsaclname = g_strdup(tlsaclname);
client->sioc = sioc;
object_ref(OBJECT(client->sioc));
client->ioc = QIO_CHANNEL(sioc);
object_ref(OBJECT(client->ioc));
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>
2017-06-08 22:26:17 +00:00
client->close_fn = close_fn;
co = qemu_coroutine_create(nbd_co_client_start, client);
qemu_coroutine_enter(co);
}