Most known NBD clients do not bother with NBD_OPT_INFO (except for clients like 'qemu-nbd --list' that don't ever intend to connect), but go straight to NBD_OPT_GO. However, it's not too hard to hack up qemu to add in an extra client step (whether info on the same name, or more interestingly, info on a different name), as a patch against qemu commit 6f214b30445: | diff --git i/nbd/client.c w/nbd/client.c | index f6733962b49b..425292ac5ea9 100644 | --- i/nbd/client.c | +++ w/nbd/client.c | @@ -1038,6 +1038,14 @@ int nbd_receive_negotiate(AioContext *aio_context, QIOChannel *ioc, | * TLS). If it is not available, fall back to | * NBD_OPT_LIST for nicer error messages about a missing | * export, then use NBD_OPT_EXPORT_NAME. */ | + if (getenv ("HACK")) | + info->name[0]++; | + result = nbd_opt_info_or_go(ioc, NBD_OPT_INFO, info, errp); | + if (getenv ("HACK")) | + info->name[0]--; | + if (result < 0) { | + return -EINVAL; | + } | result = nbd_opt_info_or_go(ioc, NBD_OPT_GO, info, errp); | if (result < 0) { | return -EINVAL; This works just fine in 1.14.0, where we call .open only once (so the INFO and GO repeat calls into the same plugin handle), but in 1.14.1 it regressed into causing an assertion failure: we are now calling .open a second time on a connection that is already opened. $ nbdkit -rfv null & $ hacked-qemu-io -f raw -r nbd://localhost -c quit ... nbdkit: null[1]: debug: null: open readonly=1 nbdkit: backend.c:179: backend_open: Assertion `h->handle == NULL' failed. Worse, on the mainline development, we have recently made it possible for plugins to actively report different information for different export names; for example, a plugin may choose to report different answers for .can_write on export A than for export B; but if we share cached handles, then an NBD_OPT_INFO on one export prevents correct answers for NBD_OPT_GO on the second export name. (The HACK envvar in my qemu modifications can be used to demonstrate cross-name requests, which are even less likely in a real client). The solution is to call .close after NBD_OPT_INFO, coupled with enough glue logic to reset cached connection handles back to the state expected by .open. This in turn means factoring out another backend_* function, but also gives us an opportunity to change backend_set_handle to no longer accept NULL. The assertion failure is, to some extent, a possible denial of service attack (one client can force nbdkit to exit by merely sending OPT_INFO before OPT_GO, preventing the next client from connecting), although this is mitigated by using TLS to weed out untrusted clients. Still, the fact that we introduced a potential DoS attack while trying to fix a traffic amplification security bug is not very nice. Sadly, as there are no known clients that easily trigger this mode of operation (OPT_INFO before OPT_GO), there is no easy way to cover this via a testsuite addition. I may end up hacking something into libnbd. Fixes: c05686f957 Signed-off-by: Eric Blake <eblake(a)redhat.com&gt; --- server/internal.h | 4 +++- server/backend.c | 13 +++++++++++++ server/connections.c | 2 +- server/filters.c | 5 +---- server/plugins.c | 4 ---- server/protocol-handshake-newstyle.c | 6 ++++++ 6 files changed, 24 insertions(+), 10 deletions(-) diff --git a/server/internal.h b/server/internal.h index c31bb340..da4fae19 100644 --- a/server/internal.h +++ b/server/internal.h @@ -334,9 +334,11 @@ extern int backend_open (struct backend *b, struct connection *conn, __attribute__((__nonnull__ (1, 2))); extern int backend_prepare (struct backend *b, struct connection *conn) __attribute__((__nonnull__ (1, 2))); +extern void backend_close (struct backend *b, struct connection *conn) + __attribute__((__nonnull__ (1, 2))); extern void backend_set_handle (struct backend *b, struct connection *conn, void *handle) - __attribute__((__nonnull__ (1, 2 /* not 3 */))); + __attribute__((__nonnull__ (1, 2, 3))); extern bool backend_valid_range (struct backend *b, struct connection *conn, uint64_t offset, uint32_t count) __attribute__((__nonnull__ (1, 2))); diff --git a/server/backend.c b/server/backend.c index 3b213bfb..64dbf7db 100644 --- a/server/backend.c +++ b/server/backend.c @@ -201,10 +201,23 @@ backend_prepare (struct backend *b, struct connection *conn) return b->prepare (b, conn, h->can_write == 0); } +void +backend_close (struct backend *b, struct connection *conn) +{ + struct b_conn_handle *h = &conn->handles[b->i]; + + debug ("%s: close", b->name); + + b->close (b, conn); + memset (h, -1, sizeof *h); + h->handle = NULL; +} + void backend_set_handle (struct backend *b, struct connection *conn, void *handle) { assert (b->i < conn->nr_handles); + assert (conn->handles[b->i].handle == NULL); conn->handles[b->i].handle = handle; } diff --git a/server/connections.c b/server/connections.c index 819f7b86..3c4296ea 100644 --- a/server/connections.c +++ b/server/connections.c @@ -369,7 +369,7 @@ free_connection (struct connection *conn) */ if (!quit && connection_get_handle (conn, 0)) { lock_request (conn); - backend->close (backend, conn); + backend_close (backend, conn); unlock_request (conn); } diff --git a/server/filters.c b/server/filters.c index 1ee62829..1091c2dd 100644 --- a/server/filters.c +++ b/server/filters.c @@ -225,12 +225,9 @@ filter_close (struct backend *b, struct connection *conn) struct backend_filter *f = container_of (b, struct backend_filter, backend); void *handle = connection_get_handle (conn, b->i); - debug ("%s: close", b->name); - if (handle && f->filter.close) f->filter.close (handle); - backend_set_handle (b, conn, NULL); - b->next->close (b->next, conn); + backend_close (b->next, conn); } /* The next_functions structure contains pointers to backend diff --git a/server/plugins.c b/server/plugins.c index 9b44c548..727fb0e0 100644 --- a/server/plugins.c +++ b/server/plugins.c @@ -273,12 +273,8 @@ plugin_close (struct backend *b, struct connection *conn) assert (connection_get_handle (conn, 0)); - debug ("close"); - if (p->plugin.close) p->plugin.close (connection_get_handle (conn, 0)); - - backend_set_handle (b, conn, NULL); } static int64_t diff --git a/server/protocol-handshake-newstyle.c b/server/protocol-handshake-newstyle.c index 87e0bcd7..06fc53ad 100644 --- a/server/protocol-handshake-newstyle.c +++ b/server/protocol-handshake-newstyle.c @@ -469,6 +469,12 @@ negotiate_handshake_newstyle_options (struct connection *conn) if (send_newstyle_option_reply (conn, option, NBD_REP_ACK) == -1) return -1; + if (option == NBD_OPT_INFO) { + if (backend->finalize (backend, conn) == -1) + return -1; + backend_close (backend, conn); + } + break; case NBD_OPT_STRUCTURED_REPLY: -- 2.21.0

The NBD spec says that if a client requests a 1-query SET_META_CONTEXT for context "base:allocation", then changes its mind and requests a 1-query SET_META_CONTEXT for "other:context", only the final query matters; in such a case, since we did not reply with a context the second time, the client must not call NBD_CMD_BLOCK_STATUS, and the server should fail it with EINVAL. If the client actually wants two contexts, it must request them in a 2-query SET_META_CONTEXT. However, our code didn't reset the boolean between two uses of the option, so we were not catching an invalid clients that requests block status in spite of their second SET_META_CONTEXT. Note that there are no known clients in the wild that can actually perform this secondary SET_META_CONTEXT request; this was found by inspection. As nbdkit does not crash in this situation, I don't see the point in hacking libnbd to make it possible to become such a client. Fixes: 26455d45 Signed-off-by: Eric Blake <eblake(a)redhat.com&gt; --- server/protocol-handshake-newstyle.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/server/protocol-handshake-newstyle.c b/server/protocol-handshake-newstyle.c index 06fc53ad..38978c67 100644 --- a/server/protocol-handshake-newstyle.c +++ b/server/protocol-handshake-newstyle.c @@ -566,10 +566,10 @@ negotiate_handshake_newstyle_options (struct connection *conn) debug ("newstyle negotiation: %s: %s count: %d", optname, option == NBD_OPT_LIST_META_CONTEXT ? "query" : "set", nr_queries); + if (option == NBD_OPT_SET_META_CONTEXT) + conn->meta_context_base_allocation = false; if (nr_queries == 0) { - if (option == NBD_OPT_SET_META_CONTEXT) - conn->meta_context_base_allocation = false; - else /* LIST */ { + if (option == NBD_OPT_LIST_META_CONTEXT) { if (send_newstyle_option_reply_meta_context (conn, option, NBD_REP_META_CONTEXT, 0, "base:allocation") == -1) -- 2.21.0

The NBD spec says that if a client requests SET_META_CONTEXT for exportA, but later requests NBD_OPT_GO/EXPORT_NAME for exportB, then the server should reject NBD_CMD_BLOCK_STATUS requests (that is, the context returned for exportA need not apply to exportB). When we originally added base:allocation, our argument was that we always ignore export names, so it was easier to just treat any two export names as being the same export, so no need to reset things. But now that we have nbdkit_export_name(), we are better off obeying the spec. Note that there are no known clients in the wild that can actually perform this cross-export-name request; this was found by inspection. I also don't see the point in hacking up libnbd to become such a client. Fixes: 4f303e44 Signed-off-by: Eric Blake <eblake(a)redhat.com&gt; --- server/internal.h | 1 + server/protocol-handshake-newstyle.c | 24 ++++++++++++++++++++++-- 2 files changed, 23 insertions(+), 2 deletions(-) diff --git a/server/internal.h b/server/internal.h index da4fae19..0603a747 100644 --- a/server/internal.h +++ b/server/internal.h @@ -182,6 +182,7 @@ struct connection { size_t nr_handles; char *exportname; + uint32_t exportnamelen; uint32_t cflags; uint16_t eflags; bool using_tls; diff --git a/server/protocol-handshake-newstyle.c b/server/protocol-handshake-newstyle.c index 785944eb..45a65487 100644 --- a/server/protocol-handshake-newstyle.c +++ b/server/protocol-handshake-newstyle.c @@ -282,6 +282,7 @@ negotiate_handshake_newstyle_options (struct connection *conn) nbdkit_error ("strndup: %m"); return -1; } + conn->exportnamelen = optlen; if (strlen (conn->exportname) != optlen) { nbdkit_error ("export name must not contain NUL bytes"); return -1; @@ -420,14 +421,21 @@ negotiate_handshake_newstyle_options (struct connection *conn) } /* As with NBD_OPT_EXPORT_NAME we print the export name and - * save it in the connection. + * save it in the connection. For NBD_OPT_GO, if an earlier + * NBD_OPT_SET_META_CONTEXT used an export name, it must match + * or else we drop the support for that context. */ + if (option == NBD_OPT_GO && conn->meta_context_base_allocation && + (exportnamelen != conn->exportnamelen || + memcmp (conn->exportname, &data[4], exportnamelen) != 0)) + conn->meta_context_base_allocation = false; free (conn->exportname); conn->exportname = strndup (&data[4], exportnamelen); if (conn->exportname == NULL) { nbdkit_error ("malloc: %m"); return -1; } + conn->exportnamelen = exportnamelen; if (strlen (conn->exportname) != exportnamelen) { nbdkit_error ("export name must not contain NUL bytes"); return -1; @@ -552,7 +560,10 @@ negotiate_handshake_newstyle_options (struct connection *conn) continue; } - /* Discard the export name, after checking that it is valid. */ + /* Remember the export name: the NBD spec says that if the client + * later uses NBD_OPT_GO on a different export, then the context + * returned here is not usable. + */ memcpy (&exportnamelen, &data[0], 4); exportnamelen = be32toh (exportnamelen); what = "checking export name"; @@ -561,6 +572,15 @@ negotiate_handshake_newstyle_options (struct connection *conn) what = "export name must not contain NUL bytes"; if (strnlen (&data[4], exportnamelen) != exportnamelen) goto opt_meta_invalid_option_len; + if (option == NBD_OPT_SET_META_CONTEXT) { + free (conn->exportname); + conn->exportname = strndup (&data[4], exportnamelen); + if (conn->exportname == NULL) { + nbdkit_error ("strndup: %m"); + return -1; + } + conn->exportnamelen = exportnamelen; + } opt_index = 4 + exportnamelen; /* Read the number of queries. */ -- 2.21.0