Message ID | 1460060981-5338-1-git-send-email-eblake@redhat.com |
---|---|
State | New |
Headers | show |
On 7 Apr 2016, at 21:29, Eric Blake <eblake@redhat.com> wrote: > Upstream NBD is documenting that servers MAY choose to operate > in a conditional mode, where it is up to the client whether to > use TLS. For qemu's case, we want to always be in FORCEDTLS > mode, because of the risk of man-in-the-middle attacks, and since > we never export more than one device; likewise, the qemu client > will ALWAYS send NBD_OPT_STARTTLS as its first option. But now > that SELECTIVETLS servers exist, it is feasible to encounter a > (non-qemu) client that does not do NBD_OPT_STARTTLS first, but > rather wants to take advantage of the conditional modes it might > find elsewhere. > > Since we require TLS, we are within our rights to drop connections > on any client that doesn't negotiate it right away, or which > attempts to negotiate it incorrectly, without violating the intent > of the NBD Protocol. However, it's better to allow the client to > continue trying, on the grounds that maybe the client will get the > hint to send NBD_OPT_STARTTLS. > > Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Alex Bligh <alex@alex.org.uk> Looks right to me - untested. > --- > > My earlier patch was arguably incomplete: > https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg01265.html > > But as it is already in a pull request, and as this one is > a bit more controversial, it's best to keep it as a separate patch. > > nbd/server.c | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/nbd/server.c b/nbd/server.c > index 7843584..2b727f0 100644 > --- a/nbd/server.c > +++ b/nbd/server.c > @@ -450,9 +450,12 @@ static int nbd_negotiate_options(NBDClient *client) > > default: > TRACE("Option 0x%x not permitted before TLS", clientflags); > + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { > + return -EIO; > + } > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_TLS_REQD, > clientflags); > - return -EINVAL; > + break; > } > } else if (fixedNewstyle) { > switch (clientflags) { > @@ -470,6 +473,9 @@ static int nbd_negotiate_options(NBDClient *client) > return nbd_negotiate_handle_export_name(client, length); > > case NBD_OPT_STARTTLS: > + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { > + return -EIO; > + } > if (client->tlscreds) { > TRACE("TLS already enabled"); > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_INVALID, > @@ -479,7 +485,7 @@ static int nbd_negotiate_options(NBDClient *client) > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_POLICY, > clientflags); > } > - return -EINVAL; > + break; > default: > TRACE("Unsupported option 0x%x", clientflags); > if (nbd_negotiate_drop_sync(client->ioc, length) != length) { > -- > 2.5.5 > >
[adding qemu-block in cc, since Paolo can't send pull request] On 04/07/2016 02:29 PM, Eric Blake wrote: > Upstream NBD is documenting that servers MAY choose to operate > in a conditional mode, where it is up to the client whether to > use TLS. For qemu's case, we want to always be in FORCEDTLS > mode, because of the risk of man-in-the-middle attacks, and since > we never export more than one device; likewise, the qemu client > will ALWAYS send NBD_OPT_STARTTLS as its first option. But now > that SELECTIVETLS servers exist, it is feasible to encounter a > (non-qemu) client that does not do NBD_OPT_STARTTLS first, but > rather wants to take advantage of the conditional modes it might > find elsewhere. > > Since we require TLS, we are within our rights to drop connections > on any client that doesn't negotiate it right away, or which > attempts to negotiate it incorrectly, without violating the intent > of the NBD Protocol. However, it's better to allow the client to > continue trying, on the grounds that maybe the client will get the > hint to send NBD_OPT_STARTTLS. > > Signed-off-by: Eric Blake <eblake@redhat.com> > --- > > My earlier patch was arguably incomplete: > https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg01265.html > > But as it is already in a pull request, and as this one is > a bit more controversial, it's best to keep it as a separate patch. > > nbd/server.c | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/nbd/server.c b/nbd/server.c > index 7843584..2b727f0 100644 > --- a/nbd/server.c > +++ b/nbd/server.c > @@ -450,9 +450,12 @@ static int nbd_negotiate_options(NBDClient *client) > > default: > TRACE("Option 0x%x not permitted before TLS", clientflags); > + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { > + return -EIO; > + } > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_TLS_REQD, > clientflags); > - return -EINVAL; > + break; > } > } else if (fixedNewstyle) { > switch (clientflags) { > @@ -470,6 +473,9 @@ static int nbd_negotiate_options(NBDClient *client) > return nbd_negotiate_handle_export_name(client, length); > > case NBD_OPT_STARTTLS: > + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { > + return -EIO; > + } > if (client->tlscreds) { > TRACE("TLS already enabled"); > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_INVALID, > @@ -479,7 +485,7 @@ static int nbd_negotiate_options(NBDClient *client) > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_POLICY, > clientflags); > } > - return -EINVAL; > + break; > default: > TRACE("Unsupported option 0x%x", clientflags); > if (nbd_negotiate_drop_sync(client->ioc, length) != length) { >
On 14 Apr 2016, at 16:25, Eric Blake <eblake@redhat.com> wrote: > [adding qemu-block in cc, since Paolo can't send pull request] > > On 04/07/2016 02:29 PM, Eric Blake wrote: >> Upstream NBD is documenting that servers MAY choose to operate >> in a conditional mode, where it is up to the client whether to >> use TLS. For qemu's case, we want to always be in FORCEDTLS >> mode, because of the risk of man-in-the-middle attacks, and since >> we never export more than one device; likewise, the qemu client >> will ALWAYS send NBD_OPT_STARTTLS as its first option. But now >> that SELECTIVETLS servers exist, it is feasible to encounter a >> (non-qemu) client that does not do NBD_OPT_STARTTLS first, but >> rather wants to take advantage of the conditional modes it might >> find elsewhere. >> >> Since we require TLS, we are within our rights to drop connections >> on any client that doesn't negotiate it right away, or which >> attempts to negotiate it incorrectly, without violating the intent >> of the NBD Protocol. However, it's better to allow the client to >> continue trying, on the grounds that maybe the client will get the >> hint to send NBD_OPT_STARTTLS. >> >> Signed-off-by: Eric Blake <eblake@redhat.com> Reviewed-by: Alex Bligh <alex@alex.org.uk> It is worth noting that this change (assuming I've read it right) in no way means that qemu would be serving resources without TLS enabled when configured with TLS, or even providing information about those resources. It's simply saying "TLS is required" (and not dropping the connection so the client can ask for TLS) if the client doesn't ask for TLS first thing. This removes the need for the client and server to communicate out of band that TLS is required (as well as conforming to the spec). Alex >> --- >> >> My earlier patch was arguably incomplete: >> https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg01265.html >> >> But as it is already in a pull request, and as this one is >> a bit more controversial, it's best to keep it as a separate patch. >> >> nbd/server.c | 10 ++++++++-- >> 1 file changed, 8 insertions(+), 2 deletions(-) >> >> diff --git a/nbd/server.c b/nbd/server.c >> index 7843584..2b727f0 100644 >> --- a/nbd/server.c >> +++ b/nbd/server.c >> @@ -450,9 +450,12 @@ static int nbd_negotiate_options(NBDClient *client) >> >> default: >> TRACE("Option 0x%x not permitted before TLS", clientflags); >> + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { >> + return -EIO; >> + } >> nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_TLS_REQD, >> clientflags); >> - return -EINVAL; >> + break; >> } >> } else if (fixedNewstyle) { >> switch (clientflags) { >> @@ -470,6 +473,9 @@ static int nbd_negotiate_options(NBDClient *client) >> return nbd_negotiate_handle_export_name(client, length); >> >> case NBD_OPT_STARTTLS: >> + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { >> + return -EIO; >> + } >> if (client->tlscreds) { >> TRACE("TLS already enabled"); >> nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_INVALID, >> @@ -479,7 +485,7 @@ static int nbd_negotiate_options(NBDClient *client) >> nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_POLICY, >> clientflags); >> } >> - return -EINVAL; >> + break; >> default: >> TRACE("Unsupported option 0x%x", clientflags); >> if (nbd_negotiate_drop_sync(client->ioc, length) != length) { >> > > -- > Eric Blake eblake redhat com +1-919-301-3266 > Libvirt virtualization library http://libvirt.org > -- Alex Bligh
On 07.04.2016 22:29, Eric Blake wrote: > Upstream NBD is documenting that servers MAY choose to operate > in a conditional mode, where it is up to the client whether to > use TLS. For qemu's case, we want to always be in FORCEDTLS > mode, because of the risk of man-in-the-middle attacks, and since > we never export more than one device; likewise, the qemu client > will ALWAYS send NBD_OPT_STARTTLS as its first option. But now > that SELECTIVETLS servers exist, it is feasible to encounter a > (non-qemu) client that does not do NBD_OPT_STARTTLS first, but > rather wants to take advantage of the conditional modes it might > find elsewhere. > > Since we require TLS, we are within our rights to drop connections > on any client that doesn't negotiate it right away, or which > attempts to negotiate it incorrectly, without violating the intent > of the NBD Protocol. However, it's better to allow the client to > continue trying, on the grounds that maybe the client will get the > hint to send NBD_OPT_STARTTLS. > > Signed-off-by: Eric Blake <eblake@redhat.com> > --- > > My earlier patch was arguably incomplete: > https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg01265.html > > But as it is already in a pull request, and as this one is > a bit more controversial, it's best to keep it as a separate patch. > > nbd/server.c | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/nbd/server.c b/nbd/server.c > index 7843584..2b727f0 100644 > --- a/nbd/server.c > +++ b/nbd/server.c > @@ -450,9 +450,12 @@ static int nbd_negotiate_options(NBDClient *client) > > default: > TRACE("Option 0x%x not permitted before TLS", clientflags); > + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { > + return -EIO; > + } > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_TLS_REQD, > clientflags); > - return -EINVAL; > + break; > } What about NBD_OPT_EXPORTNAME? The specification says that this option does not allow for errors, and so the session must be terminated if this option is sent in FORCEDTLS mode without TLS having been negotiated. Max > } else if (fixedNewstyle) { > switch (clientflags) { > @@ -470,6 +473,9 @@ static int nbd_negotiate_options(NBDClient *client) > return nbd_negotiate_handle_export_name(client, length); > > case NBD_OPT_STARTTLS: > + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { > + return -EIO; > + } > if (client->tlscreds) { > TRACE("TLS already enabled"); > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_INVALID, > @@ -479,7 +485,7 @@ static int nbd_negotiate_options(NBDClient *client) > nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_POLICY, > clientflags); > } > - return -EINVAL; > + break; > default: > TRACE("Unsupported option 0x%x", clientflags); > if (nbd_negotiate_drop_sync(client->ioc, length) != length) { >
On 04/14/2016 03:08 PM, Max Reitz wrote: > On 07.04.2016 22:29, Eric Blake wrote: >> Upstream NBD is documenting that servers MAY choose to operate >> in a conditional mode, where it is up to the client whether to >> use TLS. For qemu's case, we want to always be in FORCEDTLS >> mode, because of the risk of man-in-the-middle attacks, and since >> we never export more than one device; likewise, the qemu client >> will ALWAYS send NBD_OPT_STARTTLS as its first option. But now >> that SELECTIVETLS servers exist, it is feasible to encounter a >> (non-qemu) client that does not do NBD_OPT_STARTTLS first, but >> rather wants to take advantage of the conditional modes it might >> find elsewhere. >> >> Since we require TLS, we are within our rights to drop connections >> on any client that doesn't negotiate it right away, or which >> attempts to negotiate it incorrectly, without violating the intent >> of the NBD Protocol. However, it's better to allow the client to >> continue trying, on the grounds that maybe the client will get the >> hint to send NBD_OPT_STARTTLS. >> >> Signed-off-by: Eric Blake <eblake@redhat.com> >> --- >> +++ b/nbd/server.c >> @@ -450,9 +450,12 @@ static int nbd_negotiate_options(NBDClient *client) >> >> default: >> TRACE("Option 0x%x not permitted before TLS", clientflags); >> + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { >> + return -EIO; >> + } >> nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_TLS_REQD, >> clientflags); >> - return -EINVAL; >> + break; >> } > > What about NBD_OPT_EXPORTNAME? The specification says that this option > does not allow for errors, and so the session must be terminated if this > option is sent in FORCEDTLS mode without TLS having been negotiated. Oh, good catch. v2 coming up.
diff --git a/nbd/server.c b/nbd/server.c index 7843584..2b727f0 100644 --- a/nbd/server.c +++ b/nbd/server.c @@ -450,9 +450,12 @@ static int nbd_negotiate_options(NBDClient *client) default: TRACE("Option 0x%x not permitted before TLS", clientflags); + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { + return -EIO; + } nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_TLS_REQD, clientflags); - return -EINVAL; + break; } } else if (fixedNewstyle) { switch (clientflags) { @@ -470,6 +473,9 @@ static int nbd_negotiate_options(NBDClient *client) return nbd_negotiate_handle_export_name(client, length); case NBD_OPT_STARTTLS: + if (nbd_negotiate_drop_sync(client->ioc, length) != length) { + return -EIO; + } if (client->tlscreds) { TRACE("TLS already enabled"); nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_INVALID, @@ -479,7 +485,7 @@ static int nbd_negotiate_options(NBDClient *client) nbd_negotiate_send_rep(client->ioc, NBD_REP_ERR_POLICY, clientflags); } - return -EINVAL; + break; default: TRACE("Unsupported option 0x%x", clientflags); if (nbd_negotiate_drop_sync(client->ioc, length) != length) {
Upstream NBD is documenting that servers MAY choose to operate in a conditional mode, where it is up to the client whether to use TLS. For qemu's case, we want to always be in FORCEDTLS mode, because of the risk of man-in-the-middle attacks, and since we never export more than one device; likewise, the qemu client will ALWAYS send NBD_OPT_STARTTLS as its first option. But now that SELECTIVETLS servers exist, it is feasible to encounter a (non-qemu) client that does not do NBD_OPT_STARTTLS first, but rather wants to take advantage of the conditional modes it might find elsewhere. Since we require TLS, we are within our rights to drop connections on any client that doesn't negotiate it right away, or which attempts to negotiate it incorrectly, without violating the intent of the NBD Protocol. However, it's better to allow the client to continue trying, on the grounds that maybe the client will get the hint to send NBD_OPT_STARTTLS. Signed-off-by: Eric Blake <eblake@redhat.com> --- My earlier patch was arguably incomplete: https://lists.gnu.org/archive/html/qemu-devel/2016-04/msg01265.html But as it is already in a pull request, and as this one is a bit more controversial, it's best to keep it as a separate patch. nbd/server.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)