In the client, read the reply headers from a coroutine, switching the
read side between the "read header" coroutine and the I/O coroutine that
reads the body of the reply.
In the server, if the server can read more requests it will create a new
"read request" coroutine as soon as a request has been read. Otherwise,
the new coroutine is created in nbd_request_put.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 20170213135235.12274-8-pbonzini@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Currently the QIOTaskFunc signature takes an Object * for
the source, and an Error * for any error. We also need to
be able to provide a result pointer. Rather than continue
to add parameters to QIOTaskFunc, remove the existing
ones and simply pass the QIOTask object instead. This
has methods to access all the other data items required
in the callback impl.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The new AioPollFn io_poll() argument to aio_set_fd_handler() and
aio_set_event_handler() is used in the next patch.
Keep this code change separate due to the number of files it touches.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 20161201192652.9509-3-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Commit 7d3123e converted a single read_sync() into a while loop
that assumed that read_sync() would either make progress or give
an error. But when the server hangs up early, the client sees
EOF (a read_sync() of 0) and never makes progress, which in turn
caused qemu-iotest './check -nbd 83' to go into an infinite loop.
Rework the loop to accomodate reads cut short by EOF.
Reported-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1478551093-32757-1-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Upstream NBD protocol recently added the ability to efficiently
write zeroes without having to send the zeroes over the wire,
along with a flag to control whether the client wants to allow
a hole.
Note that when it comes to requiring full allocation, vs.
permitting optimizations, the NBD spec intentionally picked a
different sense for the flag; the rules in qemu are:
MAY_UNMAP == 0: must write zeroes
MAY_UNMAP == 1: may use holes if reads will see zeroes
while in NBD, the rules are:
FLAG_NO_HOLE == 1: must write zeroes
FLAG_NO_HOLE == 0: may use holes if reads will see zeroes
In all cases, the 'may use holes' scenario is optional (the
server need not use a hole, and must not use a hole if
subsequent reads would not see zeroes).
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-16-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
NBD commit 6d34500b clarified how clients and servers are supposed
to behave before closing a connection. It added NBD_REP_ERR_SHUTDOWN
(for the server to announce it is about to go away during option
haggling, so the client should quit sending NBD_OPT_* other than
NBD_OPT_ABORT) and ESHUTDOWN (for the server to announce it is about
to go away during transmission, so the client should quit sending
NBD_CMD_* other than NBD_CMD_DISC). It also clarified that
NBD_OPT_ABORT gets a reply, while NBD_CMD_DISC does not.
This patch merely adds the missing reply to NBD_OPT_ABORT and teaches
the client to recognize server errors. Actually teaching the server
to send NBD_REP_ERR_SHUTDOWN or ESHUTDOWN would require knowing that
the server has been requested to shut down soon (maybe we could do
that by installing a SIGINT handler in qemu-nbd, which transitions
from RUNNING to a new state that waits for the client to react,
rather than just out-right quitting - but that's a bigger task for
another day).
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-15-git-send-email-eblake@redhat.com>
[Move dummy ESHUTDOWN to include/qemu/osdep.h. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Checkpatch complains that 'return EINVAL' is usually wrong
(since we tend to favor 'return -EINVAL'). But it is a
false positive for nbd_errno_to_system_errno(). Since NBD
may add future defined wire values, refactor the code to
keep checkpatch happy.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-14-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The NBD Protocol allows the server and client to mutually agree
on a shorter handshake (omit the 124 bytes of reserved 0), via
the server advertising NBD_FLAG_NO_ZEROES and the client
acknowledging with NBD_FLAG_C_NO_ZEROES (only possible in
newstyle, whether or not it is fixed newstyle). It doesn't
shave much off the wire, but we might as well implement it.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alex Bligh <alex@alex.org.uk>
Message-Id: <1476469998-28592-13-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since we know that the maximum name we are willing to accept
is small enough to stack-allocate, rework the iteration over
NBD_OPT_LIST responses to reuse a stack buffer rather than
allocating every time. Furthermore, we don't even have to
allocate if we know the server's length doesn't match what
we are searching for.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-12-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The server has a nice helper function nbd_negotiate_drop_sync()
which lets it easily ignore fluff from the client (such as the
payload to an unknown option request). We can't quite make it
common, since it depends on nbd_negotiate_read() which handles
coroutine magic, but we can copy the idea into the client where
we have places where we want to ignore data (such as the
description tacked on the end of NBD_REP_SERVER).
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-11-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The NBD spec says that a client should send NBD_OPT_ABORT
rather than just dropping the connection, if the client doesn't
like something the server sent during option negotiation. This
is a best-effort attempt only, and can only be done in places
where we know the server is still in sync with what we've sent,
whether or not we've read everything the server has sent.
Technically, the server then has to reply with NBD_REP_ACK, but
it's not worth complicating the client to wait around for that
reply.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-10-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Rather than open-coding each option request, it's easier to
have common helper functions do the work. That in turn requires
having convenient packed types for handling option requests
and replies.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-9-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The NBD Protocol allows us to send human-readable messages
along with any NBD_REP_ERR error during option negotiation;
make use of this fact for clients that know what to do with
our message.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-8-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Rather than open-coding NBD_REP_SERVER, reuse the code we
already have by adding a length parameter. Additionally,
the refactoring will make adding NBD_OPT_GO in a later patch
easier.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-7-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Our coding convention prefers CamelCase names, and we already
have other existing structs with NBDFoo naming. Let's be
consistent, before later patches add even more structs.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-6-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We have both 'struct NBDRequest' and 'struct nbd_request'; making
it confusing to see which does what. Furthermore, we want to
rename nbd_request to align with our normal CamelCase naming
conventions. So, rename the struct which is used to associate
the data received during request callbacks, while leaving the
shorter name for the description of the request sent over the
wire in the NBD protocol.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-4-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Current upstream NBD documents that requests have a 16-bit flags,
followed by a 16-bit type integer; although older versions mentioned
only a 32-bit field with masking to find flags. Since the protocol
is in network order (big-endian over the wire), the ABI is unchanged;
but dealing with the flags as a separate field rather than masking
will make it easier to add support for upcoming NBD extensions that
increase the number of both flags and commands.
Improve some comments in nbd.h based on the current upstream
NBD protocol (https://github.com/yoe/nbd/blob/master/doc/proto.md),
and touch some nearby code to keep checkpatch.pl happy.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-3-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The NBD protocol allows servers to advertise a human-readable
description alongside an export name during NBD_OPT_LIST. Add
an option to pass through the user's string to the NBD client.
Doing this also makes it easier to test commit 200650d4, which
is the client counterpart of receiving the description.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1476469998-28592-2-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Ensure that all I/O channels created for NBD are given names
to distinguish their respective roles.
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The builtin NBD server uses its own BlockBackend now instead of reusing
the monitor/guest device one.
This means that it has its own writethrough setting now. The builtin
NBD server always uses writeback caching now regardless of whether the
guest device has WCE enabled. qemu-nbd respects the cache mode given on
the command line.
We still need to keep a reference to the monitor BB because we put an
eject notifier on it, but we don't use it for any I/O.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Rather than asserting that nbdflags is within range, just give
it the correct type to begin with :) nbdflags corresponds to
the per-export portion of NBD Protocol "transmission flags", which
is 16 bits in response to NBD_OPT_EXPORT_NAME and NBD_OPT_GO.
Furthermore, upstream NBD has never passed the global flags to
the kernel via ioctl(NBD_SET_FLAGS) (the ioctl was first
introduced in NBD 2.9.22; then a latent bug in NBD 3.1 actually
tried to OR the global flags with the transmission flags, with
the disaster that the addition of NBD_FLAG_NO_ZEROES in 3.9
caused all earlier NBD 3.x clients to treat every export as
read-only; NBD 3.10 and later intentionally clip things to 16
bits to pass only transmission flags). Qemu should follow suit,
since the current two global flags (NBD_FLAG_FIXED_NEWSTYLE
and NBD_FLAG_NO_ZEROES) have no impact on the kernel's behavior
during transmission.
CC: qemu-stable@nongnu.org
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1469129688-22848-3-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit ab7c548e added a check for invalid flags, but used an
early return on error instead of properly going through the
cleanup label.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1469129688-22848-2-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Change sector-based blk_discard(), blk_co_discard(), and
blk_aio_discard() to instead be byte-based blk_pdiscard(),
blk_co_pdiscard(), and blk_aio_pdiscard(). NBD gets a lot
simpler now that ignoring the unaligned portion of a
byte-based discard request is handled under the hood by
the block layer.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1468624988-423-6-git-send-email-eblake@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Now that NBD relies on the block layer to fragment things, we no
longer need to track an offset argument for which fragment of
a request we are actually servicing.
While at it, use true and false instead of 0 and 1 for a bool
parameter.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1468607524-19021-6-git-send-email-eblake@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
In practice the entry argument is always known at creation time, and
it is confusing that sometimes qemu_coroutine_enter is used with a
non-NULL argument to re-enter a coroutine (this happens in
block/sheepdog.c and tests/test-coroutine.c). So pass the opaque value
at creation time, for consistency with e.g. aio_bh_new.
Mostly done with the following semantic patch:
@ entry1 @
expression entry, arg, co;
@@
- co = qemu_coroutine_create(entry);
+ co = qemu_coroutine_create(entry, arg);
...
- qemu_coroutine_enter(co, arg);
+ qemu_coroutine_enter(co);
@ entry2 @
expression entry, arg;
identifier co;
@@
- Coroutine *co = qemu_coroutine_create(entry);
+ Coroutine *co = qemu_coroutine_create(entry, arg);
...
- qemu_coroutine_enter(co, arg);
+ qemu_coroutine_enter(co);
@ entry3 @
expression entry, arg;
@@
- qemu_coroutine_enter(qemu_coroutine_create(entry), arg);
+ qemu_coroutine_enter(qemu_coroutine_create(entry, arg));
@ reentry @
expression co;
@@
- qemu_coroutine_enter(co, NULL);
+ qemu_coroutine_enter(co);
except for the aforementioned few places where the semantic patch
stumbled (as expected) and for test_co_queue, which would otherwise
produce an uninitialized variable warning.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The trace format string in nbd_send_request uses PRIu16 for
request->type, but request->type is a uint32_t. This provokes
compiler warnings on the OSX clang. Use PRIu32 instead.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1466167331-17063-1-git-send-email-peter.maydell@linaro.org
Declare a constant and use that when determining if an export
name fits within the constraints we are willing to support.
Note that upstream NBD recently documented that clients MUST
support export names of 256 bytes (not including trailing NUL),
and SHOULD support names up to 4096 bytes. 4096 is a bit big
(we would lose benefits of stack-allocation of a name array),
and we already have other limits in place (for example, qcow2
snapshot names are clamped around 1024). So for now, just
stick to the required minimum, as that's easier to audit than
a full-scale support for larger names.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1463006384-7734-12-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Add some debugging to flag servers that are not compliant to
the NBD protocol. This would have flagged the server bug
fixed in commit c0301fcc.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alex Bligh <alex@alex.org.uk>
Message-Id: <1463006384-7734-11-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The kernel ioctl() interface into NBD is limited to 'unsigned long';
we MUST pass in input with that type (and not int or size_t, as
there may be platform ABIs where the wrong types promote incorrectly
through var-args). Furthermore, on 32-bit platforms, the kernel
is limited to a maximum export size of 2T (our BLKSIZE of 512 times
a SIZE_BLOCKS constrained by 32 bit unsigned long).
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1463006384-7734-8-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
NBD ioctl()s are used to manage an NBD client session where
initial handshake is done in userspace, but then the transmission
phase is handed off to the kernel through a /dev/nbdX device.
As such, all ioctls sent to the kernel on the /dev/nbdX fd belong
in client.c; nbd_disconnect() was out-of-place in server.c.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1463006384-7734-7-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The NBD protocol says that clients should not send a command flag
that has not been negotiated (whether by the client requesting an
option during a handshake, or because we advertise support for the
flag in response to NBD_OPT_EXPORT_NAME), and that servers should
reject invalid flags with EINVAL. We were silently ignoring the
flags instead. The client can't rely on our behavior, since it is
their fault for passing the bad flag in the first place, but it's
better to be robust up front than to possibly behave differently
than the client was expecting with the attempted flag.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alex Bligh <alex@alex.org.uk>
Message-Id: <1463006384-7734-6-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We have a few bugs in how we handle invalid client commands:
- A client can send an NBD_CMD_DISC where from + len overflows,
convincing us to reply with an error and stay connected, even
though the protocol requires us to silently disconnect. Fix by
hoisting the special case sooner.
- A client can send an NBD_CMD_WRITE where from + len overflows,
where we reply to the client with EINVAL without consuming the
payload; this will normally cause us to fail if the next thing
read is not the right magic, but in rare cases, could cause us
to interpret the data payload as valid commands and do things
not requested by the client. Fix by adding a complete flag to
track whether we are in sync or must disconnect.
Furthermore, we have split the checks for bogus from/len across
two functions, when it is easier to do it all at once.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1463006384-7734-5-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We should never ignore failure from nbd_negotiate_send_rep(); if
we are unable to write to the client, then it is not worth trying
to continue the negotiation. Fortunately, the problem is not
too severe - chances are that the errors being ignored here (mainly
inability to write the reply to the client) are indications of
a closed connection or something similar, which will also affect
the next attempt to interact with the client and eventually reach
a point where the errors are detected to end the loop.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1463006384-7734-4-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Clean up some debug message oddities missed earlier; this includes
some typos, and recognizing that %d is not necessarily compatible
with uint32_t. Also add a couple messages that I found useful
while debugging things.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1463006384-7734-3-git-send-email-eblake@redhat.com>
[Do not use PRIx16, clang complains. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Rather than always flushing ourselves, let the block layer
forward the FUA on to the underlying device - where all
underlying layers also understand FUA, we are now more
efficient; and where any underlying layer doesn't understand
it, now the block layer takes care of the full flush fallback
on our behalf.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1463006384-7734-2-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The cpu_to_*w() functions just compose a pointer dereference
with a byteswap. Instead use st*_p(), which handles potential
pointer misalignment and avoids the need to cast the pointer.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <1465575342-12146-1-git-send-email-peter.maydell@linaro.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The *_to_cpup() functions are not very useful, as they simply do
a pointer dereference and then a *_to_cpu(). Instead use either:
* ld*_*_p(), if the data is at an address that might not be
correctly aligned for the load
* a local dereference and *_to_cpu(), if the pointer is
the correct type and known to be correctly aligned
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <1465570836-22211-1-git-send-email-peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Similar to commit df7b97ff, we are mishandling clients that
give an unaligned NBD_CMD_TRIM request, and potentially
trimming bytes that occur before their request; which in turn
can cause potential unintended data loss (unlikely in
practice, since most clients are sane and issue aligned trim
requests). However, while we fixed read and write by switching
to the byte interfaces of blk_, we don't yet have a byte
interface for discard. On the other hand, trim is advisory, so
rounding the user's request to simply ignore the first and last
unaligned sectors (or the entire request, if it is sub-sector
in length) is just fine.
CC: qemu-stable@nongnu.org
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1464173965-9694-1-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Move it to the actual users. There are still a few includes of
qemu/bswap.h in headers; removing them is left for future work.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We have several block drivers that understand BDRV_REQ_FUA,
and emulate it in the block layer for the rest by a full flush.
But without a way to actually request BDRV_REQ_FUA during a
pass-through blk_pwrite(), FUA-aware block drivers like NBD are
forced to repeat the emulation logic of a full flush regardless
of whether the backend they are writing to could do it more
efficiently.
This patch just wires up a flags argument; followup patches
will actually make use of it in the NBD driver and in qemu-io.
Signed-off-by: Eric Blake <eblake@redhat.com>
Acked-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The NBD protocol does not (yet) force any alignment constraints
on clients. Even though qemu NBD clients always send requests
that are aligned to 512 bytes, we must be prepared for non-qemu
clients that don't care about alignment (even if it means they
are less efficient). Our use of blk_read() and blk_write() was
silently operating on the wrong file offsets when the client
made an unaligned request, corrupting the client's data (but
as the client already has control over the file we are serving,
I don't think it is a security hole, per se, just a data
corruption bug).
Note that in the case of NBD_CMD_READ, an unaligned length could
cause us to return up to 511 bytes of uninitialized trailing
garbage from blk_try_blockalign() - hopefully nothing sensitive
from the heap's prior usage is ever leaked in that manner.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Tested-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 1461249750-31928-1-git-send-email-eblake@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Upstream NBD documents (as of commit 4feebc95) 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 is programmed to talk to
such a server, and does not do NBD_OPT_STARTTLS first, but rather
wants to probe if it can use a non-encrypted export.
The NBD protocol documents that we should let such a client
continue trying, on the grounds that maybe the client will get the
hint to send NBD_OPT_STARTTLS, rather than immediately dropping
the connection.
Note that NBD_OPT_EXPORT_NAME is a special case: since it is the
only option request that can't have an error return, we have to
(continue to) drop the connection on that one; rather, what we are
fixing here is that all other replies prior to TLS initiation tell
the client NBD_REP_ERR_TLS_REQD, but keep the connection alive.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-id: 1460671343-18485-1-git-send-email-eblake@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
The NBD Protocol states that NBD_REP_SERVER may set
'length > sizeof(namelen) + namelen'; in which case the rest
of the packet is a UTF-8 description of the export. While we
don't know of any NBD servers that send this description yet,
we had better consume the data so we don't choke when we start
to talk to such a server.
Also, a (buggy/malicious) server that replies with length <
sizeof(namelen) would cause us to block waiting for bytes that
the server is not sending, and one that replies with super-huge
lengths could cause us to temporarily allocate up to 4G memory.
Sanity check things before blindly reading incorrectly.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-id: 1460077777-31004-1-git-send-email-eblake@redhat.com
Reviewed-by: Alex Bligh <alex@alex.org.uk>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Until commit 1c778ef7 ("nbd: convert to using I/O channels for actual
socket I/O", 2016-02-16), nbd_wr_sync returned -EAGAIN this scenario.
nbd_reply_ready required these semantics because it has two conflicting
requirements:
1) if a reply can be received on the socket, nbd_reply_ready needs
to read the header outside coroutine context to identify _which_
coroutine to enter to process the rest of the reply
2) on the other hand, nbd_reply_ready can find a false positive if
another thread (e.g. a VCPU thread running aio_poll) sneaks in and
calls nbd_reply_ready too. In this case nbd_reply_ready does nothing
and expects nbd_wr_syncv to return -EAGAIN.
Currently, the solution to the first requirement is to wait in the very
rare case of a read() that doesn't retrieve the reply header in its
entirety; this is what nbd_wr_syncv does by calling qio_channel_wait().
However, the unconditional call to qio_channel_wait() breaks the second
requirement. To fix this, the patch makes nbd_wr_syncv return -EAGAIN
if done is zero, similar to the code before commit 1c778ef7.
This is okay because NBD client-side negotiation is the only other case
that calls nbd_wr_syncv outside a coroutine, and it places the socket
in blocking mode. On the other hand, it is a bit unpleasant to put
this in nbd_wr_syncv(), because the function is used by both client
and server.
The full fix would be to add a counter to NbdClientSession for how
many bytes have been filled in s->reply. Then a reply can be filled
by multiple separate invocations of nbd_reply_ready and the
qio_channel_wait() call can be removed completely. Something to
consider for 2.7...
Reported-by: Changlong Xie <xiecl.fnst@cn.fujitsu.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
nbd-server.c currently fails to handle unsupported options properly.
If during option haggling the client sends an unknown request, the
server kills the connection instead of letting the client try to
fall back to something older. This is precisely what advertising
NBD_FLAG_FIXED_NEWSTYLE was supposed to fix.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1459982918-32229-1-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
nbd-client.c currently fails to handle unsupported options properly.
If during option haggling the server finds an option that is
unsupported, it returns an NBD_REP_ERR_UNSUP reply.
According to nbd's proto.md, the format for such a reply
should be:
S: 64 bits, 0x3e889045565a9 (magic number for replies)
S: 32 bits, the option as sent by the client to which this is a reply
S: 32 bits, reply type (e.g., NBD_REP_ACK for successful completion,
or NBD_REP_ERR_UNSUP to mark use of an option not known by this server
S: 32 bits, length of the reply. This may be zero for some replies,
in which case the next field is not sent
S: any data as required by the reply (e.g., an export name in the case
of NBD_REP_SERVER, or optional UTF-8 message for NBD_REP_ERR_*)
However, in nbd-client.c, the reply type was being read, and if it
contained an error, it was bailing out and issuing the next option
request without first reading the length. This meant that the
next option / handshake read had an extra 4 or more bytes of data in it.
In practice, this makes Qemu incompatible with servers that do not
support NBD_OPT_LIST.
To verify this isn't an error in the specification or my reading of
it, replies are sent by the reference implementation here:
https://github.com/yoe/nbd/blob/66dfb35/nbd-server.c#L1232
and as is evident it always sends a 'datasize' (aka length) 32 bit
word. Unsupported elements are replied to here:
https://github.com/yoe/nbd/blob/66dfb35/nbd-server.c#L1371
Signed-off-by: Alex Bligh <alex@alex.org.uk>
Message-Id: <1459882500-24316-1-git-send-email-alex@alex.org.uk>
[rework to ALWAYS consume an optional UTF-8 message from the server]
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1459961962-18771-1-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Print debug tracing messages while data is still in native
ordering, rather than after we've potentially swapped it into
network order for transmission. Also, it's nice if the server
mentions what it is replying, to correlate it to with what the
client says it is receiving.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1459913704-19949-4-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The compiler is smart enough to optimize out 'if (0)', but won't
type-check our printfs if they are hidden behind #if.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1459913704-19949-3-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The NBD Protocol requires that servers should send EPERM for
attempts to write (or trim) a read-only export. We were
correct for TRIM (blk_co_discard() gave EPERM); but were
manually setting EROFS which then got mapped to EINVAL over
the wire on writes.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <1459913704-19949-2-git-send-email-eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
