Support actually split split-virtqueues in vmd.
Initial validation was added previously. Allows guests to not need
contiguous memory for virtqueues. Adds queue size validation and
alignment.
Discussed with mlarkin@
Fix multiple issues with find_gpa_range in vmd(8).
Check for ranges that hit holes, mmio regions, and check for overflows
for regions that hit UINT64_MAX.
Input from hshoexer@. Discussed also with mlarkin@
vmm(4)/vmd(8): mp plumbing
Implements the sipi/init lapic state machine, mp IPI support and various
other changes required for mp. vmctl and man page changes for new start
and vm.conf options.
tested by dv and me on various guests
ok dv
Center timestamps in frequency regression
Subtract the first timestamp in each sample batch before accumulating the
least-squares terms. This preserves the fitted slope while avoiding
catastrophic cancellation between squared absolute Unix timestamps.
from Adam DePrince <adam.deprince at gmail.com>
ok bluhm@, deraadt@
Clean up some of the DTLS regress tests.
Some of these were workarounds needed to handle the client retransmission
cases. Others test behaviour that is soon going to be removed.
Allow DTLS clients to receive unexpected Finished messages.
DTLS servers can retransmit Finished messages, which will appear in the
current epoch. Allow DTLS clients to receive and discard these, rather
than treating them as unexpected messages.
DTLS clients should also handle last flight retransmissions, however there
are bigger issues with this code.
ok beck@ tb@
Rework handling for flights of buffered DTLS messages.
Currently outgoing messages are stored in a pqueue for retransmission.
This is completely unnecessary and requires a bunch of contortions.
Replace the pqueue with a simple linked list of buffered messages,
which form a flight. Rename dtls1_clear_record_buffer() to
dtls1_clear_flight() to more accurately reflect its purpose.
Make dtls12_create_handshake_msg() self-contained and simplify
dtls1_buffer_message(), which then allows for the removal of w_msg_hdr.
ok kenjiro@ tb@
Limit size of buffered DTLS handshake messages.
Fragments from out-of-sequence DTLS handshake messages are buffered for
reassembly. When a new out-of-sequence handshake message fragment is
received it will allocate the full message length buffer that would be
used by the fully reassembled message. This is limited to the next
10 messages in sequence, each capped to a maximum size of 128KB - a
further 16KB is allocated for each reassembly bitmap. A set of crafted
handshake message fragments can trigger allocation of 10 * 144KB
(128KB + 16KB), with 156 bytes on the wire.
Change dtls1_max_handshake_message_len() so that we limit handshake
messages to ~16KB, with the exception of a Certificate message, which
uses s->max_cert_list (128KB by default). Also ensure that we do not
currently have an identical handshake message type pending for
reassembly - this prevents resassembly of multiple Certificate messages
which would then result in the original allocations.
This caps the allocation to ~306KB (9 * ~18KB + 144KB) - the correct
[4 lines not shown]
Improve client retransmit handling in DTLS regress.
In these DTLS regress tests we can have the server complete its handshake,
while the client has not and needs to trigger retransmissions. If the
server is done but the client is not, continue to do zero-byte reads so
that we process any retransmitted messages appropriately.
ioapic and lapic code for vmm/vmd
Add i82093aa ioapic and i82489dx-compatible lapic emulation to vmd.
Also wire up the fw loading for acpi tables to support the same in
seabios via fw_cfg.
Also fix errors with the PIT and ns8250 uart emulation discovered when
they are wired to the ioapic
ok dv
The relaxation to permit TCP_NODELAY on AF_UNIX through pledge shouldn't
check the socket's protocol. This should be revisited later as it allows
slightly more than desired, but unbreaks some software.
ok deraadt
