AIO write requests for a TOE socket on a Chelsio T4+ adapter can now
DMA directly from the user-supplied buffer. This is implemented by
wiring the pages backing the user-supplied buffer and queueing special
mbufs backed by raw VM pages to the socket buffer. The TOE code
recognizes these special mbufs and builds a sglist from the VM page
array associated with the mbuf when queueing a work request to the TOE.
Because these mbufs do not have an associated virtual address, m_data
is not valid. Thus, the AIO handler does not invoke sosend() directly
for these mbufs but instead inlines portions of sosend_generic() and
tcp_usr_send().
An aiotx_buffer structure is used to describe the user buffer (e.g.
it holds the array of VM pages and a reference to the AIO job). The
special mbufs reference this structure via m_ext. Note that a single
job might be split across multiple mbufs (e.g. if it is larger than
the socket buffer size). The 'ext_arg2' member of each mbuf gives an
offset relative to the backing aiotx_buffer. The AIO job associated
with an aiotx_buffer structure is completed when the last reference to
the structure is released.
Zero-copy aio_write()'s for connections associated with a given
adapter can be enabled/disabled at runtime via the
'dev.t[45]nex.N.toe.tx_zcopy' sysctl.
MFC after: 1 month
Relnotes: yes
Sponsored by: Chelsio Communications
related to "shared" CPLs.
a) Combine t4_set_tcb_field and t4_set_tcb_field_rpl into a single
function. Allow callers to direct the response to any iq. Tidy up
set_ulp_mode_iscsi while there to use names from t4_tcb.h instead of
magic constants.
b) Remove all CPL handler tables from struct adapter. This reduces its
size by around 2KB. All handlers are now registered at MOD_LOAD instead
of attach or some kind of initialization/activation. The registration
functions do not need an adapter parameter any more.
c) Add per-iq handlers to deal with CPLs whose destination cannot be
determined solely from the opcode. There are 2 such CPLs in use right
now: SET_TCB_RPL and L2T_WRITE_RPL. The base driver continues to send
filter and L2T_WRITEs over the mgmtq and solicits the reply on fwq.
t4_tom (including the DDP code) now uses the port's ctrlq to send
L2T_WRITEs and SET_TCB_FIELDs and solicits the reply on an ofld_rxq.
fwq and ofld_rxq have different handlers that know what kind of tid to
expect in the reply. Update t4_write_l2e and callers to to support any
wrq/iq combination.
Approved by: re@ (kib@)
Sponsored by: Chelsio Communications
The interface's queues are functional after VI_INIT_DONE (which is short
of interface-up) and that's all that's needed for t4_tom to communicate
with the chip.
Approved by: re@ (gjb@)
Sponsored by: Chelsio Communications
File and disk-backed I/O requests store counts of read/written disk
blocks in each AIO job so that they can be charged to the thread that
completes an AIO request via aio_return() or aio_waitcomplete(). This
change extends AIO jobs to store counts of received/sent messages and
updates socket backends to set these counts accordingly. Note that
the socket backends are careful to only charge a single messages for
each AIO request even though a single request on a blocking socket might
invoke sosend or soreceive multiple times. This is to mimic the
resource accounting of synchronous read/write.
Adjust the UNIX socketpair AIO test to verify that the message resource
usage counts update accordingly for aio_read and aio_write.
Approved by: re (hrs)
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D6911
Inserting a full mbuf with an external cluster into the socket buffer
resulted in sbspace() returning -MLEN. However, since sb_hiwat is
unsigned, the -MLEN value was converted to unsigned in comparisons. As a
result, the socket buffer was never autosized. Note that sb_lowat is signed
to permit direct comparisons with sbspace(), but sb_hiwat is unsigned.
Follow suit with what tcp_output() does and compare the value of sbused()
with sb_hiwat instead.
Approved by: re (gjb)
Sponsored by: Chelsio Communications
This reduces the size of kaiocb slightly. I've also added some generic
fields that other backends can use in place of the BIO-specific fields.
Change the socket and Chelsio DDP backends to use 'backend3' instead of
abusing _aiocb_private.status directly. This confines the use of
_aiocb_private to the AIO internals in vfs_aio.c.
Reviewed by: kib (earlier version)
Approved by: re (gjb)
Sponsored by: Chelsio Communications
Differential Revision: https://reviews.freebsd.org/D6547
avoid panicking debug kernels.
t4_tom does not keep track of a connection once it switches to ULP mode
iWARP. If the connection falls out of ULP mode the driver/hardware seq#
etc. are out of sync. A better fix would be to figure out what the
current seq# are, update the driver's state, and perform all sanity
checks as usual.
Chelsio's TCP offload engine supports direct DMA of received TCP payload
into wired user buffers. This feature is known as Direct-Data Placement.
However, to scale well the adapter needs to prepare buffers for DDP
before data arrives. aio_read() is more amenable to this requirement than
read() as applications often call read() only after data is available in
the socket buffer.
When DDP is enabled, TOE sockets use the recently added pru_aio_queue
protocol hook to claim aio_read(2) requests instead of letting them use
the default AIO socket logic. The DDP feature supports scheduling DMA
to two buffers at a time so that the second buffer is ready for use
after the first buffer is filled. The aio/DDP code optimizes the case
of an application ping-ponging between two buffers (similar to the
zero-copy bpf(4) code) by keeping the two most recently used AIO buffers
wired. If a buffer is reused, the aio/DDP code is able to reuse the
vm_page_t array as well as page pod mappings (a kind of MMU mapping the
Chelsio NIC uses to describe user buffers). The generation of the
vmspace of the calling process is used in conjunction with the user
buffer's address and length to determine if a user buffer matches a
previously used buffer. If an application queues a buffer for AIO that
does not match a previously used buffer then the least recently used
buffer is unwired before the new buffer is wired. This ensures that no
more than two user buffers per socket are ever wired.
Note that this feature is best suited to applications sending a steady
stream of data vs short bursts of traffic.
Discussed with: np
Relnotes: yes
Sponsored by: Chelsio Communications
and t_maxseg. This dualism emerged with T/TCP, but was not properly cleaned
up after T/TCP removal. After all permutations over the years the result is
that t_maxopd stores a minimum of peer offered MSS and MTU reduced by minimum
protocol header. And t_maxseg stores (t_maxopd - TCPOLEN_TSTAMP_APPA) if
timestamps are in action, or is equal to t_maxopd otherwise. That's a very
rough estimate of MSS reduced by options length. Throughout the code it
was used in places, where preciseness was not important, like cwnd or
ssthresh calculations.
With this change:
- t_maxopd goes away.
- t_maxseg now stores MSS not adjusted by options.
- new function tcp_maxseg() is provided, that calculates MSS reduced by
options length. The functions gives a better estimate, since it takes
into account SACK state as well.
Reviewed by: jtl
Differential Revision: https://reviews.freebsd.org/D3593
Add if_requestencap() interface method which is capable of calculating
various link headers for given interface. Right now there is support
for INET/INET6/ARP llheader calculation (IFENCAP_LL type request).
Other types are planned to support more complex calculation
(L2 multipath lagg nexthops, tunnel encap nexthops, etc..).
Reshape 'struct route' to be able to pass additional data (with is length)
to prepend to mbuf.
These two changes permits routing code to pass pre-calculated nexthop data
(like L2 header for route w/gateway) down to the stack eliminating the
need for other lookups. It also brings us closer to more complex scenarios
like transparently handling MPLS nexthops and tunnel interfaces.
Last, but not least, it removes layering violation introduced by flowtable
code (ro_lle) and simplifies handling of existing if_output consumers.
ARP/ND changes:
Make arp/ndp stack pre-calculate link header upon installing/updating lle
record. Interface link address change are handled by re-calculating
headers for all lles based on if_lladdr event. After these changes,
arpresolve()/nd6_resolve() returns full pre-calculated header for
supported interfaces thus simplifying if_output().
Move these lookups to separate ether_resolve_addr() function which ether
returs error or fully-prepared link header. Add <arp|nd6_>resolve_addr()
compat versions to return link addresses instead of pre-calculated data.
BPF changes:
Raw bpf writes occupied _two_ cases: AF_UNSPEC and pseudo_AF_HDRCMPLT.
Despite the naming, both of there have ther header "complete". The only
difference is that interface source mac has to be filled by OS for
AF_UNSPEC (controlled via BIOCGHDRCMPLT). This logic has to stay inside
BPF and not pollute if_output() routines. Convert BPF to pass prepend data
via new 'struct route' mechanism. Note that it does not change
non-optimized if_output(): ro_prepend handling is purely optional.
Side note: hackish pseudo_AF_HDRCMPLT is supported for ethernet and FDDI.
It is not needed for ethernet anymore. The only remaining FDDI user is
dev/pdq mostly untouched since 2007. FDDI support was eliminated from
OpenBSD in 2013 (sys/net/if_fddisubr.c rev 1.65).
Flowtable changes:
Flowtable violates layering by saving (and not correctly managing)
rtes/lles. Instead of passing lle pointer, pass pointer to pre-calculated
header data from that lle.
Differential Revision: https://reviews.freebsd.org/D4102
Each virtual interface has its own MAC address, queues, and statistics.
The dedicated netmap interfaces (ncxgbeX / ncxlX) were already implemented
as additional VIs on each port. This change allows additional non-netmap
interfaces to be configured on each port. Additional virtual interfaces
use the naming scheme vcxgbeX or vcxlX.
Additional VIs are enabled by setting the hw.cxgbe.num_vis tunable to a
value greater than 1 before loading the cxgbe(4) or cxl(4) driver.
NB: The first VI on each port is the "main" interface (cxgbeX or cxlX).
T4/T5 NICs provide a limited number of MAC addresses for each physical port.
As a result, a maximum of six VIs can be configured on each port (including
the "main" interface and the netmap interface when netmap is enabled).
One user-visible result is that when netmap is enabled, packets received
or transmitted via the netmap interface are no longer counted in the stats
for the "main" interface, but are not accounted to the netmap interface.
The netmap interfaces now also have a new-bus device and export various
information sysctl nodes via dev.n(cxgbe|cxl).X.
The cxgbetool 'clearstats' command clears the stats for all VIs on the
specified port along with the port's stats. There is currently no way to
clear the stats of an individual VI.
Reviewed by: np
MFC after: 1 month
Sponsored by: Chelsio
- The existing TCP INP_INFO lock continues to protect the global inpcb list
stability during full list traversal (e.g. tcp_pcblist()).
- A new INP_LIST lock protects inpcb list actual modifications (inp allocation
and free) and inpcb global counters.
It allows to use TCP INP_INFO_RLOCK lock in critical paths (e.g. tcp_input())
and INP_INFO_WLOCK only in occasional operations that walk all connections.
PR: 183659
Differential Revision: https://reviews.freebsd.org/D2599
Reviewed by: jhb, adrian
Tested by: adrian, nitroboost-gmail.com
Sponsored by: Verisign, Inc.
Both are used to protect access to IP addresses lists and they can be
acquired for reading several times per packet. To reduce lock contention
it is better to use rmlock here.
Reviewed by: gnn (previous version)
Obtained from: Yandex LLC
Sponsored by: Yandex LLC
Differential Revision: https://reviews.freebsd.org/D3149
handle_ddp_close() function in t4_ddp.c as the logic is similar
to handle_ddp_data(). This allows all knowledge of the special
DDP mbufs to be private to t4_ddp.c as well.
Drivers (ULDs) and the base if_cxgbe driver.
Track the per-adapter activation of ULDs in a new "active_ulds" field.
This was done pretty arbitrarily before this change -- via TOM_INIT_DONE
in adapter->flags for TOM, and the (1 << MAX_NPORTS) bit in
adapter->offload_map for iWARP.
iWARP and hw-accelerated iSCSI rely on the TOE (supported by the TOM
ULD). The rules are:
a) If the iWARP and/or iSCSI ULDs are available when TOE is enabled then
iWARP and/or iSCSI are enabled too.
b) When the iWARP and iSCSI modules are loaded they go looking for
adapters with TOE enabled and enable themselves on that adapter.
c) You cannot deactivate or unload the TOM module from underneath iWARP
or iSCSI. Any such attempt will fail with EBUSY.
MFC after: 2 weeks
the no-longer existant sb_cc sockbuf member.
- Use sbavail() instead of sbused() in t4_soreceive_ddp() to match the
usage in soreceive_stream() on which it is based.
Discussed with: glebius (2)
sending not ready data:
o Add new flag to pru_send() flags - PRUS_NOTREADY.
o Add new protocol method pru_ready().
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
o Introduce a notion of "not ready" mbufs in socket buffers. These
mbufs are now being populated by some I/O in background and are
referenced outside. This forces following implications:
- An mbuf which is "not ready" can't be taken out of the buffer.
- An mbuf that is behind a "not ready" in the queue neither.
- If sockbet buffer is flushed, then "not ready" mbufs shouln't be
freed.
o In struct sockbuf the sb_cc field is split into sb_ccc and sb_acc.
The sb_ccc stands for ""claimed character count", or "committed
character count". And the sb_acc is "available character count".
Consumers of socket buffer API shouldn't already access them directly,
but use sbused() and sbavail() respectively.
o Not ready mbufs are marked with M_NOTREADY, and ready but blocked ones
with M_BLOCKED.
o New field sb_fnrdy points to the first not ready mbuf, to avoid linear
search.
o New function sbready() is provided to activate certain amount of mbufs
in a socket buffer.
A special note on SCTP:
SCTP has its own sockbufs. Unfortunately, FreeBSD stack doesn't yet
allow protocol specific sockbufs. Thus, SCTP does some hacks to make
itself compatible with FreeBSD: it manages sockbufs on its own, but keeps
sb_cc updated to inform the stack of amount of data in them. The new
notion of "not ready" data isn't supported by SCTP. Instead, only a
mechanical substitute is done: s/sb_cc/sb_ccc/.
A proper solution would be to take away struct sockbuf from struct
socket and allow protocols to implement their own socket buffers, like
SCTP already does. This was discussed with rrs@.
Sponsored by: Netflix
Sponsored by: Nginx, Inc.
sb_cc member of struct sockbuf to a couple of inline functions:
sbavail() and sbused()
Right now they are equal, but once notion of "not ready socket buffer data",
will be checked in, they are going to be different.
Sponsored by: Netflix
Sponsored by: Nginx, Inc.
initial MPA exchange must be tracked this way so that t4_tom's state for
the tid is all clean at the time the tid transitions to RDMA mode. Once
it does, t4_tom is out of the way and iw_cxgbe uses the qp endpoints
directly.
Sponsored by: Chelsio Communications
cannot be sent to the chip because a prerequisite L2 resolution
failed.
Submitted by: Hariprasad at chelsio dot com (original version)
MFC after: 2 weeks.
