[PATCH net-next 00/13] rxrpc: Increasing SACK size and moving away from softirq, part 5

From: David Howells
Date: Tue Jan 31 2023 - 12:13:51 EST


Here's the fifth part of patches in the process of moving rxrpc from doing
a lot of its stuff in softirq context to doing it in an I/O thread in
process context and thereby making it easier to support a larger SACK
table.

The full description is in the description for the first part[1] which is
now upstream. The second and third parts are also upstream[2]. A subset
of the original fourth part[3] got applied as a fix for a race[4].

The fifth part includes some cleanups:

(1) Miscellaneous trace header cleanups: fix a trace string, display the
security index in rx_packet rather than displaying the type twice,
remove some whitespace to make checkpatch happier and remove some
excess tabulation.

(2) Convert ->recvmsg_lock to a spinlock as it's only ever locked
exclusively.

(3) Make ->ackr_window and ->ackr_nr_unacked non-atomic as they're only
used in the I/O thread.

(4) Don't use call->tx_lock to access ->tx_buffer as that is only accessed
inside the I/O thread. sendmsg() loads onto ->tx_sendmsg and the I/O
thread decants from that to the buffer.

(5) Remove local->defrag_sem as DATA packets are transmitted serially by
the I/O thread.

(6) Remove the service connection bundle is it was only used for its
channel_lock - which has now gone.

And some more significant changes:

(7) Add a debugging option to allow a delay to be injected into packet
reception to help investigate the behaviour over longer links than
just a few cm.

(8) Generate occasional PING ACKs to probe for RTT information during a
receive heavy call.

(9) Simplify the SACK table maintenance and ACK generation. Now that both
parts are done in the same thread, there's no possibility of a race
and no need to try and be cunning to avoid taking a BH spinlock whilst
copying the SACK table (which in the future will be up to 2K) and no
need to rotate the copy to fit the ACK packet table.

(10) Use SKB_CONSUMED when freeing received DATA packets (stop dropwatch
complaining).

The patches are tagged here:

git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git tags/rxrpc-next-20230131

And can be found on this branch:

http://git.kernel.org/cgit/linux/kernel/git/dhowells/linux-fs.git/log/?h=rxrpc-next

David

Link: https://lore.kernel.org/r/166794587113.2389296.16484814996876530222.stgit@xxxxxxxxxxxxxxxxxxxxxx/ [1]
Link: https://lore.kernel.org/r/166994010342.1732290.13771061038178613124.stgit@xxxxxxxxxxxxxxxxxxxxxx/ [2]
Link: https://lore.kernel.org/r/167034231605.1105287.1693064952174322878.stgit@xxxxxxxxxxxxxxxxxxxxxx/ [3]
Link: https://lore.kernel.org/r/167308517118.1538866.3440481802366869065.stgit@xxxxxxxxxxxxxxxxxxxxxx/ [4]

David Howells (13):
rxrpc: Fix trace string
rxrpc: Remove whitespace before ')' in trace header
rxrpc: Shrink the tabulation in the rxrpc trace header a bit
rxrpc: Convert call->recvmsg_lock to a spinlock
rxrpc: Allow a delay to be injected into packet reception
rxrpc: Generate extra pings for RTT during heavy-receive call
rxrpc: De-atomic call->ackr_window and call->ackr_nr_unacked
rxrpc: Simplify ACK handling
rxrpc: Don't lock call->tx_lock to access call->tx_buffer
rxrpc: Remove local->defrag_sem
rxrpc: Show consumed and freed packets as non-dropped in dropwatch
rxrpc: Change rx_packet tracepoint to display securityIndex not type
twice
rxrpc: Kill service bundle

include/trace/events/rxrpc.h | 480 +++++++++++++++++++----------------
net/rxrpc/Kconfig | 9 +
net/rxrpc/af_rxrpc.c | 2 +-
net/rxrpc/ar-internal.h | 15 +-
net/rxrpc/call_accept.c | 2 +-
net/rxrpc/call_event.c | 15 +-
net/rxrpc/call_object.c | 7 +-
net/rxrpc/conn_service.c | 7 -
net/rxrpc/input.c | 60 ++---
net/rxrpc/io_thread.c | 48 +++-
net/rxrpc/local_object.c | 7 +-
net/rxrpc/misc.c | 7 +
net/rxrpc/output.c | 69 ++---
net/rxrpc/proc.c | 4 +-
net/rxrpc/recvmsg.c | 18 +-
net/rxrpc/skbuff.c | 4 +-
net/rxrpc/sysctl.c | 17 +-
net/rxrpc/txbuf.c | 12 +-
18 files changed, 438 insertions(+), 345 deletions(-)