| Message ID | 20240306000655.1100294-1-dhowells@redhat.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | rxrpc: Miscellaneous changes and make use of MSG_SPLICE_PAGES | expand |
On Wed, Mar 06, 2024 at 12:06:30AM +0000, David Howells wrote: > Here are some changes to AF_RXRPC: > > (1) Cache the transmission serial number of ACK and DATA packets in the > rxrpc_txbuf struct and log this in the retransmit tracepoint. > > (2) Don't use atomics on rxrpc_txbuf::flags[*] and cache the intended wire > header flags there too to avoid duplication. > > (3) Cache the wire checksum in rxrpc_txbuf to make it easier to create > jumbo packets in future (which will require altering the wire header > to a jumbo header and restoring it back again for retransmission). > > (4) Fix the protocol names in the wire ACK trailer struct. > > (5) Strip all the barriers and atomics out of the call timer tracking[*]. > > (6) Remove atomic handling from call->tx_transmitted and > call->acks_prev_seq[*]. > > (7) Don't bother resetting the DF flag after UDP packet transmission. To > change it, we now call directly into UDP code, so it's quick just to > set it every time. > > (8) Merge together the DF/non-DF branches of the DATA transmission to > reduce duplication in the code. > > (9) Add a kvec array into rxrpc_txbuf and start moving things over to it. > This paves the way for using page frags. > > (10) Split (sub)packet preparation and timestamping out of the DATA > transmission function. This helps pave the way for future jumbo > packet generation. > > (11) In rxkad, don't pick values out of the wire header stored in > rxrpc_txbuf, buf rather find them elsewhere so we can remove the wire > header from there. > > (12) Move rxrpc_send_ACK() to output.c so that it can be merged with > rxrpc_send_ack_packet(). > > (13) Use rxrpc_txbuf::kvec[0] to access the wire header for the packet > rather than directly accessing the copy in rxrpc_txbuf. This will > allow that to be removed to a page frag. > > (14) Switch from keeping the transmission buffers in rxrpc_txbuf allocated > in the slab to allocating them using page fragment allocators. There > are separate allocators for DATA packets (which persist for a while) > and control packets (which are discarded immediately). > > We can then turn on MSG_SPLICE_PAGES when transmitting DATA and ACK > packets. > > We can also get rid of the RCU cleanup on rxrpc_txbufs, preferring > instead to release the page frags as soon as possible. > > (15) Parse received packets before handling timeouts as the former may > reset the latter. > > (16) Make sure we don't retransmit DATA packets after all the packets have > been ACK'd. > > (17) Differentiate traces for PING ACK transmission. > > (18) Switch to keeping timeouts as ktime_t rather than a number of jiffies > as the latter is too coarse a granularity. Only set the call timer at > the end of the call event function from the aggregate of all the > timeouts, thereby reducing the number of timer calls made. In future, > it might be possible to reduce the number of timers from one per call > to one per I/O thread and to use a high-precision timer. > > (19) Record RTT probes after successful transmission rather than recording > it before and then cancelling it after if unsuccessful[*]. This > allows a number of calls to get the current time to be removed. > > (20) Clean up the resend algorithm as there's now no need to walk the > transmission buffer under lock[*]. DATA packets can be retransmitted > as soon as they're found rather than being queued up and transmitted > when the locked is dropped. > > (21) When initially parsing a received ACK packet, extract some of the > fields from the ack info to the skbuff private data. This makes it > easier to do path MTU discovery in the future when the call to which a > PING RESPONSE ACK refers has been deallocated. > > > [*] Possible with the move of almost all code from softirq context to the > I/O thread. > > The patches are tagged here: > > git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git tags/rxrpc-iothread-20240305 > > And can be found on this branch: > > http://git.kernel.org/cgit/linux/kernel/git/dhowells/linux-fs.git/log/?h=rxrpc-iothread > > David > > Link: https://lore.kernel.org/r/20240301163807.385573-1-dhowells@redhat.com/ # v1 > Link: https://lore.kernel.org/r/20240304084322.705539-1-dhowells@redhat.com/ # v2 > > Changes > ======= > ver #3) > - Use passed-in gfp in rxkad_alloc_txbuf() rather than GFP_KRNEL. Hi David, Thanks for the update above. For the record, I don't have anything to flag in this revision of the patchset. > - Adjust rxkad_alloc_txbuf()'s txb check to put return in if-statement. > > ver #2) > - Removed an unused variable. > - Use ktime_to_us() rather than dividing a ktime by 1000 in tracepoints. ...
Hello: This series was applied to netdev/net-next.git (main) by David Howells <dhowells@redhat.com>: On Wed, 6 Mar 2024 00:06:30 +0000 you wrote: > Here are some changes to AF_RXRPC: > > (1) Cache the transmission serial number of ACK and DATA packets in the > rxrpc_txbuf struct and log this in the retransmit tracepoint. > > (2) Don't use atomics on rxrpc_txbuf::flags[*] and cache the intended wire > header flags there too to avoid duplication. > > [...] Here is the summary with links: - [net-next,v3,01/21] rxrpc: Record the Tx serial in the rxrpc_txbuf and retransmit trace https://git.kernel.org/netdev/net-next/c/ba132d841d56 - [net-next,v3,02/21] rxrpc: Convert rxrpc_txbuf::flags into a mask and don't use atomics https://git.kernel.org/netdev/net-next/c/12bdff73a147 - [net-next,v3,03/21] rxrpc: Note cksum in txbuf https://git.kernel.org/netdev/net-next/c/41b8debba79c - [net-next,v3,04/21] rxrpc: Fix the names of the fields in the ACK trailer struct https://git.kernel.org/netdev/net-next/c/17469ae0582a - [net-next,v3,05/21] rxrpc: Strip barriers and atomics off of timer tracking https://git.kernel.org/netdev/net-next/c/d73f3a748875 - [net-next,v3,06/21] rxrpc: Remove atomic handling on some fields only used in I/O thread https://git.kernel.org/netdev/net-next/c/693f9c13ec89 - [net-next,v3,07/21] rxrpc: Do lazy DF flag resetting https://git.kernel.org/netdev/net-next/c/d32636982ce9 - [net-next,v3,08/21] rxrpc: Merge together DF/non-DF branches of data Tx function https://git.kernel.org/netdev/net-next/c/1ac6a8536c2c - [net-next,v3,09/21] rxrpc: Add a kvec[] to the rxrpc_txbuf struct https://git.kernel.org/netdev/net-next/c/ff342bdc59f4 - [net-next,v3,10/21] rxrpc: Split up the DATA packet transmission function https://git.kernel.org/netdev/net-next/c/44125d5aadda - [net-next,v3,11/21] rxrpc: Don't pick values out of the wire header when setting up security https://git.kernel.org/netdev/net-next/c/a1c9af4d4467 - [net-next,v3,12/21] rxrpc: Move rxrpc_send_ACK() to output.c with rxrpc_send_ack_packet() https://git.kernel.org/netdev/net-next/c/99afb28c676c - [net-next,v3,13/21] rxrpc: Use rxrpc_txbuf::kvec[0] instead of rxrpc_txbuf::wire https://git.kernel.org/netdev/net-next/c/8985f2b09b33 - [net-next,v3,14/21] rxrpc: Do zerocopy using MSG_SPLICE_PAGES and page frags https://git.kernel.org/netdev/net-next/c/49489bb03a50 - [net-next,v3,15/21] rxrpc: Parse received packets before dealing with timeouts https://git.kernel.org/netdev/net-next/c/3e0b83ee535d - [net-next,v3,16/21] rxrpc: Don't permit resending after all Tx packets acked https://git.kernel.org/netdev/net-next/c/a711d976e1cd - [net-next,v3,17/21] rxrpc: Differentiate PING ACK transmission traces. https://git.kernel.org/netdev/net-next/c/12a66e77c499 - [net-next,v3,18/21] rxrpc: Use ktimes for call timeout tracking and set the timer lazily https://git.kernel.org/netdev/net-next/c/153f90a066dd - [net-next,v3,19/21] rxrpc: Record probes after transmission and reduce number of time-gets https://git.kernel.org/netdev/net-next/c/4d267ad6fd56 - [net-next,v3,20/21] rxrpc: Clean up the resend algorithm https://git.kernel.org/netdev/net-next/c/37473e416234 - [net-next,v3,21/21] rxrpc: Extract useful fields from a received ACK to skb priv data https://git.kernel.org/netdev/net-next/c/4b68137a20bc You are awesome, thank you!
Here are some changes to AF_RXRPC: (1) Cache the transmission serial number of ACK and DATA packets in the rxrpc_txbuf struct and log this in the retransmit tracepoint. (2) Don't use atomics on rxrpc_txbuf::flags[*] and cache the intended wire header flags there too to avoid duplication. (3) Cache the wire checksum in rxrpc_txbuf to make it easier to create jumbo packets in future (which will require altering the wire header to a jumbo header and restoring it back again for retransmission). (4) Fix the protocol names in the wire ACK trailer struct. (5) Strip all the barriers and atomics out of the call timer tracking[*]. (6) Remove atomic handling from call->tx_transmitted and call->acks_prev_seq[*]. (7) Don't bother resetting the DF flag after UDP packet transmission. To change it, we now call directly into UDP code, so it's quick just to set it every time. (8) Merge together the DF/non-DF branches of the DATA transmission to reduce duplication in the code. (9) Add a kvec array into rxrpc_txbuf and start moving things over to it. This paves the way for using page frags. (10) Split (sub)packet preparation and timestamping out of the DATA transmission function. This helps pave the way for future jumbo packet generation. (11) In rxkad, don't pick values out of the wire header stored in rxrpc_txbuf, buf rather find them elsewhere so we can remove the wire header from there. (12) Move rxrpc_send_ACK() to output.c so that it can be merged with rxrpc_send_ack_packet(). (13) Use rxrpc_txbuf::kvec[0] to access the wire header for the packet rather than directly accessing the copy in rxrpc_txbuf. This will allow that to be removed to a page frag. (14) Switch from keeping the transmission buffers in rxrpc_txbuf allocated in the slab to allocating them using page fragment allocators. There are separate allocators for DATA packets (which persist for a while) and control packets (which are discarded immediately). We can then turn on MSG_SPLICE_PAGES when transmitting DATA and ACK packets. We can also get rid of the RCU cleanup on rxrpc_txbufs, preferring instead to release the page frags as soon as possible. (15) Parse received packets before handling timeouts as the former may reset the latter. (16) Make sure we don't retransmit DATA packets after all the packets have been ACK'd. (17) Differentiate traces for PING ACK transmission. (18) Switch to keeping timeouts as ktime_t rather than a number of jiffies as the latter is too coarse a granularity. Only set the call timer at the end of the call event function from the aggregate of all the timeouts, thereby reducing the number of timer calls made. In future, it might be possible to reduce the number of timers from one per call to one per I/O thread and to use a high-precision timer. (19) Record RTT probes after successful transmission rather than recording it before and then cancelling it after if unsuccessful[*]. This allows a number of calls to get the current time to be removed. (20) Clean up the resend algorithm as there's now no need to walk the transmission buffer under lock[*]. DATA packets can be retransmitted as soon as they're found rather than being queued up and transmitted when the locked is dropped. (21) When initially parsing a received ACK packet, extract some of the fields from the ack info to the skbuff private data. This makes it easier to do path MTU discovery in the future when the call to which a PING RESPONSE ACK refers has been deallocated. [*] Possible with the move of almost all code from softirq context to the I/O thread. The patches are tagged here: git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git tags/rxrpc-iothread-20240305 And can be found on this branch: http://git.kernel.org/cgit/linux/kernel/git/dhowells/linux-fs.git/log/?h=rxrpc-iothread David Link: https://lore.kernel.org/r/20240301163807.385573-1-dhowells@redhat.com/ # v1 Link: https://lore.kernel.org/r/20240304084322.705539-1-dhowells@redhat.com/ # v2 Changes ======= ver #3) - Use passed-in gfp in rxkad_alloc_txbuf() rather than GFP_KRNEL. - Adjust rxkad_alloc_txbuf()'s txb check to put return in if-statement. ver #2) - Removed an unused variable. - Use ktime_to_us() rather than dividing a ktime by 1000 in tracepoints. David Howells (21): rxrpc: Record the Tx serial in the rxrpc_txbuf and retransmit trace rxrpc: Convert rxrpc_txbuf::flags into a mask and don't use atomics rxrpc: Note cksum in txbuf rxrpc: Fix the names of the fields in the ACK trailer struct rxrpc: Strip barriers and atomics off of timer tracking rxrpc: Remove atomic handling on some fields only used in I/O thread rxrpc: Do lazy DF flag resetting rxrpc: Merge together DF/non-DF branches of data Tx function rxrpc: Add a kvec[] to the rxrpc_txbuf struct rxrpc: Split up the DATA packet transmission function rxrpc: Don't pick values out of the wire header when setting up security rxrpc: Move rxrpc_send_ACK() to output.c with rxrpc_send_ack_packet() rxrpc: Use rxrpc_txbuf::kvec[0] instead of rxrpc_txbuf::wire rxrpc: Do zerocopy using MSG_SPLICE_PAGES and page frags rxrpc: Parse received packets before dealing with timeouts rxrpc: Don't permit resending after all Tx packets acked rxrpc: Differentiate PING ACK transmission traces. rxrpc: Use ktimes for call timeout tracking and set the timer lazily rxrpc: Record probes after transmission and reduce number of time-gets rxrpc: Clean up the resend algorithm rxrpc: Extract useful fields from a received ACK to skb priv data include/trace/events/rxrpc.h | 198 ++++++++-------- net/rxrpc/af_rxrpc.c | 12 +- net/rxrpc/ar-internal.h | 88 ++++--- net/rxrpc/call_event.c | 327 ++++++++++++-------------- net/rxrpc/call_object.c | 56 ++--- net/rxrpc/conn_client.c | 4 +- net/rxrpc/conn_event.c | 16 +- net/rxrpc/conn_object.c | 4 + net/rxrpc/input.c | 116 +++++---- net/rxrpc/insecure.c | 11 +- net/rxrpc/io_thread.c | 11 + net/rxrpc/local_object.c | 3 + net/rxrpc/misc.c | 8 +- net/rxrpc/output.c | 441 +++++++++++++++++------------------ net/rxrpc/proc.c | 10 +- net/rxrpc/protocol.h | 6 +- net/rxrpc/rtt.c | 36 +-- net/rxrpc/rxkad.c | 57 ++--- net/rxrpc/sendmsg.c | 63 ++--- net/rxrpc/sysctl.c | 16 +- net/rxrpc/txbuf.c | 174 +++++++++++--- 21 files changed, 853 insertions(+), 804 deletions(-)