[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <157323719180.10408.3472322881536070517.stgit@firesoul>
Date: Fri, 08 Nov 2019 19:20:17 +0100
From: Jesper Dangaard Brouer <brouer@...hat.com>
To: unlisted-recipients:; (no To-header on input)
Cc: Toke Høiland-Jørgensen <toke@...hat.com>,
netdev@...r.kernel.org,
Ilias Apalodimas <ilias.apalodimas@...aro.org>,
Jesper Dangaard Brouer <brouer@...hat.com>,
Saeed Mahameed <saeedm@...lanox.com>,
Matteo Croce <mcroce@...hat.com>,
Jonathan Lemon <jonathan.lemon@...il.com>,
Lorenzo Bianconi <lorenzo@...nel.org>,
Tariq Toukan <tariqt@...lanox.com>
Subject: [net-next v1 PATCH 0/2] Change XDP lifetime guarantees for
page_pool objects
This patchset change XDP lifetime guarantees for page_pool objects
(registered via the xdp_rxq_info_reg_mem_model API). The new guarantee
is that page_pool objects stay valid (are not free'ed) until all
inflight pages are returned.
It was commit d956a048cd3f (“xdp: force mem allocator removal and
periodic warning”) that introduce the force removal of page_pool
objects (after 120 second). While working on extending page_pool
recycling to cover SKBs[1], we[2] realised that this force removal
approach was a mistake.
Tested and monitored via bpftrace scripts provide here[3].
[1] https://github.com/xdp-project/xdp-project/tree/master/areas/mem
[2] we == Ilias, Jonathan, Tariq, Saeed and me
[3] https://github.com/xdp-project/xdp-project/tree/master/areas/mem/bpftrace
---
Jesper Dangaard Brouer (2):
xdp: revert forced mem allocator removal for page_pool
page_pool: make inflight returns more robust via blocking alloc cache
include/net/page_pool.h | 2 ++
include/trace/events/xdp.h | 35 +++--------------------------------
net/core/page_pool.c | 32 +++++++++++++++++++++++++-------
net/core/xdp.c | 36 +++++++++++++-----------------------
4 files changed, 43 insertions(+), 62 deletions(-)
--
Powered by blists - more mailing lists