From: Alexander Lobakin <alexandr.lobakin@intel.com>
To: "David S. Miller" <davem@davemloft.net>,
Eric Dumazet <edumazet@google.com>,
Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>
Cc: Alexander Lobakin <alexandr.lobakin@intel.com>,
Yury Norov <yury.norov@gmail.com>,
Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
Michal Swiatkowski <michal.swiatkowski@linux.intel.com>,
Rasmus Villemoes <linux@rasmusvillemoes.dk>,
Nikolay Aleksandrov <razor@blackwall.org>,
netdev@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: [PATCH net-next 0/4] netlink: add 'bitmap' attribute type and API
Date: Thu, 21 Jul 2022 17:59:46 +0200 [thread overview]
Message-ID: <20220721155950.747251-1-alexandr.lobakin@intel.com> (raw)
Add a new type of Netlink attribute -- bitmap.
Lots of different bitfields in the kernel grow through time,
sometimes significantly. They can consume a u8 at fist, then require
16 bits, 32... To support such bitfields without rewriting code each
time, kernel have bitmaps. For now, that is purely in-kernel type
and API, and to communicate with userspace, they need to be
converted to some primitive at first (e.g. __u32 etc.), dealing with
that sometimes bitfields will run out of the size set for the
corresponding Netlink attribute. Those can be netdev features,
linkmodes and so on.
Internally, in-kernel bitmaps are represented as arrays of unsigned
longs. This provides optimal performance and memory usage; however,
bitness dependent types can't be used to communicate between kernel
and userspace -- for example, userapp can be 32-bit on a 64-bit
system. So, to provide reliable communication data type, 64-bit
arrays are now used. Netlink core takes care of converting them
from/to unsigned longs when sending or receiving Netlink messages;
although, on LE and 64-bit systems conversion is a no-op. They also
can have explicit byteorder -- core code also handles this (both
kernel and userspace must know in advance the byteorder of a
particular attribute), as well as cases when the userspace and the
kernel assume different number of bits (-> different number of u64s)
for an attribute.
Basic consumer functions/macros are:
* nla_put_bitmap and nla_get_bitmap families -- to easily put a
bitmap to an skb or get it from a received message (only pointer
to an unsigned long bitmap and the number of bits in it are
needed), with optional explicit byteorder;
* nla_total_size_bitmap() -- to provide estimate size in bytes to
Netlink needed to store a bitmap;
* {,__}NLA_POLICY_BITMAP() -- to declare a Netlink policy for a
bitmap attribute.
Netlink policy for a bitmap can have an optional bitmap mask of bits
supported by the code -- for example, to filter out obsolete bits
removed some time ago. Without it, Netlink will make sure no bits
past the passed number are set. Both variants can be requested from
the userspace and the kernel will put a mask into a new policy
attribute (%NL_POLICY_TYPE_ATTR_BITMAP_MASK).
BTW, Ethtool bitsets provide similar functionality, but it operates
with u32s (u64 is more convenient and optimal on most platforms) and
Netlink bitmaps is a generic interface providing policies and data
verification (Ethtool bitsets are declared simply as %NLA_BINARY),
generic getters/setters etc.
An example of using this API can be found in my IP tunnel tree[0]
(planned for submission later), the actual average number of locs
to start both sending and receiving bitmaps in one subsys is ~10[1].
And it looks like that some of the already existing APIs could be
later converted to Netlink bitmaps or expanded as well.
[0] https://github.com/alobakin/linux/commits/ip_tunnel
[1] https://github.com/alobakin/linux/commit/26d2f2cf13fd
Alexander Lobakin (4):
bitmap: add converting from/to 64-bit arrays of explicit byteorder
bitmap: add a couple more helpers to work with arrays of u64s
lib/test_bitmap: cover explicitly byteordered arr64s
netlink: add 'bitmap' attribute type (%NL_ATTR_TYPE_BITMAP /
%NLA_BITMAP)
include/linux/bitmap.h | 80 ++++++++++++++++--
include/net/netlink.h | 159 ++++++++++++++++++++++++++++++++++-
include/uapi/linux/netlink.h | 5 ++
lib/bitmap.c | 143 +++++++++++++++++++++++++++----
lib/nlattr.c | 43 +++++++++-
lib/test_bitmap.c | 22 +++--
net/netlink/policy.c | 44 ++++++++++
7 files changed, 461 insertions(+), 35 deletions(-)
base-commit: 3a2ba42cbd0b669ce3837ba400905f93dd06c79f
---
Targeted for net-next, but uses base-commit from the bitmap tree
(a merge will be needed) and the API is for kernel-wide/generic
usage rather than networking-specific.
--
2.36.1
next reply other threads:[~2022-07-21 16:02 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-21 15:59 Alexander Lobakin [this message]
2022-07-21 15:59 ` [PATCH net-next 1/4] bitmap: add converting from/to 64-bit arrays of explicit byteorder Alexander Lobakin
2022-07-21 15:59 ` [PATCH net-next 2/4] bitmap: add a couple more helpers to work with arrays of u64s Alexander Lobakin
2022-07-21 15:59 ` [PATCH net-next 3/4] lib/test_bitmap: cover explicitly byteordered arr64s Alexander Lobakin
2022-07-21 15:59 ` [PATCH net-next 4/4] netlink: add 'bitmap' attribute type (%NL_ATTR_TYPE_BITMAP / %NLA_BITMAP) Alexander Lobakin
2022-07-21 18:13 ` [PATCH net-next 0/4] netlink: add 'bitmap' attribute type and API Jakub Kicinski
2022-07-22 14:55 ` Alexander Lobakin
2022-07-22 18:19 ` Jakub Kicinski
2022-07-23 15:52 ` Jakub Kicinski
2022-07-25 13:02 ` Alexander Lobakin
2022-07-25 18:53 ` Jakub Kicinski
2022-07-26 10:41 ` Alexander Lobakin
[not found] ` <CAHp75Ve7oXjNyc0GD5x9ZW=DVgCqmLOBfCP4O2cDi2DG=4SiwQ@mail.gmail.com>
2022-07-25 10:24 ` Alexander Lobakin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20220721155950.747251-1-alexandr.lobakin@intel.com \
--to=alexandr.lobakin@intel.com \
--cc=andriy.shevchenko@linux.intel.com \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux@rasmusvillemoes.dk \
--cc=michal.swiatkowski@linux.intel.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=razor@blackwall.org \
--cc=yury.norov@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®