We reintroduce hashmap.{h,c}, list.h and set.h verbatim from upstream,
before we punt dead code. The following is the upstream message:
This is a rewrite of the hashmap implementation. Its advantage is lower
memory usage.
It uses open addressing (entries are stored in an array, as opposed to
linked lists). Hash collisions are resolved with linear probing and
Robin Hood displacement policy. See the references in hashmap.c.
Some fun empirical findings about hashmap usage in systemd on my laptop:
- 98 % of allocated hashmaps are Sets.
- Sets contain 78 % of all entries, plain Hashmaps 17 %, and
OrderedHashmaps 5 %.
- 60 % of allocated hashmaps contain only 1 entry.
- 90 % of allocated hashmaps contain 5 or fewer entries.
- 75 % of all entries are in hashmaps that use trivial_hash_ops.
Clearly it makes sense to:
- store entries in distinct entry types. Especially for Sets - their
entries are the most numerous and they require the least information
to store an entry.
- have a way to store small numbers of entries directly in the hashmap
structs, and only allocate the usual entry arrays when the direct
storage is full.
The implementation has an optional debugging feature (enabled by
defining the ENABLE_HASHMAP_DEBUG macro), where it:
- tracks all allocated hashmaps in a linked list so that one can
easily find them in gdb,
- tracks which function/line allocated a given hashmap, and
- checks for invalid mixing of hashmap iteration and modification.
Since entries are not allocated one-by-one anymore, mempools are not
used for entries. Originally I meant to drop mempools entirely, but it's
still worth it to use them for the hashmap structs. My testing indicates
that it makes loading of units about 5 % faster (a test with 10000 units
where more than 200000 hashmaps are allocated - pure malloc: 449±4 ms,
mempools: 427±7 ms).
Here are some memory usage numbers, taken on my laptop with a more or
less normal Fedora setup after booting with SELinux disabled (SELinux
increases systemd's memory usage significantly):
systemd (PID 1) Original New Change
dirty memory (from pmap -x 1) [KiB] 2152 1264 -41 %
total heap allocations (from gdb-heap) [KiB] 1623 756 -53 %
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
This is a combination of upstream commits
dced15575f7bf172e3c9d798c43477539525558c
7bed7f0e3b1f9596aedb64657e432e5b267adebb
by
Simon McVittie <simon.mcvittie@collabora.co.uk>
7bed7f0e3b1f9596aedb64657e432e5b267adebb
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
The way the kernel namespaces have been implemented breaks assumptions
udev made regarding uevent sequence numbers. Creating devices in a
namespace "steals" uevents and its sequence numbers from the host. It
confuses the "udevadmin settle" logic, which might block until util a
timeout is reached, even when no uevent is pending.
Remove any assumptions about sequence numbers and deprecate libudev's
API exposing these numbers; none of that can reliably be used anymore
when namespaces are involved.
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
These commits were authored by
Zbigniew Jędrzejewski-Szmek
Tom Gundersen
Kay Sievers
Lennart Poettering
Shawn Landden
Daniel Buch
Martin Pitt
Karel Zak
Yang Zhiyong
Note: udev_builtin_net_setup_link has *not* been imported. Also
still missing from udev-builtin is udev_builtin_uaccess.
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
This reverts commit aa417a4d83.
Preface: The kmod+tmpfiles static dev-node creation requires two commands to
be executed at runtime -- it is not something that will automatically occur
without a system's setup being explicitly designed or changed so that these
commands are executed.
Preface2: In order for the kmod+tmpfiles static dev-node creation to work
properly, that -must- be executed at startup before {systemd-,}udevd starts.
The reason for this is because udevd will only set permissions on those files
at startup, and so if udevd starts beforehand then these nodes will exist with
permissions that are (probably) too restrictive.
The function in udevd which creates static-nodes is non-fatal and only updates
mtime on the devnodes if they already exist. As such, if a system is configured
to execute kmod+tmpfiles to create static-nodes, because that must occur first,
eudev's udevd will not conflict. Also, if a system does not execute kmod+tmpfiles,
then eudev will still create the static devnodes, even if kmod-14 or higher is
installed.
There *may* be a conflict if kmod+tmpfiles is executed after udevd starts, but
as per "preface2" this is not a supported configuration.
Signed-off-by: Ian Stakenvicius <axs@gentoo.org>
This address upstream commit edeb68c53f1cdc452016b4c8512586a70b1262e3
and https://bugs.gentoo.org/show_bug.cgi?id=477890. If eudev is
configured with --enable-libkmod then we check for kmod >= 14 and
ifdef out the code removed in the upstream commit. Otherwise we
retain it for modutils.
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
This commit imports the new internal keyboard handling from upstream.
This is a combination of many upstream commits, including those
that added code, removed old code, and updated the hwdb.
Some commits (hwdb ones specifically) were unrelated but brought
in anyways to keep the whole hwdb consistent. Each upstream
commit included is as follows:
9d7d42bc406a2ac04639674281ce3ff6beeda790 - internal keymap support
0c959b39175b126fdb70ae00de37ca6d9c8ca3a1 - hwdb: keyboard -- add file
e8193554925a22b63bef0e77b8397b56d63a91ff - hwdb: keyboard -- update comments
c79d894d590fc9df4861738555cc43c477e33376 - hwdb: import data
aedc2eddd16e48d468e6ad0aea2caf00c7d37365 - hwdb: keyboard update
97a9313cafccf772ce03f5ebd36fe4d9d8412583 - hwdb: drop non-existant Samsung 900XC3 from keymap
ddc77f62244bb41d5c8261517e2e1ff1b763fc94 - switch from udev keymaps to hwdb
0c3815773331b263713f4f7b9d80bc1ca159338e - also remove keymaps-force-release directory
1b6bce89b3383904d0dab619dd38bff673f7286e - keymap: re-add Logitech USB corded/cordless models
bf89b99c5a39115112c2eda4c2103e2db54988d2 - 60-keyboard.hwdb: Fix syntax error
ce39bb6909578017aa10031638e724e038f0b859 - hwdb: data update, upstream
884c86812c51479496edd50b278383d7bb67baf0 - rules: keyboard - use builtin command
All code from each of the above commits is attributed to the original
authors.
There were some adjustments made in order to support the code differences
between upstream and eudev, which was done by myself.
Also of note is that the code can still be disabled via the --disable-keymaps
configure option, which was removed from upstream.
Signed-off-by: Ian Stakenvicius <axs@gentoo.org>
This code was dropped while importing from upstream. It is needed
to allow the legacy renaming of interfaces like eth* or wlan*.
X-Eudev-Bug: 56
X-Eudev-Bug-URL: https://github.com/gentoo/eudev/issues/56
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
Previously, rootprefix was set to the actual value of the default prefix,
which caused issues when running 'make distcheck' since prefix is set
to a non-standard value but rootprefix is unset. By assigning the variable
rather than its contents, this should now be avoided.
Both scripts had hard-coded paths to 'rule_generator.functions'; these were changed
to be substituted in by configure so the paths will match where files are
installed.øy
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
This restores the rule generator scripts for the persistent
network and optical device rule generator scripts that were
removed after udev-171, and re-introduces their installation
to the build system.
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
Distros that whish to support old kernels should set
--with-firmware-dirs="/usr/lib/firmware/updates:/usr/lib/firmware"
to retain the old behaviour.
Signed-off-by: Anthony G. Basile <blueness@gentoo.org>
Signed-off-by: Ian Stakenvicius <axs@gentoo.org>
ie, fix another regression in our build system ; gentoo bug 455326
it's common for eg. mips systems to not have symlink from the directory
{lib32 or lib64} to /lib. only the libraries belong to /$(get_libdir) and
everything in /lib/udev goes to, well, /lib/udev, NOT /$(get_libdir)/udev
-- ssuominen
Signed-off-by: Ian Stakenvicius <axs@gentoo.org>
