The array applet_nameofs consumes two bytes per applet. It encodes
nofork/noexec flags
suid flags
the offset of the applet name in the applet_name string
Change the applet_table build tool to store the flags in two separate
arrays (applet_flags and applet_suid). Replace applet_nameofs[] with a
smaller version that only stores a limited number of offsets.
This requires changes to the macros APPLET_IS_NOFORK, APPLET_IS_NOEXEC
and APPLET_SUID.
According to Valgrind the original find_applet_by_name required
353 cycles per call, averaged over all names. Adjusting the number
of known offsets allows space to be traded off against execution time:
KNOWN_OFFSETS cycles bytes (wrt KNOWN_OFFSETS = 0)
0 9057 -
2 4604 32
4 2407 75
8 1342 98
16 908 130
32 884 194
This patch uses KNOWN_OFFSETS = 8.
v2:
Remove some dead code from the applet_table tool;
Treat the applet in the middle of the table as a special case.
v3:
Use the middle applet to adjust the start of the linear search as
well as the last applet found.
v4:
Use an augmented linear search in find_applet_by_name.
Drop the special treatment of the middle name from get_applet_name:
most of the advantage now derives from the last stored value.
v5:
Don't store index in applet_nameofs, it can be calculated.
v6:
Tweaks by Denys
function old new delta
find_applet_by_name 25 125 +100
applet_suid - 92 +92
run_applet_no_and_exit 452 460 +8
run_applet_and_exit 695 697 +2
applet_name_compare 31 - -31
applet_nameofs 734 14 -720
------------------------------------------------------------------------------
(add/remove: 1/1 grow/shrink: 3/1 up/down: 202/-751) Total: -549 bytes
text data bss dec hex filename
925464 906 17160 943530 e65aa busybox_old
924915 906 17160 942981 e6385 busybox_unstripped
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
As reported in bug 8506:
$ X=abcdÉfghÍjklmnÓpqrstÚvwcyz
$ echo ${#X}
abcd26
The result should be 26.
This regression was introduced by:
<d68d1fb> 2015-05-18 [Ron Yorston] ash: code shrink around varvalue
The length in characters was being used to discard the contents of
the variable instead of the length in bytes.
URL: https://bugs.busybox.net/8506
Reported-by: Martijn Dekker <martijn@inlv.org>
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
In coreutils/ls.c, 1.19 introduced commit
2f7d9e8903, removing the variable tabstops and
hard coding the column separation to 2 characters, but was not done correctly.
The column_width assumes a gap of 1 character, so the computed number of
columns exceeds the terminal width when many small files are encountered.
A minor problem but surprisingly annoying.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Otherwise, "-t 0" usage may end up sending them forever
if server does not respond.
function old new delta
udhcpc_main 2846 2836 -10
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
RFC2131 paragraph 4.1 states DHCP messages broadcast by a client prior to
that client obtaining its IP address must have the source IP address
field in the header set to 0.
Request messages transmitted in renewing and rebinding state need to use
the obtained IP address as source IP address in the header; this behavior
lines up with other implementations like ISC dhcp client.
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Let udhcpd retain the information about expired leases when restarting
so that the leases are reserved until they possibly become the oldest
expired lease.
This reduces the frequency of IP address changes for example when the
DHCP server and a group of clients, who do not store and request their
previously offered IP address across restarts, are collectively restarted
and the startup order of the clients are not guaranteed.
Signed-off-by: Christian Lindeberg <christian.lindeberg@axis.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This patch allow to have multiple interface definitions, much like
Debian's ifupdown. More specifically, it removes the check for a
duplicate definition, so the impact on binary size should be fairly
minimal.
This configuration:
iface eth0 inet static
address 192.168.0.15
netmask 255.255.0.0
gateway 192.168.0.1
iface eth0 inet static
address 10.0.0.1
netmask 255.255.255.0
Will add two addresses to eth0 if ip is used. If ifconfig is used,
the standards methods will likely not stack, but the administrator may
still use the manual method. The DHCP method may work depending on the
DHCP client in use.
This is a fairly advanced feature for power users who knows what they
are doing. There are not many other network configuration systems that
allows multiple addresses on an interface.
Signed-off-by: Nicolas Cavallari <nicolas.cavallari@green-communications.fr>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
The non-fancy version of the from_cpuset uses CPU_SETSIZE as if it
represents the number of bytes in the cpuset, while it is actually
the number of bits. This leads to out-of-bounds accesses on the
cpu_set_t in the big-endian case. Basically all uses of CPU_SETSIZE
have to be divided by 8. This is done correctly in the fancy version
of from_cpuset.
In addition, the big-endian case is completely wrong to begin with.
All standard C libraries that I know of implement cpu_set_t as an
unsigned long array, so both for big and little endian, the least
significant bits are in the beginning of the array. Therefore, the
approach taken for the little endian case is equally valid. We only
need special handling for big endian when CPU_SETSIZE is large and
we use an unsigned long long to get more bits out.
Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This matches behavior with kmod which has been the standard for a long
time at this point.
URL: https://bugs.busybox.net/8021
Reported-by: Jö <jorrit@jorrit.de>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Building & running with ASAN is super helpful, so add a dedicated config
knob for it. This way people don't have to guess at the right compiler
settings in order to get a good build. We can just tell people to enable
this one option.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
When launched as PID 1 and after parsing its arguments, init wipes all
all of them except argv[0] and rewrites argv[0] to contain only "init",
so that its command-line appears solely as "init" in tools such as ps.
This patch adds the FEATURE_INIT_MODIFY_CMDLINE which, if set to n, will
make init preserve all its arguments including argv[0], be they parsed or
ignored.
The original command-line used to launch init can then be retrieved in
/proc/1/cmdline on Linux, for example.
Signed-off-by: Nicolas Carrier <nicolas.carrier@parrot.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
After commit 8e74adab01
("libbb: make INET[6]_rresolve use sockaddr2{host,dotted}_noport")
INET_sprint6 uses more than just sin6_addr, it also tries to display the
scope id, which is uninitialized when called from ife_print6.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This patch fixes compiling busybox with FEATURE_UTMP and _WTMP enabled.
musl, while not really support utmp/wtmp, provides stub functions, as well
as variables such as _PATH_UTMP, so that programs using utmp or wtmp can
still compile fine.
My reasoning for this patch is that on Exherbo, I'm currently trying to get
us to be able to use the same busybox config file for both glibc and musl
systems, using utmp/wtmp on systems that support it, and using the stubs
on musl without needing two different configs.
As of latest musl git, it provides all utmp functions needed; 1.1.12 doesn't,
but I sent a patch to Rich to add the utmp{,x}name functions expected to
exist, which was merged into musl upstream.
Signed-off-by: Kylie McClain <somasissounds@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
