3293bc1469
Fixes:
commit 52a515d18724bbb34e3ccbbb0218efcc4eccc0a8
"udhcp: use poll() instead of select()"
Feb 16 2017
udhcp_sp_read() is meant to check whether signal pipe indeed has some data to read.
In the above commit, it was changed as follows:
- if (!FD_ISSET(signal_pipe.rd, rfds))
+ if (!pfds[0].revents)
return 0;
The problem is, the check was working for select() purely by accident.
Caught signal interrupts select()/poll() syscalls, they return with EINTR
(regardless of SA_RESTART flag in sigaction). _Then_ signal handler is invoked.
IOW: they can't see any changes to fd state caused by signal haldler
(in our case, signal handler makes signal pipe ready to be read).
For select(), it means that rfds[] bit array is unmodified, bit of signal
pipe's read fd is still set, and the above check "works": it thinks select()
says there is data to read.
This accident does not work for poll(): .revents stays clear, and we do not
try reading signal pipe as we should. In udhcpd, we fall through and block
in socket read. Further SIGTERM signals simply cause socket read to be
interrupted and then restarted (since SIGTERM handler has SA_RESTART=1).
Fixing this as follows: remove the check altogether. Set signal pipe read fd
to nonblocking mode. Always read it in udhcp_sp_read().
If read fails, assume it's EAGAIN and return 0 ("no signal seen").
udhcpd avoids reading signal pipe on every recvd packet by looping if EINTR
(using safe_poll()) - thus ensuring we have correct .revents for all fds -
and calling udhcp_sp_read() only if pfds[0].revents!=0.
udhcpc performs much fewer reads (typically it sleeps >99.999% of the time),
there is no need to optimize it: can call udhcp_sp_read() after each poll
unconditionally.
To robustify socket reads, unconditionally set pfds[1].revents=0
in udhcp_sp_fd_set() (which is before poll), and check it before reading
network socket in udhcpd.
TODO:
This might still fail: if pfds[1].revents=POLLIN, socket read may still block.
There are rare cases when select/poll indicates that data can be read,
but then actual read still blocks (one such case is UDP packets with
wrong checksum). General advise is, if you use a poll/select loop,
keep all your fds nonblocking.
Maybe we should also do that to our network sockets?
function old new delta
udhcp_sp_setup 55 65 +10
udhcp_sp_fd_set 54 60 +6
udhcp_sp_read 46 36 -10
udhcpd_main 1451 1437 -14
udhcpc_main 2723 2708 -15
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/3 up/down: 16/-39) Total: -23 bytes
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Please see the LICENSE file for details on copying and usage.
Please refer to the INSTALL file for instructions on how to build.
What is busybox:
BusyBox combines tiny versions of many common UNIX utilities into a single
small executable. It provides minimalist replacements for most of the
utilities you usually find in bzip2, coreutils, dhcp, diffutils, e2fsprogs,
file, findutils, gawk, grep, inetutils, less, modutils, net-tools, procps,
sed, shadow, sysklogd, sysvinit, tar, util-linux, and vim. The utilities
in BusyBox often have fewer options than their full-featured cousins;
however, the options that are included provide the expected functionality
and behave very much like their larger counterparts.
BusyBox has been written with size-optimization and limited resources in
mind, both to produce small binaries and to reduce run-time memory usage.
Busybox is also extremely modular so you can easily include or exclude
commands (or features) at compile time. This makes it easy to customize
embedded systems; to create a working system, just add /dev, /etc, and a
Linux kernel. Busybox (usually together with uClibc) has also been used as
a component of "thin client" desktop systems, live-CD distributions, rescue
disks, installers, and so on.
BusyBox provides a fairly complete POSIX environment for any small system,
both embedded environments and more full featured systems concerned about
space. Busybox is slowly working towards implementing the full Single Unix
Specification V3 (http://www.opengroup.org/onlinepubs/009695399/), but isn't
there yet (and for size reasons will probably support at most UTF-8 for
internationalization). We are also interested in passing the Linux Test
Project (http://ltp.sourceforge.net).
----------------
Using busybox:
BusyBox is extremely configurable. This allows you to include only the
components and options you need, thereby reducing binary size. Run 'make
config' or 'make menuconfig' to select the functionality that you wish to
enable. (See 'make help' for more commands.)
The behavior of busybox is determined by the name it's called under: as
"cp" it behaves like cp, as "sed" it behaves like sed, and so on. Called
as "busybox" it takes the second argument as the name of the applet to
run (I.E. "./busybox ls -l /proc").
The "standalone shell" mode is an easy way to try out busybox; this is a
command shell that calls the built-in applets without needing them to be
installed in the path. (Note that this requires /proc to be mounted, if
testing from a boot floppy or in a chroot environment.)
The build automatically generates a file "busybox.links", which is used by
'make install' to create symlinks to the BusyBox binary for all compiled in
commands. This uses the CONFIG_PREFIX environment variable to specify
where to install, and installs hardlinks or symlinks depending
on the configuration preferences. (You can also manually run
the install script at "applets/install.sh").
----------------
Downloading the current source code:
Source for the latest released version, as well as daily snapshots, can always
be downloaded from
http://busybox.net/downloads/
You can browse the up to the minute source code and change history online.
http://git.busybox.net/busybox/
Anonymous GIT access is available. For instructions, check out:
http://www.busybox.net/source.html
For those that are actively contributing and would like to check files in,
see:
http://busybox.net/developer.html
The developers also have a bug and patch tracking system
(https://bugs.busybox.net) although posting a bug/patch to the mailing list
is generally a faster way of getting it fixed, and the complete archive of
what happened is the git changelog.
Note: if you want to compile busybox in a busybox environment you must
select CONFIG_DESKTOP.
----------------
Getting help:
when you find you need help, you can check out the busybox mailing list
archives at http://busybox.net/lists/busybox/ or even join
the mailing list if you are interested.
----------------
Bugs:
if you find bugs, please submit a detailed bug report to the busybox mailing
list at busybox@busybox.net. a well-written bug report should include a
transcript of a shell session that demonstrates the bad behavior and enables
anyone else to duplicate the bug on their own machine. the following is such
an example:
to: busybox@busybox.net
from: diligent@testing.linux.org
subject: /bin/date doesn't work
package: busybox
version: 1.00
when i execute busybox 'date' it produces unexpected results.
with gnu date i get the following output:
$ date
fri oct 8 14:19:41 mdt 2004
but when i use busybox date i get this instead:
$ date
illegal instruction
i am using debian unstable, kernel version 2.4.25-vrs2 on a netwinder,
and the latest uclibc from cvs.
-diligent
note the careful description and use of examples showing not only what
busybox does, but also a counter example showing what an equivalent app
does (or pointing to the text of a relevant standard). Bug reports lacking
such detail may never be fixed... Thanks for understanding.
----------------
Portability:
Busybox is developed and tested on Linux 2.4 and 2.6 kernels, compiled
with gcc (the unit-at-a-time optimizations in version 3.4 and later are
worth upgrading to get, but older versions should work), and linked against
uClibc (0.9.27 or greater) or glibc (2.2 or greater). In such an
environment, the full set of busybox features should work, and if
anything doesn't we want to know about it so we can fix it.
There are many other environments out there, in which busybox may build
and run just fine. We just don't test them. Since busybox consists of a
large number of more or less independent applets, portability is a question
of which features work where. Some busybox applets (such as cat and rm) are
highly portable and likely to work just about anywhere, while others (such as
insmod and losetup) require recent Linux kernels with recent C libraries.
Earlier versions of Linux and glibc may or may not work, for any given
configuration. Linux 2.2 or earlier should mostly work (there's still
some support code in things like mount.c) but this is no longer regularly
tested, and inherently won't support certain features (such as long files
and --bind mounts). The same is true for glibc 2.0 and 2.1: expect a higher
testing and debugging burden using such old infrastructure. (The busybox
developers are not very interested in supporting these older versions, but
will probably accept small self-contained patches to fix simple problems.)
Some environments are not recommended. Early versions of uClibc were buggy
and missing many features: upgrade. Linking against libc5 or dietlibc is
not supported and not interesting to the busybox developers. (The first is
obsolete and has no known size or feature advantages over uClibc, the second
has known bugs that its developers have actively refused to fix.) Ancient
Linux kernels (2.0.x and earlier) are similarly uninteresting.
In theory it's possible to use Busybox under other operating systems (such as
MacOS X, Solaris, Cygwin, or the BSD Fork Du Jour). This generally involves
a different kernel and a different C library at the same time. While it
should be possible to port the majority of the code to work in one of
these environments, don't be surprised if it doesn't work out of the box. If
you're into that sort of thing, start small (selecting just a few applets)
and work your way up.
In 2005 Shaun Jackman has ported busybox to a combination of newlib
and libgloss, and some of his patches have been integrated.
Supported hardware:
BusyBox in general will build on any architecture supported by gcc. We
support both 32 and 64 bit platforms, and both big and little endian
systems.
Under 2.4 Linux kernels, kernel module loading was implemented in a
platform-specific manner. Busybox's insmod utility has been reported to
work under ARM, CRIS, H8/300, x86, ia64, x86_64, m68k, MIPS, PowerPC, S390,
SH3/4/5, Sparc, and v850e. Anything else probably won't work.
The module loading mechanism for the 2.6 kernel is much more generic, and
we believe 2.6.x kernel module loading support should work on all
architectures supported by the kernel.
----------------
Please feed suggestions, bug reports, insults, and bribes back to the busybox
mailing list:
busybox@busybox.net
and/or maintainer:
Denys Vlasenko
<vda.linux@googlemail.com>