gmane.linux.busybox

bb 1.17 causing subsequent driver problems

Peter Scheie — 2012-05-25T16:27:22

I've been using bb 1.16.2 to build a custom initramfs, which we use to
replace the stock initrd in Centos 5 (we do some special things we need
in the initramfs).  Now I need mkfs_ext2, so I grabbed the bb 1.17.2
that I use in another project.  But I'm getting some odd behavior.  The
initramfs loads fine and does all the stuff it's supposed to, after
which it does a switch_root (so the initramfs & bb are out of the
picture) and mounts the file systems from the local disk, runs the
Centos rc scripts, all which seems to go fine.  But with 1.17, when udev
is started by Centos's rc.sysinit, I get an error that "DM multipath
kernel driver not loaded", and then later a modprobe warning, "Could not
open 'kernel/net/netfilter/x_tables.ko': no such file or directory"
followed by another modprobe error, "FATAL: Error inserting ip_tables
(kernel/net/ipv4/netfilter.ko: no such file or directory", even though
both those files are on the disk.  And of course, iptables then won't
run.

Any ideas as to what changed between 1

Problems with --sort-section=alignment

Rich Felker — 2012-05-24T00:21:51

Busybox is linked with --sort-section=alignment, and this seems to be
causing an obscure bugs one of my users experienced building busybox
on ARM/musl target. I've filed a bug report with binutils, but I'm not
sure where the issue should be addressed. See here:

http://sourceware.org/bugzilla/show_bug.cgi?id=14156

Basically, --sort-section=alignment wants to reorder the .init
section, which is not possible. The problem does not show up on other
targets, but I can't see anything that inhibits reordering .init, so
it seems to be working only by chance...

Rich

swapon: implementing discard option.

Andy Lee — 2012-05-23T01:02:59

Hi,

This makes swapon to use discard option.
Enable support for discarding freed pages before they are reused.

Cheers,
Andy

---
 util-linux/Config.src  |    7 +++++++
 util-linux/swaponoff.c |   41 +++++++++++++++++++++++++++++++++--------
 2 files changed, 40 insertions(+), 8 deletions(-)

diff --git a/util-linux/Config.src b/util-linux/Config.src
index 57a52ce..84652eb 100644
--- a/util-linux/Config.src
+++ b/util-linux/Config.src
< at >< at > -606,6 +606,13 < at >< at > config FEATURE_SWAPON_PRI
 help
   Enable support for setting swap device priority in swapon.

+config FEATURE_SWAPON_DISCARD
+bool "Support discard option -d"
+default y
+depends on SWAPONOFF
+help
+  Enable support for discarding freed pages before they are reused.
+
 config SWITCH_ROOT
 bool "switch_root"
 default y
diff --git a/util-linux/swaponoff.c b/util-linux/swaponoff.c
index 54867ec..d911b31 100644
--- a/util-linux/swaponoff.c
+++ b/util-linux/swaponoff.c
< at >< at > -8,10 +8,13 < at >< at >
  */

 //usage:#define swapon_trivial_usage
-//usage:       "[-a]

[PATCH] unzip: ignore chmod errors

Natanael Copa — 2012-05-22T14:56:22

This makes unzip to FAT filesystems not exit with error.

This similar to how the "normal" unzip works.
---
 archival/unzip.c       |    2 +-
 include/libbb.h        |    1 +
 libbb/make_directory.c |    7 ++++++-
 3 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/archival/unzip.c b/archival/unzip.c
index 3c76cda..c1b945a 100644
--- a/archival/unzip.c
+++ b/archival/unzip.c
< at >< at > -596,7 +596,7 < at >< at > int unzip_main(int argc, char **argv)
 printf("   creating: %s\n", dst_fn);
 }
 unzip_create_leading_dirs(dst_fn);
-if (bb_make_directory(dst_fn, dir_mode, 0)) {
+if (bb_make_directory(dst_fn, dir_mode, FILEUTILS_IGNORE_CHMOD_ERR)) {
 xfunc_die();
 }
 } else {
diff --git a/include/libbb.h b/include/libbb.h
index f12800f..5e5c8c7 100644
--- a/include/libbb.h
+++ b/include/libbb.h
< at >< at > -333,6 +333,7 < at >< at > enum {/* DO NOT CHANGE THESE VALUES!  cp.c, mv.c, install.c depend on them. */
 FILEUTILS_PRESERVE_SECURITY_CONTEXT = 1 << 9, /* -c */
 FILEUTILS_SET_SECURITY_CONTEXT

patches for building with Android NDK

Tias — 2012-05-21T22:03:54

Dear Busybox developers,

Since Rob Walker's patch[1], building busybox with the Android NDK has 
really become dead simple; no special hacks or external libraries are 
needed anymore.

However, a large number of applets can currently not be built.

I started documenting why applets don't build, as well as collecting 
patches from different sources, that fix the issues:
https://github.com/tias/android-busybox-ndk
https://github.com/tias/android-busybox-ndk/tree/master/patches

I would love to see the ideas behind these patches integrated, so that a 
standard busybox build can be used in the many android open-source 
projects that depend on it (and currently use a statically built 
third-party blob).


However, I am unfamiliar with the coding practices of the busybox 
community, and many of these patches were written to fix the build and 
not necessarily to get merged upstream.
I consider my work in collecting these patches as just a first step.
Hence, I would like to solicit feedback on what can and can not

[PATCH] Allow XZ streaming format in tar.

Boris Reisig — 2012-05-21T19:24:17

Here's a patch that will allow tar to uncompress XZ file formats when it 
can't properly detect XZ data formats when the input data comes from the 
standard input.  I have added the (-J) option to force tar to recognize 
XZ file format. Without this patch, it won't work with data streams.

Example: wget -O - http://192.168.0.1/1.xz | tar -xvJf -

diff -uNr busybox-original/archival/tar.c busybox/archival/tar.c
--- busybox-original/archival/tar.c     2012-05-21 00:19:38.876360998 -0500
+++ busybox/archival/tar.c      2012-05-21 00:05:08.299222265 -0500
< at >< at > -693,6 +693,7 < at >< at >
  //usage:       "-[" IF_FEATURE_TAR_CREATE("c") "xt"
  //usage:       IF_FEATURE_SEAMLESS_Z("Z")
  //usage:       IF_FEATURE_SEAMLESS_GZ("z")
+//usage:       IF_FEATURE_SEAMLESS_XZ("J")
  //usage:       IF_FEATURE_SEAMLESS_BZ2("j")
  //usage:       IF_FEATURE_SEAMLESS_LZMA("a")
  //usage:       IF_FEATURE_TAR_CREATE("h")
< at >< at > -719,6 +720,9 < at >< at >
  //usage:       IF_FEATURE_SEAMLESS_GZ(
  //usage:     "\n       z       (De)compress using gzip"

/sbin/reboot does SIGINT and not SIGTERM?

Michael J. Hammel — 2012-05-17T03:06:44

I was trying to get rcK to run to shutdown gracefully and noticed that
the inittab should use SHUTDOWN for the action.  If I run /sbin/reboot
this causes a SIGINT, which doesn't cause the rcK to get run.  If I do
kill -15 1 then rcK gets run and I get a reboot.

Is this the intended action for /sbin/reboot or am I doing something
wrong here?  poweroff also doesn't do SIGTERM for init.

I'm using Busybox 1.19.4.  My relevant inittab lines are:

# Stuff to do for the 3-finger salute
::ctrlaltdel:/sbin/reboot

# Stuff to do before rebooting
::shutdown:/etc/init.d/rcK
::shutdown:/bin/umount -a -r
::shutdown:/sbin/swapoff -a

[PATCH] ext4 detection in blkid applet

Tanguy Pruvot — 2012-05-15T23:22:08

Available here :

http://review.cyanogenmod.com/16046

--
Tanguy Pruvot                         email : tanguy.pruvot< at >gmail.com

[PATCH] Fix link time error for functions calling check_signature16()

Anthony G. Basile — 2012-05-15T11:36:00

From: Anthony G. Basile <blueness< at >gentoo.org>

archival/libarchive/decompress_bunzip2.c, decompress_gunzip.c and
decompress_uncompress.c all make calls to check_signature16() which
is defined in open_transformer.c.  However, Kbuild.src does not
properly respect this dependency leading to a link time error.
This patch addresses this problem.

Signed-off-by: Anthony G. Basile <blueness< at >gentoo.org>
---
 archival/libarchive/Kbuild.src |   10 +++++-----
 1 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/archival/libarchive/Kbuild.src b/archival/libarchive/Kbuild.src
index e2134be..72dc733 100644
--- a/archival/libarchive/Kbuild.src
+++ b/archival/libarchive/Kbuild.src
< at >< at > -39,17 +39,17 < at >< at > DPKG_FILES:= \
 INSERT
 
 lib-$(CONFIG_AR)                        += get_header_ar.o unpack_ar_archive.o
-lib-$(CONFIG_BUNZIP2)                   += decompress_bunzip2.o
+lib-$(CONFIG_BUNZIP2)                   += open_transformer.o decompress_bunzip2.o
 lib-$(CONFIG_UNLZMA)                    += decompress_unlzma.o

mdev automount questions

Walter Dnes — 2012-05-14T07:45:46

  This really should be a separate topic.  Sorry for derailing the wifi
question.

On Sun, May 13, 2012 at 12:16:37PM +0200, Laurent Bercot wrote


  A couple of options...

1) The comments in /lib/mdev/usbdisk_link say
# NOTE: since mdev -s only provide $MDEV, don't depend on any hotplug vars.
Can I use the lack of an ACTION var to assume that this call is from
"mdev -s" (i.e. bootup), whereas if ACTION is present, assume that it's
a hotplug
if [ "X${ACTION}" == "X" ]; then
   assume it's a hard drive coming up at boot time (mdev -s)
else
   it's from the command in /proc/sys/kernel/hotplug (userspace hotplug)
fi

2) Get info from /sys

cat /sys/block/sda/removable
0

cat /sys/block/sdb/removable 
1

MAC address deny in udhcpd

Tom Isaacson — 2012-05-14T00:08:34

In the example udhcpd.conf it shows how to declare a static IP address 
for a specific MAC address:

|   # Static leases map
   static_lease 00:60:08:11:CE:4E 192.168.0.54
|

You can also use a MAC address range by using the wildcard:

|   static_lease 00:21:5a:e7:xx:xx 10.10.1.254

But it doesn't seem possible to deny an IP address to a MAC address or range, which is possible in dhcpd.conf. I was thinking this could be done in the same way static leases are done:
||   static_lease 00:21:5a:e7:xx:xx 0.0.0.0|
Alternatively we could add a new command:
   deny|00:21:5a:e7:xx:xx|

What are your thoughts?

|Tom

|

Possible change to /etc/mdev.conf default?

Walter Dnes — 2012-05-13T22:14:22

  I was spelunking through my /etc/mdev.conf, looking at the syntax,
when something occured to me looking at...

sr[0-9]*        root:cdrom 660 < at >ln -sf $MDEV cdrom

  What happens if you have multiple devices, e.g. "sr0" and "sr1"?  Does
the system get confused?  Would the following work better...

sr[0-9]*        root:cdrom 660 < at >ln -sf $MDEV cdrom%1

Yank multiple character bug in VI

Wakko Warner — 2012-05-12T14:24:35

When editing a file, typing #yl where # is > 1, only 1 byte is yanked.  Same
is true for #dl except that only 1 character is deleted.  Replacing "l" with
"h" works for yanking/deleting X amount of characters left of the cursor.

vim-tiny 2:7.3.429-2 in debian works as expected.

Tested with busybox v1.20.0 and
git with commit a5ee090e8651692545514a81a16c6cde3a2dc577 (Which was the
latest when I cloned today)

mdev with USB wifi adapters

Michael J. Hammel — 2012-05-11T20:02:59

I've googled on this but didn't find much, though there was a long
discussion in the BusyBox archives on the extension of mdev event
variables (and other things) that helped.

I'm looking at using mdev to recognize a USB wifi adapter at boot time.
I've finally figured out that a line like the following will at least
get my handler script called:

usb.* - - < at >/usr/bin/spew.sh

(spew.sh is just a script echoing the event variables passed to it)

Using a device such as "net.*" doesn't seem to work for USB connected
wifi adapters.  I believe using event variables should also work but the
output from spew.sh showed that none of them were getting set except
MDEV, SUBSYSTEM and PWD.  I couldn't find a regex for SUBSYTEM that
identified just the wifi adapter. 

I was thinking the best way to use this would be to have a handler
script that uses the MDEV variable to identify the USB bus and device or
major/minor numbers, then use those to find the vendor and product ID
and match those with a configuration file to deter

Massive eh_frame bloat in busybox (stripped)

Rich Felker — 2012-05-10T02:44:39

The busybox build system first builds busybox_unstripped with debug
info, then uses the strip utility to obtain a non-debug busybox binary
that's not so bloated. However, ever since GCC switched to using
DWARF2 for debugging data, it puts a .eh_frame section in the output
binary which is NOT stripped by the strip command. See:

$ size -A busybox
section       size        addr
.init           11   134512788
.text       533493   134512800
.fini            6   135046293
.rodata     146886   135046304
.eh_frame    72232   135193192
...

It's possible that strip -R .eh_frame could be used to remove this
bloat, but see my bug report here:

http://sourceware.org/bugzilla/show_bug.cgi?id=14037

I'm fairly confident this bug in strip can't break static binaries,
but I'm unsure if it's safe for dynamic-linked ones. Some checking
should definitely be done if it's going to be added to busybox's strip
command...

Rich

[PATCH] syslogd should use _PATH_LOG when available

Cédric Cabessa — 2012-05-08T14:51:34

---
 sysklogd/syslogd.c |   11 +++++++----
 1 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/sysklogd/syslogd.c b/sysklogd/syslogd.c
index fc380d9..f7fbc16 100644
--- a/sysklogd/syslogd.c
+++ b/sysklogd/syslogd.c
< at >< at > -55,6 +55,9 < at >< at >
 #define SYSLOG_NAMES_CONST
 #include <syslog.h>
 */
+#ifndef _PATH_LOG
+#define _PATH_LOG"/dev/log"
+#endif
 
 #include <sys/un.h>
 #include <sys/uio.h>
< at >< at > -768,8 +771,8 < at >< at > static NOINLINE int create_socket(void)
 
 /* Unlink old /dev/log or object it points to. */
 /* (if it exists, bind will fail) */
-strcpy(sunx.sun_path, "/dev/log");
-dev_log_name = xmalloc_follow_symlinks("/dev/log");
+strcpy(sunx.sun_path, _PATH_LOG);
+dev_log_name = xmalloc_follow_symlinks(_PATH_LOG);
 if (dev_log_name) {
 safe_strncpy(sunx.sun_path, dev_log_name, sizeof(sunx.sun_path));
 free(dev_log_name);
< at >< at > -778,7 +781,7 < at >< at > static NOINLINE int create_socket(void)
 
 sock_fd = xsocket(AF_UNIX, SOCK_DGRAM, 0);
 xbind(sock_fd, (struct sockaddr *) &sunx, sizeof(sunx));
-chmod("/

[PATCH] Only write leasefile if anything changed

Phil Sutter — 2012-05-08T14:09:34

Also, move the check for changed leases to before the select(), so
changes will occur immediately instead only after select() returns the
next time.

Signed-off-by: Phil Sutter <phil.sutter< at >viprinet.com>
Signed-off-by: Nico Erfurth <nico.erfurth< at >viprinet.com>
---
 networking/udhcp/dhcpd.c |   10 +++++++++-
 1 files changed, 9 insertions(+), 1 deletions(-)

diff --git a/networking/udhcp/dhcpd.c b/networking/udhcp/dhcpd.c
index dd55e70..356c81f 100644
--- a/networking/udhcp/dhcpd.c
+++ b/networking/udhcp/dhcpd.c
< at >< at > -298,6 +298,7 < at >< at > int udhcpd_main(int argc UNUSED_PARAM, char **argv)
 {
 int server_socket = -1, retval, max_sock;
 uint8_t *state;
+uint8_t leases_changed = 0;
 unsigned timeout_end;
 unsigned num_ips;
 unsigned opt;
< at >< at > -389,6 +390,11 < at >< at > int udhcpd_main(int argc UNUSED_PARAM, char **argv)
 uint32_t static_lease_nip;
 struct dyn_lease *lease, fake_lease;
 
+if (leases_changed) {
+write_leases();
+leases_changed = 0;
+}
+
 if (server_socket < 0) {
 server_socket = udhcp_liste

(unknown)

Sven-Göran Bergh — 2012-05-08T06:59:14

Hi,


Some thoughts on the defaults:

1) OTP should not be enabled by default in bb. It should require an
active config, like CONFIG_FEATURE_LOGIN_OTP. (I guess all agree on
this one)

2) Even when compiled with OTP I think it should be easy to disable
OTP for some users. Some examples:

    # cat /etc/otp
    $UID1:$SHAREDSECRET:$DELIVERY
    $UID2:$SHAREDSECRET:

   a) $UID1 will get the PIN delivered by the configured $DELIVERY.
   b) $UID2 will get the PIN within login (the default).
   c) $UID3 will not benefit from OTP at all (not present in /etc/otp).

3) When compiled with OTP, default for the special case with no
/etc/otp could be argued, but here are my 2 cents:
   a) No /etc/opt => enable the default for all users (PIN in login)
   b) Empty /etc/otp => OTP disabled for all users (consistent with 2)

/Sven

Bug in busybox sed - this one is real

Rich Felker — 2012-05-08T03:12:09

$ echo hello | busybox sed 's/l*/< at >/g'
< at >he< at >o
$ echo hello | sed 's/l*/< at >/g'
< at >h< at >e< at >o< at >

The latter is of course correct. My naive diagnosis is that Busybox
sed's 'g' option has an off-by-one error in its next-match search,
i.e. advanced by one character too many before searching for the next
match.

Rich

[BUG?] busybox sed fails to process ncurses scripts

John Spencer — 2012-05-07T04:16:16

when compiling ncurses 5.9 with busybox sed,
it'll produce invalid C files

../ncurses/lib_gen.c:27:3: error: expected identifier or '(' before 'int'

the file contains:
/* DO NOT EDIT THIS FILE BY HAND!
* It is generated by ./base/MKlib_gen.sh generated. */

the line that errs is
* int waddch(stdscr,const chtype z) { ... }

which clearly is invalid C

it is generated by ncurses-5.9/ncurses/./base/MKlib_gen.sh which does a 
lot of sed...

since i'm not very familiar with sed, it would be nice if someone could 
look at the issue.

using gnu sed instead fixes the build.

--JS

OTP feature for /bin/login

Guylhem — 2012-05-04T22:37:28

Hello

I have prepared a OTP implementation for /bin/login, where a password
can optionally be replaced by an MOTP (md5 of epoch+pin+shared secret)
if ENABLE_FEATURE_LOGINOTP is defined.

This could be useful for example when using telnetd. The code can
easily be tweaked to support other things besides MOTP.

Shall I just post the code on the list? (it's a full replacement of
correct_password.c, GPLv2)

Guylhem