udhcpc[6]: on log level 1, three messages about raw socket is overkill

Move first two messages to log2 level:

08:46:32.23824 udhcpc: opening raw socket on ifindex 2
08:46:32.23825 udhcpc: got raw socket fd
08:46:32.26354 udhcpc: created raw socket

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ping: better config help text for FEATURE_FANCY_PING

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

config: deindent all help texts

Those two spaces after tab have no effect, and always a nuisance when editing.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libarchive: FEATURE_LZMA_FAST should be visible if FEATURE_SEAMLESS_LZMA

While at it, tweak some config help texts.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

unzip: sanitize filename length: malloc(1234mb) is not funny

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

unzip: implement -j, closes 9126

function                                             old     new   delta
unzip_main                                          2642    2703     +61
packed_usage                                       31747   31770     +23
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 84/0)               Total: 84 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

unzip: robustify overwrite checks

function                                             old     new   delta
get_lstat_mode                                         -      55     +55
unzip_main                                          2667    2642     -25
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 0/1 up/down: 55/-25)             Total: 30 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

unzip: support symlinks. Closes 10031

function                                             old     new   delta
unzip_main                                          2519    2667    +148

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

bzcat,zcat: simplify code if gunzip/bunzip2 is not selected

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

shell: optional support for read -t N.NNN, closes 10101

function                                             old     new   delta
shell_builtin_read                                  1097    1277    +180
dump_procs                                           353     359      +6

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

bootchartd: fix typo in comment

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

bootchartd: prevent premature stop by device mapper

Kernel thread kdmflush was recognized as display manager kdm,
check for string length.

Signed-off-by: Uwe Geuder <busybox2017-ugeuder@snkmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sparc: needs -fPIC

Pinted out by Thomas Petazzoni.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

swapon/swapoff: trim config help text

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Update remaining menuconfig items with approximate applet sizes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

od_bloaty: fix floating point output

Currently od_bloaty does this:

   $ /bin/echo -en '\x00\x00\x00\x00\x00\x00\x00\x00' | busybox od -f
   od: invalid character 'F' in type string 'fF'
   $ /bin/echo -en '\x00\x00\x00\x00\x00\x00\x00\x00' | busybox od -t fD
   od: invalid character 'D' in type string 'fD'
   $ /bin/echo -en '\x00\x00\x00\x00\x00\x00\x00\x00' | busybox od -t f
   0000000
   0000010

The first two occur because the alphabetic length specifier isn't being
properly skipped.  The third is due to the empty length specifier being
treated as alphabetic so we fall off the end of the FDL_sizeof array with
undetermined consequences.  Coreutils defaults to printing a double in
this case.

With this patch the output is:

   $ /bin/echo -en '\x00\x00\x00\x00\x00\x00\x00\x00' | ./busybox od -f
   0000000   0.0000000e+00   0.0000000e+00
   0000010
   $ /bin/echo -en '\x00\x00\x00\x00\x00\x00\x00\x00' | ./busybox od -t fD
   0000000   0.000000000000000e+00
   0000010
   $ /bin/echo -en '\x00\x00\x00\x00\x00\x00\x00\x00' | ./busybox od -t f
   0000000   0.000000000000000e+00
   0000010

I guess nobody uses BusyBox od to print floating point numbers.

Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hdparm: remove contradicting size info in config help

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: remove contradicting size info in config help

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: remove contradicting size info in config help

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Update menuconfig items with approximate applet sizes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Tweaks from mass recompile for 1.27.1 release

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: tweak "help" output

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

printf: fix format string sanity check

One of the tests for printf checks for an invalid bare '%' in the
format string:

   $ busybox printf '%' a b c
   printf: %: invalid format

On x86_64 a slightly different test doesn't work correctly:

   $ busybox printf '%' d e f
   printf: invalid number 'd'
   printf: invalid number 'e'
   printf: invalid number 'f'

On other platforms the test fails randomly depending on how the
arguments are laid out in memory.

There are two places in the code where strchr is used to determine if
a character in the format string is valid.  However, strchr also returns
a valid pointer if the character being searched for is the null terminator
thus causing the code to incorrectly suppose that a valid character has
been found.

Add explicit checks for the null terminator.

Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: fix readonly2.tests failure

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: another testcase for "READONLY_VAR=VAL BLTIN ..."

Currently fails.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: do not assign to readonly VAR in "VAR=VAL CMD" syntax too

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: add readonly testcase, fix fallout

function                                             old     new   delta
helper_export_local                                  185     214     +29
run_pipe                                            1549    1560     +11
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 40/0)               Total: 40 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: convert exp/ro/local parameters to bitfields in one flag param

function                                             old     new   delta
helper_export_local                                  174     185     +11
set_local_var                                        424     420      -4
run_list                                            1048    1044      -4
set_vars_and_save_old                                 88      83      -5
set_local_var_from_halves                             27      22      -5
run_pipe                                            1554    1549      -5
builtin_export                                       173     168      -5
set_pwd_var                                           40      34      -6
builtin_readonly                                      70      64      -6
expand_one_var                                      1625    1618      -7
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/9 up/down: 11/-47)            Total: -36 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: HUSH_READONLY depends on HUSH

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: forgot to emit error on (failing) second "readonly VAR=VAL"

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: smaller code in !READONLY configs

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: implement "readonly" builtin

function                                             old     new   delta
builtin_readonly                                       -      70     +70
helper_export_local                                  152     174     +22
bltins1                                              348     360     +12
expand_one_var                                      1620    1625      +5
builtin_export                                       168     173      +5
set_pwd_var                                           36      40      +4
set_local_var                                        410     414      +4
set_vars_and_save_old                                 85      88      +3
set_local_var_from_halves                             24      27      +3
run_pipe                                            1551    1554      +3
run_list                                            1046    1048      +2
builtin_type                                         116     114      -2
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 10/1 up/down: 133/-2)           Total: 131 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: more general format ${var:EXPR:EXPR}

function                                             old     new   delta
subevalvar                                          1171    1202     +31
localcmd                                             364     366      +2

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: "you disabled math" is wrong: user did not disable it, builder of ash did

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: small fix to last commit

die_if_script() indeed dies only in scripts! Must handle the case where it continues.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: support ${VAR:N:-M}

function                                             old     new   delta
expand_one_var                                      1602    1615     +13
builtin_type                                         114     116      +2

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

shells: expand TODO comments, no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ttysize: if stdin is not tty, try stdout, then stderr

function                                             old     new   delta
ttysize_main                                         135     175     +40
packed_usage                                       31686   31672     -14

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: fix "cmd1 && cmd2 &" handling on NOMMU

function                                             old     new   delta
done_pipe                                            234     238      +4

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

arch: new applet (same as unmae -m). ~30 bytes.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

shell: and_or_and_backgrounding.tests is no longer "UNFIXED BUG"

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: fix and_or_and_backgrounding.tests failure

function                                             old     new   delta
done_pipe                                            133     218     +85

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

config: reorder items in "Busybox Settings", improve help

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

config: merge "Busybox Settings" and "Busybox Library Tuning" into one menu

Tweak a few help texts while at it

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Move get_unaligned_le32() macros to platform.h

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

fix more instances of ": $((a++))" in shell scripts

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

make_single_applets: fix ": $((fail++))" expansion error

$((fail++)) is not a required expression in POSIX, and in "dash" it
could produce an error like this:

    ./make_single_applets.sh: 61: arithmetic expression: expecting primary: "fail++"

Replace this with something portable: fail=$((fail+1)) would work.

Signed-off-by: Kang-Che Sung <explorer09@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

config: fix tab-damaged help text

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

tls: remove last int16 local variables in pstm code

function                                             old     new   delta
pstm_mul_comba                                       439     447      +8
pstm_sqr_comba                                       475     478      +3
pstm_montgomery_reduce                               399     381     -18
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/1 up/down: 11/-18)             Total: -7 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

tls: fix pstm asm constraint problem

function                                             old     new   delta
pstm_sqr_comba                                       551     475     -76

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

build system: FEATURE_LIBBUSYBOX_STATIC - try to pull libc/libm into libbusybox

It variously fails with different toolchains I tried...

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

blkdiscard: provide our own BLK[SEC]DISCARD if necessary

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

tls: fix build problem on non-static i386

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Makefile: include per-arch Makefile before Makefile.flags

Makefile.flags contains:

ARCH_FPIC ?= -fpic
ARCH_FPIE ?= -fpie

However, arch/$(ARCH)/Makefile gets included *after* Makefile.flags,
and therefore doesn't get the chance to provide its own value.

Fix this by including arch/$(ARCH)/Makefile *before* Makefile.flags.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

copyfd: guard use of munmap() with #if (windows builds need this)

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: protect WIFSTOPPED use with #if JOBS

This change fixes the build in setups where there are
no headers defining WIFSTOPPED and WSTOPSIG (where JOBS has to be
set to 0).

This partially reverts 4700fb5be (ash: make dowait() a bit more
readable. Logic is unchanged, 2015-10-09).

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libbb: hide getnetbyaddr() inside "#if ENABLE_FEATURE_ETC_NETWORKS" block

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: make "wait %1" work even if the job is dead

Example script:

sleep 1 | (sleep 1;exit 3) &
sleep 2
echo Zero:$?
wait %1
echo Three:$?

function                                             old     new   delta
clean_up_last_dead_job                                 -      24     +24
process_wait_result                                  426     447     +21
builtin_wait                                         285     293      +8
insert_job_into_table                                264     269      +5
builtin_jobs                                          68      73      +5
remove_job_from_table                                 59      57      -2
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 4/1 up/down: 63/-2)              Total: 61 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: remove superfluous comparison

function                                             old     new   delta
builtin_wait                                         291     285      -6

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

uuencode: allow space instead of "`" as padding char. Closes 10046

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

xxd: allow "-" as file name meaning stdin

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libbb/dump: allow skipping over stdin too

function                                             old     new   delta
next                                                 294     265     -29

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libbb/dump: fix a few broken commits and shrink code

function                                             old     new   delta
next                                                 310     294     -16

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libbb: safe_write should not return EINTR

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

dd: fix status=none. Closes 10066

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: add support for "set -e"

function                                             old     new   delta
run_list                                             978    1046     +68
o_opt_strings                                         24      32      +8
reset_traps_to_defaults                              136     142      +6
pick_sighandler                                       57      60      +3
packed_usage                                       31772   31770      -2
hush_main                                            983     961     -22
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 4/2 up/down: 85/-24)             Total: 61 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

cat: fix "cat -An" ignoring -n; make numbering go througn all files

function                                             old     new   delta
cat_main                                             418     428     +10

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

add-remove-shell: fix crash when shell is already added

Avoid dereferencing 'don_add' in strcmp since it is invalid
pointer.

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

cat: allow compiling out -n and -b

When these options were introduced in d88f94a5df3a2edb8ba56fab5c13674b452f87ab
it provides no config options to compile them out. Now provide one.

Introduce config FEATURE_CATN.

Signed-off-by: Kang-Che Sung <explorer09@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

unzip: TODO for symlink support

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

inetd,mount: do not die if uclibc without RPC is detected

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: add TODO for "set -e"

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: simplify insert_job_into_table() a bit

function                                             old     new   delta
done_word                                            767     761      -6
insert_job_into_table                                325     264     -61
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/2 up/down: 59/-126)           Total: -67 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: rename a few functions

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libbb: do not die if setgid/setuid(real_id) on startup fails

Based on a patch from Steven McDonald <steven@steven-mcdonald.id.au>:

This makes 'unshare --user' work correctly in the case where the user's
shell is provided by busybox itself.

'unshare --user' creates a new user namespace without any uid mappings.
As a result, /bin/busybox is setuid nobody:nogroup within the
namespace, as that is the only user. However, since no uids are mapped,
attempting to call setgid/setuid fails, even though this would do
nothing:

  $ unshare --user ./busybox.broken ash
  ash: setgid: Invalid argument

'unshare --map-root-user' still works, but because Linux only allows
uid/gid mappings to be set up once, creating a root mapping makes such
a namespace useless for creating multi-user containers.

With this patch, setgid and setuid will not be called in the case where
they would do nothing, which is always the case inside a new user
namespace because all uids are effectively mapped to nobody:

  $ id -u
  1000
  $ ls -lh busybox.fixed
  -rwsr-xr-x    1 root     root      826.2K May 21 00:33 busybox.fixed
  $ unshare --user ./busybox.fixed ash
  $ id -u
  65534

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

shuf: fix random line selection. Closes 9971

"""
For example, given input file:

    foo
    bar
    baz

after shuffling the input file, foo will never end up back on the first line.
This came to light when I ran into a use-case where someone was selecting
a random line from a file using shuf | head -n 1, and the results on busybox
were showing a statistical anomaly (as in, the first line would never ever
be picked) vs the same process running on environments that had gnu coreutils
installed.

On line https://git.busybox.net/busybox/tree/coreutils/shuf.c#n56 it uses
r %= i, which will result in 0 <= r < i, while the algorithm specifies
0 <= r <= i.
"""

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

crond: code shrink

function                                             old     new   delta
load_crontab                                         936     925     -11

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

crond: move misplaced comment

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

crond: support @daily etc

function                                             old     new   delta
start_jobs                                             -     348    +348
load_crontab                                         766     936    +170
static.SpecAry                                         -      96     +96
crond_main                                          1424    1134    -290
------------------------------------------------------------------------------
(add/remove: 2/0 grow/shrink: 1/1 up/down: 614/-290)          Total: 324 bytes

Based on patch by Jonathan Kolb <kolbyjack@gmail.com>

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

watchdog: do not use argc, other cleanups

function                                             old     new   delta
watchdog_main                                        298     291      -7

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

watchdog: stop watchdog first on startup

Some watchdog implementations may do things other than issue
a reboot on a watchdog timeout.  In this case, there's the
possibility of restarting this program from the state of
the watchdog device not being properly stopped (done by writing
a 'V' and closing the device).  Since it wasn't stopped, the
driver may not be able to restart the watchdog when this program
reopens it and starts pinging it.

To fix this, the code will always first issue the stop when it
starts up.

function                                             old     new   delta
shutdown_on_signal                                     -      32     +32
watchdog_main                                        268     298     +30
shutdown_watchdog                                      -      25     +25
watchdog_shutdown                                     41       -     -41
------------------------------------------------------------------------------
(add/remove: 2/1 grow/shrink: 1/0 up/down: 87/-41)             Total: 46 bytes

Signed-off-by: Matt Spinler <spinler@us.ibm.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

beep: disallow FEATURE_BEEP_FREQ = 0 in configuration

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: explain why wait5.tests is failing

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: fix "(sleep 1; exit 3) & sleep 2; echo $?; wait $!; echo $?"

function                                             old     new   delta
process_wait_result                                  414     426     +12
builtin_wait                                         283     291      +8
run_list                                             974     978      +4
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 3/0 up/down: 24/0)               Total: 24 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: fix two redirection testcase failures

function                                             old     new   delta
save_fds_on_redirect                                 183     256     +73
fcntl_F_DUPFD                                          -      46     +46
restore_redirects                                     74      96     +22
xdup_and_close                                        51      72     +21
setup_redirects                                      196     200      +4
hush_main                                            988     983      -5
static.C                                              12       -     -12
run_pipe                                            1595    1551     -44
------------------------------------------------------------------------------
(add/remove: 1/1 grow/shrink: 4/2 up/down: 166/-61)           Total: 105 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

main: fix the case where user has "halt" as login shell. Closes 9986

halt::0:0::/:/sbin/halt

function                                             old     new   delta
run_applet_and_exit                                  748     751      +3
run_applet_no_and_exit                               467     459      -8

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: remove dependency on libcap headers

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: factor out capability name printing

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: code shrink, and grouping capability code together

function                                             old     new   delta
static.versions                                        -       3      +3
getcaps                                              174     149     -25

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: allow modifying ambient capabilities

With Linux 4.3, a new set of capabilities has been introduced with the
ambient capabilities. These aim to solve the problem that it was
impossible to grant run programs with elevated privileges across
non-root users. Quoting from capabilities(7):

    This is a set of capabilities that are preserved across an execve(2)
    of a program that is not privileged.  The ambient capability set
    obeys the invariant that no capability can ever be ambient if it is
    not both permitted and inheritable.

With this new set of capabilities it is now possible to run an
executable with elevated privileges as a different user, making it much
easier to do proper privilege separation.

Note though that the `--ambient-caps` switch is not part of any released
version of util-linux, yet. It has been applied in 0c92194ee (setpriv:
support modifying the set of ambient capabilities, 2017-06-24) and will
probably be part of v2.31.

function                                             old     new   delta
parse_cap                                              -     174    +174
setpriv_main                                        1246    1301     +55
.rodata                                           146307  146347     +40
static.setpriv_longopts                               40      55     +15
packed_usage                                       32092   32079     -13

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: allow modifying inheritable caps

The main use case of setpriv is to modify the current state of
privileges available to the calling process and spawn a new executable
with the modified, new state. Next to the already supported case of
modifying the no-new-privs flag, util-linux also supports to modify
capability sets.

This commit introduces to add or drop capabilities from the set of
inheritable capabilities. Quoting from capabilities(7):

    This is a set of capabilities preserved across an execve(2).
    Inheritable capabilities remain inheritable when executing any
    program, and inheritable capabilities are added to the permitted set
    when executing a program that has the corresponding bits set in the
    file inheritable set.

As such, inheritable capabilities enable executing files with certain
privileges if the file itself has these privileges set. Note though that
inheritable capabilities are dropped across execve when running as a
non-root user.

function                                             old     new   delta
getcaps                                                -     237    +237
setpriv_main                                        1129    1246    +117
.rodata                                           146198  146307    +109
static.setpriv_longopts                               29      40     +11
packed_usage                                       32107   32092     -15

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: dump ambient capabilities

As with the previous commit, this commit introduces the ability to dump
the set of ambient capabilities.

function                                             old     new   delta
setpriv_main                                         982    1129    +147
.rodata                                           146148  146198     +50

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: dump capability bounding set

As with the previous commit, this one implements the ability to dump the
capability bounding set.

function                                             old     new   delta
setpriv_main                                         838     982    +144
.rodata                                           146101  146148     +47

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: dump inheritable capability set

The setpriv executable from util-linux also dumps out information on the
different capability sets known by the kernel. By default, these are the
inheritable capabilities, bounding capabilities and (not yet released)
the ambient capabilities, which have been introduced with Linux 4.3.
This patch introduces the ability to dump the set of inheritable
capabilities.

By default, setpriv(1) identifies capabilities by their human-readable
name, for example 'net_admin'. For unknown capabilities, though, it does
instead use the capability's value, for example 'cap_12', which is
equivalent to 'net_admin'. As there is no kernel interface to retrieve
capability names by their index, we have to declare these ourselves,
which adds to setpriv's size.

To counteract, using the human-readble name has been made configurable.
The following sizes are with the 'FEATURE_SETPRIV_CAPABILITY_NAMES'
enabled:

function                                             old     new   delta
.rodata                                           145969  146405    +436
setpriv_main                                         467     842    +375
capabilities                                           -     304    +304

And with 'FEATURE_SETPRIV_CAPABILITY_NAMES' disabled:

function                                             old     new   delta
setpriv_main                                         467     838    +371
.rodata                                           145969  146101    +132

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setpriv: dump no-new-privs info

Introduce the ability to dump the state of the no-new-privs flag, which
states whethere it is allowed to grant new privileges.

function                                             old     new   delta
setpriv_main                                         419     467     +48
.rodata                                           145926  145969     +43

Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

shell: sync redir/* tests

Note: hush-redir/redir_to_bad_fd.tests currently fails

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: rename redir5.tests (hush has redir5.tests which is different)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: add a TODO about redir3.tests failure

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: rename hush-redir/redir3.tests (ash has redir3.tests which id different)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: Print error messages on shift -1

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>