Merge util-linux (2.40-8) import into refs/heads/workingbranch

lsns: fix netns use

 # ip netns add vpn
 # lsns -T -t net
 Segmentation fault (core dumped)

The function interpolate_missing_namespaces() reads data from /proc.
However, in the case of a persistent namespace, there is no procfs
entry for the namespace. Therefore, this function should ignore it.

Fixes: https://github.com/util-linux/util-linux/issues/2982
Signed-off-by: Karel Zak <kzak@redhat.com>
(cherry picked from commit 42880f54894c1d7fd113552ff8474566a2595a0d)

Gbp-Pq: Topic upstream
Gbp-Pq: Name lsns-fix-netns-use.patch

libmount: fix comment typo for mnt_fs_get_comment()

Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
(cherry picked from commit c630dbc38f7cd1f838e65e2c43d8375bce1ec20f)

Gbp-Pq: Topic upstream
Gbp-Pq: Name libmount-fix-comment-typo-for-mnt_fs_get_comment.patch

libmount: Fix access check for utab in context

The function mnt_has_regular_utab() properly detects that the utab is
not writable, but this is ignored by the high-level context API. As a
result, the library later attempts to update the file and ends up with
a warning in mount(8):

 $ mkdir sys
 $ unshare --map-root-user --mount
 $ mount --rbind /sys sys
 $ umount --lazy sys; echo $?
 umount: /home/user/sys: filesystem was unmounted, but failed to update userspace mount table.
 16

In this case, the utab should be ignored.

Fixes: https://github.com/util-linux/util-linux/issues/2981
Signed-off-by: Karel Zak <kzak@redhat.com>
(cherry picked from commit 97f7bfc0fdf74b6a6e220ba9d2f620386e660b29)

Gbp-Pq: Topic upstream
Gbp-Pq: Name libmount-Fix-access-check-for-utab-in-context.patch

lsblk: simplify SOURCES code

This complicated implementation originated from an early version of
zero-separated items for libsmartcols. It is no longer necessary.

References: https://github.com/util-linux/util-linux/pull/2983
Signed-off-by: Karel Zak <kzak@redhat.com>
(cherry picked from commit af4c7dac191abadc33aae75916583c890cd2ebc3)

Gbp-Pq: Topic upstream
Gbp-Pq: Name lsblk-simplify-SOURCES-code.patch

findmnt: always zero-terminate SOURCES data

libsmartcols expects it's data fields to be zero terminated.
See the call to strlen() in scols_column_greatest_wrap().
ul_buffer however does not guarantee that termination,
ul_buffer_append_strings() discard the zero-termination.

Always zero-terminate in get_data_col_sources() and drop the now
unnecessary variable "i".

Closes: https://github.com/util-linux/util-linux/issues/2980
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
(cherry picked from commit 6594679e9af805075c282da35b1b2a8d767e403b)

Gbp-Pq: Topic upstream
Gbp-Pq: Name findmnt-always-zero-terminate-SOURCES-data.patch

agetty: Don't override TERM passed by the user

Before 4869b259d68f65ea88df625ce8df9c0177d55a01, any TERM passed
on the agetty command line would be used instead of the default TERM.
After 4869b259d68f65ea88df625ce8df9c0177d55a01, the default TERM is
used unconditionally.

Fix the regression by checking if the user passed a custom TERM.

Fixes: 4869b259d68f65ea88df625ce8df9c0177d55a01
(cherry picked from commit af354e92111769d57e43eb4f1825c0d99c894ddb)

Gbp-Pq: Topic upstream
Gbp-Pq: Name agetty-Don-t-override-TERM-passed-by-the-user.patch

libsmartcols: reset wrap after calculation

Fixes: https://github.com/util-linux/util-linux/issues/2956
Signed-off-by: Karel Zak <kzak@redhat.com>
(cherry picked from commit 62f64717ec134b10c5a670403c2d8c43b608e671)

Gbp-Pq: Topic upstream
Gbp-Pq: Name libsmartcols-reset-wrap-after-calculation.patch

lslocks: remove a unused local variable

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
(cherry picked from commit c8c85e8e54aa201766f2420e2a7c86ccfa8d5df0)

Gbp-Pq: Topic upstream
Gbp-Pq: Name lslocks-remove-a-unused-local-variable.patch

lslocks: don't abort gathering per-process information even if opening a /proc/[0-9]* fails

If a process ($pid) taking a lock is gone while running lslocks,
gathering per-process information for the process may fail in opening
/proc/$pid. Though lslocks should work with incomplete information,
the original code stopped gathering per-process information for the
other processes.

As a result of the original behavior, tests/ts/lslocks/lslocks really
failed in an environment where multiple test cases ran simultaneously.

Close #2624 again.
See also #2633.

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
(cherry picked from commit 20df923cc3d2da12a5296223c3ff0de7484cc587)

Gbp-Pq: Topic upstream
Gbp-Pq: Name lslocks-don-t-abort-gathering-per-process-information-eve.patch

lsns: tolerate lsns_ioctl(fd, NS_GET_{PARENT,USERNS}) failing with ENOSYS

With the original code, "lsns/filedesc" test case failed on
"build (qemu-user, s390x)" and "build (qemu-user, riscv64)".

On the platforms, lsns_ioctl(fd, NS_GET_{PARENT,USERNS}) failed
with ENOSYS. The error stoped the iteration for gathering
information from /proc/[0-9]+. As a result, lsns printed
nothing. We don't expect this behavior.

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
(cherry picked from commit f2a8b20d9c63f771d1fddd639ea1ec3fe034dc6d)

Gbp-Pq: Topic upstream
Gbp-Pq: Name lsns-tolerate-lsns_ioctl-fd-NS_GET_-PARENT-USERNS-failing.patch

lsns: report with warnx if a namespace related ioctl fails with ENOSYS

Qemu userspace emulation reports ENOSYS if it doesn't support a given
ioctl command.

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
(cherry picked from commit 0a7a8fbc7a82dfbd9e925d1b4d4936bf4e011bc1)

Gbp-Pq: Topic upstream
Gbp-Pq: Name lsns-report-with-warnx-if-a-namespace-related-ioctl-fails.patch

Fix misplaced else in mnt_update_already_done

See 477401f0de

(cherry picked from commit 6b316e0c9f1feb61bef3bb14b1356be8fbefcc17)

Gbp-Pq: Topic upstream
Gbp-Pq: Name Fix-misplaced-else-in-mnt_update_already_done.patch

findmnt: revise the code for -I and -D option

Fixes #2913.

3dd79293b5b655da9d913dedd8facb08959a7826 added -I option.
However, the code used bit flags (FL_DF and FL_DF_INODES) wrongly;
the code broke the output of -D option.

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
(cherry picked from commit aa0c45e8686df509c169121473d2cfb4fe18c3ea)

Gbp-Pq: Topic upstream
Gbp-Pq: Name findmnt-revise-the-code-for-I-and-D-option.patch

libblkid: topology/ioctl: simplify ioctl handling

Coverity complains about the data copy within the union.
Instead unroll the loop which is less code and easier to follow.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
(cherry picked from commit 2ab95860db0b7423906911b8b9e9f231332c2c11)

Gbp-Pq: Topic upstream
Gbp-Pq: Name libblkid-topology-ioctl-simplify-ioctl-handling.patch

libblkid: topology/ioctl: correctly handle kernel types

Commit 5d71d711d07a ("libblkid: topolicy/ioctl: use union for multiple data types")
incorrectly assumed that set_ulong and set_int refer to the type
returned by the kernel. Instead the different function pointer names
refer to the types of the function pointers.
However all ioctls, except for the later added BLKGETDISKSEQ, return
32bit integers.
This made libblkid also interpret the upper 32bits too, leading to
garbage values.

Introduce a new member 'kernel_size' to also handle the 64bit
BLKGETDISKSEQ.

Drop data.ul as it is no actually used.

Closes #2904
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
(cherry picked from commit a10081a52c48a5232db30d8e0a7a7570cc8163ae)

Gbp-Pq: Topic upstream
Gbp-Pq: Name libblkid-topology-ioctl-correctly-handle-kernel-types.patch

pam_lastlog2: link against liblastlog

While at it also drop the duplicated include path.

Fixes #2897
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
(cherry picked from commit 2722b40012740372daf3585b7c071fcbede13359)

Gbp-Pq: Topic upstream
Gbp-Pq: Name pam_lastlog2-link-against-liblastlog.patch

libblkid: Fix segfault when blkid.conf doesn't exist

* Move 'line' and 'uevent' to the beginning of the LIBECONF code.
* Remove unwanted space between function name and arguments.
* Check for 'line' pointer before dereferencing.

References: https://github.com/util-linux/util-linux/pull/2883
Signed-off-by: Karel Zak <kzak@redhat.com>
(cherry picked from commit 7e357241b413a01c37b0b4d064bc0a47e3259361)

Gbp-Pq: Topic upstream
Gbp-Pq: Name libblkid-Fix-segfault-when-blkid.conf-doesn-t-exist.patch

Revert "lib/pager: Apply pager-specific fixes only when needed"

Less is not always detectable, as it can be hidden behind symlinks or
wrapper as for example in a default Debian installation.
Also testing for the literal string "less" does not match full path
specifications like "/usr/bin/less".

Instead always apply the fixes.

This reverts commit c10ad975895372122f72c8d9da089b6ea69f778b.
Closes #2951
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Debian-Bug: 1068831

Gbp-Pq: Topic upstream
Gbp-Pq: Name Revert-lib-pager-Apply-pager-specific-fixes-only-when-nee.patch

audit-arch.h: add defines for m68k, sh

John Paul Adrian Glaubitz points out that libseccomp needs to release 2.6.0,
too.

Upstream commit 5ebf0edb0a0531cba801c7791543c610725bf944.

Signed-off-by: Chris Hofstaedtler <zeha@debian.org>
Gbp-Pq: Topic upstream
Gbp-Pq: Name audit-arch.h-add-defines-for-m68k-sh.patch

tests: (lsfd::mkfds-multiplexing) skip if /proc/$pid/syscall is broken

Close #2867
Close #2887

We should skip the test case on the platforms where /proc/$pid/syscall
doesn't report correct system call number. On such platforms,

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name tests-lsfd-mkfds-multiplexing-skip-if-proc-pid-syscall-is.patch

tests: mark fadvise/drop as known failing on buildds

buildds build on a tmpfs, but this is not visible inside the build environment.

Gbp-Pq: Topic debian
Gbp-Pq: Name tests-mark-fadvise-drop-as-known-failing-on-buildds.patch

tests: (test_mkfds::sockdiag) verify the recived message

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name tests-test_mkfds-sockdiag-verify-the-recived-message.patch

lsfd: add LSFD_DEBUG env var for debugging

Signed-off-by: Masatake YAMATO <yamato@redhat.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name lsfd-add-LSFD_DEBUG-env-var-for-debugging.patch

tests: mark lsfd mkfds-socketpair as failing in sbuild

Gbp-Pq: Topic debian
Gbp-Pq: Name tests-mark-lsfd-tests-failing-in-sbuild.patch

Use sensible-pager

Closes: #1014368
Gbp-Pq: Topic debian
Gbp-Pq: Name sensible-pager.patch

Install lsfd into /usr/bin

Gbp-Pq: Topic debian
Gbp-Pq: Name lsfd-usrbin.patch

hardlink tests: set known failed

sha256 cannot be initialized on some archs / buildds.

Gbp-Pq: Topic debian
Gbp-Pq: Name hardlink-tests-known-failed.patch

verbose-tests

Enable verbose tests and print content of failing tests output file.

Gbp-Pq: Topic debian
Gbp-Pq: Name verbose-tests.patch

util-linux: getopt(1) should point to where examples are on a Debian system

Debian BTS #913049

Gbp-Pq: Topic debian
Gbp-Pq: Name man-getopt-examples.patch

util-linux (2.40-8) unstable; urgency=medium

  [ Helmut Grohne ]
  * Also cover ctrlaltdel in /usr-move mitigation (Closes: #1069923)

[dgit import unpatched util-linux 2.40-8]

Import util-linux_2.40-8.debian.tar.xz

[dgit import tarball util-linux 2.40-8 util-linux_2.40-8.debian.tar.xz]

Import util-linux_2.40.orig.tar.xz

[dgit import orig util-linux_2.40.orig.tar.xz]

Merge util-linux (2.39.3-6) import into refs/heads/workingbranch

build: only build test_enosys if an audit arch exists

Link: https://lore.kernel.org/util-linux/9211accf8670f28778166a1acdc186e8dd28f2e8.camel@physik.fu-berlin.de/
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name build-only-build-test_enosys-if-an-audit-arch-exists.patch

fincore: (tests) also use nosize error file

Link: https://lore.kernel.org/util-linux/b8cafd0d89878949705782312dfe800ec2779035.camel@physik.fu-berlin.de/
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fincore-tests-also-use-nosize-error-file.patch

cfdisk: fix menu behavior after writing changes

Florian wrote:
  after a successful write, cfdisk remains on the "Write" cursor and
  furthermore when navigating to "Quit" will continue to  show
  "...without writing changes", despite there were writes. This patch
  addresses that.

Based on patch from Florian Zimmermann <florian.zimmermann@gmail.com>

Signed-off-by: Karel Zak <kzak@redhat.com>
Closes: #927041
Gbp-Pq: Topic upstream
Gbp-Pq: Name cfdisk-fix-menu-behavior-after-writing-changes.patch

setarch: add riscv64/riscv32 support

Signed-off-by: Michal Biesek <michalbiesek@gmail.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name setarch-add-riscv64-riscv32-support.patch

fadvise: (test) test with 64k blocks

This will allow the tests to also pass on systems with 64k pagesizes.

Closes #2249
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-test-with-64k-blocks.patch

fadvise: (test) don't compare fincore page counts

These depend on the machines pagesize and are therefore not a good
comparision.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-don-t-compare-fincore-page-counts.patch

fadvise: (tests) factor out calls to "fincore"

This will make it easier to pass argument later.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-tests-factor-out-calls-to-fincore.patch

fadvise: (test) dynamically calculate expected test values

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-dynamically-calculate-expected-test-values.patch

enosys: fix build on hppa

The following patch fixes the definition of SECCOMP_ARCH_NATIVE
on hppa.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Gbp-Pq: Topic upstream
Gbp-Pq: Name enosys-fix-build-on-hppa.patch

fincore: (tests) fix double log output

The fincore tests call ts_log_both inside an output redirection of both stdout
and stderr, leading to the ts_log_both output ending up twice in both stdout
and stderr.
Call ts_log_both before setting up the output redirection.

Signed-off-by: Chris Hofstaedtler <zeha@debian.org>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fincore-tests-fix-double-log-output.patch

Use sensible-pager

Closes: #1014368
Gbp-Pq: Topic debian
Gbp-Pq: Name sensible-pager.patch

Install lsfd into /usr/bin

Gbp-Pq: Topic debian
Gbp-Pq: Name lsfd-usrbin.patch

hardlink tests: set known failed

sha256 cannot be initialized on some archs / buildds.

Gbp-Pq: Topic debian
Gbp-Pq: Name hardlink-tests-known-failed.patch

verbose-tests

Enable verbose tests and print content of failing tests output file.

Gbp-Pq: Topic debian
Gbp-Pq: Name verbose-tests.patch

util-linux: getopt(1) should point to where examples are on a Debian system

Debian BTS #913049

Gbp-Pq: Topic debian
Gbp-Pq: Name man-getopt-examples.patch

util-linux (2.39.3-6) unstable; urgency=medium

  * Upload to unstable.

[dgit import unpatched util-linux 2.39.3-6]

Import util-linux_2.39.3-6.debian.tar.xz

[dgit import tarball util-linux 2.39.3-6 util-linux_2.39.3-6.debian.tar.xz]

Merge util-linux (2.39.3-2) import into refs/heads/workingbranch

fincore: (tests) also use nosize error file

Link: https://lore.kernel.org/util-linux/b8cafd0d89878949705782312dfe800ec2779035.camel@physik.fu-berlin.de/
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fincore-tests-also-use-nosize-error-file.patch

cfdisk: fix menu behavior after writing changes

Florian wrote:
  after a successful write, cfdisk remains on the "Write" cursor and
  furthermore when navigating to "Quit" will continue to  show
  "...without writing changes", despite there were writes. This patch
  addresses that.

Based on patch from Florian Zimmermann <florian.zimmermann@gmail.com>

Signed-off-by: Karel Zak <kzak@redhat.com>
Closes: #927041
Gbp-Pq: Topic upstream
Gbp-Pq: Name cfdisk-fix-menu-behavior-after-writing-changes.patch

setarch: add riscv64/riscv32 support

Signed-off-by: Michal Biesek <michalbiesek@gmail.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name setarch-add-riscv64-riscv32-support.patch

fadvise: (test) test with 64k blocks

This will allow the tests to also pass on systems with 64k pagesizes.

Closes #2249
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-test-with-64k-blocks.patch

fadvise: (test) don't compare fincore page counts

These depend on the machines pagesize and are therefore not a good
comparision.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-don-t-compare-fincore-page-counts.patch

fadvise: (tests) factor out calls to "fincore"

This will make it easier to pass argument later.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-tests-factor-out-calls-to-fincore.patch

fadvise: (test) dynamically calculate expected test values

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-dynamically-calculate-expected-test-values.patch

enosys: fix build on hppa

The following patch fixes the definition of SECCOMP_ARCH_NATIVE
on hppa.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Gbp-Pq: Topic upstream
Gbp-Pq: Name enosys-fix-build-on-hppa.patch

fincore: (tests) fix double log output

The fincore tests call ts_log_both inside an output redirection of both stdout
and stderr, leading to the ts_log_both output ending up twice in both stdout
and stderr.
Call ts_log_both before setting up the output redirection.

Signed-off-by: Chris Hofstaedtler <zeha@debian.org>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fincore-tests-fix-double-log-output.patch

Use sensible-pager

Closes: #1014368
Gbp-Pq: Topic debian
Gbp-Pq: Name sensible-pager.patch

Install lsfd into /usr/bin

Gbp-Pq: Topic debian
Gbp-Pq: Name lsfd-usrbin.patch

hardlink tests: set known failed

sha256 cannot be initialized on some archs / buildds.

Gbp-Pq: Topic debian
Gbp-Pq: Name hardlink-tests-known-failed.patch

verbose-tests

Enable verbose tests and print content of failing tests output file.

Gbp-Pq: Topic debian
Gbp-Pq: Name verbose-tests.patch

util-linux: getopt(1) should point to where examples are on a Debian system

Debian BTS #913049

Gbp-Pq: Topic debian
Gbp-Pq: Name man-getopt-examples.patch

util-linux (2.39.3-2) unstable; urgency=medium

  [ Gioele Barabucci ]
  * Install manpages via dh_installman instead of dh_install
  * d/util-linux-locales.install: Do not install manpages when building with <nodoc>
  * d/control: Support <nodoc> build profile

[dgit import unpatched util-linux 2.39.3-2]

Import util-linux_2.39.3-2.debian.tar.xz

[dgit import tarball util-linux 2.39.3-2 util-linux_2.39.3-2.debian.tar.xz]

Import util-linux_2.39.3.orig.tar.xz

[dgit import orig util-linux_2.39.3.orig.tar.xz]

Merge util-linux (2.39.2-6) import into refs/heads/workingbranch

build-sys: fix libmount/src/hooks.c use

Reported-by: Samuel Thibault <samuel.thibault@aquilenet.fr>
Signed-off-by: Karel Zak <kzak@redhat.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name build-sys-fix-libmount-src-hooks-c-use.patch

setterm: avoid restoring flags from uninitialized memory

Depending on the used compiler and flags, previously either F_SETFL was called
with 0 or with a random value. Never with the intended previous flags.

[kzak@redhat.com: - tiny coding style change]

Signed-off-by: Chris Hofstaedtler <zeha@debian.org>
Tested-by: Emanuele Rocca <ema@debian.org>
Signed-off-by: Karel Zak <kzak@redhat.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name setterm-resize-uninit-flags.patch

cfdisk: fix menu behavior after writing changes

Florian wrote:
  after a successful write, cfdisk remains on the "Write" cursor and
  furthermore when navigating to "Quit" will continue to  show
  "...without writing changes", despite there were writes. This patch
  addresses that.

Based on patch from Florian Zimmermann <florian.zimmermann@gmail.com>

Signed-off-by: Karel Zak <kzak@redhat.com>
Closes: #927041
Gbp-Pq: Topic upstream
Gbp-Pq: Name cfdisk-fix-menu-behavior-after-writing-changes.patch

setarch: add riscv64/riscv32 support

Signed-off-by: Michal Biesek <michalbiesek@gmail.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name setarch-add-riscv64-riscv32-support.patch

libmount: Fix regression when mounting with atime

A regression was introduced in v2.39 that causes mounting with the atime
option to fail:

  $ mkfs.ext4 -F /dev/sdi
  $ mount -o atime /dev/sdi /mnt/sdi
  mount: /mnt/sdi: not mount point or bad option.
         dmesg(1) may have more information after failed mount system call.

The failure comes from the mount_setattr(2) call returning -EINVAL. This
is because we pass an invalid value for the attr_clr argument. From a
strace capture we have:

  mount_setattr(4, "", AT_EMPTY_PATH, {attr_set=0, attr_clr=MOUNT_ATTR_NOATIME, propagation=0 /* MS_??? */, userns_fd=0}, 32) = -1 EINVAL (Invalid argument)

We can't pass MOUNT_ATTR_NOATIME to mount_setattr(2) through the attr_clr
argument because all atime options are exclusive, so in order to set atime
one has to pass MOUNT_ATTR__ATIME to attr_clr and leave attr_set as
MOUNT_ATTR_RELATIME (which is defined as a value of 0).

This can be read from the man page for mount_setattr(2) and also from the
kernel source:

  $ cat fs/namespace.c
  static int build_mount_kattr(const struct mount_attr *attr, size_t usize,
                               struct mount_kattr *kattr, unsigned int flags)
  {
      (...)
      /*
       * Since the MOUNT_ATTR_<atime> values are an enum, not a bitmap,
       * users wanting to transition to a different atime setting cannot
       * simply specify the atime setting in @attr_set, but must also
       * specify MOUNT_ATTR__ATIME in the @attr_clr field.
       * So ensure that MOUNT_ATTR__ATIME can't be partially set in
       * @attr_clr and that @attr_set can't have any atime bits set if
       * MOUNT_ATTR__ATIME isn't set in @attr_clr.
       */
      if (attr->attr_clr & MOUNT_ATTR__ATIME) {
          if ((attr->attr_clr & MOUNT_ATTR__ATIME) != MOUNT_ATTR__ATIME)
              return -EINVAL;

              /*
               * Clear all previous time settings as they are mutually
               * exclusive.
               */
              kattr->attr_clr |= MNT_RELATIME | MNT_NOATIME;
              switch (attr->attr_set & MOUNT_ATTR__ATIME) {
              case MOUNT_ATTR_RELATIME:
                  kattr->attr_set |= MNT_RELATIME;
                  break;
              case MOUNT_ATTR_NOATIME:
                  kattr->attr_set |= MNT_NOATIME;
                  break;
              case MOUNT_ATTR_STRICTATIME:
                  break;
              default:
                  return -EINVAL;
              }
    (...)

So fix this by setting attr_clr MOUNT_ATTR__ATIME if we want to clear any
atime related option.

Signed-off-by: Filipe Manana <fdmanana@kernel.org>
Closes: #1042714
Gbp-Pq: Topic upstream
Gbp-Pq: Name libmount-Fix-regression-when-mounting-with-atime.patch

fadvise: (test) test with 64k blocks

This will allow the tests to also pass on systems with 64k pagesizes.

Closes #2249
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-test-with-64k-blocks.patch

fadvise: (test) don't compare fincore page counts

These depend on the machines pagesize and are therefore not a good
comparision.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-don-t-compare-fincore-page-counts.patch

fadvise: (tests) factor out calls to "fincore"

This will make it easier to pass argument later.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-tests-factor-out-calls-to-fincore.patch

fadvise: (test) dynamically calculate expected test values

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-dynamically-calculate-expected-test-values.patch

enosys: fix build on hppa

The following patch fixes the definition of SECCOMP_ARCH_NATIVE
on hppa.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Gbp-Pq: Topic upstream
Gbp-Pq: Name enosys-fix-build-on-hppa.patch

fincore: (tests) fix double log output

The fincore tests call ts_log_both inside an output redirection of both stdout
and stderr, leading to the ts_log_both output ending up twice in both stdout
and stderr.
Call ts_log_both before setting up the output redirection.

Signed-off-by: Chris Hofstaedtler <zeha@debian.org>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fincore-tests-fix-double-log-output.patch

Use sensible-pager

Closes: #1014368
Gbp-Pq: Topic debian
Gbp-Pq: Name sensible-pager.patch

Install lsfd into /usr/bin

Gbp-Pq: Topic debian
Gbp-Pq: Name lsfd-usrbin.patch

hardlink tests: set known failed

sha256 cannot be initialized on some archs / buildds.

Gbp-Pq: Topic debian
Gbp-Pq: Name hardlink-tests-known-failed.patch

verbose-tests

Enable verbose tests and print content of failing tests output file.

Gbp-Pq: Topic debian
Gbp-Pq: Name verbose-tests.patch

util-linux: getopt(1) should point to where examples are on a Debian system

Debian BTS #913049

Gbp-Pq: Topic debian
Gbp-Pq: Name man-getopt-examples.patch

util-linux (2.39.2-6) unstable; urgency=medium

  [ Samuel Thibault ]
  * d/util-linux.manpages,util-linux-locales.install: Generalize hurd-i386 into
    hurd-any.

  [ Chris Hofstaedtler ]
  * Fix FTBFS on hurd-i386 using upstreamed patch.
    Thanks to Samuel Thibault <sthibault@debian.org> (Closes: #1055875)

[dgit import unpatched util-linux 2.39.2-6]

Import util-linux_2.39.2-6.debian.tar.xz

[dgit import tarball util-linux 2.39.2-6 util-linux_2.39.2-6.debian.tar.xz]

Merge util-linux (2.39.2-5) import into refs/heads/workingbranch

setterm: avoid restoring flags from uninitialized memory

Depending on the used compiler and flags, previously either F_SETFL was called
with 0 or with a random value. Never with the intended previous flags.

Signed-off-by: Chris Hofstaedtler <zeha@debian.org>
Gbp-Pq: Topic debian
Gbp-Pq: Name setterm-resize-uninit-flags.patch

cfdisk: fix menu behavior after writing changes

Florian wrote:
  after a successful write, cfdisk remains on the "Write" cursor and
  furthermore when navigating to "Quit" will continue to  show
  "...without writing changes", despite there were writes. This patch
  addresses that.

Based on patch from Florian Zimmermann <florian.zimmermann@gmail.com>

Signed-off-by: Karel Zak <kzak@redhat.com>
Closes: #927041
Gbp-Pq: Topic upstream
Gbp-Pq: Name cfdisk-fix-menu-behavior-after-writing-changes.patch

setarch: add riscv64/riscv32 support

Signed-off-by: Michal Biesek <michalbiesek@gmail.com>
Gbp-Pq: Topic upstream
Gbp-Pq: Name setarch-add-riscv64-riscv32-support.patch

libmount: Fix regression when mounting with atime

A regression was introduced in v2.39 that causes mounting with the atime
option to fail:

  $ mkfs.ext4 -F /dev/sdi
  $ mount -o atime /dev/sdi /mnt/sdi
  mount: /mnt/sdi: not mount point or bad option.
         dmesg(1) may have more information after failed mount system call.

The failure comes from the mount_setattr(2) call returning -EINVAL. This
is because we pass an invalid value for the attr_clr argument. From a
strace capture we have:

  mount_setattr(4, "", AT_EMPTY_PATH, {attr_set=0, attr_clr=MOUNT_ATTR_NOATIME, propagation=0 /* MS_??? */, userns_fd=0}, 32) = -1 EINVAL (Invalid argument)

We can't pass MOUNT_ATTR_NOATIME to mount_setattr(2) through the attr_clr
argument because all atime options are exclusive, so in order to set atime
one has to pass MOUNT_ATTR__ATIME to attr_clr and leave attr_set as
MOUNT_ATTR_RELATIME (which is defined as a value of 0).

This can be read from the man page for mount_setattr(2) and also from the
kernel source:

  $ cat fs/namespace.c
  static int build_mount_kattr(const struct mount_attr *attr, size_t usize,
                               struct mount_kattr *kattr, unsigned int flags)
  {
      (...)
      /*
       * Since the MOUNT_ATTR_<atime> values are an enum, not a bitmap,
       * users wanting to transition to a different atime setting cannot
       * simply specify the atime setting in @attr_set, but must also
       * specify MOUNT_ATTR__ATIME in the @attr_clr field.
       * So ensure that MOUNT_ATTR__ATIME can't be partially set in
       * @attr_clr and that @attr_set can't have any atime bits set if
       * MOUNT_ATTR__ATIME isn't set in @attr_clr.
       */
      if (attr->attr_clr & MOUNT_ATTR__ATIME) {
          if ((attr->attr_clr & MOUNT_ATTR__ATIME) != MOUNT_ATTR__ATIME)
              return -EINVAL;

              /*
               * Clear all previous time settings as they are mutually
               * exclusive.
               */
              kattr->attr_clr |= MNT_RELATIME | MNT_NOATIME;
              switch (attr->attr_set & MOUNT_ATTR__ATIME) {
              case MOUNT_ATTR_RELATIME:
                  kattr->attr_set |= MNT_RELATIME;
                  break;
              case MOUNT_ATTR_NOATIME:
                  kattr->attr_set |= MNT_NOATIME;
                  break;
              case MOUNT_ATTR_STRICTATIME:
                  break;
              default:
                  return -EINVAL;
              }
    (...)

So fix this by setting attr_clr MOUNT_ATTR__ATIME if we want to clear any
atime related option.

Signed-off-by: Filipe Manana <fdmanana@kernel.org>
Closes: #1042714
Gbp-Pq: Topic upstream
Gbp-Pq: Name libmount-Fix-regression-when-mounting-with-atime.patch

fadvise: (test) test with 64k blocks

This will allow the tests to also pass on systems with 64k pagesizes.

Closes #2249
Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-test-with-64k-blocks.patch

fadvise: (test) don't compare fincore page counts

These depend on the machines pagesize and are therefore not a good
comparision.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-don-t-compare-fincore-page-counts.patch

fadvise: (tests) factor out calls to "fincore"

This will make it easier to pass argument later.

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-tests-factor-out-calls-to-fincore.patch

fadvise: (test) dynamically calculate expected test values

Signed-off-by: Thomas Weißschuh <thomas@t-8ch.de>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fadvise-test-dynamically-calculate-expected-test-values.patch

enosys: fix build on hppa

The following patch fixes the definition of SECCOMP_ARCH_NATIVE
on hppa.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Gbp-Pq: Topic upstream
Gbp-Pq: Name enosys-fix-build-on-hppa.patch

fincore: (tests) fix double log output

The fincore tests call ts_log_both inside an output redirection of both stdout
and stderr, leading to the ts_log_both output ending up twice in both stdout
and stderr.
Call ts_log_both before setting up the output redirection.

Signed-off-by: Chris Hofstaedtler <zeha@debian.org>
Gbp-Pq: Topic upstream
Gbp-Pq: Name fincore-tests-fix-double-log-output.patch

Use sensible-pager

Closes: #1014368
Gbp-Pq: Topic debian
Gbp-Pq: Name sensible-pager.patch