Re: [gentoo-dev] [PATCH v4] mount-boot.eclass: Check if /boot is sane, but don't try to mount it.

2019-12-10 Thread Ulrich Mueller
> On Sat, 07 Dec 2019, Ulrich Müller wrote:

> The eclass failed to remount a read-only mounted /boot, because package
> collision sanity checks in recent Portage versions prevented it from
> reaching pkg_preinst() at all. Furthermore, with the "mount-sandbox"
> feature enabled, the mount won't be propagated past pkg_preinst() and
> installed files would end up under the (shadowed) mount point.

> Therefore don't even attempt to mount /boot ourselves, but error out
> if it isn't mounted read/write and ask the user to mount /boot.

> Also clean up and simplify. (For example, awk is a grown-up program
> which doesn't need any help from egrep or sed. :-)

> Closes: https://bugs.gentoo.org/532264
> See-also: https://bugs.gentoo.org/274130#c5
> Signed-off-by: Ulrich Müller 
> ---
> v3: Exit awk commands on first match.

> v4: Added die statements after awk commands
> Fixed typo in mount-boot_is_disabled function documentation
> Reverted renaming of I_KNOW_WHAT_I_AM_DOING variable

Pushed (with one small spelling change in an error message).

Ulrich


signature.asc
Description: PGP signature


Re: [gentoo-dev] [PATCH v4] mount-boot.eclass: Check if /boot is sane, but don't try to mount it.

2019-12-09 Thread Jaco Kroon
Hi,

>> What about just checking "${EROOT}/boot" instead?
> For what, existence? There may well be a "boot" directory present under
> EROOT. (And we could check ${EROOT}/etc/fstab, but I don't think we
> should open that can of worms. There's no reliable way to guess the
> user's exact configuration for non-trivial ROOT.)

I was hoping there is a somewhat reliable way ... but if not I guess as
is will be the best we can do.

IMHO:  Ship it.

Even if there is issues still, the new state will be better than current.

Kind Regards,
Jaco




Re: [gentoo-dev] [PATCH v4] mount-boot.eclass: Check if /boot is sane, but don't try to mount it.

2019-12-09 Thread Ulrich Mueller
> On Mon, 09 Dec 2019, Jaco Kroon wrote:

> Acked-by: Jaco Kroon 

OK, I can add this.

>> +    if [[ ${EROOT:-/} != / ]] ; then
>>          return 0
>>      fi

> I don't use spaces in path names ... but what happens here if ROOT or
> EPREFIX (and by implication EROOT) contains a space?

No problem here, the test would still succeed and return 0. (But I
predict that EROOT containing spaces will cause a lot of trouble
elsewhere. :)

> What about just checking "${EROOT}/boot" instead?

For what, existence? There may well be a "boot" directory present under
EROOT. (And we could check ${EROOT}/etc/fstab, but I don't think we
should open that can of worms. There's no reliable way to guess the
user's exact configuration for non-trivial ROOT.)

Ulrich


signature.asc
Description: PGP signature


Re: [gentoo-dev] [PATCH v4] mount-boot.eclass: Check if /boot is sane, but don't try to mount it.

2019-12-09 Thread Jaco Kroon
Hi Ulrich,

I'm happy with this "as is", but there may be a few improvements still.

By the way:  This improves the situation for mounted ro /boot by moving
the check from preinst to pretend.

For noauto /boot (I believe the default and recommended) this fixes things.

This is the reason I decided to rather go with mounting /boot but as ro
instead of not mounting at all.

May I also suggest we start recommended read-only /boot instead of not
mounted at all in order to avoid similar issues from recurring?

Kind Regards,
Jaco

On 2019/12/07 11:10, Ulrich Müller wrote:

> The eclass failed to remount a read-only mounted /boot, because package
> collision sanity checks in recent Portage versions prevented it from
> reaching pkg_preinst() at all. Furthermore, with the "mount-sandbox"
> feature enabled, the mount won't be propagated past pkg_preinst() and
> installed files would end up under the (shadowed) mount point.
>
> Therefore don't even attempt to mount /boot ourselves, but error out
> if it isn't mounted read/write and ask the user to mount /boot.
>
> Also clean up and simplify. (For example, awk is a grown-up program
> which doesn't need any help from egrep or sed. :-)
>
> Closes: https://bugs.gentoo.org/532264
> See-also: https://bugs.gentoo.org/274130#c5
> Signed-off-by: Ulrich Müller 
Acked-by: Jaco Kroon 
>
> ---
> v3: Exit awk commands on first match.
>
> v4: Added die statements after awk commands
> Fixed typo in mount-boot_is_disabled function documentation
> Reverted renaming of I_KNOW_WHAT_I_AM_DOING variable
>
>  eclass/mount-boot.eclass | 144 +--
>  1 file changed, 48 insertions(+), 96 deletions(-)
>
> diff --git a/eclass/mount-boot.eclass b/eclass/mount-boot.eclass
> index 938df6732f43..ca27aca7efbd 100644
> --- a/eclass/mount-boot.eclass
> +++ b/eclass/mount-boot.eclass
> @@ -1,156 +1,108 @@
> -# Copyright 1999-2015 Gentoo Foundation
> +# Copyright 1999-2019 Gentoo Authors
>  # Distributed under the terms of the GNU General Public License v2
>  
>  # @ECLASS: mount-boot.eclass
>  # @MAINTAINER:
>  # base-sys...@gentoo.org
>  # @BLURB: functions for packages that install files into /boot
>  # @DESCRIPTION:
>  # This eclass is really only useful for bootloaders.
>  #
>  # If the live system has a separate /boot partition configured, then this
>  # function tries to ensure that it's mounted in rw mode, exiting with an
> -# error if it can't. It does nothing if /boot isn't a separate partition.
> +# error if it can't.  It does nothing if /boot isn't a separate
partition.
> +
> +case ${EAPI:-0} in
> +    4|5|6|7) ;;
> +    *) die "${ECLASS}: EAPI ${EAPI:-0} not supported" ;;
> +esac
>  
>  EXPORT_FUNCTIONS pkg_pretend pkg_preinst pkg_postinst pkg_prerm
pkg_postrm
>  
> -# @FUNCTION: mount-boot_disabled
> +# @FUNCTION: mount-boot_is_disabled
>  # @INTERNAL
>  # @DESCRIPTION:
>  # Detect whether the current environment/build settings are such that
we do not
>  # want to mess with any mounts.
>  mount-boot_is_disabled() {
> -    # Since this eclass only deals with /boot, skip things when ROOT
is active.
> -    if [[ "${ROOT:-/}" != "/" ]] ; then
> +    # Since this eclass only deals with /boot, skip things when EROOT
is active.
> +    if [[ ${EROOT:-/} != / ]] ; then
>          return 0
>      fi

I don't use spaces in path names ... but what happens here if ROOT or
EPREFIX (and by implication EROOT) contains a space?

What about just checking "${EROOT}/boot" instead?

Would that even be possible ... ?

>
>  
>      # If we're only building a package, then there's no need to check
things.
> -    if [[ "${MERGE_TYPE}" == "buildonly" ]] ; then
> +    if [[ ${MERGE_TYPE} == buildonly ]] ; then
>          return 0
>      fi
>  
>      # The user wants us to leave things be.
>      if [[ -n ${DONT_MOUNT_BOOT} ]] ; then
>          return 0
>      fi
>  
>      # OK, we want to handle things ourselves.
>      return 1
>  }
>  
>  # @FUNCTION: mount-boot_check_status
>  # @INTERNAL
>  # @DESCRIPTION:
> -# Figure out what kind of work we need to do in order to have /boot
be sane.
> -# Return values are:
> -# 0 - Do nothing at all!
> -# 1 - It's mounted, but is currently ro, so need to remount rw.
> -# 2 - It's not mounted, so need to mount it rw.
> +# Check if /boot is sane, i.e., mounted read/write if on a separate
> +# partition.  Die if conditions are not fulfilled.
>  mount-boot_check_status() {
>      # Get out fast if possible.
> -    mount-boot_is_disabled && return 0
> +    mount-boot_is_disabled && return
>  
>      # note that /dev/BOOT is in the Gentoo default /etc/fstab file
> -    local fstabstate=$(awk '!/^#|^[[:blank:]]+#|^\/dev\/BOOT/ {print
$2}' /etc/fstab | egrep "^/boot$" )
> -    local procstate=$(awk '$2 ~ /^\/boot$/ {print $2}' /proc/mounts)
> -    local proc_ro=$(awk '{ print $2 " ," $4 "," }' /proc/mounts | sed
-n '/^\/boot .*,ro,/p')
> -
> -    if [ -n "${fstabstate}" ] && [ -n "${procstate}" ] ; then
> -        if [ -n "${proc_ro}" ] ; th

Re: [gentoo-dev] [PATCH v4] mount-boot.eclass: Check if /boot is sane, but don't try to mount it.

2019-12-07 Thread Ulrich Mueller
> On Sat, 07 Dec 2019, Thomas Deutschmann wrote:

> just wondering if you have seen https://bugs.gentoo.org/532264#c24.
> If this is still valid, is your change really needed?

I have. IIUC it is a special case for one ebuild and cannot be
generalised.

Ulrich


signature.asc
Description: PGP signature


Re: [gentoo-dev] [PATCH v4] mount-boot.eclass: Check if /boot is sane, but don't try to mount it.

2019-12-07 Thread Thomas Deutschmann
Hi,

just wondering if you have seen https://bugs.gentoo.org/532264#c24. If
this is still valid, is your change really needed?


-- 
Regards,
Thomas Deutschmann / Gentoo Linux Developer
C4DD 695F A713 8F24 2AA1 5638 5849 7EE5 1D5D 74A5



signature.asc
Description: OpenPGP digital signature