[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Gaël PORTAY]

Pseudo consists of:
* a client (cli),
* a server (daemon) and
* a database managed by this daemon
  which makes a relation between inode, filename, type/mode...

Without -f/-d argument, the client forks a process, sets LD_PRELOAD,
execve's the command given as argument and waits for its completion.
The LD_PRELOAD'ed library tries to connect to the daemon and spawns it
if it isn't running.

Neither the client itself nor the LD_PRELOAD'ed library waits for the
daemon to terminate. As a consequence, the daemon may still be alive
after the completion of the command executed through the client.

Under some circumstances (running build in a docker), the daemon (which
may be still alive after the build) is killed with SIGKILL, thus it
does not let it the time to properly sync its database.

The database is rendered inconsistent for the next build. In this case,
makedevs will complain about the file type/mode, see error below.

makedevs: line xx: node (...)/output/target/dev/console exists but is of wrong type

Note that this error was introduced by c85cd189.

Because the database did not have the time to sync (ie. /dev/console
record is missing), makedevs tests the existing and real /dev/console
file mode (created by the previous build) which is a regular file to
what should have been reported by pseudo, i.e. a character device.
Those two modes are different and makedevs exits with error.

To solve this issue, this patch make the wrapper control the life of
the daemon. It spawns the server before running the script inside the
pseudo context, tells it to stop with "pseudo -S", and waits for the
server to exit before the wrapper exits.

Signed-off-by: Ga?l PORTAY <gael.portay@savoirfairelinux.com>
Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
---
 package/pseudo/pseudo-wrapper | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/package/pseudo/pseudo-wrapper b/package/pseudo/pseudo-wrapper
index 9c8dbdb..194c88f 100644
--- a/package/pseudo/pseudo-wrapper
+++ b/package/pseudo/pseudo-wrapper
@@ -8,7 +8,6 @@ if [ "${0##*/}" = "fakeroot" ]; then
 fi
 
 export PSEUDO_PREFIX="$(dirname "${0%/*}")"
-export PSEUDO_OPTS="-t0"
 if [ -n "${TARGET_DIR}" ]; then
     export PSEUDO_PASSWD="${TARGET_DIR}"
 fi
@@ -16,4 +15,18 @@ if [ -n "${BASE_DIR}" ]; then
     export PSEUDO_LOCALSTATEDIR="${BASE_DIR}/build/.pseudodb"
 fi
 
-exec "${0%/*}/pseudo" "${@}"
+# spawn server manually...
+"${0%/*}/pseudo" -f &
+pid="${!}"
+
+# ... run script/command...
+"${0%/*}/pseudo" "${@}"
+rc="${?}"
+
+# ... terminate server...
+"${0%/*}/pseudo" -S
+
+# ... and wait for its completion to make sure database is synced before
+# returning.
+wait "${pid}"
+exit "${rc}"
-- 
2.10.2

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Yann E. MORIN]

Ga?l, All,

On 2016-11-23 12:36 -0500, Ga?l PORTAY spake thusly:
> Pseudo consists of:
> * a client (cli),
> * a server (daemon) and
> * a database managed by this daemon
>   which makes a relation between inode, filename, type/mode...
> 
> Without -f/-d argument, the client forks a process, sets LD_PRELOAD,
> execve's the command given as argument and waits for its completion.
> The LD_PRELOAD'ed library tries to connect to the daemon and spawns it
> if it isn't running.
> 
> Neither the client itself nor the LD_PRELOAD'ed library waits for the
> daemon to terminate. As a consequence, the daemon may still be alive
> after the completion of the command executed through the client.
> 
> Under some circumstances (running build in a docker), the daemon (which
> may be still alive after the build) is killed with SIGKILL, thus it
> does not let it the time to properly sync its database.
> 
> The database is rendered inconsistent for the next build. In this case,
> makedevs will complain about the file type/mode, see error below.
> 
> makedevs: line xx: node (...)/output/target/dev/console exists but is of wrong type
> 
> Note that this error was introduced by c85cd189.
> 
> Because the database did not have the time to sync (ie. /dev/console
> record is missing), makedevs tests the existing and real /dev/console
> file mode (created by the previous build) which is a regular file to
> what should have been reported by pseudo, i.e. a character device.
> Those two modes are different and makedevs exits with error.
> 
> To solve this issue, this patch make the wrapper control the life of
> the daemon. It spawns the server before running the script inside the
> pseudo context, tells it to stop with "pseudo -S", and waits for the
> server to exit before the wrapper exits.
> 
> Signed-off-by: Ga?l PORTAY <gael.portay@savoirfairelinux.com>
> Reviewed-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
> ---
>  package/pseudo/pseudo-wrapper | 17 +++++++++++++++--
>  1 file changed, 15 insertions(+), 2 deletions(-)
> 
> diff --git a/package/pseudo/pseudo-wrapper b/package/pseudo/pseudo-wrapper
> index 9c8dbdb..194c88f 100644
> --- a/package/pseudo/pseudo-wrapper
> +++ b/package/pseudo/pseudo-wrapper
> @@ -8,7 +8,6 @@ if [ "${0##*/}" = "fakeroot" ]; then
>  fi
>  
>  export PSEUDO_PREFIX="$(dirname "${0%/*}")"
> -export PSEUDO_OPTS="-t0"
>  if [ -n "${TARGET_DIR}" ]; then
>      export PSEUDO_PASSWD="${TARGET_DIR}"
>  fi
> @@ -16,4 +15,18 @@ if [ -n "${BASE_DIR}" ]; then
>      export PSEUDO_LOCALSTATEDIR="${BASE_DIR}/build/.pseudodb"
>  fi
>  
> -exec "${0%/*}/pseudo" "${@}"
> +# spawn server manually...
> +"${0%/*}/pseudo" -f &
> +pid="${!}"

There still is a little corner case I'm not too happy about... The above
returns immediately, but we don;t know if the daemon is ready. It may
take a bit of time for it to start compeltely (i.e. reload the DB and
repair its now-internal state)...

> +# ... run script/command...
> +"${0%/*}/pseudo" "${@}"

But then we (almost) immediately start a client, which may try to
connect to the daemon while it is not yet ready, and thus the client may
not detect it yet, and thus wmay spawn his own daemon.

And we'd be back to square one...

Of course, we could run:

    pseudo -d

which is guaranteed to return after the daemon is ready...

> +rc="${?}"
> +
> +# ... terminate server...
> +"${0%/*}/pseudo" -S
> +
> +# ... and wait for its completion to make sure database is synced before
> +# returning.
> +wait "${pid}"

But then we would not have a PID to wait on... :-/

So, the question is: are we sure the client will talk to the daemon that
we spawn?
If the answer is "yes, we're sure", then all is good; id the answer is
"no, we're not sure", then it's a problem...

Thoughts?

Now, my position on all this mess is :

  - we revert back to using fakeroot for 2016.11;

  - long term, we want to definitely switch to using pseudo:
    - it is actively maintained,,
    - it has more features than fakeroot,
    - some of those features are very interesting for us.

Regards,
Yann E. MORIN.

> +exit "${rc}"
> -- 
> 2.10.2
> 

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Thomas Petazzoni]

Hello,

On Wed, 23 Nov 2016 19:28:43 +0100, Yann E. MORIN wrote:

>   - long term, we want to definitely switch to using pseudo:
>     - it is actively maintained,,
>     - it has more features than fakeroot,
>     - some of those features are very interesting for us.

Could you explain which additional features of pseudo are interesting
for us?

Thanks,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Yann E. MORIN]

Thomas, All,

On 2016-11-23 21:57 +0100, Thomas Petazzoni spake thusly:
> On Wed, 23 Nov 2016 19:28:43 +0100, Yann E. MORIN wrote:
> >   - long term, we want to definitely switch to using pseudo:
> >     - it is actively maintained,,
> >     - it has more features than fakeroot,
> >     - some of those features are very interesting for us.
> 
> Could you explain which additional features of pseudo are interesting
> for us?

pseudo can intercept and redirect acceses to /etc/paswd (and group,
shadow, gshadow) to the ones in the target. With that, we could probably
greatly simplify our handling of users.

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Thomas Petazzoni]

Hello,

On Wed, 23 Nov 2016 22:34:03 +0100, Yann E. MORIN wrote:

> On 2016-11-23 21:57 +0100, Thomas Petazzoni spake thusly:
> > On Wed, 23 Nov 2016 19:28:43 +0100, Yann E. MORIN wrote:  
> > >   - long term, we want to definitely switch to using pseudo:
> > >     - it is actively maintained,,
> > >     - it has more features than fakeroot,
> > >     - some of those features are very interesting for us.  
> > 
> > Could you explain which additional features of pseudo are interesting
> > for us?  
> 
> pseudo can intercept and redirect acceses to /etc/paswd (and group,
> shadow, gshadow) to the ones in the target. With that, we could probably
> greatly simplify our handling of users.

Is that the "additional features that are very interesting for us"? If
it is, then I don't think it's that interesting. At least it
definitely doesn't offset the drawback of having to build host-sqlite
and host-attr.

Our handling of users is not complicated, it's very explicit, works
fine, and is actually probably better than some black magic that
intercepts accesses to /etc/passwd.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Arnout Vandecappelle]

On 23-11-16 19:28, Yann E. MORIN wrote:
> Ga?l, All,
> 
> On 2016-11-23 12:36 -0500, Ga?l PORTAY spake thusly:
[snip]
>> -exec "${0%/*}/pseudo" "${@}"
>> +# spawn server manually...
>> +"${0%/*}/pseudo" -f &
>> +pid="${!}"
> 
> There still is a little corner case I'm not too happy about... The above
> returns immediately, but we don;t know if the daemon is ready. It may
> take a bit of time for it to start compeltely (i.e. reload the DB and
> repair its now-internal state)...
> 
>> +# ... run script/command...
>> +"${0%/*}/pseudo" "${@}"
> 
> But then we (almost) immediately start a client, which may try to
> connect to the daemon while it is not yet ready, and thus the client may
> not detect it yet, and thus wmay spawn his own daemon.
> 
> And we'd be back to square one...
> 
> Of course, we could run:
> 
>     pseudo -d
> 
> which is guaranteed to return after the daemon is ready...
> 
>> +rc="${?}"
>> +
>> +# ... terminate server...
>> +"${0%/*}/pseudo" -S
>> +
>> +# ... and wait for its completion to make sure database is synced before
>> +# returning.
>> +wait "${pid}"
> 
> But then we would not have a PID to wait on... :-/

 PID is not a problem, we can get it from LOCALSTATEDIR. The problem is that the
wait builtin only works for direct children of the shell.

> 
> So, the question is: are we sure the client will talk to the daemon that
> we spawn?
> If the answer is "yes, we're sure", then all is good; id the answer is
> "no, we're not sure", then it's a problem...

 I don't think we're really sure, but I do believe that pseudo protects against
this situation, i.e. when two servers start up concurrently, only one of them
will really run. The problem is that the one that commits suicide may be the one
that we start from the pseudo wrapper, so we end up wait'ing for the wrong pid.

> 
> Thoughts?
> 
> Now, my position on all this mess is :
> 
>   - we revert back to using fakeroot for 2016.11;

 Yeah, I think we need to do that. Too bad for the Fedora users.

 Regards,
 Arnout

> 
>   - long term, we want to definitely switch to using pseudo:
>     - it is actively maintained,,
>     - it has more features than fakeroot,
>     - some of those features are very interesting for us.
> 
> Regards,
> Yann E. MORIN.
> 
>> +exit "${rc}"
>> -- 
>> 2.10.2
>>
> 

-- 
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Gaël PORTAY]

Hi all,

This patch *should* be abandonned.

The issue has been fixed and mainlined by the Seebs (the maintainer of pseudo),
see commit d6eb2df3fe63b765f35d62332add4d0e4e9c6a39.

Now we can update the wrapper script to run:
exec "${0%/*}/pseudo" -S "${@}"

The -S argument tells server to shutdown and "waits" for the server to complete.

Seebs has picked up the solution proposed by Arnout. It uses the socket to
determine when the server has terminated. The socket is closed right before
exiting.

On Wed, Nov 23, 2016 at 10:47:26PM +0100, Arnout Vandecappelle wrote:
> 
> > 
> > So, the question is: are we sure the client will talk to the daemon that
> > we spawn?
> > If the answer is "yes, we're sure", then all is good; id the answer is
> > "no, we're not sure", then it's a problem...
> 
>  I don't think we're really sure, but I do believe that pseudo protects against
> this situation, i.e. when two servers start up concurrently, only one of them
> will really run. The problem is that the one that commits suicide may be the one
> that we start from the pseudo wrapper, so we end up wait'ing for the wrong pid.

There is a lock, thus only a single instance of a server could run.

> > 
> > Thoughts?
> > 
> > Now, my position on all this mess is :
> > 
> >   - we revert back to using fakeroot for 2016.11;
> 
>  Yeah, I think we need to do that. Too bad for the Fedora users.

Agree. Even if this issue can be easily fixed, it remains the issue reported by
Jerome.

Regards,
Gael

[Buildroot] [PATCH v3] pseudo-wrapper: fix console issue

[Arnout Vandecappelle]

On 24-11-16 00:59, Ga?l PORTAY wrote:
> This patch *should* be abandonned.

 In such a situation, you should mark the patch as "Changes Requested" in
patchwork. You can do that yourself for your own patches.

 I've now done that.

 Regards,
 Arnout
-- 
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF