[PATCH] qio: fix command spawn RDONLY/WRONLY

marcandre.lureau@redhat.com posted 1 patch 3 years, 5 months ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20220901101120.2741109-1-marcandre.lureau@redhat.com
Maintainers: "Daniel P. Berrangé" <berrange@redhat.com>
io/channel-command.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
[PATCH] qio: fix command spawn RDONLY/WRONLY
Posted by marcandre.lureau@redhat.com 3 years, 5 months ago
From: Marc-André Lureau <marcandre.lureau@redhat.com>

The in/out handling is inverted, although nothing seemed to notice that yet.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 io/channel-command.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/io/channel-command.c b/io/channel-command.c
index 9f2f4a1793..ed17b44f74 100644
--- a/io/channel-command.c
+++ b/io/channel-command.c
@@ -79,10 +79,10 @@ qio_channel_command_new_spawn(const char *const argv[],
     flags = flags & O_ACCMODE;
 
     if (flags == O_RDONLY) {
-        stdinnull = true;
+        stdoutnull = true;
     }
     if (flags == O_WRONLY) {
-        stdoutnull = true;
+        stdinnull = true;
     }
 
     if (stdinnull || stdoutnull) {
-- 
2.37.2


Re: [PATCH] qio: fix command spawn RDONLY/WRONLY
Posted by Daniel P. Berrangé 3 years, 5 months ago
On Thu, Sep 01, 2022 at 02:11:20PM +0400, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> The in/out handling is inverted, although nothing seemed to notice that yet.

On the contrary, it is correct, and the unit tests validate this.

> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> ---
>  io/channel-command.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/io/channel-command.c b/io/channel-command.c
> index 9f2f4a1793..ed17b44f74 100644
> --- a/io/channel-command.c
> +++ b/io/channel-command.c
> @@ -79,10 +79,10 @@ qio_channel_command_new_spawn(const char *const argv[],
>      flags = flags & O_ACCMODE;
>  
>      if (flags == O_RDONLY) {
> -        stdinnull = true;
> +        stdoutnull = true;
>      }
>      if (flags == O_WRONLY) {
> -        stdoutnull = true;
> +        stdinnull = true;
>      }

This change breaks the unit tests.

The confusion is because there are two parties involves. The 'flags'
variable is from the POV of the parent process, while stdinnull/stdoutnull
are from the POV of the child process.

IOW, if the parent process is reading from the child (O_RDONLY),
then the child needs a stdout to write to the parent, but not
any stdin to read from the parent, hence we set stdin to /dev/null
in the child.

With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|


Re: [PATCH] qio: fix command spawn RDONLY/WRONLY
Posted by Marc-André Lureau 3 years, 5 months ago
Hi

On Thu, Sep 1, 2022 at 2:32 PM Daniel P. Berrangé <berrange@redhat.com>
wrote:

> On Thu, Sep 01, 2022 at 02:11:20PM +0400, marcandre.lureau@redhat.com
> wrote:
> > From: Marc-André Lureau <marcandre.lureau@redhat.com>
> >
> > The in/out handling is inverted, although nothing seemed to notice that
> yet.
>
> On the contrary, it is correct, and the unit tests validate this.
>

> > Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > ---
> >  io/channel-command.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/io/channel-command.c b/io/channel-command.c
> > index 9f2f4a1793..ed17b44f74 100644
> > --- a/io/channel-command.c
> > +++ b/io/channel-command.c
> > @@ -79,10 +79,10 @@ qio_channel_command_new_spawn(const char *const
> argv[],
> >      flags = flags & O_ACCMODE;
> >
> >      if (flags == O_RDONLY) {
> > -        stdinnull = true;
> > +        stdoutnull = true;
> >      }
> >      if (flags == O_WRONLY) {
> > -        stdoutnull = true;
> > +        stdinnull = true;
> >      }
>
> This change breaks the unit tests.
>
>
Does it really test it then? we are talking about test-io-channel-command ?
It works before and after for me. Other tests as well.

The confusion is because there are two parties involves. The 'flags'
> variable is from the POV of the parent process, while stdinnull/stdoutnull
> are from the POV of the child process.
>
> IOW, if the parent process is reading from the child (O_RDONLY),
> then the child needs a stdout to write to the parent, but not
> any stdin to read from the parent, hence we set stdin to /dev/null
> in the child.
>

Ok, thanks for the clarification!


-- 
Marc-André Lureau
Re: [PATCH] qio: fix command spawn RDONLY/WRONLY
Posted by Daniel P. Berrangé 3 years, 5 months ago
On Thu, Sep 01, 2022 at 02:47:17PM +0400, Marc-André Lureau wrote:
> Hi
> 
> On Thu, Sep 1, 2022 at 2:32 PM Daniel P. Berrangé <berrange@redhat.com>
> wrote:
> 
> > On Thu, Sep 01, 2022 at 02:11:20PM +0400, marcandre.lureau@redhat.com
> > wrote:
> > > From: Marc-André Lureau <marcandre.lureau@redhat.com>
> > >
> > > The in/out handling is inverted, although nothing seemed to notice that
> > yet.
> >
> > On the contrary, it is correct, and the unit tests validate this.
> >
> 
> > > Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
> > > ---
> > >  io/channel-command.c | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/io/channel-command.c b/io/channel-command.c
> > > index 9f2f4a1793..ed17b44f74 100644
> > > --- a/io/channel-command.c
> > > +++ b/io/channel-command.c
> > > @@ -79,10 +79,10 @@ qio_channel_command_new_spawn(const char *const
> > argv[],
> > >      flags = flags & O_ACCMODE;
> > >
> > >      if (flags == O_RDONLY) {
> > > -        stdinnull = true;
> > > +        stdoutnull = true;
> > >      }
> > >      if (flags == O_WRONLY) {
> > > -        stdoutnull = true;
> > > +        stdinnull = true;
> > >      }
> >
> > This change breaks the unit tests.
> >
> >
> Does it really test it then? we are talking about test-io-channel-command ?
> It works before and after for me. Other tests as well.

Yes, it certainly fails for me:

$ ./build/tests/unit/test-io-channel-command -p /io/channel/command/fifo/sync
# random seed: R02S5134a426f643aeefdbe73921046b66da
# Start of io tests
# Start of channel tests
# Start of command tests
# Start of fifo tests
**
ERROR:../tests/unit/io-channel-helpers.c:152:qio_channel_test_validate: assertion failed: (test->readerr == NULL)
Bail out! ERROR:../tests/unit/io-channel-helpers.c:152:qio_channel_test_validate: assertion failed: (test->readerr == NULL)
Aborted (core dumped)

With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|