On 9/5/22 20:33, Marc-André Lureau wrote:
Hi

On Mon, Sep 5, 2022 at 4:28 PM Guoyi Tu <t...@chinatelecom.cn <mailto:t...@chinatelecom.cn>> wrote:



    On 9/5/22 19:19, Marc-André Lureau wrote:
     > Hi
     >
     > On Tue, Aug 23, 2022 at 12:00 PM <t...@chinatelecom.cn
    <mailto:t...@chinatelecom.cn>
     > <mailto:t...@chinatelecom.cn <mailto:t...@chinatelecom.cn>>> wrote:
     >
     >     From: Guoyi Tu <t...@chinatelecom.cn
    <mailto:t...@chinatelecom.cn> <mailto:t...@chinatelecom.cn
    <mailto:t...@chinatelecom.cn>>>
     >
     >     Introduce qemu_socketpair() to create socket pair fd, and
     >     set the close-on-exec flag at default as with the other type
     >     of socket does.
     >
     >     besides, the live update feature is developing, so it's necessary
     >     to do that.
     >
     >     Guoyi Tu (2):
     >        oslib-posix: Introduce qemu_socketpair()
     >        vhost-user: Call qemu_socketpair() instead of socketpair()
     >
     >
     > Looks like a good idea to me. We will eventually extend the
    support for
     > win32 (as discussed in "[PATCH 19/51] tests/qtest: Build
     > test-filter-{mirror, redirector} cases for posix only").
     >
     > There are other places where you can replace existing
    socketpair() calls
     > in the code base. Why not do it?
     >
    yeah, Thanks for reminding me, Maybe i can replace all the existing
    socketpair() calls in a separate patch if this patch can be merged to
    upstream.


It needs to be reviewed though, all may not want to set CLOEXEC, but most should. Else, we should probably consider switching to GSpawn which does dup or unset CLOEXEC for explicitely passed fds.

Got it. Special care will be taken when replace those functions.

Thanks again.

--
Guoyi
    --
    Guoyi

     > Current patches lgtm
     > Reviewed-by: Marc-André Lureau <marcandre.lur...@redhat.com
    <mailto:marcandre.lur...@redhat.com>
     > <mailto:marcandre.lur...@redhat.com
    <mailto:marcandre.lur...@redhat.com>>>
     >
     > --
     > Marc-André Lureau



--
Marc-André Lureau

Reply via email to