Hello Marco,
2011/11/6 Marco d'Itri <m...@linux.it>:
> On Nov 06, Émeric Maschino <emeric.masch...@gmail.com> wrote:
>
>> Starting with udev 170 (well, IIRC!), console is flooded at startup with:
>> udevd[XXX]: unable to receive ctrl connection: Function not implemented
> Your lacks support for SOCK_CLOEXEC on accept, why?
Well,
emeric@longspeak:~$ rgrep SOCK_CLOEXEC /usr/src/linux-source-2.6.38
most notably gives:
/usr/src/linux-source-2.6.38/include/linux/net.h:#define
SOCK_CLOEXEC O_CLOEXEC
Looking at net.h, I can see that SOCK_CLOEXEC is an unconditional
definition. So, no problem there.
What about O_CLOEXEC then?
emeric@longspeak:~$ rgrep O_CLOEXEC /usr/src/linux-source-2.6.38
most notably gives:
/usr/src/linux-source-2.6.38/include/asm-generic/fcntl.h:#define
O_CLOEXEC 02000000 /* set close_on_exec */
/usr/src/linux-source-2.6.38/arch/alpha/include/asm/fcntl.h:#define
O_CLOEXEC 010000000 /* set close_on_exec */
/usr/src/linux-source-2.6.38/arch/mips/include/asm/socket.h:#define
SOCK_CLOEXECO_CLOEXEC
/usr/src/linux-source-2.6.38/arch/parisc/include/asm/fcntl.h:#define
O_CLOEXEC 010000000 /* set close_on_exec */
/usr/src/linux-source-2.6.38/arch/sparc/include/asm/fcntl.h:#define
O_CLOEXEC 0x400000
Here again, O_CLOEXEC is unconditionally defined to 02000000 in
fcntl.h and redefined for alpha, mips, parisc and sparc architectures
only.
It seems that SOCK_CLOEXEC is thus correctly defined, don't you think so?
> It is supposed to be available since 2.6.27.
This is with latest "installable on ia64" Wheezy kernel 2.6.38.
Is there a way to check for "SOCK_CLOEXEC symbol" in the currently
running kernel? This should give a definitive answer regardless
SOCK_CLOEXEC support.
> Does the old_cloexec patch included up to release 164-4 fix this for you?
Yes, udev 164-4 is fine (no error message, /dev correctly populated).
Émeric
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
