[PATCH][BZ 16852] Do not clobber recvmmsg argument.

[PATCH][BZ 16852] Do not clobber recvmmsg argument.

[Ondřej Bílka]

Hi, here Rich asked about intended behaviour of recvmmsg timeout
argument. A kernel overwrites it by remaining time but we declare it
with const argument. So who is correct?

If we want to make timeout constant following patch should work.

Comments?

	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
	timeout argument.

diff --git a/sysdeps/unix/sysv/linux/recvmmsg.c b/sysdeps/unix/sysv/linux/recvmmsg.c
index 57ddf31..56690ff 100644
--- a/sysdeps/unix/sysv/linux/recvmmsg.c
+++ b/sysdeps/unix/sysv/linux/recvmmsg.c
@@ -37,6 +37,10 @@ int
 recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
 	  const struct timespec *tmo)
 {
+  struct timespec dummy;
+  memcpy (&dummy, tmo, sizeof (struct timespec));
+  tmo = &dummy;
+
   if (SINGLE_THREAD_P)
     return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);
 
@@ -61,6 +65,10 @@ int
 recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
 	  const struct timespec *tmo)
 {
+  struct timespec dummy;
+  memcpy (&dummy, tmo, sizeof (struct timespec));
+  tmo = &dummy;
+
   if (__glibc_likely (have_recvmmsg >= 0))
     {
       int ret = __internal_recvmmsg (fd, vmessages, vlen, flags, tmo);

Re: [PATCH][BZ 16852] Do not clobber recvmmsg argument.

[Andreas Schwab]

Ondřej Bílka <neleai@seznam.cz> writes:

> +  struct timespec dummy;
> +  memcpy (&dummy, tmo, sizeof (struct timespec));

What's wrong with an assignment?

>    if (SINGLE_THREAD_P)
>      return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);

You still pass a const where a non-const is expected.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

[PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Ondřej Bílka]

On Mon, Apr 28, 2014 at 05:40:56PM +0200, Andreas Schwab wrote:
> Ondřej Bílka <neleai@seznam.cz> writes:
> 
> > +  struct timespec dummy;
> > +  memcpy (&dummy, tmo, sizeof (struct timespec));
> 
> What's wrong with an assignment?
>
My first idea was memcpy, thanks. 
> >    if (SINGLE_THREAD_P)
> >      return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);
> 
> You still pass a const where a non-const is expected.
>
fixed. 
> Andreas.
> 

Here is v2.

	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
	timeout argument.


diff --git a/sysdeps/unix/sysv/linux/recvmmsg.c b/sysdeps/unix/sysv/linux/recvmmsg.c
index 57ddf31..04a065f 100644
--- a/sysdeps/unix/sysv/linux/recvmmsg.c
+++ b/sysdeps/unix/sysv/linux/recvmmsg.c
@@ -35,14 +35,16 @@
 #ifdef __NR_recvmmsg
 int
 recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
-	  const struct timespec *tmo)
+	  const struct timespec *_tmo)
 {
+  struct timespec tmo = *_tmo;
+
   if (SINGLE_THREAD_P)
-    return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);
+    return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, &tmo);
 
   int oldtype = LIBC_CANCEL_ASYNC ();
 
-  int result = INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);
+  int result = INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, &tmo);
 
   LIBC_CANCEL_RESET (oldtype);
 
@@ -52,18 +54,20 @@ recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
 # ifndef __ASSUME_RECVMMSG_SOCKETCALL
 extern int __internal_recvmmsg (int fd, struct mmsghdr *vmessages,
 				unsigned int vlen, int flags,
-				const struct timespec *tmo)
+				struct timespec *tmo)
      attribute_hidden;
 
 static int have_recvmmsg;
 
 int
 recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
-	  const struct timespec *tmo)
+	  const struct timespec *_tmo)
 {
+  struct timespec tmo = *_tmo;
+
   if (__glibc_likely (have_recvmmsg >= 0))
     {
-      int ret = __internal_recvmmsg (fd, vmessages, vlen, flags, tmo);
+      int ret = __internal_recvmmsg (fd, vmessages, vlen, flags, &tmo);
       /* The kernel returns -EINVAL for unknown socket operations.
 	 We need to convert that error to an ENOSYS error.  */
       if (__builtin_expect (ret < 0, 0)

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[David Miller]

From: Ondřej Bílka <neleai@seznam.cz>
Date: Mon, 28 Apr 2014 18:04:20 +0200

> 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> 	timeout argument.

It is extremely unfortunate if we've defined this argument as const,
now you are making it so that the user has no mechanism to get the
updated timeval other than to define their own syscall stubs.

This is doubly unfortunately since there is absolutely no reason for
us to have defined the interface different from what the kernel
actually provides.

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Rich Felker]

On Mon, Apr 28, 2014 at 06:04:20PM +0200, Ondřej Bílka wrote:
> diff --git a/sysdeps/unix/sysv/linux/recvmmsg.c b/sysdeps/unix/sysv/linux/recvmmsg.c
> index 57ddf31..04a065f 100644
> --- a/sysdeps/unix/sysv/linux/recvmmsg.c
> +++ b/sysdeps/unix/sysv/linux/recvmmsg.c
> @@ -35,14 +35,16 @@
>  #ifdef __NR_recvmmsg
>  int
>  recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> -	  const struct timespec *tmo)
> +	  const struct timespec *_tmo)
>  {
> +  struct timespec tmo = *_tmo;
> +

This is invalid because a null tmo argument is allowed (and is in fact
the normal usage case). You have to check for it.

Rich

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Michael Kerrisk (man-pages)]

On 04/28/2014 07:01 PM, David Miller wrote:
> From: Ondøej Bílka <neleai@seznam.cz>
> Date: Mon, 28 Apr 2014 18:04:20 +0200
> 
>> 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
>> 	timeout argument.
> 
> It is extremely unfortunate if we've defined this argument as const,
> now you are making it so that the user has no mechanism to get the
> updated timeval other than to define their own syscall stubs.
> 
> This is doubly unfortunately since there is absolutely no reason for
> us to have defined the interface different from what the kernel
> actually provides.

Agreed. Surely, the right fix here is simply to remove the "const" 
qualifier from the glibc API?

Cheers,

Michael

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Ondřej Bílka]

On Wed, Apr 30, 2014 at 10:28:53AM +0200, Michael Kerrisk (man-pages) wrote:
> On 04/28/2014 07:01 PM, David Miller wrote:
> > From: Ondřej Bílka <neleai@seznam.cz>
> > Date: Mon, 28 Apr 2014 18:04:20 +0200
> > 
> >> 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> >> 	timeout argument.
> > 
> > It is extremely unfortunate if we've defined this argument as const,
> > now you are making it so that the user has no mechanism to get the
> > updated timeval other than to define their own syscall stubs.
> > 
> > This is doubly unfortunately since there is absolutely no reason for
> > us to have defined the interface different from what the kernel
> > actually provides.
> 
> Agreed. Surely, the right fix here is simply to remove the "const" 
> qualifier from the glibc API?
> 
Now I am also for removing const. Initially I looked to documentation
and this was undocumented feature so I was not sure about it. Could we
also document this in manpages?

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Michael Kerrisk (man-pages)]

Hello Ondřej,

On 04/30/2014 11:07 AM, Ondřej Bílka wrote:
> On Wed, Apr 30, 2014 at 10:28:53AM +0200, Michael Kerrisk (man-pages) wrote:
>> On 04/28/2014 07:01 PM, David Miller wrote:
>>> From: Ondřej Bílka <neleai@seznam.cz>
>>> Date: Mon, 28 Apr 2014 18:04:20 +0200
>>>
>>>> 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
>>>> 	timeout argument.
>>>
>>> It is extremely unfortunate if we've defined this argument as const,
>>> now you are making it so that the user has no mechanism to get the
>>> updated timeval other than to define their own syscall stubs.
>>>
>>> This is doubly unfortunately since there is absolutely no reason for
>>> us to have defined the interface different from what the kernel
>>> actually provides.
>>
>> Agreed. Surely, the right fix here is simply to remove the "const" 
>> qualifier from the glibc API?
>>
> Now I am also for removing const. Initially I looked to documentation
> and this was undocumented feature so I was not sure about it. 

Yes, my mistake. 

> Could we
> also document this in manpages?

Yes, sure, I'll do that. Do you think we also need to document the
glibc 'const' issue under a BUGS section?

Cheers,

Michael


-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Michael Kerrisk (man-pages)]

On Wed, Apr 30, 2014 at 1:17 PM, Michael Kerrisk (man-pages)
<mtk.manpages@gmail.com> wrote:
> Hello Ondřej,
>
> On 04/30/2014 11:07 AM, Ondřej Bílka wrote:
>> On Wed, Apr 30, 2014 at 10:28:53AM +0200, Michael Kerrisk (man-pages) wrote:
>>> On 04/28/2014 07:01 PM, David Miller wrote:
>>>> From: Ondřej Bílka <neleai@seznam.cz>
>>>> Date: Mon, 28 Apr 2014 18:04:20 +0200
>>>>
>>>>>    * sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
>>>>>    timeout argument.
>>>>
>>>> It is extremely unfortunate if we've defined this argument as const,
>>>> now you are making it so that the user has no mechanism to get the
>>>> updated timeval other than to define their own syscall stubs.
>>>>
>>>> This is doubly unfortunately since there is absolutely no reason for
>>>> us to have defined the interface different from what the kernel
>>>> actually provides.
>>>
>>> Agreed. Surely, the right fix here is simply to remove the "const"
>>> qualifier from the glibc API?
>>>
>> Now I am also for removing const. Initially I looked to documentation
>> and this was undocumented feature so I was not sure about it.
>
> Yes, my mistake.

Sigh! Now I remember why that timeout behavior didn't get documented:

http://thread.gmane.org/gmane.linux.man/3477/focus=3500

The timeout argument is completely broken, AFAICT. I never did get a
reply from Arnaldo (but, as you can see in the thread, others agreed
tha there's a problem), and then I got distracted :-{.

Cheers,

Michael


-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Rich Felker]

On Wed, Apr 30, 2014 at 03:35:40PM +0200, Michael Kerrisk (man-pages) wrote:
> On Wed, Apr 30, 2014 at 1:17 PM, Michael Kerrisk (man-pages)
> <mtk.manpages@gmail.com> wrote:
> > Hello Ondřej,
> >
> > On 04/30/2014 11:07 AM, Ondřej Bílka wrote:
> >> On Wed, Apr 30, 2014 at 10:28:53AM +0200, Michael Kerrisk (man-pages) wrote:
> >>> On 04/28/2014 07:01 PM, David Miller wrote:
> >>>> From: Ondřej Bílka <neleai@seznam.cz>
> >>>> Date: Mon, 28 Apr 2014 18:04:20 +0200
> >>>>
> >>>>>    * sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> >>>>>    timeout argument.
> >>>>
> >>>> It is extremely unfortunate if we've defined this argument as const,
> >>>> now you are making it so that the user has no mechanism to get the
> >>>> updated timeval other than to define their own syscall stubs.
> >>>>
> >>>> This is doubly unfortunately since there is absolutely no reason for
> >>>> us to have defined the interface different from what the kernel
> >>>> actually provides.
> >>>
> >>> Agreed. Surely, the right fix here is simply to remove the "const"
> >>> qualifier from the glibc API?
> >>>
> >> Now I am also for removing const. Initially I looked to documentation
> >> and this was undocumented feature so I was not sure about it.
> >
> > Yes, my mistake.
> 
> Sigh! Now I remember why that timeout behavior didn't get documented:
> 
> http://thread.gmane.org/gmane.linux.man/3477/focus=3500
> 
> The timeout argument is completely broken, AFAICT. I never did get a
> reply from Arnaldo (but, as you can see in the thread, others agreed
> tha there's a problem), and then I got distracted :-{.

Indeed, this looks like a big problem and I don't see any viable
solution. I think the documented behavior, for now, should be that the
behavior is unspecified unless the timeout is a null pointer.

Rich

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Ondřej Bílka]

On Mon, Apr 28, 2014 at 01:01:22PM -0400, David Miller wrote:
> From: Ondřej Bílka <neleai@seznam.cz>
> Date: Mon, 28 Apr 2014 18:04:20 +0200
> 
> > 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> > 	timeout argument.
> 
> It is extremely unfortunate if we've defined this argument as const,
> now you are making it so that the user has no mechanism to get the
> updated timeval other than to define their own syscall stubs.
> 
> This is doubly unfortunately since there is absolutely no reason for
> us to have defined the interface different from what the kernel
> actually provides.

Ok, here is second alternative that drops constness.


	* socket/recvmmsg.c (recvmmsg): Drop const argument.
	* socket/sys/socket.h: Likewise
	* sysdeps/unix/sysv/linux/recvmmsg.c: Likewise.

---
 socket/recvmmsg.c                  | 2 +-
 socket/sys/socket.h                | 2 +-
 sysdeps/unix/sysv/linux/recvmmsg.c | 6 +++---
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/socket/recvmmsg.c b/socket/recvmmsg.c
index ed0c369..3daa501 100644
--- a/socket/recvmmsg.c
+++ b/socket/recvmmsg.c
@@ -23,7 +23,7 @@
    Returns the number of bytes read or -1 for errors.  */
 int
 recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
-	  const struct timespec *tmo)
+	  struct timespec *tmo)
 {
   __set_errno (ENOSYS);
   return -1;
diff --git a/socket/sys/socket.h b/socket/sys/socket.h
index 95ee26a..091b08c 100644
--- a/socket/sys/socket.h
+++ b/socket/sys/socket.h
@@ -209,7 +209,7 @@ extern ssize_t recvmsg (int __fd, struct msghdr *__message, int __flags);
    __THROW.  */
 extern int recvmmsg (int __fd, struct mmsghdr *__vmessages,
 		     unsigned int __vlen, int __flags,
-		     const struct timespec *__tmo);
+		     struct timespec *__tmo);
 #endif
 
 
diff --git a/sysdeps/unix/sysv/linux/recvmmsg.c b/sysdeps/unix/sysv/linux/recvmmsg.c
index 57ddf31..6c9ca44 100644
--- a/sysdeps/unix/sysv/linux/recvmmsg.c
+++ b/sysdeps/unix/sysv/linux/recvmmsg.c
@@ -35,7 +35,7 @@
 #ifdef __NR_recvmmsg
 int
 recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
-	  const struct timespec *tmo)
+	  struct timespec *tmo)
 {
   if (SINGLE_THREAD_P)
     return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);
@@ -52,14 +52,14 @@ recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
 # ifndef __ASSUME_RECVMMSG_SOCKETCALL
 extern int __internal_recvmmsg (int fd, struct mmsghdr *vmessages,
 				unsigned int vlen, int flags,
-				const struct timespec *tmo)
+				struct timespec *tmo)
      attribute_hidden;
 
 static int have_recvmmsg;
 
 int
 recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
-	  const struct timespec *tmo)
+	  struct timespec *tmo)
 {
   if (__glibc_likely (have_recvmmsg >= 0))
     {
-- 
1.8.4.rc3

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Michael Kerrisk (man-pages)]

On Wed, Apr 30, 2014 at 6:04 PM, Rich Felker <dalias@libc.org> wrote:
> On Wed, Apr 30, 2014 at 03:35:40PM +0200, Michael Kerrisk (man-pages) wrote:
>> On Wed, Apr 30, 2014 at 1:17 PM, Michael Kerrisk (man-pages)
>> <mtk.manpages@gmail.com> wrote:
>> > Hello Ondřej,
>> >
>> > On 04/30/2014 11:07 AM, Ondřej Bílka wrote:
>> >> On Wed, Apr 30, 2014 at 10:28:53AM +0200, Michael Kerrisk (man-pages) wrote:
>> >>> On 04/28/2014 07:01 PM, David Miller wrote:
>> >>>> From: Ondřej Bílka <neleai@seznam.cz>
>> >>>> Date: Mon, 28 Apr 2014 18:04:20 +0200
>> >>>>
>> >>>>>    * sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
>> >>>>>    timeout argument.
>> >>>>
>> >>>> It is extremely unfortunate if we've defined this argument as const,
>> >>>> now you are making it so that the user has no mechanism to get the
>> >>>> updated timeval other than to define their own syscall stubs.
>> >>>>
>> >>>> This is doubly unfortunately since there is absolutely no reason for
>> >>>> us to have defined the interface different from what the kernel
>> >>>> actually provides.
>> >>>
>> >>> Agreed. Surely, the right fix here is simply to remove the "const"
>> >>> qualifier from the glibc API?
>> >>>
>> >> Now I am also for removing const. Initially I looked to documentation
>> >> and this was undocumented feature so I was not sure about it.
>> >
>> > Yes, my mistake.
>>
>> Sigh! Now I remember why that timeout behavior didn't get documented:
>>
>> http://thread.gmane.org/gmane.linux.man/3477/focus=3500
>>
>> The timeout argument is completely broken, AFAICT. I never did get a
>> reply from Arnaldo (but, as you can see in the thread, others agreed
>> tha there's a problem), and then I got distracted :-{.
>
> Indeed, this looks like a big problem and I don't see any viable
> solution. I think the documented behavior, for now, should be that the
> behavior is unspecified unless the timeout is a null pointer.

I got the formulation of the program slightly wrong, but the
fundamental problem is the same:

http://thread.gmane.org/gmane.linux.man/5677/focus=5702

Cheers,

Michael



-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/

[PING][PATCH v2][BZ 16852] Fix recvmmsg prototype

[Ondřej Bílka]

ping
On Sun, May 04, 2014 at 05:34:24PM +0200, Ondřej Bílka wrote:
> On Mon, Apr 28, 2014 at 01:01:22PM -0400, David Miller wrote:
> > From: Ondřej Bílka <neleai@seznam.cz>
> > Date: Mon, 28 Apr 2014 18:04:20 +0200
> > 
> > > 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> > > 	timeout argument.
> > 
> > It is extremely unfortunate if we've defined this argument as const,
> > now you are making it so that the user has no mechanism to get the
> > updated timeval other than to define their own syscall stubs.
> > 
> > This is doubly unfortunately since there is absolutely no reason for
> > us to have defined the interface different from what the kernel
> > actually provides.
> 
> Ok, here is second alternative that drops constness.
> 
> 
> 	* socket/recvmmsg.c (recvmmsg): Drop const argument.
> 	* socket/sys/socket.h: Likewise
> 	* sysdeps/unix/sysv/linux/recvmmsg.c: Likewise.
> 
> ---
>  socket/recvmmsg.c                  | 2 +-
>  socket/sys/socket.h                | 2 +-
>  sysdeps/unix/sysv/linux/recvmmsg.c | 6 +++---
>  3 files changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/socket/recvmmsg.c b/socket/recvmmsg.c
> index ed0c369..3daa501 100644
> --- a/socket/recvmmsg.c
> +++ b/socket/recvmmsg.c
> @@ -23,7 +23,7 @@
>     Returns the number of bytes read or -1 for errors.  */
>  int
>  recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> -	  const struct timespec *tmo)
> +	  struct timespec *tmo)
>  {
>    __set_errno (ENOSYS);
>    return -1;
> diff --git a/socket/sys/socket.h b/socket/sys/socket.h
> index 95ee26a..091b08c 100644
> --- a/socket/sys/socket.h
> +++ b/socket/sys/socket.h
> @@ -209,7 +209,7 @@ extern ssize_t recvmsg (int __fd, struct msghdr *__message, int __flags);
>     __THROW.  */
>  extern int recvmmsg (int __fd, struct mmsghdr *__vmessages,
>  		     unsigned int __vlen, int __flags,
> -		     const struct timespec *__tmo);
> +		     struct timespec *__tmo);
>  #endif
>  
>  
> diff --git a/sysdeps/unix/sysv/linux/recvmmsg.c b/sysdeps/unix/sysv/linux/recvmmsg.c
> index 57ddf31..6c9ca44 100644
> --- a/sysdeps/unix/sysv/linux/recvmmsg.c
> +++ b/sysdeps/unix/sysv/linux/recvmmsg.c
> @@ -35,7 +35,7 @@
>  #ifdef __NR_recvmmsg
>  int
>  recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> -	  const struct timespec *tmo)
> +	  struct timespec *tmo)
>  {
>    if (SINGLE_THREAD_P)
>      return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);
> @@ -52,14 +52,14 @@ recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
>  # ifndef __ASSUME_RECVMMSG_SOCKETCALL
>  extern int __internal_recvmmsg (int fd, struct mmsghdr *vmessages,
>  				unsigned int vlen, int flags,
> -				const struct timespec *tmo)
> +				struct timespec *tmo)
>       attribute_hidden;
>  
>  static int have_recvmmsg;
>  
>  int
>  recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> -	  const struct timespec *tmo)
> +	  struct timespec *tmo)
>  {
>    if (__glibc_likely (have_recvmmsg >= 0))
>      {
> -- 
> 1.8.4.rc3

-- 

parallel processors running perpendicular today

Re: [PING][PATCH v2][BZ 16852] Fix recvmmsg prototype

[Ondřej Bílka]

ping
On Fri, May 23, 2014 at 02:02:02PM +0200, Ondřej Bílka wrote:
> ping
> On Sun, May 04, 2014 at 05:34:24PM +0200, Ondřej Bílka wrote:
> > On Mon, Apr 28, 2014 at 01:01:22PM -0400, David Miller wrote:
> > > From: Ondřej Bílka <neleai@seznam.cz>
> > > Date: Mon, 28 Apr 2014 18:04:20 +0200
> > > 
> > > > 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> > > > 	timeout argument.
> > > 
> > > It is extremely unfortunate if we've defined this argument as const,
> > > now you are making it so that the user has no mechanism to get the
> > > updated timeval other than to define their own syscall stubs.
> > > 
> > > This is doubly unfortunately since there is absolutely no reason for
> > > us to have defined the interface different from what the kernel
> > > actually provides.
> > 
> > Ok, here is second alternative that drops constness.
> > 
> > 
> > 	* socket/recvmmsg.c (recvmmsg): Drop const argument.
> > 	* socket/sys/socket.h: Likewise
> > 	* sysdeps/unix/sysv/linux/recvmmsg.c: Likewise.
> > 
> > ---
> >  socket/recvmmsg.c                  | 2 +-
> >  socket/sys/socket.h                | 2 +-
> >  sysdeps/unix/sysv/linux/recvmmsg.c | 6 +++---
> >  3 files changed, 5 insertions(+), 5 deletions(-)
> > 
> > diff --git a/socket/recvmmsg.c b/socket/recvmmsg.c
> > index ed0c369..3daa501 100644
> > --- a/socket/recvmmsg.c
> > +++ b/socket/recvmmsg.c
> > @@ -23,7 +23,7 @@
> >     Returns the number of bytes read or -1 for errors.  */
> >  int
> >  recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> > -	  const struct timespec *tmo)
> > +	  struct timespec *tmo)
> >  {
> >    __set_errno (ENOSYS);
> >    return -1;
> > diff --git a/socket/sys/socket.h b/socket/sys/socket.h
> > index 95ee26a..091b08c 100644
> > --- a/socket/sys/socket.h
> > +++ b/socket/sys/socket.h
> > @@ -209,7 +209,7 @@ extern ssize_t recvmsg (int __fd, struct msghdr *__message, int __flags);
> >     __THROW.  */
> >  extern int recvmmsg (int __fd, struct mmsghdr *__vmessages,
> >  		     unsigned int __vlen, int __flags,
> > -		     const struct timespec *__tmo);
> > +		     struct timespec *__tmo);
> >  #endif
> >  
> >  
> > diff --git a/sysdeps/unix/sysv/linux/recvmmsg.c b/sysdeps/unix/sysv/linux/recvmmsg.c
> > index 57ddf31..6c9ca44 100644
> > --- a/sysdeps/unix/sysv/linux/recvmmsg.c
> > +++ b/sysdeps/unix/sysv/linux/recvmmsg.c
> > @@ -35,7 +35,7 @@
> >  #ifdef __NR_recvmmsg
> >  int
> >  recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> > -	  const struct timespec *tmo)
> > +	  struct timespec *tmo)
> >  {
> >    if (SINGLE_THREAD_P)
> >      return INLINE_SYSCALL (recvmmsg, 5, fd, vmessages, vlen, flags, tmo);
> > @@ -52,14 +52,14 @@ recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> >  # ifndef __ASSUME_RECVMMSG_SOCKETCALL
> >  extern int __internal_recvmmsg (int fd, struct mmsghdr *vmessages,
> >  				unsigned int vlen, int flags,
> > -				const struct timespec *tmo)
> > +				struct timespec *tmo)
> >       attribute_hidden;
> >  
> >  static int have_recvmmsg;
> >  
> >  int
> >  recvmmsg (int fd, struct mmsghdr *vmessages, unsigned int vlen, int flags,
> > -	  const struct timespec *tmo)
> > +	  struct timespec *tmo)
> >  {
> >    if (__glibc_likely (have_recvmmsg >= 0))
> >      {
> > -- 
> > 1.8.4.rc3
> 
> -- 
> 
> parallel processors running perpendicular today

-- 

PCMCIA slave driver

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Mike Frysinger]

[-- Attachment #1: Type: text/plain, Size: 828 bytes --]

On Sun 04 May 2014 17:34:24 Ondřej Bílka wrote:
> On Mon, Apr 28, 2014 at 01:01:22PM -0400, David Miller wrote:
> > From: Ondřej Bílka <neleai@seznam.cz>
> > Date: Mon, 28 Apr 2014 18:04:20 +0200
> > 
> > > 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> > > 	timeout argument.
> > 
> > It is extremely unfortunate if we've defined this argument as const,
> > now you are making it so that the user has no mechanism to get the
> > updated timeval other than to define their own syscall stubs.
> > 
> > This is doubly unfortunately since there is absolutely no reason for
> > us to have defined the interface different from what the kernel
> > actually provides.
> 
> Ok, here is second alternative that drops constness.

i think everyone agreed this was the best way to go for now
-mike

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Ondřej Bílka]

On Sat, Aug 02, 2014 at 10:16:40AM -0400, Mike Frysinger wrote:
> On Sun 04 May 2014 17:34:24 Ondřej Bílka wrote:
> > On Mon, Apr 28, 2014 at 01:01:22PM -0400, David Miller wrote:
> > > From: Ondřej Bílka <neleai@seznam.cz>
> > > Date: Mon, 28 Apr 2014 18:04:20 +0200
> > > 
> > > > 	* sysdeps/unix/sysv/linux/recvmmsg.c (recvmmsg): Do not clobber
> > > > 	timeout argument.
> > > 
> > > It is extremely unfortunate if we've defined this argument as const,
> > > now you are making it so that the user has no mechanism to get the
> > > updated timeval other than to define their own syscall stubs.
> > > 
> > > This is doubly unfortunately since there is absolutely no reason for
> > > us to have defined the interface different from what the kernel
> > > actually provides.
> > 
> > Ok, here is second alternative that drops constness.
> 
> i think everyone agreed this was the best way to go for now
> -mike

As I decided to relax on holiday this waited a while. I pushed this now.

Re: [PATCH v2][BZ 16852] Do not clobber recvmmsg argument.

[Andreas Schwab]

Ondřej Bílka <neleai@seznam.cz> writes:

> 	* socket/recvmmsg.c (recvmmsg): Drop const argument.
                                                  ^from
> 	* socket/sys/socket.h: Likewise
> 	* sysdeps/unix/sysv/linux/recvmmsg.c: Likewise.

Ok.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."