This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH v4] Make bindresvport() function to multithread-safe


On Fri, Sep 28, 2012 at 5:15 AM, Peng Haitao <penght@cn.fujitsu.com> wrote:
> bindresvport() uses a static variable port which is not protected.
> It is not safe when in multithread circumstance.
>
> bindresvport() select a port number from the range 512 to 1023, when in
> multithread circumstance, the port may be 1024. So the static variable
> will be protected.

Roland, Andreas,

Do you have any comments on this?

Looks like performance goes *up* when we switch to a __thread variable
and use gettid, but I've asked Peng to justify this.

http://sourceware.org/ml/libc-alpha/2012-09/msg00744.html


> Signed-off-by: Peng Haitao <penght@cn.fujitsu.com>
> ---
>  ChangeLog           |  6 ++++++
>  sunrpc/bindrsvprt.c |  8 +++++---
>  sunrpc/bindrsvprt.h | 37 +++++++++++++++++++++++++++++++++++++
>  3 files changed, 48 insertions(+), 3 deletions(-)
>  create mode 100644 sunrpc/bindrsvprt.h
>
> diff --git a/ChangeLog b/ChangeLog
> index 3004989..8db0157 100644
> --- a/ChangeLog
> +++ b/ChangeLog
> @@ -1,3 +1,9 @@
> +2012-09-27  Peng Haitao  <penght@cn.fujitsu.com>
> +
> +       [BZ #13763]
> +       * sunrpc/bindrsvprt.c: Add __thread when define static's port,
> +       getpid() is changed to gettid() in Linux.
> +
>  2012-09-27  David S. Miller  <davem@davemloft.net>
>
>         [BZ #14376]
> diff --git a/sunrpc/bindrsvprt.c b/sunrpc/bindrsvprt.c
> index d493c9f..ab835a1 100644
> --- a/sunrpc/bindrsvprt.c
> +++ b/sunrpc/bindrsvprt.c
> @@ -1,5 +1,5 @@
>  /*
> - * Copyright (c) 2010, Oracle America, Inc.
> + * Copyright (c) 2010-2012, Oracle America, Inc.

Is this correct? I don't know that we have a ruling on changing Copyright
headers for copyright that the FSF doesn't own.

Have we established a rule here?

>   *
>   * Redistribution and use in source and binary forms, with or without
>   * modification, are permitted provided that the following conditions are
> @@ -36,13 +36,15 @@
>  #include <sys/socket.h>
>  #include <netinet/in.h>
>
> +#include "bindrsvprt.h"

Can you make this #include <bindrsvprt.h> and allow the normal sysdep
scanning to pick it up?

By using <> you allow other machines to override this if required.

> +
>  /*
>   * Bind a socket to a privileged IP port
>   */
>  int
>  bindresvport (int sd, struct sockaddr_in *sin)
>  {
> -  static short port;
> +  static __thread short port;

I'm curious how our relocation count changes with this?

Net neutral or did we gain extra relocations by changing this from
static to static __thread?

Could you have a look into that?

>    struct sockaddr_in myaddr;
>    int i;
>
> @@ -66,7 +68,7 @@ bindresvport (int sd, struct sockaddr_in *sin)
>
>    if (port == 0)
>      {
> -      port = (__getpid () % NPORTS) + STARTPORT;
> +      port = (get_port () % NPORTS) + STARTPORT;
>      }
>
>    /* Initialize to make gcc happy.  */
> diff --git a/sunrpc/bindrsvprt.h b/sunrpc/bindrsvprt.h
> new file mode 100644
> index 0000000..221f811
> --- /dev/null
> +++ b/sunrpc/bindrsvprt.h
> @@ -0,0 +1,37 @@
> +/* Copyright (C) 2012 Free Software Foundation, Inc.
> +   This file is part of the GNU C Library.
> +
> +   The GNU C Library is free software; you can redistribute it and/or
> +   modify it under the terms of the GNU Lesser General Public
> +   License as published by the Free Software Foundation; either
> +   version 2.1 of the License, or (at your option) any later version.
> +
> +   The GNU C Library is distributed in the hope that it will be useful,
> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> +   Lesser General Public License for more details.
> +
> +   You should have received a copy of the GNU Lesser General Public
> +   License along with the GNU C Library; if not, see
> +   <http://www.gnu.org/licenses/>.  */
> +
> +#ifdef __linux__
> +#ifdef INTERNAL_SYSCALL
> +#define get_port()                            \

Call this __get_uid(), since that's what you're doing.

You are getting a unique id for the process or thread
and using that in the hash function to find a good port
to start the bind attempt.

> + ({                                           \
> +  INTERNAL_SYSCALL_DECL (err);                \
> +  short __port;                               \
> +  __port = INTERNAL_SYSCALL (gettid, err, 0); \
> +  __port;                                     \
> + })
> +#else
> +#define get_port()                            \
> + ({                                           \
> +  short __port;                               \
> +  __port = INLINE_SYSCALL (gettid, 0);        \
> +  __port;                                     \
> + })
> +#endif /* INTERNAL_SYSCALL */
> +#else
> +#define get_port() __getpid()
> +#endif /* __linux__ */
> --
> 1.7.11.4
>

Cheers,
Carlos.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]