qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Michael Roth <mdroth@linux.vnet.ibm.com>
To: "Denis V. Lunev" <den@openvz.org>
Cc: Olga Krishtal <okrishtal@virtuozzo.com>,
	Stefan Weil <sw@weilnetz.de>,
	qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH 2/2] qga: set file descriptors in qmp_guest_file_open non-blocking on Win32
Date: Wed, 21 Oct 2015 18:29:49 -0500	[thread overview]
Message-ID: <20151021232949.10420.46582@loki> (raw)
In-Reply-To: <1444913471-15132-3-git-send-email-den@openvz.org>

Quoting Denis V. Lunev (2015-10-15 07:51:11)
> From: Olga Krishtal <okrishtal@virtuozzo.com>
> 
> Set fd non-blocking to avoid common use cases (like reading from a
> named pipe) from hanging the agent. This was missed in the original
> code.
> 
> Restore compatibility with Posix implementation.
> 
> The patch adds Win32 specific implementation of qemu_set_nonblock.
> 
> On Windows OS there is a separate API for changing flags of file, pipes
> and sockets. Portable way to change file descriptor flags requires
> to detect file descriptor type and proper actions depending of that
> type. The patch adds wrapper qemu_set_fd_nonblocking into Windows specific
> code to handle this stuff properly.
> 
> The only problem is that qemu_set_nonblock is void but this should not
> be a big deal.
> 
> Despite the fact that qemu_fd_register() is used only once here and now
> it is difficult to drop this function as it uses AioContext internals
> which we do not want to involve here.
> 
> Signed-off-by: Olga Krishtal <okrishtal@virtuozzo.com>
> Signed-off-by: Denis V. Lunev <den@openvz.org>
> Reviewed-by: Yuri Pudgorodskiy <yur@virtuozzo.com>
> CC: Michael Roth <mdroth@linux.vnet.ibm.com>
> CC: Stefan Weil <sw@weilnetz.de>
> ---
>  qga/commands-win32.c |  6 ++++++
>  util/oslib-win32.c   | 57 ++++++++++++++++++++++++++++++++++++++++++++++------
>  2 files changed, 57 insertions(+), 6 deletions(-)
> 
> diff --git a/qga/commands-win32.c b/qga/commands-win32.c
> index 41bdd3f..745bddf 100644
> --- a/qga/commands-win32.c
> +++ b/qga/commands-win32.c
> @@ -34,6 +34,7 @@
>  #include "qapi/qmp/qerror.h"
>  #include "qemu/queue.h"
>  #include "qemu/host-utils.h"
> +#include "qemu/sockets.h"
> 
>  #ifndef SHTDN_REASON_FLAG_PLANNED
>  #define SHTDN_REASON_FLAG_PLANNED 0x80000000
> @@ -156,6 +157,11 @@ int64_t qmp_guest_file_open(const char *path, bool has_mode,
>          return -1;
>      }
> 
> +    /* set fd non-blocking to avoid common use cases (like reading from a
> +     * named pipe) from hanging the agent
> +     */
> +    qemu_set_nonblock(fileno(fh));

This doesn't seem to build for me:

  CC    qga/channel-win32.o
  /home/mdroth/w/qemu3.git/qga/commands-win32.c: In function
  ‘qmp_guest_file_open’:
  /home/mdroth/w/qemu3.git/qga/commands-win32.c:165: warning:
  dereferencing ‘void *’ pointer
  /home/mdroth/w/qemu3.git/qga/commands-win32.c:165: error: request for
  member ‘_file’ in something not a structure or union

That's with ubuntu 14.04 and running:

  ../qemu3.git/configure --cross-prefix=i586-mingw32msvc-
  --target-list=x86_64-softmmu --extra-cflags=-Wall && make -j4

fileno() expects a FILE *, but fh is actually a HANDLE. Is there an
environment where this is not the case? I think what you want is
_open_osfhandle().

Note that it looks like you then might need to track the resulting fd and
use _close() in place of CloseHandle() in guest-file-close:

  https://msdn.microsoft.com/en-us/library/bdts1c9x.aspx

That also make me a bit concerned that read/write operations on the
HANDLE would then need to be replaced with read()/write() etc... the
documentation doesn't seem very clear on this.

I'm starting to think it best we just introduce a
qemu_set_nonblock_w32_handle(HANDLE ...) to do this, and leave the
existing qemu_set_nonblock() as it was before (winsock FDs only).

If people want to risk whatever wierd stuff might come about by
switching between HANDLE/fd they can use
_open_osfhandle()/_get_osfhandle() prior to calling
qemu_set_nonblock*(). But since we started off with a HANDLE in the
first place, there's no reason to us to switch to fd, then back to
HANDLE.

> +
>      fd = guest_file_handle_add(fh, errp);
>      if (fd < 0) {
>          CloseHandle(&fh);
> diff --git a/util/oslib-win32.c b/util/oslib-win32.c
> index 08f5a9c..e1580c8 100644
> --- a/util/oslib-win32.c
> +++ b/util/oslib-win32.c
> @@ -121,18 +121,63 @@ struct tm *localtime_r(const time_t *timep, struct tm *result)
>  }
>  #endif /* CONFIG_LOCALTIME_R */
> 
> -void qemu_set_block(int fd)
> +static void qemu_set_fd_nonblocking(int fd, bool nonblocking)
>  {
> -    unsigned long opt = 0;
> -    WSAEventSelect(fd, NULL, 0);
> +    HANDLE handle;
> +    DWORD file_type, pipe_state;
> +    unsigned long opt = (unsigned long)nonblocking;
> +
> +    handle = (HANDLE)_get_osfhandle(fd);
> +    if (handle == INVALID_HANDLE_VALUE) {
> +        return;
> +    }
> +
> +    file_type = GetFileType(handle);
> +    if (file_type != FILE_TYPE_PIPE) {
> +        return;
> +    }
> +
> +    /* If file_type == FILE_TYPE_PIPE, according to msdn
> +     * the specified file is socket or named pipe */
> +    if (GetNamedPipeHandleState(handle, &pipe_state, NULL,
> +                                NULL, NULL, NULL, 0)) {
> +        /* The fd is named pipe fd */
> +        if (!nonblocking == !(pipe_state & PIPE_NOWAIT)) {
> +            /* In this case we do not need perform any operation, because
> +             * nonblocking = true and PIPE_NOWAIT is already set or
> +             * nonblocking = false and PIPE_NOWAIT is not set */
> +            return;
> +        }
> +
> +        if (nonblocking) {
> +            pipe_state |= PIPE_NOWAIT;
> +        } else {
> +            pipe_state &= ~PIPE_NOWAIT;
> +        }
> +
> +        SetNamedPipeHandleState(handle, &pipe_state, NULL, NULL);
> +        return;
> +    }
> +
> +    /* The fd is socket fd */
> +    if (!nonblocking) {
> +        WSAEventSelect(fd, NULL, 0);
> +    }
>      ioctlsocket(fd, FIONBIO, &opt);
> +    if (nonblocking) {
> +        qemu_fd_register(fd);
> +    }
> +    return;
> +}
> +
> +void qemu_set_block(int fd)
> +{
> +    qemu_set_fd_nonblocking(fd, false);
>  }
> 
>  void qemu_set_nonblock(int fd)
>  {
> -    unsigned long opt = 1;
> -    ioctlsocket(fd, FIONBIO, &opt);
> -    qemu_fd_register(fd);
> +    qemu_set_fd_nonblocking(fd, true);
>  }
> 
>  int socket_set_fast_reuse(int fd)
> -- 
> 2.1.4
> 

  reply	other threads:[~2015-10-21 23:30 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-10-15 12:51 [Qemu-devel] [PATCH v2 0/2] qga: non-blocking fd cleanups Denis V. Lunev
2015-10-15 12:51 ` [Qemu-devel] [PATCH 1/2] qga: drop hand-made guest_file_toggle_flags helper Denis V. Lunev
2015-10-15 12:51 ` [Qemu-devel] [PATCH 2/2] qga: set file descriptors in qmp_guest_file_open non-blocking on Win32 Denis V. Lunev
2015-10-21 23:29   ` Michael Roth [this message]
  -- strict thread matches above, loose matches on Subject: below --
2015-10-07 10:59 [Qemu-devel] [PATCH 0/2] qga: non-blocking fd cleanups Denis V. Lunev
2015-10-07 10:59 ` [Qemu-devel] [PATCH 2/2] qga: set file descriptors in qmp_guest_file_open non-blocking on Win32 Denis V. Lunev
2015-10-12 17:38   ` Michael Roth
2015-10-12 17:47     ` Denis V. Lunev
2015-10-12 18:39     ` Stefan Weil
2015-10-12 19:14       ` Denis V. Lunev

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20151021232949.10420.46582@loki \
    --to=mdroth@linux.vnet.ibm.com \
    --cc=den@openvz.org \
    --cc=okrishtal@virtuozzo.com \
    --cc=qemu-devel@nongnu.org \
    --cc=sw@weilnetz.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).