qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Stefan Hajnoczi <stefanha@gmail.com>
To: Sam Bobroff <sam.bobroff@au1.ibm.com>
Cc: qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH 1/1] Protect stderr from non-blocking mode
Date: Thu, 28 Jan 2016 12:11:18 +0000	[thread overview]
Message-ID: <20160128121118.GI7206@stefanha-x1.localdomain> (raw)
In-Reply-To: <0b88a3432f2fcfee1e57abfa9420ca6c5c386b49.1453954698.git.sam.bobroff@au1.ibm.com>

[-- Attachment #1: Type: text/plain, Size: 2324 bytes --]

On Thu, Jan 28, 2016 at 03:18:23PM +1100, Sam Bobroff wrote:
> On Linux, if QEMU is run from a shell with -d to enable debug logging
> but without directing it to a file (e.g. -D is not used, and
> qemu_logfile is set to stderr), and no shell redirection is used, it
> is possible for log messages to be lost under load.
> 
> This is caused by a combination of several factors:
> 
> * The shell (e.g. bash) may provide stdin, stdout and stderr as
>   duplicates of a single open file, so they share file status flags
>   including O_NONBLOCK.
> 
> * As character devices are registered (see qemu_chr_open_stdio()),
>   stdin and stdout are set non-blocking.
> 
> * The printf() family of functions, including fprintf(), are not
>   "non-blocking" aware: if they receive EAGAIN after partially
>   writing their output, they immediately return EAGAIN with no way to
>   discover how much output was written, if any.
> 
> So O_NONBLOCK is set on stdin/stdout, which causes O_NONBLOCK to be
> set on stderr, and qemu_logfile is set to stderr. Then, under load,
> fprintf()s to qemu_logfile return EAGAIN, which is ignored and data is
> lost. This can't be fixed by handling EAGAIN because an unknown amount
> of data has been written.
> 
> This patch works around the issue by re-opening the underlying tty
> file, which is available in Linux as /proc/self/fd/2, and duplicating
> it into fd 2 which causes stderr to refer to a new, unshared, "file
> description". Additionally, if we end up logging to a non-blocking
> file descriptor on a non-Linux platform, where no workaround has been
> implemented, we display a warning message.
> 
> Signed-off-by: Sam Bobroff <sam.bobroff@au1.ibm.com>
> ---
> This is somewhat of an RFC as the issue probably hasn't been noticed
> very often (ever?) and there are several approaches to addressing it.
> The discussion was a while ago, here:
> 
> http://lists.nongnu.org/archive/html/qemu-devel/2015-12/msg00761.html
> 
>  include/sysemu/os-posix.h |  2 +-
>  include/sysemu/os-win32.h |  2 +-
>  os-posix.c                | 30 +++++++++++++++++++++++++++++-
>  os-win32.c                |  2 +-
>  vl.c                      |  2 +-
>  5 files changed, 33 insertions(+), 5 deletions(-)

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

      reply	other threads:[~2016-01-28 12:11 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-01-28  4:18 [Qemu-devel] [PATCH 1/1] Protect stderr from non-blocking mode Sam Bobroff
2016-01-28 12:11 ` Stefan Hajnoczi [this message]

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=20160128121118.GI7206@stefanha-x1.localdomain \
    --to=stefanha@gmail.com \
    --cc=qemu-devel@nongnu.org \
    --cc=sam.bobroff@au1.ibm.com \
    /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).