qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Ani Sinha <anisinha@redhat.com>
To: John Snow <jsnow@redhat.com>
Cc: qemu-devel <qemu-devel@nongnu.org>,
	"Cédric Le Goater" <clg@kaod.org>,
	"Hanna Reitz" <hreitz@redhat.com>,
	"Cleber Rosa" <crosa@redhat.com>,
	"Wainer dos Santos Moschetta" <wainersm@redhat.com>,
	qemu-arm <qemu-arm@nongnu.org>,
	Qemu-block <qemu-block@nongnu.org>,
	"Andrew Jeffery" <andrew@aj.id.au>,
	"Joel Stanley" <joel@jms.id.au>, "Kevin Wolf" <kwolf@redhat.com>,
	"Philippe Mathieu-Daudé" <philmd@linaro.org>,
	"Beraldo Leal" <bleal@redhat.com>,
	"Daniel Berrange" <berrange@redhat.com>,
	"Peter Maydell" <peter.maydell@linaro.org>,
	"Michael S. Tsirkin" <mst@redhat.com>
Subject: Re: [PATCH v2 4/6] python/machine: use socketpair() for console connections
Date: Thu, 27 Jul 2023 12:19:11 +0530	[thread overview]
Message-ID: <C1FA63AE-D031-445C-A872-745DBBC505AD@redhat.com> (raw)
In-Reply-To: <4EE2965D-F72B-4002-99BF-732F5F9734C6@redhat.com>



> On 27-Jul-2023, at 11:22 AM, Ani Sinha <anisinha@redhat.com> wrote:
> 
> 
> 
>> On 26-Jul-2023, at 10:51 PM, John Snow <jsnow@redhat.com> wrote:
>> 
>> 
>> 
>> On Wed, Jul 26, 2023, 6:50 AM Ani Sinha <anisinha@redhat.com> wrote:
>> 
>> 
>>> On 25-Jul-2023, at 11:33 PM, John Snow <jsnow@redhat.com> wrote:
>>> 
>>> Create a socketpair for the console output. This should help eliminate
>>> race conditions around console text early in the boot process that might
>>> otherwise have been dropped on the floor before being able to connect to
>>> QEMU under "server,nowait".
>>> 
>>> Signed-off-by: John Snow <jsnow@redhat.com>
>> 
>> Thanks for doing this. I recall we spoke about this late last year in the context of fixing my bios-bits avocado test and adding a console output there.
>> 
>> Yep! I think you need a few more changes to do what you wanted. IIRC, you also want to be able to drain the console log while waiting for the vm to terminate of its own accord, which I don't support yet.
>> 
>> (If you use console socket's self draining mode, it should be possible to forego the early termination of the console socket and allow this behavior. Maybe I can work that in now...)
> 
> yeah we want to collect all the console logs while the VM is running until it self terminates. Maybe you can add a flag for this behavior to not early terminate the socket. I think we need to add mathods to keep reading the socket and write to a file until the socket is closed. Maybe QemuMachine needs to be enhanced. 

I see something like

      console_drainer = datadrainer.LineLogger(self.vm.console_socket.fileno(),
                                                 logger=self.log.getChild('console'))
        console_drainer.start()

in LinuxTest. Maybe I would be able to use something similar.

> 
>> 
>> Anything else I'm forgetting ...?
>> 
>> Except the concern below,
>> 
>> Reviewed-by: Ani Sinha <anisinha@redhat.com>
>> 
>> Thanks 😊 
>> 
>> 
>> 
>>> ---
>>> python/qemu/machine/machine.py | 30 +++++++++++++++++++++++++++---
>>> 1 file changed, 27 insertions(+), 3 deletions(-)
>>> 
>>> diff --git a/python/qemu/machine/machine.py b/python/qemu/machine/machine.py
>>> index 26f0fb8a81..09f214c95c 100644
>>> --- a/python/qemu/machine/machine.py
>>> +++ b/python/qemu/machine/machine.py
>>> @@ -159,6 +159,8 @@ def __init__(self,
>>> 
>>>        self._name = name or f"{id(self):x}"
>>>        self._sock_pair: Optional[Tuple[socket.socket, socket.socket]] = None
>>> +        self._cons_sock_pair: Optional[
>>> +            Tuple[socket.socket, socket.socket]] = None
>>>        self._temp_dir: Optional[str] = None
>>>        self._base_temp_dir = base_temp_dir
>>>        self._sock_dir = sock_dir
>>> @@ -315,8 +317,9 @@ def _base_args(self) -> List[str]:
>>>        for _ in range(self._console_index):
>>>            args.extend(['-serial', 'null'])
>>>        if self._console_set:
>>> -            chardev = ('socket,id=console,path=%s,server=on,wait=off' %
>>> -                       self._console_address)
>>> +            assert self._cons_sock_pair is not None
>>> +            fd = self._cons_sock_pair[0].fileno()
>>> +            chardev = f"socket,id=console,fd={fd}"
>>>            args.extend(['-chardev', chardev])
>>>            if self._console_device_type is None:
>>>                args.extend(['-serial', 'chardev:console'])
>>> @@ -351,6 +354,10 @@ def _pre_launch(self) -> None:
>>>                nickname=self._name
>>>            )
>>> 
>>> +        if self._console_set:
>>> +            self._cons_sock_pair = socket.socketpair()
>>> +            os.set_inheritable(self._cons_sock_pair[0].fileno(), True)
>>> +
>>>        # NOTE: Make sure any opened resources are *definitely* freed in
>>>        # _post_shutdown()!
>>>        # pylint: disable=consider-using-with
>>> @@ -368,6 +375,9 @@ def _pre_launch(self) -> None:
>>>    def _post_launch(self) -> None:
>>>        if self._sock_pair:
>>>            self._sock_pair[0].close()
>>> +        if self._cons_sock_pair:
>>> +            self._cons_sock_pair[0].close()
>>> +
>>>        if self._qmp_connection:
>>>            if self._sock_pair:
>>>                self._qmp.connect()
>>> @@ -518,6 +528,11 @@ def _early_cleanup(self) -> None:
>>>            self._console_socket.close()
>>>            self._console_socket = None
>>> 
>>> +        if self._cons_sock_pair:
>>> +            self._cons_sock_pair[0].close()
>>> +            self._cons_sock_pair[1].close()
>>> +            self._cons_sock_pair = None
>>> +
>>>    def _hard_shutdown(self) -> None:
>>>        """
>>>        Perform early cleanup, kill the VM, and wait for it to terminate.
>>> @@ -878,10 +893,19 @@ def console_socket(self) -> socket.socket:
>>>        Returns a socket connected to the console
>>>        """
>>>        if self._console_socket is None:
>>> +            if not self._console_set:
>>> +                raise QEMUMachineError(
>>> +                    "Attempt to access console socket with no connection")
>>> +            assert self._cons_sock_pair is not None
>>> +            # os.dup() is used here for sock_fd because otherwise we'd
>>> +            # have two rich python socket objects that would each try to
>>> +            # close the same underlying fd when either one gets garbage
>>> +            # collected.
>>>            self._console_socket = console_socket.ConsoleSocket(
>>> -                self._console_address,
>>> +                sock_fd=os.dup(self._cons_sock_pair[1].fileno()),
>>>                file=self._console_log_path,
>>>                drain=self._drain_console)
>>> +            self._cons_sock_pair[1].close()
>> 
>> I am not 100% sure but should we save the new sock_fd here? Like
>> self._cons_sock_pair[1] = sock_fd ;
>> 
>> Then next time console_socket() is invoked, the correct fd will be duped.
>> 
>> It should be cached to self._console_socket, so it should return the same instance every time, no second call to os.dup().
> 
> yeah I missed your patch 3. All good now.
> 
>> 
>> self._console_socket takes ownership of the duplicated fd and we retain ownership of _cons_sock_pair[1] which we then close right after.
>> 
>> All three sockets are closed and None'd if applicable during _early_cleanup().
>> 
>> 
>>>        return self._console_socket
>>> 
>>>    @property
>>> -- 
>>> 2.41.0



  reply	other threads:[~2023-07-27  7:05 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-07-25 18:03 [PATCH v2 0/6] python/machine: use socketpair() for console socket John Snow
2023-07-25 18:03 ` [PATCH v2 1/6] python/machine: move socket setup out of _base_args property John Snow
2023-07-25 18:16   ` Daniel P. Berrangé
2023-07-26  7:10   ` Ani Sinha
2023-07-25 18:03 ` [PATCH v2 2/6] python/machine: close sock_pair in cleanup path John Snow
2023-07-25 18:18   ` Daniel P. Berrangé
2023-07-26  7:23     ` Ani Sinha
2023-07-25 18:03 ` [PATCH v2 3/6] python/console_socket: accept existing FD in initializer John Snow
2023-07-25 18:19   ` Daniel P. Berrangé
2023-07-27  5:41   ` Ani Sinha
2023-07-25 18:03 ` [PATCH v2 4/6] python/machine: use socketpair() for console connections John Snow
2023-07-25 18:21   ` Daniel P. Berrangé
2023-07-26 10:50   ` Ani Sinha
2023-07-26 17:21     ` John Snow
2023-07-27  5:52       ` Ani Sinha
2023-07-27  6:49         ` Ani Sinha [this message]
2023-07-27 10:41         ` Daniel P. Berrangé
2023-07-25 18:03 ` [PATCH v2 5/6] python/machine: use socketpair() for qtest connection John Snow
2023-07-25 18:24   ` Daniel P. Berrangé
2023-07-25 18:03 ` [PATCH v2 6/6] python/machine: remove unused sock_dir argument John Snow
2023-07-25 18:26   ` Daniel P. Berrangé
2023-07-25 18:33     ` John Snow
2023-07-25 18:36       ` Daniel P. Berrangé
2023-07-27 13:05 ` [PATCH v2 0/6] python/machine: use socketpair() for console socket Peter Maydell

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=C1FA63AE-D031-445C-A872-745DBBC505AD@redhat.com \
    --to=anisinha@redhat.com \
    --cc=andrew@aj.id.au \
    --cc=berrange@redhat.com \
    --cc=bleal@redhat.com \
    --cc=clg@kaod.org \
    --cc=crosa@redhat.com \
    --cc=hreitz@redhat.com \
    --cc=joel@jms.id.au \
    --cc=jsnow@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=mst@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=philmd@linaro.org \
    --cc=qemu-arm@nongnu.org \
    --cc=qemu-block@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    --cc=wainersm@redhat.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).