From: Amos Kong <akong@redhat.com>
To: Orit Wasserman <owasserm@redhat.com>
Cc: qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH v3 3/3] Fix address handling in inet_nonblocking_connect
Date: Mon, 24 Sep 2012 17:48:35 +0800 [thread overview]
Message-ID: <50602C73.2080503@redhat.com> (raw)
In-Reply-To: <505EAD8A.90701@redhat.com>
On 23/09/12 14:34, Orit Wasserman wrote:
> On 09/20/2012 06:16 PM, Amos Kong wrote:
>> ----- Original Message -----
>>> On 09/19/2012 11:31 AM, Amos Kong wrote:
>>>> On 14/09/12 02:58, Orit Wasserman wrote:
>>>>> getaddrinfo can give us a list of addresses, but we only try to
>>>>> connect to the first one. If that fails we never proceed to
>>>>> the next one. This is common on desktop setups that often have
>>>>> ipv6
>>>>> configured but not actually working.
>>>>>
>>>>> To fix this make inet_connect_nonblocking retry connection with a
>>>>> different
>>>>> address.
>>>>> callers on inet_nonblocking_connect register a callback function
>>>>> that will
>>>>> be called when connect opertion completes, in case of failure the
>>>>> fd will have
>>>>> a negative value
>>>>
>>>>
>>>> Hi Orit,
>>>>
>>>> We almost get to the destination :)
>>>>
>>>>> Signed-off-by: Orit Wasserman <owasserm@redhat.com>
>>>>> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
>>>>> ---
>>>>> migration-tcp.c | 29 +++++-----------
>>>>> qemu-char.c | 2 +-
>>>>> qemu-sockets.c | 95
>>>>> +++++++++++++++++++++++++++++++++++++++++++++++-------
>>>>> qemu_socket.h | 13 ++++++--
>>>>> 4 files changed, 102 insertions(+), 37 deletions(-)
>>>>>
>>>>
>>>> ...
>>>>
>>>>> diff --git a/qemu-sockets.c b/qemu-sockets.c
>>>>> index 212075d..d321c58 100644
>>>>> --- a/qemu-sockets.c
>>>>> +++ b/qemu-sockets.c
>>>>> @@ -24,6 +24,7 @@
>>>>>
>>>>> #include "qemu_socket.h"
>>>>> #include "qemu-common.h" /* for qemu_isdigit */
>>>>> +#include "main-loop.h"
>>>>>
>>>>> #ifndef AI_ADDRCONFIG
>>>>> # define AI_ADDRCONFIG 0
>>>>> @@ -217,11 +218,69 @@ listen:
>>>>> ((rc) == -EINPROGRESS)
>>>>> #endif
>>>>>
>>>>> +/* Struct to store connect state for non blocking connect */
>>>>> +typedef struct ConnectState {
>>>>> + int fd;
>>>>> + struct addrinfo *addr_list;
>>>>> + struct addrinfo *current_addr;
>>>>> + ConnectHandler *callback;
>>>>> + void *opaque;
>>>>
>>>>> + Error *errp;
>>>>
>>>> This member is not used.
>>> I will remove it.
>>>>
>>>>> +} ConnectState;
>>>>
>>>> # make (gcc-4.4.6-4.el6.x86_64)
>>>> CC qemu-sockets.o
>>>> qemu-sockets.c:229: error: redefinition of typedef ‘ConnectState’
>>>> qemu_socket.h:46: note: previous declaration of ‘ConnectState’ was
>>>> here
>>>> make: *** [qemu-sockets.o] Error 1
>>>>
>>>>
>>>> struct ConnectState {
>>>> int fd;
>>>> struct addrinfo *addr_list;
>>>> struct addrinfo *current_addr;
>>>> ConnectHandler *callback;
>>>> void *opaque;
>>>> };
>>>>
>>>>
>>>>> +
>>>>> static int inet_connect_addr(struct addrinfo *addr, bool block,
>>>>> - bool *in_progress)
>>>>> + bool *in_progress, ConnectState
>>>>> *connect_state);
>>>>> +
>>>>> +static void wait_for_connect(void *opaque)
>>>>> +{
>>>>> + ConnectState *s = opaque;
>>>>> + int val = 0, rc = 0;
>>>>> + socklen_t valsize = sizeof(val);
>>>>> + bool in_progress = false;
>>>>> +
>>>>> + qemu_set_fd_handler2(s->fd, NULL, NULL, NULL, NULL);
>>>>> +
>>>>> + do {
>>>>> + rc = getsockopt(s->fd, SOL_SOCKET, SO_ERROR, (void *)
>>>>> &val, &valsize);
>>>>> + } while (rc == -1 && socket_error() == EINTR);
>>>>
>>>> # man getsockopt
>>>> RETURN VALUE
>>>> On success, zero is returned. On error, -1 is returned, and
>>>> errno is set appropriately.
>>>>
>>>> .. original code:
>>>> ..
>>>> .. do {
>>>> .. ret = getsockopt(s->fd, SOL_SOCKET, SO_ERROR, (void *)
>>>> &val, &valsize);
>>>> .. } while (ret == -1 && (socket_error()) == EINTR);
>>>> ..
>>>> .. if (ret < 0) { // ret equals to -1, and socket_error() !=
>>>> EINTR
>>>>
>>>> If ret < 0, just set the error state, but the callback function is
>>>> not set to NULL.
>>>> Then tcp_wait_for_connect() will be called again, and retry to
>>>> checking current
>>>> socket by getsockopt().
>>>>
>>>> But in this patch, we will abandon current socket, and connect next
>>>> address on the list.
>>>> We should reserve that block, and move
>>>> qemu_set_fd_handler2(...NULL..) below it.
>>> I'm sorry I don't get you point, it makes no difference where is
>>> qemu_set_fd_handlers(...NULL...).
>>> rc < 0 happens in two cases:
>>> 1. val != 0 - connect error
>>
>> It seems rc is 0 in this case.
>>
>>> 2. getsockopt failed for reason different than EINTR .
>>> In both cases we can only move to the next addr in the list,
>>> inet_connect_addr will call qemu_set_fd_handlers with the new socket
>>> fd.
>>
>> In the past, if rc < 0, we will return without setting fd handler to NULL,
>> tcp_wait_for_connect() will be re-called. the original socket (current addr)
>> will be re-checked by getsockopt()
>>
>> But in your patch, it will retry next addr in the list.
>>
> In the past we didn't have the address list so we could not continue to the next one,
> now we can.
Yes, we should try to connect next addr if current address is not available.
But rc < 0 doesn't mean current is not available.
"rc < 0" means fail to get socket option, we need to retry to _get_
socket option.
--
Amos.
next prev parent reply other threads:[~2012-09-24 9:48 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-09-13 18:58 [Qemu-devel] [PATCH v3 0/3] nonblocking connect address handling cleanup Orit Wasserman
2012-09-13 18:58 ` [Qemu-devel] [PATCH v3 1/3] Refactor inet_connect_opts function Orit Wasserman
2012-09-14 8:58 ` Juan Quintela
2012-09-19 8:33 ` Amos Kong
2012-09-20 2:33 ` Amos Kong
2012-09-20 11:22 ` Orit Wasserman
2012-09-20 12:38 ` Markus Armbruster
2012-09-20 11:21 ` Orit Wasserman
2012-09-13 18:58 ` [Qemu-devel] [PATCH v3 2/3] Separate inet_connect into inet_connect (blocking) and inet_nonblocking_connect Orit Wasserman
2012-09-14 9:07 ` Juan Quintela
2012-09-20 12:44 ` Markus Armbruster
2012-09-20 14:56 ` Orit Wasserman
2012-09-13 18:58 ` [Qemu-devel] [PATCH v3 3/3] Fix address handling in inet_nonblocking_connect Orit Wasserman
2012-09-14 9:17 ` Juan Quintela
2012-09-19 8:31 ` Amos Kong
2012-09-20 11:20 ` Orit Wasserman
2012-09-20 15:16 ` Amos Kong
2012-09-23 6:34 ` Orit Wasserman
2012-09-24 9:48 ` Amos Kong [this message]
2012-09-24 10:40 ` Orit Wasserman
2012-09-24 11:41 ` Amos Kong
2012-09-20 6:03 ` Michael S. Tsirkin
2012-09-20 8:57 ` Orit Wasserman
2012-09-20 9:37 ` Michael S. Tsirkin
2012-09-20 10:00 ` Orit Wasserman
2012-09-20 13:14 ` Markus Armbruster
2012-09-20 14:53 ` Orit Wasserman
2012-09-21 8:03 ` Markus Armbruster
2012-09-23 7:31 ` Orit Wasserman
2012-09-20 13:19 ` [Qemu-devel] [PATCH v3 0/3] nonblocking connect address handling cleanup Markus Armbruster
2012-09-20 14:55 ` Orit Wasserman
2012-09-20 15:12 ` Paolo Bonzini
2012-09-20 15:19 ` Michael S. Tsirkin
2012-09-20 15:15 ` Michael S. Tsirkin
2012-09-21 8:03 ` Markus Armbruster
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=50602C73.2080503@redhat.com \
--to=akong@redhat.com \
--cc=owasserm@redhat.com \
--cc=qemu-devel@nongnu.org \
/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).