IPv6 support for php-fpm (fixes: #55508)#631
Conversation
sapi/fpm/php-fpm.conf.in
Outdated
There was a problem hiding this comment.
Minor, but could you fix these silly space issues? :)
|
fixed and rebased |
|
+1 please, php-fpm is the last component in our internal network that does not support ipv6, we would love to kill ipv4 internally :) |
|
Yes I agree; we have the exact same problem. Everything runs over ipv6 but our firewalls need to make 1 exception because fpm cannot handle ipv6. The only alternative is running an instance of socat to convert ipv6 to ipv4.... |
|
+1000 this would kill the last IPv4 dep |
|
+1! |
|
Yes please! |
|
Is somebody going to take a look at this before PHP 5.6? |
|
💯 What happens when one enters only port? Will it listen on both IPv4 and IPv6? |
|
Ping @Tyrael, what do you think? Possible for 5.6? |
|
We are too deep into the release release (we already had an RC2, feature freeze since beta1) to have this in 5.6, but we I would like to see this in master. Sorry that we missed this, I will try to do a better job with keeping up with the pull requests, but if you see an rfc not getting the deserved attention, feel free to drop a reminder email to [email protected] |
|
@Tyrael This looks like a minor, self-contained feature addition. Can't we add that in 5.6.1? Unless I misunderstand the impact, this could also go to 5.5. |
|
Bump. |
Surprisingly, this can be added in RC4: "Add SO_REUSEPORT + SO_BROADCAST support via socket stream context option" a51bf0c No references to a pull request, bug report or whatsoever. An individual PHP develeoper with git access can do whatever they want. A pull request for IPv6 support which should have been added 5 years ago, with multiple votes and issues on bugs.php.net cannot be merged "because of feature freeze". Strange... |
|
indeed this is pretty absurd... |
|
I'm all right to have this added to 5.5 |
There was a problem hiding this comment.
This looks like replacing ']' by \0 which seems strange to me
There was a problem hiding this comment.
@jpauli if the user enters an ipv6 address with the [ip]:port format those [ and ] have to be removed. This line you pointed just deletes the ] in the last char and the next line addr++; deletes the [ in the first char
|
@arjenschol Thanks for catching this, I've just sent an email to the author to revert |
|
+1 lets get rid of legacy IP. |
|
Now 5.6.0 is released, can this be pulled into the 5.6 branch for release in 5.6.1? |
|
From the thread above it seems you should have it into 5.5 and merge it On Fri, Aug 29, 2014 at 12:01 PM, Arjen [email protected] wrote:
|
|
I would be fine having this in 5.6.1, as it shouldn't have any impact for existing fpm deployments, but people can opt-in to listen on the v6 address with the new config option. |
|
Yes sure, it can go to 5.5 as well |
|
this must be a joke or something... |
|
@jpauli so what are we waiting for merging this? |
|
This is sad. So much ignorance... 😟 |
|
Also, noone yet answered my question:
|
|
@Majkl578 i can't guarantee that without testing it but looking at the code it should open connections for both ipv4 and ipv6 if you only set the port :) |
|
Bump again, ping @Tyrael + @jpauli. This is really getting awkward... @acasademont: Thanks, that's how it should work, I was just curious whether the implementation is aware of that. :) |
|
Bump. |
|
hi, I've merged the patch into 5.5 and upwards. |
|
thank you all! this is awesome news :D |
|
Hey @globin , I consistently see sapi/fpm/tests/003.phpt fails with: |
|
This change (or another FPM related change in version 5.5.19 and 5.6.3) seems to break PHP-FPM when using the network stack. It can be reproduced e.g. on Ubuntu 14.04 by creating 5 different pools (port numbers 9001 to 9005) and then starting the PHP-FPM service. I would expect to have a listener on each of the five ports, instead only the first pool is actually loaded. The others fail to load (can e.g. be checked with I had to downgrade to version 5.5.18/5.6.2 to fix the issue. |
|
@leofeyer I have reported a similar bug for that @ https://bugs.php.net/bug.php?id=68420. I don't think posting bugs in the comments of PR is a good idea |
|
@SjonHortensius Thanks, but this bug seems to handle a different issue, so I have created a separate one: https://bugs.php.net/bug.php?id=68423 |

This removes hardcoded IPv4 code and adds support for IPv6