[Bug 8332] Multiple applications fail to send ICMP requests on Linux (failure to open IPPROTO_ICMP sockets)

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=8332

--- Comment #39 from Gabriel Ivăncescu <gabrielopcode at gmail.com> ---
(In reply to oiaohm from comment #38)
> (In reply to Zebediah Figura from comment #37)
> > I'm adjusting the title, since it's misleading. As far as I understand
> > SOCK_DGRAM + IPPROTO_ICMP actually does sufficiently work on Linux (Damjan
> > claims it is "completely broken", which seems odd seeing as `wine ping`
> > works fine.) The problem is, as far as I'm aware, this is *also* forbidden
> > by default on some distributions, including Debian, via the
> > "net.ipv4.ping_group_range" sysctl.
> > 
> > Should we just require that sysctl to be set for Wine? Is it reasonable to
> > try to fall back to ping(1)? I don't see any other way to achieve what we
> > want...
> > 
> > The linked download no longer works on Windows (i.e. it gives the same "no
> > internet detected" message; probably the service died), so I'm removing that.
> 
> https://serverfault.com/questions/1002225/how-do-i-limit-ping-icmp-responses-
> on-a-debian-10-server
> 
> There are a lot of different changeable values 
> 
> net.ipv4.ping_group_range is is straight up forbid.   There is also the
> problem that a icmp send or receive rate limit can be set like mentioned
> above and other setting.
> 
> https://social.msdn.microsoft.com/Forums/en-US/74fd0d8e-b75a-4aa5-bb50-
> 140e606950b6/icmp-error-processing-on-udp-socket?forum=wsk
> 
> Also I am sorry but Damjan Jovanovic is wrong in a fatal way.   Windows ICMP
> error handling changes with windows version and configuration as well.   
> --Windows doesn't seem to provide any ICMP error data anyway:-- is right and
> wrong as the same time.
> 
> Mac OS and Linux both don't match what different versions of Windows will
> provide applications with at times.
> 
> The Linux should provide enough data to reconstruct all the different wacky
> ways windows can respond to applications in case of ICMP error.
> 
> Also to be really horrible windows ICMP error response altered based on
> Windows firewalls as well.   These differences can be why a program works
> under XP then fails under Windows 7 and newer.  Or only works with Windows 7
> and newer when particular firewall settings are set.
> 

I think you missed the point. The failure happens when either creating the
socket (because no permissions), or when sending it AFAIK, not the error
received, although that's also another one of the issues between
implementations.

Anyway, Huw has done a rewrite of icmp on top of nsi. Damjan, can you test with
latest wine git and see if it's still broken?

-- 
Do not reply to this email, post in Bugzilla using the
above URL to reply.
You are receiving this mail because:
You are watching all bug changes.