[PATCH net] ipv6: bring NLM_DONE out to a separate recv() again

[PATCH net] ipv6: bring NLM_DONE out to a separate recv() again

[Jakub Kicinski]

Commit under Fixes optimized the number of recv() calls
needed during RTM_GETROUTE dumps, but we got multiple
reports of applications hanging on recv() calls.
Applications expect that a route dump will be terminated
with a recv() reading an individual NLM_DONE message.

Coalescing NLM_DONE is perfectly legal in netlink,
but even tho reporters fixed the code in respective
projects, chances are it will take time for those
applications to get updated. So revert to old behavior
(for now)?

This is an IPv6 version of commit 460b0d33cf10 ("inet: bring
NLM_DONE out to a separate recv() again").

Reported-by: Maciej Żenczykowski <zenczykowski@gmail.com>
Link: https://lore.kernel.org/all/CANP3RGc1RG71oPEBXNx_WZFP9AyphJefdO4paczN92n__ds4ow@mail.gmail.com
Reported-by: Stefano Brivio <sbrivio@redhat.com>
Link: https://lore.kernel.org/all/20240315124808.033ff58d@elisabeth
Reported-by: Ilya Maximets <i.maximets@ovn.org>
Link: https://lore.kernel.org/all/02b50aae-f0e9-47a4-8365-a977a85975d3@ovn.org
Fixes: 5fc68320c1fb ("ipv6: remove RTNL protection from inet6_dump_fib()")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
---
CC: dsahern@kernel.org
CC: donald.hunter@gmail.com
CC: maze@google.com
---
 net/ipv6/ip6_fib.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
index d9c9a542a414..eb111d20615c 100644
--- a/net/ipv6/ip6_fib.c
+++ b/net/ipv6/ip6_fib.c
@@ -2514,7 +2514,8 @@ int __init fib6_init(void)
 		goto out_kmem_cache_create;
 
 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE, NULL,
-				   inet6_dump_fib, RTNL_FLAG_DUMP_UNLOCKED);
+				   inet6_dump_fib, RTNL_FLAG_DUMP_UNLOCKED |
+				   RTNL_FLAG_DUMP_SPLIT_NLM_DONE);
 	if (ret)
 		goto out_unregister_subsys;
 
-- 
2.45.2

Re: [PATCH net] ipv6: bring NLM_DONE out to a separate recv() again

[Ilya Maximets]

On 6/18/24 21:39, Jakub Kicinski wrote:
> Commit under Fixes optimized the number of recv() calls
> needed during RTM_GETROUTE dumps, but we got multiple
> reports of applications hanging on recv() calls.
> Applications expect that a route dump will be terminated
> with a recv() reading an individual NLM_DONE message.
> 
> Coalescing NLM_DONE is perfectly legal in netlink,
> but even tho reporters fixed the code in respective
> projects, chances are it will take time for those
> applications to get updated. So revert to old behavior
> (for now)?
> 
> This is an IPv6 version of commit 460b0d33cf10 ("inet: bring
> NLM_DONE out to a separate recv() again").
> 
> Reported-by: Maciej Żenczykowski <zenczykowski@gmail.com>
> Link: https://lore.kernel.org/all/CANP3RGc1RG71oPEBXNx_WZFP9AyphJefdO4paczN92n__ds4ow@mail.gmail.com
> Reported-by: Stefano Brivio <sbrivio@redhat.com>
> Link: https://lore.kernel.org/all/20240315124808.033ff58d@elisabeth
> Reported-by: Ilya Maximets <i.maximets@ovn.org>
> Link: https://lore.kernel.org/all/02b50aae-f0e9-47a4-8365-a977a85975d3@ovn.org
> Fixes: 5fc68320c1fb ("ipv6: remove RTNL protection from inet6_dump_fib()")
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> ---
> CC: dsahern@kernel.org
> CC: donald.hunter@gmail.com
> CC: maze@google.com
> ---
>  net/ipv6/ip6_fib.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
> index d9c9a542a414..eb111d20615c 100644
> --- a/net/ipv6/ip6_fib.c
> +++ b/net/ipv6/ip6_fib.c
> @@ -2514,7 +2514,8 @@ int __init fib6_init(void)
>  		goto out_kmem_cache_create;
>  
>  	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE, NULL,
> -				   inet6_dump_fib, RTNL_FLAG_DUMP_UNLOCKED);
> +				   inet6_dump_fib, RTNL_FLAG_DUMP_UNLOCKED |
> +				   RTNL_FLAG_DUMP_SPLIT_NLM_DONE);
>  	if (ret)
>  		goto out_unregister_subsys;
>  

Thanks, Jakub!  LGTM.

Libreswan in OVS tests is working fine with this change on top of 6.10-rc4.
The rest of the test suite is also working fine (except for FTP conntrack,
but that's an unrelated issue).

Tested-by: Ilya Maximets <i.maximets@ovn.org>

Re: [PATCH net] ipv6: bring NLM_DONE out to a separate recv() again

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Tue, 18 Jun 2024 12:39:14 -0700 you wrote:
> Commit under Fixes optimized the number of recv() calls
> needed during RTM_GETROUTE dumps, but we got multiple
> reports of applications hanging on recv() calls.
> Applications expect that a route dump will be terminated
> with a recv() reading an individual NLM_DONE message.
> 
> Coalescing NLM_DONE is perfectly legal in netlink,
> but even tho reporters fixed the code in respective
> projects, chances are it will take time for those
> applications to get updated. So revert to old behavior
> (for now)?
> 
> [...]

Here is the summary with links:
  - [net] ipv6: bring NLM_DONE out to a separate recv() again
    https://git.kernel.org/netdev/net/c/02a176d42a88

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html