ping: revert "ping: do not bind to device when destination IP is on d… #396
Conversation
…evice" The issue should be fixed in kernell side, see: https://lore.kernel.org/all/ea03066bc7256ab86df8d3501f3440819305be57.1644988852.git.lucien.xin@gmail.com/ Otherwise it would cause regressions in some kernel selftests. Signed-off-by: Xin Long <lucien.xin@gmail.com>
|
@kerolasa |
|
@lxin FYI we consider reverting cc44f4c as you did in this PR (see #405 (comment)), but IMHO nobody from the community knows all the corner cases which need to be tested. Given you fixed it in the kernel:
and these fixes has been backported to all stable / LTS, most of the users which install updates have problem solved. It's just a question whether all enterprise distros backport these fixes and how many users will continue using unfixed kernels. Also users with unfixed kernel will probably not update iputils either. To be clear, I'm not against fixing it also in userspace, but we must be sure we avoid regression. nit: it'd be better to run |
|
Hi, Petr
I'm not sure. cc44f4c has also caused some regression though it's only seen
in the kselftests, which is also why I made the kernel fix. I don't know
which regression is more serious regardless of the patch's correctness:).
Note that, there's a workaround for the original issue, not using the PING
socket by:
# sysctl -w net.ipv4.ping_group_range="1 0"
Also, we should also be aware of how the iputils version and kernel version
are related in the distro. like the latest iptuils may require the
kernel version to be newer than kernel-x.y, or we can put some changelog to
say that this revert should be backported together with the kernel commit.
Thanks.
…On Thu, Nov 3, 2022 at 4:05 AM Petr Vorel ***@***.***> wrote:
@lxin <https://github.com/lxin> patch in kernel you reference was merged
into kernel mainline in
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=35a79e64de29e8d57a5989aac57611c0cd29e13e
(v5.17-rc5, released in v5.17). Do you still think this should be reverted?
That would fix older kernels which does not have fix, but OTOH most of
linux distros use LTS kernels whic got this fix.
—
Reply to this email directly, view it on GitHub
<#396 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABJHL5Z3GFOXGIDGYBKLCPLWGNW6JANCNFSM5QCEC7BQ>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
David Ahern, the IPv4/IPv6 kernel maintainer also agreed with reverting this |
…evice"
The issue should be fixed in kernell side, see:
https://lore.kernel.org/all/ea03066bc7256ab86df8d3501f3440819305be57.1644988852.git.lucien.xin@gmail.com/
Otherwise it would cause regressions in some kernel selftests.
Signed-off-by: Xin Long lucien.xin@gmail.com