LWN: Comments on "OpenSSH 8.5 released" https://lwn.net/Articles/848048/ This is a special feed containing comments posted to the individual LWN article titled "OpenSSH 8.5 released". en-us Tue, 04 Nov 2025 21:22:53 +0000 Tue, 04 Nov 2025 21:22:53 +0000 https://www.rssboard.org/rss-specification lwn@lwn.net My use case: one hundred systems with the same ssh host key https://lwn.net/Articles/848711/ https://lwn.net/Articles/848711/ emmi3 <div class="FormattedComment"> I have the following setup: nearly one hundred thin clients for home office use (&quot;Telearbeit&quot; / tele work) running from the same live linux image.<br> <p> The (cutomized) images are built using live-build form debian-live. Normally live-build would delete the ssh host key during build time and live-config would create a new ssh host key on every startup. This was undesirable since ssh would complain about the changed host key after every reboot of the thin client. Therefore I baked one predefined host key directly into the image.<br> <p> The thin clients are connected to our university environment via wireguard using a 10-something private subnet. Thus we have nearly one hundred different physical hosts (with different but fixed IPs and hostnames) using the same ssh host key.<br> <p> I don&#x27;t see anything wrong with this setup and I think this is a valid use case. If my ssh client starts complaining about all those hosts having the same host key, I will have to start creating separate keys for every client and distributing them like I do with the wireguard preshared keys and other client specific data right now. No big deal, but I don&#x27;t really see any benefit from this.<br> </div> Mon, 08 Mar 2021 17:13:10 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848634/ https://lwn.net/Articles/848634/ josh <div class="FormattedComment"> Cloud providers don&#x27;t typically do this in their DHCP servers. (And I think it makes sense that they don&#x27;t, for a variety of reasons, not least of which that it&#x27;s better to show people very quickly that IPs will change, rather than let them experience breakage later on.)<br> </div> Mon, 08 Mar 2021 00:22:51 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848600/ https://lwn.net/Articles/848600/ Cyberax <div class="FormattedComment"> <font class="QuotedText">&gt; Then you&#x27;ll have a lot less of this happening, as each VM will end up using the same address virtually all the time.</font><br> Then you&#x27;ll run out of addresses, since VMs are disposable and each new VM gets a new MAC.<br> </div> Sun, 07 Mar 2021 15:12:23 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848590/ https://lwn.net/Articles/848590/ vadim <div class="FormattedComment"> You can configure a DHCP server to hand out leases for a long time, like a month or even a year.<br> <p> Then you&#x27;ll have a lot less of this happening, as each VM will end up using the same address virtually all the time.<br> <p> </div> Sun, 07 Mar 2021 12:21:04 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848311/ https://lwn.net/Articles/848311/ josh <div class="FormattedComment"> Virtual machine instances that are regularly shut down and brought back up, and don&#x27;t have or need a static IP. Start instance, get IP for instance, SSH to instance, work with instance, shut down instance.<br> </div> Thu, 04 Mar 2021 22:06:43 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848310/ https://lwn.net/Articles/848310/ Cyberax <div class="FormattedComment"> Uh, the hosts will have different keys but the same IP. So you get the dreaded &quot;WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED&quot; message from SSH each time you try to connect.<br> </div> Thu, 04 Mar 2021 22:06:22 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848267/ https://lwn.net/Articles/848267/ nye <div class="FormattedComment"> Now I&#x27;m *really* curious. What&#x27;s the application here? No worries if it&#x27;s something you can&#x27;t/don&#x27;t want to go into.<br> </div> Thu, 04 Mar 2021 17:43:21 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848254/ https://lwn.net/Articles/848254/ josh <div class="FormattedComment"> Same virtual machine, same host key, no hostname, different IPs.<br> </div> Thu, 04 Mar 2021 16:16:54 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848187/ https://lwn.net/Articles/848187/ nye <div class="FormattedComment"> Well unless those hosts are reusing the same host key then there won&#x27;t be any &quot;other host names/addresses already associated with the key&quot;, so you can&#x27;t end up with a list containing hundreds of entries.<br> <p> (And if they *are* reusing the same key, then you still won&#x27;t end up with such a list unless you connect via a new throwaway DNS name for each one instead of using a fixed hostname or the unchanging IP address.)<br> </div> Thu, 04 Mar 2021 12:21:42 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848186/ https://lwn.net/Articles/848186/ Cyberax <div class="FormattedComment"> No, reusing the same IP for different throwaway hosts.<br> </div> Thu, 04 Mar 2021 11:37:32 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848185/ https://lwn.net/Articles/848185/ nye <div class="FormattedComment"> As in you&#x27;re re-using the same host key on hundreds of machines, or you&#x27;re connecting to the same machine via hundreds of aliases? Both of these seem like pretty niche use cases that I&#x27;d only expect to see in some kind of automated environment (probably involving throwaway test systems in the first case, given the risk involved in reusing a key).<br> </div> Thu, 04 Mar 2021 11:33:11 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848184/ https://lwn.net/Articles/848184/ johill <div class="FormattedComment"> I think they meant *host* key, not *client* key, here? At least that&#x27;s how I read it? Hmm, maybe not?<br> </div> Thu, 04 Mar 2021 10:36:39 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848157/ https://lwn.net/Articles/848157/ josh <div class="FormattedComment"> I&#x27;d expect the common case for me to be in the hundreds. That doesn&#x27;t seem unreasonable, depending on how it&#x27;s presented.<br> </div> Thu, 04 Mar 2021 07:11:57 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848146/ https://lwn.net/Articles/848146/ djm <div class="FormattedComment"> <font class="QuotedText">&gt; And this one, though it could produce a massive amount of output in some cases.</font><br> <p> yeah, if this turns out to be a problem in practice then let us know and we&#x27;ll add a limit.<br> </div> Thu, 04 Mar 2021 01:03:20 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848143/ https://lwn.net/Articles/848143/ unixbhaskar <div class="FormattedComment"> <font class="QuotedText">&gt; * ssh(1): when prompting the user to accept a new hostkey, display</font><br> <font class="QuotedText">&gt; any other host names/addresses already associated with the key.</font><br> <p> This one would be really interesting!<br> </div> Wed, 03 Mar 2021 23:12:00 +0000 OpenSSH 8.5 released https://lwn.net/Articles/848106/ https://lwn.net/Articles/848106/ josh <div class="FormattedComment"> <font class="QuotedText">&gt; * ssh(1): disable CheckHostIP by default. It provides insignificant</font><br> <font class="QuotedText">&gt; benefits while making key rotation significantly more difficult,</font><br> <font class="QuotedText">&gt; especially for hosts behind IP-based load-balancers.</font><br> <p> I&#x27;m excited to see this change.<br> <p> <font class="QuotedText">&gt; * ssh(1): when prompting the user to accept a new hostkey, display</font><br> <font class="QuotedText">&gt; any other host names/addresses already associated with the key.</font><br> <p> And this one, though it could produce a massive amount of output in some cases.<br> </div> Wed, 03 Mar 2021 18:13:13 +0000