I remember when you were actually able to keep up with CVEs and you could (at least somewhat) prioritize patches while disregarding those which felt irrelevant.
But now.. are we simply supposed to use our f̶a̶v̶o̶r̶i̶t̶e̶ package manager and batch update our systems in a loop?
while true; do apt-get update && apt-get upgrade -y; sleep 1d; done
wait. erhm... and probably an automatic reboot/kernel reload every 48h; THAT WILL DO IT! </sarcasm>
Especially in environments where manglement mandate something like Qualys, leading to demands that systems must be patched to address the critical vulnerabilities that it's reported despite said vulnerabilities being unexploitable.
One that I trot out periodically as an example of this is a CVE that would only be exploitable if running on an IBM s390 with EBCDIC codepages. Our security team nevertheless wanted it patched, because Qualys said it was a vulnerability.
Had that in /old job/ years ago, every security report would list off a load of packages that needed updating. Problem was, not only were they not actual vulnerabilities but they were all for some legacy software that was in the process of being removed.
Was always fun to purge a load of systems from old shit, and watch the counter drop.
This is a confusing comment because Debian/Ubuntu/etc (you referenced apt) have separate security repositories that exist for this purpose. They have only security fixes and don't try to unnecessarily fix "other" stuff. I have not heard of the security repos pushing out a breaking update before.
`unattended-upgrade` uses only a subset of repos (ie security) and does this for you. If you actually wanted to schedule reboots, it can do that too.
perhaps we better add a job to automatically order new IoT devices from amazon too; once a month is probably good enough not to suffer. I think we just invented Upgrades-As-A-Service.. only $50/month for freshly not-yet-known-to-be-b0rked firmware!
There are pro/enterprise features that can upgrade the linux kernel without downtime, like livepatch or kpatch. (The non-proprietary klp-build is still missing tooling for most distros I think.)
Theoretically they should slow down. I suspect people will be much pickier about which apps to use in production. Which, ironically, will be directly at odds with more and more products being pumped out..
>Alternatively, you might shrug and wait to see which
of these will get a logo and catchy name in the next
few weeks and then try to focus on those.
hehehe
>Another approach might be to sit back, have a nice cup
of zen, and just always pull all updates and then
update your entire fleet of systems on a weekly basis,
which, yes, I sure would like to be able to do, but
reality keeps getting in my way.
Am I the only one that gets concerned about the laissez-faire approach to security that a LOT of people take. I get that some exploits are not like the others, but some of the shit I've seen in places....
I remember when you were actually able to keep up with CVEs and you could (at least somewhat) prioritize patches while disregarding those which felt irrelevant.
But now.. are we simply supposed to use our f̶a̶v̶o̶r̶i̶t̶e̶ package manager and batch update our systems in a loop?
wait. erhm... and probably an automatic reboot/kernel reload every 48h; THAT WILL DO IT! </sarcasm>Especially in environments where manglement mandate something like Qualys, leading to demands that systems must be patched to address the critical vulnerabilities that it's reported despite said vulnerabilities being unexploitable.
One that I trot out periodically as an example of this is a CVE that would only be exploitable if running on an IBM s390 with EBCDIC codepages. Our security team nevertheless wanted it patched, because Qualys said it was a vulnerability.
Had that in /old job/ years ago, every security report would list off a load of packages that needed updating. Problem was, not only were they not actual vulnerabilities but they were all for some legacy software that was in the process of being removed.
Was always fun to purge a load of systems from old shit, and watch the counter drop.
This is a confusing comment because Debian/Ubuntu/etc (you referenced apt) have separate security repositories that exist for this purpose. They have only security fixes and don't try to unnecessarily fix "other" stuff. I have not heard of the security repos pushing out a breaking update before.
`unattended-upgrade` uses only a subset of repos (ie security) and does this for you. If you actually wanted to schedule reboots, it can do that too.
The comment in question is a joke; it was never intended to be read as actual advice.
Wire the script to a pressure sensor in your seat, save-all/update/reboot every time you get out of your chair.
Bonus points for relaunching programs open during initial execution of script.
perhaps we better add a job to automatically order new IoT devices from amazon too; once a month is probably good enough not to suffer. I think we just invented Upgrades-As-A-Service.. only $50/month for freshly not-yet-known-to-be-b0rked firmware!
are you gonna spot the LLM bill or should I?
... or that the kernel can't be hot reloaded is itself a vulnerability?
Debian based systems can be queried if they need a reboot because of dpkg. But this was the same then 20 years ago....
There are pro/enterprise features that can upgrade the linux kernel without downtime, like livepatch or kpatch. (The non-proprietary klp-build is still missing tooling for most distros I think.)
I still don't trust that sort of thing. And I don't think I ever will.
"Shit why is everything dead"
LLMs gave that one last push: Linux is not sustainable.
We need to start looking at seL4 as the base, instead.
Theoretically they should slow down. I suspect people will be much pickier about which apps to use in production. Which, ironically, will be directly at odds with more and more products being pumped out..
>Alternatively, you might shrug and wait to see which of these will get a logo and catchy name in the next few weeks and then try to focus on those.
hehehe
>Another approach might be to sit back, have a nice cup of zen, and just always pull all updates and then update your entire fleet of systems on a weekly basis, which, yes, I sure would like to be able to do, but reality keeps getting in my way.
Am I the only one that gets concerned about the laissez-faire approach to security that a LOT of people take. I get that some exploits are not like the others, but some of the shit I've seen in places....