udev: don't spread variables through all nodes#3097
Conversation
howl
force-pushed
the
dont-spread-variable
branch
from
a0c6021 to
54ba976
Compare
|
Could you elaborate 'we don't want'? Who is we and why? Just trying to figure out what the problem is. Maybe you could share some logs or explain what's going wrong. Thanks! |
|
@igaw you need to ask? Is anything but not normal to set variables unconditionally to absolutely all the nodes in dev. Just reporting it into systemd @yuwata set this to v260 milestone and look the issue was here. No need for logs just do yourself sudo udevadm trigger and then udevadm info yourself the device you want as is spreaded over absolutely all nodes, I suggest /dev/zero, /dev/null, /dev/full, /dev/kmesg you shouldn't do that by no mean. I have set it to the just the only same subsystem this project checks that variable and removed unnecessary conditionals. |
|
Not everyone understands udev and all the dark corners. Having a proper commit message is important. And yes I have to ask, I am not an expert. |
NVME_HOST_IFACE=none is being set to every existing node device but this shouldn't be done, we don't want this in /dev/peanutbutter. To quick check this error just launch "udevadm trigger" and then "udevadm info" any dev node you want, the variable will be set for all the nodes. The easier way to understand this /dev/zero, /dev/null, /dev/random has nothing to do with this variable. To solve this a condition to only apply for the nvme subsystem is added as is the only subsystem later check for this variable to be present. While at it remove all the action change conditions that are always true given the if it's any other kind of action a goto end in the beginning. Signed-off-by: David Santamaría Rogado <howl.nsp@gmail.com>
howl
force-pushed
the
dont-spread-variable
branch
from
54ba976 to
d912707
Compare
I have just edited the commit message to explain even more a very trivial change. Did you read the commit message or just the PR message? Because you complain about commit message but replied to things only in the PR and as far I can see misunderstanding the intention of it . Take a look for /dev/mordor, it exists in some systems, peanutbutter just means what it means, is meaningless to alter every device node unconditionally for this variable. |
|
Thanks for taking the time to explain it to me and make the commit message a bit clearer. Really appreciated. |
|
@igaw and sorry my nearly bad manners, I assumed a just because I want it but you haven't the full context. Thanks for taking this issue into account. |
3 participants
This has been discussed here systemd/systemd#40670
NVME_HOST_IFACE=none is being set to every existing node device but this
shouldn't be done, we don't want this in /dev/peanutbutter.
Set it only for nvme subsystem devices.
While at it remove all the action change conditions that are always true
given the if it's any other kind of action a goto end in the beginning.