marketage has quit [Remote host closed the connection]
TimF has joined #spectrum
TimFroehlich has joined #spectrum
TimF has quit [Ping timeout: 240 seconds]
<hypokeimenon[m]>
Exactly what a spammy virus would say!
<Scrblue[m]>
If only I had a per-application VM to check out suspicious links in
<qyliss>
:D
<Scrblue[m]>
Kinda wonder what the coupon is for, though
<Scrblue[m]>
They have a DIY CNC router kit on Aliexpress and I wouldn't mind a discount on it
<MichaelRaskin>
Pretty sure container is good enough in case of an aliexpress link might want to enter domain manually first, though
<Scrblue[m]>
I'm still figuring out NixOS so setting up a container sounds like hell right now :(
<MichaelRaskin>
Well, I have an nsjail setup and it does not use NixOS mainline… what stops me is more about not wanting to boost the link visit counts
Ke has quit [Ping timeout: 245 seconds]
hiroshi[m] has quit [Ping timeout: 245 seconds]
Ke has joined #spectrum
hiroshi[m] has joined #spectrum
TimF has joined #spectrum
<qyliss>
I seem to have run into a weird kernel state where it seems to both think that a PCI device is and is not bound to the vfio-pci driver, in a way that makes it impossible to do anything with it
TimFroehlich has quit [Ping timeout: 252 seconds]
<V>
Schrödinger's PCI device
<MichaelRaskin>
Murphy's PCI device: bound or not bound, depending on what would error out the current operation
<qyliss>
it's entirely possible that I'm just misunderstanding how this is supposed to work, of course, but this hasn't happened before
<qyliss>
aha! I was able to get it back with remove_id followed by new_id
<qyliss>
I suppose this probably comes from me not understanding the difference between new_id and bind, and remove_id and unbind.
<qyliss>
"The bind/unbind attributes tries to attach/detach a driver to/from a device supported by that driver. The new_id/remove_id attributes adds/removes device IDs to/from the list of devices supported by the driver."
<qyliss>
So I think what happened was that my device was in the list of devices supported by vfio-pci, but not actually bound to any driver
<qyliss>
and I was relying on the binding happening as a side effect of adding it to the supported list
<qyliss>
which doesn't work when it's already there
<qyliss>
so instead what I probably need to do is manually bind after attempting to add it to the supported list
<qyliss>
but that means adding it to the supported list is allowed to fail, and binding it is also allowed to fail, but they're not both allowed to fail
<qyliss>
but I can just let them both fail and then check if the device is bound correctly
<qyliss>
I'm probably not making much sense to anybody trying to follow this lol
<qyliss>
I'll write it up better in the patch body
<MichaelRaskin>
I dunno, makes perfect sense to me
<qyliss>
oh, nice :)
<MichaelRaskin>
And I did not know about remove_id/new_id in advance!