Skip to contentSkip to conversation

Can't load modules with parameters given

DISCUSSION Benjamin Block 2 participants 3 messages Dec 19, 2013 – Dec 22, 2013

lore
Share & tools
lore
#0 lore raw

Hi,

I have a rather strange problem currently. I am trying to load a module with some parameters (bnx2x with disable_tpa=1) and it fails because this parameter is unknown according to dmesg. But modinfo lists this exact paramter and it is also included in the source. And while this is not an in-tree driver, I could also replicate the same behavior with in-tree device drivers (like bonding with the parameter tx_queues):

$ modprobe bonding # works fine
$ rmmod bonding
$ modprobe bonding tx_queues=32
modprobe: ERROR: could not insert 'bonding': Unknown symbol in module,
or unknown parameter (see dmesg)

and dmesg tells me:
bonding: Unknown parameter `tx_queues'
and nothing else (as with bnx2x, this parameter is listed by modinfo).

Is there some obvious kernel-config-option I could have messed up or something else like this?

I am using a Gentoo-Userspace kmod (v15) with a Ubuntu-Kernel (3.8.13.13
- this is a version with some ubuntu-patches applied) - I know, the combination is rather strange. Any ideas?

best regards,
- Benjamin

#1 lore raw

Hi Benjamin,

Quoted 3 lines · Benjamin Block

Hi,

I have a rather strange problem currently. I am trying to load a module

strange indeed

Quoted 4 lines

with some parameters (bnx2x with disable_tpa=1) and it fails because this parameter is unknown according to dmesg. But modinfo lists this exact paramter and it is also included in the source. And while this is not an in-tree driver, I could also replicate the same behavior with

are you sure modinfo and modprobe are reading the same module? maybe you updated your modules between the operations?

Quoted 8 lines

in-tree device drivers (like bonding with the parameter tx_queues):

$ modprobe bonding # works fine
$ rmmod bonding
$ modprobe bonding tx_queues=32
modprobe: ERROR: could not insert 'bonding': Unknown symbol in module,
or unknown parameter (see dmesg)
Quoted 6 lines

and dmesg tells me:
bonding: Unknown parameter `tx_queues'
and nothing else (as with bnx2x, this parameter is listed by modinfo).

Is there some obvious kernel-config-option I could have messed up or something else like this?

So... this changed in recent kernels (3.11 -- see the commit message that introduced this behavior below)... we don't fail to load a module anymore due to unknown option. So if/when you upgrade you won't see this problem anymore.

Anyway, it shouldn't fail for you like it did, but it might as well be a problem of the particularly patched kernel you are using... not sure.

Quoted 4 lines

I am using a Gentoo-Userspace kmod (v15) with a Ubuntu-Kernel (3.8.13.13
- this is a version with some ubuntu-patches applied) - I know, the combination is rather strange. Any ideas?

the combination is not a problem and should work just fine.

Lucas De Marchi

---
commit 54041d8a73337411b485ff76957fb106cb5d40d0

Author: Rusty Russell <[email protected]>
Date:   Tue Jul 2 15:35:12 2013 +0930

modules: don't fail to load on unknown parameters.

Although parameters are supposed to be part of the kernel API, experimental parameters are often removed. In addition, downgrading a kernel might cause previously-working modules to fail to load.

On balance, it's probably better to warn, and load the module anyway. This may let through a typo, but at least the logs will show it.

#2 lore raw
Quoted 10 lines · Lucas De Marchi

Hi Benjamin,

Quoted 3 lines · Benjamin Block

Hi,

I have a rather strange problem currently. I am trying to load a module

strange indeed

Quoted 8 lines
Quoted 4 lines

with some parameters (bnx2x with disable_tpa=1) and it fails because this parameter is unknown according to dmesg. But modinfo lists this exact paramter and it is also included in the source. And while this is not an in-tree driver, I could also replicate the same behavior with

are you sure modinfo and modprobe are reading the same module? maybe you updated your modules between the operations?

pretty sure. It shows the exact path to the .ko, because this is a out-of-tree driver that is not located under /lib/modules/.. And in case of the bonding-example, there is no other module anywhere else with the same name.

Quoted 27 lines
Quoted 9 lines

in-tree device drivers (like bonding with the parameter tx_queues):

$ modprobe bonding # works fine
$ rmmod bonding
$ modprobe bonding tx_queues=32
modprobe: ERROR: could not insert 'bonding': Unknown symbol in module,
or unknown parameter (see dmesg)
Quoted 6 lines

and dmesg tells me:
bonding: Unknown parameter `tx_queues'
and nothing else (as with bnx2x, this parameter is listed by modinfo).

Is there some obvious kernel-config-option I could have messed up or something else like this?

So... this changed in recent kernels (3.11 -- see the commit message that introduced this behavior below)... we don't fail to load a module anymore due to unknown option. So if/when you upgrade you won't see this problem anymore.

Anyway, it shouldn't fail for you like it did, but it might as well be a problem of the particularly patched kernel you are using... not sure.

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

I am using a Gentoo-Userspace kmod (v15) with a Ubuntu-Kernel (3.8.13.13
- this is a version with some ubuntu-patches applied) - I know, the combination is rather strange. Any ideas?

the combination is not a problem and should work just fine.

I am now certain that this is a specific and/or combination of kernel-configuration-options which is messing this up here. I have a debug- and a release-configuration with the only difference that the debug-config contains several kernel-hacking-features enabled that the release doesn't include (compiled with symbols, lockdep, tracing and stuff like this). And in the release-config I could load the module with the parameter given.

When I'm at work again I will figure out which config messes this up exactly (I don't have access to the machines right now).

- Benjamin