Re: [gentoo-user] Loading a Firmware Module By hand?

[Floyd Anderson]

On Mon, 18 Dec 2017 05:11:20 +
Hunter Jozwiak  wrote:

On 12/17/17, Andrey Utkin  wrote:

On Sun, Dec 17, 2017 at 12:34:14AM -0500, Hunter Jozwiak wrote:

Hi,

I have an ath10k_pci device that I'm trying to get hooked to the
Internet, but I'm having some strange issues. It is trying to load the
2.1 firmware, but I don't think that is the proper firmware for the
interface to have; I think it ought to be loading the 3.0 module, but
am not quite sure on that either, or how I could go about injecting
that into the modprobe; I wasn't able to pinpoint the firmware blob
the ISO was using, so that wasn't much of a pointer in the right
direction either. I see that the 3.0 blob does exist in
/lib/firmware/ath10k/QCABLEFAGD/HW3.0, but there are many bin files,


I have little to no idea about your actual case... But could it be that
you have a recent linux-firmware package (which provides /lib/firmware/
files) and not recent enough kernel? I think kernel is what decides
which firmware file to load.





!!! rearranged top-posting !!!


Hmm. I have kernel 4.14.7 and linux-firmware 20171206. I tried version
9 as well, but that didn't help matters, either. Nor did
compiling the firmware into the kernel; either 4.14 is too old, or it
is too new. I tried copying the firmware my live iso was using, but
that didn't help either.



I think you are a little bit too vague in your given info. If you don’t 
show your firmware related kernel settings (those lines posted by Mick 
earlier) nor what dmesg said about the firmware loading success of your 
specified config, people tends to think you know what you are doing and 
therefore may eliminate any errors based on syntactical mistakes or 
similar from their thoughts.


For instance, you wrote “It is trying to load the 2.1 firmware” but 
because trying != loaded successfully, nobody knows if 2.1 works or has 
been failed (and why), so you have a need to try the 3.0 blob.


As Mick pointed out, look what is in your dmesg log and communicate that 
(not only your own interpretation of it). Maybe there’s another module 
configured that also supports and loads the 2.1 blob so it must be 
blacklisted [1] or not built at all. Also the firmware must be available 
on boot and module load, so a probably used initramfs must include it as 
long as it is not built into the kernel.



[1] 


--
Regards,
floyd

Re: [gentoo-user] Loading a Firmware Module By hand?

[Hunter Jozwiak]

Hmm. I have kernel 4.14.7 and linux-firmware 20171206. I tried version
9 as well, but that didn't help matters, either. Nor did
compiling the firmware into the kernel; either 4.14 is too old, or it
is too new. I tried copying the firmware my live iso was using, but
that didn't help either.

On 12/17/17, Andrey Utkin  wrote:
> On Sun, Dec 17, 2017 at 12:34:14AM -0500, Hunter Jozwiak wrote:
>> Hi,
>>
>> I have an ath10k_pci device that I'm trying to get hooked to the
>> Internet, but I'm having some strange issues. It is trying to load the
>> 2.1 firmware, but I don't think that is the proper firmware for the
>> interface to have; I think it ought to be loading the 3.0 module, but
>> am not quite sure on that either, or how I could go about injecting
>> that into the modprobe; I wasn't able to pinpoint the firmware blob
>> the ISO was using, so that wasn't much of a pointer in the right
>> direction either. I see that the 3.0 blob does exist in
>> /lib/firmware/ath10k/QCABLEFAGD/HW3.0, but there are many bin files,
>
> I have little to no idea about your actual case... But could it be that
> you have a recent linux-firmware package (which provides /lib/firmware/
> files) and not recent enough kernel? I think kernel is what decides
> which firmware file to load.
>

Re: [gentoo-user] Choice of TLD for internal network

[R0b0t1]

On Sun, Dec 17, 2017 at 8:05 PM, Peter Humphrey  wrote:
> Hello list,
>
> I've been running Linux systems since 1994, calling my private LAN mynet
> (bowdlerised). Now I come to install neth server on one machine, it insists
> that I tell it a domain name with at least two dots in it. But I don't have
> a standard TLD.
>
> What do you all call your local LANs? Following Google hints, it looks as
> though I may have to change all .mynet references to .mynet.internal. Is
> this really necessary, and is it a good idea?
>
> I can't possibly be the first to stumble over this one, surely.
>

What is the exact message? Search the codebase for it, and if it is
some hardcoded check I would submit a bug report.

I would call them "name" similar to you. I know of no special reason
to call them anything else.

Cheers,
 R0b0t1

[gentoo-user] Choice of TLD for internal network

[Peter Humphrey]

Hello list,

I've been running Linux systems since 1994, calling my private LAN mynet 
(bowdlerised). Now I come to install neth server on one machine, it insists 
that I tell it a domain name with at least two dots in it. But I don't have 
a standard TLD.

What do you all call your local LANs? Following Google hints, it looks as 
though I may have to change all .mynet references to .mynet.internal. Is 
this really necessary, and is it a good idea?

I can't possibly be the first to stumble over this one, surely.

-- 
Regards,
Peter.
Linux Counter 5290

Re: [gentoo-user] cross compiling arm with 17 profiles.

[R0b0t1]

Hello,

On Sun, Dec 17, 2017 at 8:14 AM, Mart Raudsepp  wrote:
> On P, 2017-12-17 at 16:50 +0800, Bill Kenworthy wrote:
>> Something I cant figure out:
>>
>> ARM is still on the 13 profiles - should an amd64 system used to
>> cross
>> compile for arm (Raspberry Pi's) be left on the 13 profiles or 17
>> will
>> work fine?
>
> ARM profiles are delayed to potentially fix CHOSTs together with the
> profile update. Though no-one is actively doing the work to my
> knowledge right now.
>
> I guess it could cause trouble from default PIE vs no PIE from native
> compiler, but I don't know enough about that field to know for sure.
>

If you know anything at all that is more than myself, so can you link
to past discussions that you are aware of?

> If you pay attention to any future CHOST changes and handle them
> yourself at the right time, you could manually choose the appropriate
> 17.0 arm profile as your symlink (it doesn't show up in eselect profile
> due to no profiles.desc entry, but should be there in profiles/). If
> changes are done, you might be caught a bit off-guard though at the
> time they are done though and I'm not sure what the effects of that
> would be either (probably not too bad).
>

My experience with ARM(64) is that it is mature enough that you can
expect @system to work unless proven otherwise. Lots of other packages
have failures.

Cheers,
 R0b0t1

Re: [gentoo-user] Re: Again, emerge -e @world related questions...

[Michael Orlitzky]

On 12/17/2017 06:45 AM, Melleus wrote:

> After some digging in the files I commented out v3.4 line in
> /etc/python-exec/python-exec.conf by hand and eselect then begins to
> work as I expect. The question is that I think that I should not edit
> that file by hand. So is it a bug or might I done something wrong?
> 
> Thank you for pointing me in right direction.

This is being discussed in

  https://bugs.gentoo.org/635678

and

  https://bugs.gentoo.org/639578

It looks like the resolution has two parts. First, eselect was updated
to mark uninstalled pythons as being uninstalled. So I think they'll
still be shown in the list, but it will be obvious that you shouldn't
select them.

The second phase of the fix would be to update that python-exec.conf
file automatically, when the last version of e.g. python-3.4.x is
removed. I don't think that's been done yet.

Either way, it's a bug and nothing you did wrong.

Re: [gentoo-user] Canary Pies

[Michael Orlitzky]

On 12/16/2017 10:43 PM, tu...@posteo.de wrote:
> Hi,
> 
> Currently I am scanning directories of my system with checksec to
> identify relevant files of haveing "No PIE" or "No canary found" set.
> 
> Is there any technical reason for which such files cannot be compiled
> in a way so they have "PIE" and "Canary found" set ?

Some packages with hand-written assembly will fail to compile with the
stack-smashing protection enabled. That should be rare, though. For PIE
I'm not sure.


> How "dangerous" is that ?

Not very, but it depends on the package. If it's a game, who cares. If
it's a library used by firefox, you probably want the extra protection.

Re: [gentoo-user] cross compiling arm with 17 profiles.

[Mart Raudsepp]

On P, 2017-12-17 at 16:50 +0800, Bill Kenworthy wrote:
> Something I cant figure out:
> 
> ARM is still on the 13 profiles - should an amd64 system used to
> cross
> compile for arm (Raspberry Pi's) be left on the 13 profiles or 17
> will
> work fine?

ARM profiles are delayed to potentially fix CHOSTs together with the
profile update. Though no-one is actively doing the work to my
knowledge right now.

I guess it could cause trouble from default PIE vs no PIE from native
compiler, but I don't know enough about that field to know for sure.

If you pay attention to any future CHOST changes and handle them
yourself at the right time, you could manually choose the appropriate
17.0 arm profile as your symlink (it doesn't show up in eselect profile
due to no profiles.desc entry, but should be there in profiles/). If
changes are done, you might be caught a bit off-guard though at the
time they are done though and I'm not sure what the effects of that
would be either (probably not too bad).

[gentoo-user] Re: Again, emerge -e @world related questions...

[Melleus]

Michael Orlitzky  writes:

> On 12/08/2017 09:53 AM, Melleus wrote:
>> I had moved to v 17.0 profile mostly painless, though it was a time
>> consuming event. But I got one point anyway. Python in my system was
>> updated from 3.4 to 3.5 and after 3.4 was removed with depclean, the
>> option for v 3.4 in eselect python remains. It looks a bit weird to me
>> when I can choose with eselect the version of python that is not
>> currently present in the system. Is this intended behavior?
>
> Guessing: no. (What happens if you select it?)
It selects, but when attempting to run Python it falls back to v3.5

> There might be some python-3.4 stuff left on your system that tricks
> eselect into thinking that python-3.4 is installed. For example, in
> eselect-php we do,
>
>   find_targets() {
> cd "@LIBDIR@" && echo php*.*
>   }
>
> and that is easily fooled by creating any file in /usr/lib/php-x.y.
I could not find any remnants of Python v3.4 in my system. Though I'm
more academician than a IT guru. I moved to Gentoo as the last
mainstream distribution free from systemd. I like its flexibility, but
the maintenance of a binary distribution would be much less burden for
me and for my quite old hardware.

> You might have to dig through eselect-python to see how it works, or ask
> somebody who knows.
After some digging in the files I commented out v3.4 line in
/etc/python-exec/python-exec.conf by hand and eselect then begins to
work as I expect. The question is that I think that I should not edit
that file by hand. So is it a bug or might I done something wrong?

Thank you for pointing me in right direction.

Regards,
Anatoly.

Re: [gentoo-user] Loading a Firmware Module By hand?

[Andrey Utkin]

On Sun, Dec 17, 2017 at 12:34:14AM -0500, Hunter Jozwiak wrote:
> Hi,
> 
> I have an ath10k_pci device that I'm trying to get hooked to the
> Internet, but I'm having some strange issues. It is trying to load the
> 2.1 firmware, but I don't think that is the proper firmware for the
> interface to have; I think it ought to be loading the 3.0 module, but
> am not quite sure on that either, or how I could go about injecting
> that into the modprobe; I wasn't able to pinpoint the firmware blob
> the ISO was using, so that wasn't much of a pointer in the right
> direction either. I see that the 3.0 blob does exist in
> /lib/firmware/ath10k/QCABLEFAGD/HW3.0, but there are many bin files,

I have little to no idea about your actual case... But could it be that
you have a recent linux-firmware package (which provides /lib/firmware/
files) and not recent enough kernel? I think kernel is what decides
which firmware file to load.


signature.asc
Description: Digital signature

Re: [gentoo-user] Loading a Firmware Module By hand?

[Mick]

On Sunday, 17 December 2017 05:34:14 GMT Hunter Jozwiak wrote:
> Hi,
> 
> I have an ath10k_pci device that I'm trying to get hooked to the
> Internet, but I'm having some strange issues. It is trying to load the
> 2.1 firmware, but I don't think that is the proper firmware for the
> interface to have; I think it ought to be loading the 3.0 module, but
> am not quite sure on that either, or how I could go about injecting
> that into the modprobe; 

Usually the kernel will probe the device and load the appropriate firmware the 
device needs.  The output of dmesg will show if loading the firmware was 
successful, or complain if different firmware was needed and not found in your 
"/lib/firmware/" or whatever you have set up in your kernel as the directory 
for firmware blobs.


> I wasn't able to pinpoint the firmware blob
> the ISO was using, so that wasn't much of a pointer in the right
> direction either. I see that the 3.0 blob does exist in
> /lib/firmware/ath10k/QCABLEFAGD/HW3.0, but there are many bin files,
> so choosing the right one is a bit tricky by the looks of things.
> Earlier today, I had read the Gentoo wiki on the topic, which
> suggested that I compile the blob into the kernel itself, but the link
> they gave only described the advantages and disadvantages of modular
> kernels and how to drop kmod if you're using a moduleless kernel. Does
> anybody have any insight on this matter?
> 
> Thanks,
> 
> Hunter

Have a look in:

# Generic Driver Options

and then:

CONFIG_PREVENT_FIRMWARE_BUILD=y
CONFIG_FW_LOADER=y
CONFIG_FIRMWARE_IN_KERNEL=y
CONFIG_EXTRA_FIRMWARE="Put your firmware files in here, space separated"
CONFIG_EXTRA_FIRMWARE_DIR="/lib/firmware/"

I don't have your NIC to know the specifics, but the above ought to work.
-- 
Regards,
Mick

signature.asc
Description: This is a digitally signed message part.

[gentoo-user] cross compiling arm with 17 profiles.

[Bill Kenworthy]

Something I cant figure out:

ARM is still on the 13 profiles - should an amd64 system used to cross
compile for arm (Raspberry Pi's) be left on the 13 profiles or 17 will
work fine?

BillK