No networking after resume from suspend

[Richmond]

From: Richmond <rich...@criptext.com>
Newsgroups: linux.debian.user
Subject: No networking after resume from suspend
Date: Thu, 22 Apr 2021 22:34:07 +0100
Organization: Frantic
Message-ID: <84r1j2k...@example.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)
X-Draft-From: ("linux.debian.user")
X-No-Archive: No
Cancel-Lock: sha1:f+poUKVPx2E0cnaVpAmVy2aJg84=
--text follows this line--
When I resume from suspend there is no networking. I cannot find a way
to restart it. I tried these various commands.

systemctl restart network
/etc/init.d/networking restart
systemctl reset-failed
systemctl restart networking.service
systemctl restart network-online.target
systemctl restart network-manager.service
systemctl start network-manager.service
systemctl stop network-manager.service

lshw says:

description: Ethernet interface
product: 88E8071 PCI-E Gigabit Ethernet Controller
vendor: Marvell Technology Group Ltd.

What can I do?

[Dan Ritter]

Richmond wrote:
> When I resume from suspend there is no networking. I cannot find a way
> to restart it. I tried these various commands.
>
> systemctl restart network
> /etc/init.d/networking restart
> systemctl reset-failed
> systemctl restart networking.service
> systemctl restart network-online.target
> systemctl restart network-manager.service
> systemctl start network-manager.service
> systemctl stop network-manager.service
>
> lshw says:
>
> description: Ethernet interface
> product: 88E8071 PCI-E Gigabit Ethernet Controller
> vendor: Marvell Technology Group Ltd.

Let's try from the bottom up?

ip link show
will show you the interfaces recognized by the kernel. If this
works, it might show you an eth0, an en0, or something like a
enp22s0 device. Let's call it "SAM".

If any part of SAM reads "down", do this:

sudo ip link set up SAM

and make sure the cable is plugged in.

Now let's set an address.

ip addr show SAM
that might show you an IPv4 address. If not, or if you don't
think it looks right for your network, you can set one by
hand:

sudo ip addr set 192.168.0.43/24 dev SAM

or you can ask DHCP for one:

sudo dhclient SAM

Finally, try pinging your gateway router.

Let us know how far you get with this, and exactly what the
error messages are.

-dsr-

[Cindy Sue Causey]

On 4/22/21, Richmond <rich...@criptext.com> wrote:
> When I resume from suspend there is no networking. I cannot find a way
> to restart it. I tried these various commands.
>
> systemctl restart network
> /etc/init.d/networking restart
> systemctl reset-failed
> systemctl restart networking.service
> systemctl restart network-online.target
> systemctl restart network-manager.service
> systemctl start network-manager.service
> systemctl stop network-manager.service
>
> lshw says:
>
> description: Ethernet interface
> product: 88E8071 PCI-E Gigabit Ethernet Controller
> vendor: Marvell Technology Group Ltd.
>

> What can I do?


Questions where answers might help come to mind. Primarily, has this
always occurred, or did it just start up in the last couple days?

I'm on a new old secondhand laptop today. Just started using it
yesterday. It's doing similar, but I can trigger it back on via
"wicd-curses". A quick CTRL+R (refresh) for no particular reason and
then CTRL+C (connect), and I'm back up and running.

WHEN it happens again, I'm going to try to remember to try your
various commands to see if I receive similar results. I'm going to
copy those right now so they're available offline when it's time.

A PS to this is that a couple of us were having a reconnect problem a
while back. Mine would just suddenly stop and then refuse to
reconnect. Always took a reboot (and sometimes 2) before it would
start working again.

PPS What's your eth0 being called? Mine's a moving target depending on
which (older) laptop I'm using. I've used 4 different names in the
last two weeks.

Two of those renames were the long version that Users also see for
wireless dongles. That was appropriate because I was using two
different ethernet to USB port adapters. Those long values are tied to
the specific products we're using.

The other two values now flip between "enp1s0" and "eno1". Those are
both for internal (onboard) ports. So far, it appears to be that those
stay specific to two different laptops. I like that eno1 one. I STILL
can't confidently remember that other naming pattern every time it
comes up.

Those values are entered under Preferences in the wicd-curses
interface. I had a problem earlier today where it wouldn't connect
back up. I had just learned that /etc/wicd/manager-settings.conf is
relevant so I peeked at it.

That value was wrong. It didn't refresh after I used the terminal
interface. I changed it, and mine connected back up by itself as fast
as that changed file was saved.

Which brings me to ANOTHER observation through all of this. Sometimes
sitting here counting to 10 or 15 seconds helps after clicking once to
reconnect.

That was an accidental find, a "patience pays" kind of deal. I'm
presuming that Users having similar experiences may be doing the same
thing I did.

My old method was to keep repeatedly clicking a reconnect button every
couple of seconds then finally close the program in frustration. These
last few days, just sitting here staring at the screen while counting
seconds gives it time to churn out a successful connection.

Cindy :)
--
Cindy-Sue Causey
Talking Rock, Pickens County, Georgia, USA

* runs with birdseed *

[Richmond]

Thanks for your reply.

Cindy Sue Causey <butterf...@gmail.com> writes:

> Questions where answers might help come to mind. Primarily, has this
> always occurred, or did it just start up in the last couple days?

It has occured since installing debian (10). Prior to that I was using
opensuse.

>
> I'm on a new old secondhand laptop today. Just started using it
> yesterday. It's doing similar, but I can trigger it back on via
> "wicd-curses". A quick CTRL+R (refresh) for no particular reason and
> then CTRL+C (connect), and I'm back up and running.

I don't think I am using Wicked, if that is the equivalent of the
opensuse Wicked. I think I am using network manager.

There is nothing in

/etc/network/interfaces.d/

I haven't messed around with things I just installed debian 10 with Mate
and it mostly works.

[Richmond]

Dan Ritter <d...@randomstring.org> writes:

> Richmond wrote:
>> When I resume from suspend there is no networking. I cannot find a way
>> to restart it. I tried these various commands.
>>
>> systemctl restart network
>> /etc/init.d/networking restart
>> systemctl reset-failed
>> systemctl restart networking.service
>> systemctl restart network-online.target
>> systemctl restart network-manager.service
>> systemctl start network-manager.service
>> systemctl stop network-manager.service
>>
>> lshw says:
>>
>> description: Ethernet interface
>> product: 88E8071 PCI-E Gigabit Ethernet Controller
>> vendor: Marvell Technology Group Ltd.
>
> Let's try from the bottom up?
>
> ip link show
> will show you the interfaces recognized by the kernel. If this
> works, it might show you an eth0, an en0, or something like a
> enp22s0 device. Let's call it "SAM".

Thanks for your reply.

enp2s0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000

I omited the second line which contained what looks like a mac address.

>
> If any part of SAM reads "down", do this:
>
> sudo ip link set up SAM

ip link set up enp2s0

No output here but the link is still down.

>
> and make sure the cable is plugged in.

:)

>
> Now let's set an address.
>
> ip addr show SAM
> that might show you an IPv4 address. If not, or if you don't
> think it looks right for your network, you can set one by
> hand:

No address here, it repeats that the link is down.

>
> sudo ip addr set 192.168.0.43/24 dev SAM
>
> or you can ask DHCP for one:
>
> sudo dhclient SAM
>

This paused for a long time but didn't set an address.

So I will try to set one manually as above...

It didn't like 'set', so I used 'add'

The address it is supposed to be is 192.168.1.43, you came close to
guessing it!

ip addr add 192.168.1.43/24 dev enp2s0

This produced a message in the GUI saying a connection was established
to enp2s0 but the link is still down.

> Finally, try pinging your gateway router.
>
> Let us know how far you get with this, and exactly what the
> error messages are.
>

ping 192.168.1.1
PING 192.168.1.1 (192.168.1.1) 56(84) bytes of data.
>From 192.168.1.43 icmp_seq=1 Destination Host Unreachable
... etc

I tried this before the other commands too, and got this:

ping 192.168.1.1
connect: Network is unreachable

[Andrei POPESCU]

On Vi, 23 apr 21, 10:46:21, Richmond wrote:

> Dan Ritter <d...@randomstring.org> writes:
> >
> > If any part of SAM reads "down", do this:
> >
> > sudo ip link set up SAM
>
> ip link set up enp2s0
>
> No output here but the link is still down.

This suggests to me the problem is at a lower level, i.e. the kernel
module (driver).

Try unloading / re-loading it ('modprobe -r' and 'modprobe').

You should be able to find the module in use with 'lspci -k', possibly
with '-nn' as well to help identify the correct device[1], or by
examining the kernel log (dmesg).

[1] can't test this as the hardware in use here doesn't have PCI buses.

Kind regards,
Andrei
--
http://wiki.debian.org/FAQsFromDebianUser

[Dan Ritter]

Richmond wrote:
> > Let's try from the bottom up?
> >
> > ip link show
> > will show you the interfaces recognized by the kernel. If this
> > works, it might show you an eth0, an en0, or something like a
> > enp22s0 device. Let's call it "SAM".
>
> Thanks for your reply.
>
> enp2s0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast state DOWN mode DEFAULT group default qlen 1000

> > and make sure the cable is plugged in.
>
> :)

Ah, you laugh, but: the NO-CARRIER...UP... state DOWN means that
the kernel recognizes the device and there's a cable issue.

Go trace the cable, replug in both ends, and if that doesn't
work, replace the cable. If the far end has a free port, try
swapping to a different port - this one might be toasted.

-dsr-

[Richmond]

Looks like this bug:

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1798921

[Dan Ritter]

Richmond wrote:
> Looks like this bug:
>
> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1798921

Could be.

In which case, this will be solved in the new Stable, most
likely out in May or June.

You could try a backports kernel before that.

-dsr-

[David Wright]

On Fri 23 Apr 2021 at 10:49:00 (+0100), Richmond wrote:
> Cindy Sue Causey <butterf...@gmail.com> writes:
>
> > Questions where answers might help come to mind. Primarily, has this
> > always occurred, or did it just start up in the last couple days?
>
> It has occured since installing debian (10). Prior to that I was using
> opensuse.
>
> > I'm on a new old secondhand laptop today. Just started using it
> > yesterday. It's doing similar, but I can trigger it back on via
> > "wicd-curses". A quick CTRL+R (refresh) for no particular reason and
> > then CTRL+C (connect), and I'm back up and running.
>
> I don't think I am using Wicked, if that is the equivalent of the
> opensuse Wicked. I think I am using network manager.

JFTR there's no connection (pun, sorry) between Wicd¹ and Wicked.²
Confusion is compounded by their identical pronunciation.

Wicked would claim to be a superset of NetworkManager. Wicd is a
simple connection manager, handling just one connection at a time.

In any case, wicd's days are numbered unless and until it moves
forward from Python2. There's a Python3 version in experimental,
but the hiatus might not be good for sustaining its popularity.

¹ acronym, Wireless Interface Connection Daemon

² "No REST for the wicked", motivational joke after SUSE rejected
using REST, a REpresentational State Transfer interface.

Cheers,
David.

[Cindy Sue Causey]

On 4/23/21, David Wright <deb...@lionunicorn.co.uk> wrote:
> On Fri 23 Apr 2021 at 10:49:00 (+0100), Richmond wrote:
>> Cindy Sue Causey <butterf...@gmail.com> writes:
>>
>> > Questions where answers might help come to mind. Primarily, has this
>> > always occurred, or did it just start up in the last couple days?
>>
>> It has occured since installing debian (10). Prior to that I was using
>> opensuse.
>>
>> > I'm on a new old secondhand laptop today. Just started using it
>> > yesterday. It's doing similar, but I can trigger it back on via
>> > "wicd-curses". A quick CTRL+R (refresh) for no particular reason and
>> > then CTRL+C (connect), and I'm back up and running.
>>
>> I don't think I am using Wicked, if that is the equivalent of the
>> opensuse Wicked. I think I am using network manager.
>
> JFTR there's no connection (pun, sorry) between Wicd¹ and Wicked.²
> Confusion is compounded by their identical pronunciation.
>
> Wicked would claim to be a superset of NetworkManager. Wicd is a
> simple connection manager, handling just one connection at a time.
>
> In any case, wicd's days are numbered unless and until it moves
> forward from Python2. There's a Python3 version in experimental,
> but the hiatus might not be good for sustaining its popularity.

Experimental is the version I'm using. That was an empowering moment
there. It helped me finally understand all the chatter about backports
and the irreplaceable value to Users looking for missing programs that
didn't make it in time to be included in any given release.

> ¹ acronym, Wireless Interface Connection Daemon
>
> ² "No REST for the wicked", motivational joke after SUSE rejected
> using REST, a REpresentational State Transfer interface.

Thank you! I always wondered but never pursued what the wicd acronym
was. I assumed it was an acronym, anyway, because of that "d" that
appears in other packages, too.

#1 I'm so sorry, I made a major mistake last night. To operate
wicd-curses, the commands are SHIFT+R for refresh and SHIFT+C to
connect. There's no CTRL used there. That was the first thought I had
this morning after waking up.

The tl;dr for the rest of this is that systemctl worked for me this
morning, but I'm using wicd instead of e.g. network-manager. They're
both still available on Debian because there are differences in how
they function. They don't appear to be duplicate overkill by both
being installable. My memory recall is that they don't play nice
together.

Ok, so I tried those various systemctl lines. When I actually read
them, I realized this isn't an apples to apples comparison. That
"network-manager" was the sign.

In addition to that, I'm on a wired connection in case this was about
wireless. For the moment, I can only assume wicd might handle wireless
connections similarly.

With respect to those terminal commands, I first tried wicd-curses
instead of network-manager. That didn't work because it's an
interface.

So I tried "systemctl restart wicd" but couldn't tell that anything
was happening beyond that it didn't throw any errors. I tried
"systemctl stop wicd", and it understandably killed my open
wicd-curses interface in the terminal.

Next, "systemctl start wicd" was entered. I reopened wicd-curses...
and it was already connected online without any further intervention.
It was cool to see that interaction, am just waiting for my brain to
have the ah-ha moment about exactly what happened.

In the meantime, that's my "learn something new" today. One my
earliest "thank you" posts about Debian was about the fact that there
are frequently at least two ways to accomplish many tasks. That kept
coming to mind over the last month of horrific hardware failures.

Several of those alternates were used repeatedly the last few weeks.
In particular, the right click on the desktop to access the
Applications menu was invaluable. I was working on a badly damaged
screen that blocked access to the usual method of clicking the word
"Applications" on the top left in XFCE4.

N.B. Apparently the dogs or I stepped on my Dell Duo when it was on
the floor beside me here. The leaking damage on the screen looked just
like a giant footprint, lol.

AND I just now remembered that I could have moved the desktop panel
that holds that Application menu. Hopefully there will never be a next
time where that needs to occur.

[Cindy Sue Causey]

Those ethernet to USB port adapter products have been irreplaceable
for me. The only ones that have NEVER worked are the ones that come as
a two-for-one package. They don't seem to work for most other Users,
either. I tested the one I have just a few days ago AGAIN. Zero
interaction.

Maybe I'll try it yet AGAIN today. I'd like to solve that so that it's
one more successful option for mainly poverty folks with computers.
The package I bought was surely under $10. I think I've probably paid
at most maybe $5 or $6 for the single ones.

All of the adapters that come as a single item have worked great. You
just have to track down the new identifier, e.g. one of mine is
"enx00909e9dd1ee". That long value goes wherever one normally types in
eth0, enp1s0, eno1, etc.

Dmesg often reports that value. I grep /var/log/kern* for the word
"renamed" because that's what I always remember first. Except that
then there's having to open kern.log in a text editor occasionally so
yeah, dmesg rocks if one can remember to go that route.

Grep works with dmesg, too. I just tested it. Only one line to read as
output. Very cognitively friendly! :)

[Richmond]

I tried kernels 5.9 and 5.10 but neither cured it. So perhaps I should report it
as a bug in debian?

[Dan Ritter]

To the kernel packagers group, yup.

-dsr-

[David Wright]

On Fri 23 Apr 2021 at 13:23:31 (-0400), Cindy Sue Causey wrote:
>
> All of the adapters that come as a single item have worked great. You
> just have to track down the new identifier, e.g. one of mine is
> "enx00909e9dd1ee". That long value goes wherever one normally types in
> eth0, enp1s0, eno1, etc.
>
> Dmesg often reports that value. I grep /var/log/kern* for the word
> "renamed" because that's what I always remember first. Except that
> then there's having to open kern.log in a text editor occasionally so
> yeah, dmesg rocks if one can remember to go that route.
>
> Grep works with dmesg, too. I just tested it. Only one line to read as
> output. Very cognitively friendly! :)

That seems like a lot of typing. You should be able to read your
interface name with 5 keystrokes:

$ ip a

and l(ink) will work in place of a(ddress).

OTOH my startup files set Mywiredifname for scripts to use, where:

Mywiredifname=$(ip -o link show | sed -e '/^[0-9]\+: [^e]/d;s/[0-9]\+: \([^:]\+\): .*/\1/;q')

$ echo $Mywiredifname
enp3s0
$

(My days of running multiple ethernet cards are long gone,
so sed will quit after one match.)

Cheers,
David.

[Greg Wooledge]

On Fri, Apr 23, 2021 at 09:59:09PM -0500, David Wright wrote:
> OTOH my startup files set Mywiredifname for scripts to use, where:
>
> Mywiredifname=$(ip -o link show | sed -e '/^[0-9]\+: [^e]/d;s/[0-9]\+: \([^:]\+\): .*/\1/;q')
>
> $ echo $Mywiredifname
> enp3s0

unicorn:~$ ip -o link show | awk -F': ' '{print $2}'
lo
lan0

You're relying on the "default" names which ensure that an ethernet
interface begins with an 'e', so we could fold that in as well:

ip -o link show | awk -F': ' '$2 ~ /^e/ {print $2}'

It's slightly prettier than yours, and doesn't rely on GNUisms (which are
probably not a big issue since a system with "ip" probably has GNU sed).

> (My days of running multiple ethernet cards are long gone,
> so sed will quit after one match.)

Sounds like a naive assumption. Some motherboards have dual NICs built in,
don't they?

[Dan Ritter]

A fair number of consumer motherboards, and then I have server
motherboards with up to five NICs each, though 3 is more common.

-dsr-

[David Wright]

On Fri 23 Apr 2021 at 23:11:58 (-0400), Greg Wooledge wrote:
> On Fri, Apr 23, 2021 at 09:59:09PM -0500, David Wright wrote:
> > OTOH my startup files set Mywiredifname for scripts to use, where:
> >
> > Mywiredifname=$(ip -o link show | sed -e '/^[0-9]\+: [^e]/d;s/[0-9]\+: \([^:]\+\): .*/\1/;q')
> >
> > $ echo $Mywiredifname
> > enp3s0
>
> unicorn:~$ ip -o link show | awk -F': ' '{print $2}'
> lo
> lan0

Yes —— though one might wonder why someone would use this little
script to discover the interface name on a system where, I assume,
they chose the name themselves.

> You're relying on the "default" names which ensure that an ethernet
> interface begins with an 'e', so we could fold that in as well:
>
> ip -o link show | awk -F': ' '$2 ~ /^e/ {print $2}'

One certainly has to eliminate 'w', at the very least, to eliminate
any wireless card. And {print $2; exit} guarantees only one match,
to correspond with my 'q'.

> It's slightly prettier than yours, and doesn't rely on GNUisms (which are
> probably not a big issue since a system with "ip" probably has GNU sed).
>
> > (My days of running multiple ethernet cards are long gone,
> > so sed will quit after one match.)
>
> Sounds like a naive assumption. Some motherboards have dual NICs built in,
> don't they?

I look forward to the day. I've got this far in life without ever
buying a computer, so I may well never see one. It /is/ an assumption,
hence my mentioning it. I don't want to overpromise—the substance of
my post was contained in 5 keystrokes: ip a
for Cindy to type, instead of greping kern.log and dmesg.

My assumption that all modern PCs would have one ethernet interface
has already fallen by the wayside—some (most?) laptops now have zero.

But thanks for the awk suggestions. All the (trivial) awk code
in my startup files is just copied from elsewhere. I think the reason
I never got into awk is simply that I came from mainframes/spitbol
via MSDOS/perl/grep/sed to linux/python.

Cheers,
David.

[Anssi Saari]

Sure you can do that. I'd assume you find behavior problematic so you
could also figure out a workaround. For example, if unloading and
reloading the sky2 module after suspend works (like in that Ubuntu bug
report), then why not just automate that?