No network after S3

No network after S3

[Peter.Zeipelt-zqRNUXuvxA0b1SvskN2V4Q]

Hello list

I hope this is the right place for my question.
Using an ABIT KT7-RAID with SuSE 8.2prof and the kernel 2.6.1 or 2.6.2
STR seems to work fine.
But trying to use network nothing really happens.
I tried out several NICs (3c905c, 100PRO/S and one with a 8139b).
When i try to ping another  host in the local net, all get is an
Destination Host unreachable.
The 8139b says "no buffer space available" .
A dmesg with the 3c905 says "IRQ blocked by another device?"
But there isn't any other device on this IRQ.
Un- and reloading the module doesn't help.
Trying the STR on an ASUS A7N8X with driver for the nforce NIC in the
kernel brings the same result.
Another strange thing to me is, that S3 and S4 are working better on
the old ABIT  board with Via KT133
chipset then on my newer nforce boards.
The ASUS didn't fall in S4 mode, cause it seems that the disks are
powered down before writing the RAM
on it.
An Enmic 8RDA+ doesn't boot after S3.
But for now my main problem is to bring the network to work after the
S3.
Any hints?

Thanks
Peter


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Len Brown]

Peter,
Same NIC works with pci=noacpi or acpi=off?
If yes, then this is an ACPI bug.  If no, then the problem lies
elsewhere.

I'd defer dealing with the suspend/resume issues until more basic
functionality is working properly.

thanks,
-Len


On Sat, 2004-02-14 at 09:39, Peter.Zeipelt-zqRNUXuvxA0b1SvskN2V4Q@public.gmane.org wrote:
> Hello list
> 
> I hope this is the right place for my question.
> Using an ABIT KT7-RAID with SuSE 8.2prof and the kernel 2.6.1 or 2.6.2
> STR seems to work fine.
> But trying to use network nothing really happens.
> I tried out several NICs (3c905c, 100PRO/S and one with a 8139b).
> When i try to ping another  host in the local net, all get is an
> Destination Host unreachable.
> The 8139b says "no buffer space available" .
> A dmesg with the 3c905 says "IRQ blocked by another device?"
> But there isn't any other device on this IRQ.
> Un- and reloading the module doesn't help.
> Trying the STR on an ASUS A7N8X with driver for the nforce NIC in the
> kernel brings the same result.
> Another strange thing to me is, that S3 and S4 are working better on
> the old ABIT  board with Via KT133
> chipset then on my newer nforce boards.
> The ASUS didn't fall in S4 mode, cause it seems that the disks are
> powered down before writing the RAM
> on it.
> An Enmic 8RDA+ doesn't boot after S3.
> But for now my main problem is to bring the network to work after the
> S3.
> Any hints?
> 
> Thanks
> Peter
> 
> 
> -------------------------------------------------------
> SF.Net is sponsored by: Speed Start Your Linux Apps Now.
> Build and deploy apps & Web services for Linux with
> a free DVD software kit from IBM. Click Now!
> http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
> _______________________________________________
> Acpi-devel mailing list
> Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/acpi-devel



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Len Brown]

dunno why i didn't realize that you were talking about the NIC failing
after resuming from S3, guess I didn't read the subject line;-)

Yes, we've got a number of bugzilla entries that show that somehow
interrupts are toast after resume.  Perhaps this failure is the same as
those?

-Len

On Sun, 2004-02-15 at 00:36, Len Brown wrote:
> Peter,
> Same NIC works with pci=noacpi or acpi=off?
> If yes, then this is an ACPI bug.  If no, then the problem lies
> elsewhere.
> 
> I'd defer dealing with the suspend/resume issues until more basic
> functionality is working properly.
> 
> thanks,
> -Len
> 
> 
> On Sat, 2004-02-14 at 09:39, Peter.Zeipelt-zqRNUXuvxA0b1SvskN2V4Q@public.gmane.org wrote:
> > Hello list
> > 
> > I hope this is the right place for my question.
> > Using an ABIT KT7-RAID with SuSE 8.2prof and the kernel 2.6.1 or 2.6.2
> > STR seems to work fine.
> > But trying to use network nothing really happens.
> > I tried out several NICs (3c905c, 100PRO/S and one with a 8139b).
> > When i try to ping another  host in the local net, all get is an
> > Destination Host unreachable.
> > The 8139b says "no buffer space available" .
> > A dmesg with the 3c905 says "IRQ blocked by another device?"
> > But there isn't any other device on this IRQ.
> > Un- and reloading the module doesn't help.
> > Trying the STR on an ASUS A7N8X with driver for the nforce NIC in the
> > kernel brings the same result.
> > Another strange thing to me is, that S3 and S4 are working better on
> > the old ABIT  board with Via KT133
> > chipset then on my newer nforce boards.
> > The ASUS didn't fall in S4 mode, cause it seems that the disks are
> > powered down before writing the RAM
> > on it.
> > An Enmic 8RDA+ doesn't boot after S3.
> > But for now my main problem is to bring the network to work after the
> > S3.
> > Any hints?
> > 
> > Thanks
> > Peter
> > 
> > 
> > -------------------------------------------------------
> > SF.Net is sponsored by: Speed Start Your Linux Apps Now.
> > Build and deploy apps & Web services for Linux with
> > a free DVD software kit from IBM. Click Now!
> > http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
> > _______________________________________________
> > Acpi-devel mailing list
> > Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> > https://lists.sourceforge.net/lists/listinfo/acpi-devel
> 
> 
> 
> -------------------------------------------------------
> SF.Net is sponsored by: Speed Start Your Linux Apps Now.
> Build and deploy apps & Web services for Linux with
> a free DVD software kit from IBM. Click Now!
> http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
> _______________________________________________
> Acpi-devel mailing list
> Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/acpi-devel



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Peter.Zeipelt-zqRNUXuvxA0b1SvskN2V4Q]

Len,

thank `s for your advice.
I tried pci=noacpi. No diference.
If I understand you right, I should wait a little bit before dealing
with this
powersaving stuff.
You mentioned the bugzilla entries.
Can you tell me please where I can find them?
So I can see, if the problems there are similar to mine.

Peter

Len Brown schrieb:
> dunno why i didn't realize that you were talking about
> the NIC failing
> after resuming from S3, guess I didn't read the subject
> line;-)
> 
> Yes, we've got a number of bugzilla entries that show
> that somehow
> interrupts are toast after resume.  Perhaps this failure
> is the same as
> those?
> 
> -Len
> 
> On Sun, 2004-02-15 at 00:36, Len Brown wrote:
> > Peter,
> > Same NIC works with pci=noacpi or acpi=off?
> > If yes, then this is an ACPI bug.  If no, then the
> problem lies
> > elsewhere.
> > 
> > I'd defer dealing with the suspend/resume issues until
> more basic
> > functionality is working properly.
> > 
> > thanks,
> > -Len
> > 
> > 
> > On Sat, 2004-02-14 at 09:39, Peter.Zeipelt-zqRNUXuvxA0b1SvskN2V4Q@public.gmane.org
> wrote:
> > > Hello list
> > > 
> > > I hope this is the right place for my question.
> > > Using an ABIT KT7-RAID with SuSE 8.2prof and the
> kernel 2.6.1 or 2.6.2
> > > STR seems to work fine.
> > > But trying to use network nothing really happens.
> > > I tried out several NICs (3c905c, 100PRO/S and one
> with a 8139b).
> > > When i try to ping another  host in the local net,
> all get is an
> > > Destination Host unreachable.
> > > The 8139b says "no buffer space available" .
> > > A dmesg with the 3c905 says "IRQ blocked by another
> device?"
> > > But there isn't any other device on this IRQ.
> > > Un- and reloading the module doesn't help.
> > > Trying the STR on an ASUS A7N8X with driver for the
> nforce NIC in the
> > > kernel brings the same result.
> > > Another strange thing to me is, that S3 and S4 are
> working better on
> > > the old ABIT  board with Via KT133
> > > chipset then on my newer nforce boards.
> > > The ASUS didn't fall in S4 mode, cause it seems that
> the disks are
> > > powered down before writing the RAM
> > > on it.
> > > An Enmic 8RDA+ doesn't boot after S3.
> > > But for now my main problem is to bring the network
> to work after the
> > > S3.
> > > Any hints?
> > > 
> > > Thanks
> > > Peter
> > > 
> > > 
> > >
> -------------------------------------------------------
> > > SF.Net is sponsored by: Speed Start Your Linux Apps
> Now.
> > > Build and deploy apps & Web services for Linux with
> > > a free DVD software kit from IBM. Click Now!
> > >
> http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
> > > _______________________________________________
> > > Acpi-devel mailing list
> > > Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> > >
> https://lists.sourceforge.net/lists/listinfo/acpi-devel
> > 
> > 
> > 
> >
> -------------------------------------------------------
> > SF.Net is sponsored by: Speed Start Your Linux Apps
> Now.
> > Build and deploy apps & Web services for Linux with
> > a free DVD software kit from IBM. Click Now!
> > http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
> > _______________________________________________
> > Acpi-devel mailing list
> > Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> >
> https://lists.sourceforge.net/lists/listinfo/acpi-devel
> 
>


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

RE: No network after S3

[Brown, Len]

>I tried pci=noacpi. No diference.
>If I understand you right, I should wait a little bit before dealing
>with this powersaving stuff.

If the NIC were now working before attempting suspend/resume, then we'd
need to get that fixed first.  However, I think you're saying that the
NIC _does_ work under normal conditions, and the failure is only after
returning from S3. yes?

So this seems suspend/resume specific, and your use of pci=noacpi
further suggests it has nothing to do with the initial interrupt
assignments.

We've got a number of post-resume failures, and this issue is rising on
the priority list -- so I expect we'll spend some time looking at it
soon.

Here's a bugzilla query of component=ACPI with the word "suspend".  I
apologize in advance if outlook wraps this url on you...

http://bugzilla.kernel.org/buglist.cgi?short_desc_type=allwordssubstr&sh
ort_desc=&component=ACPI&long_desc_type=allwordssubstr&long_desc=suspend
&kernel_version_type=allwordssubstr&kernel_version=&bug_status=NEW&bug_s
tatus=ASSIGNED&bug_status=RESOLVED&emailassigned_to1=1&emailtype1=substr
ing&email1=&emailassigned_to2=1&emailreporter2=1&emailcc2=1&emailtype2=s
ubstring&email2=&bugidtype=include&bug_id=&changedin=&chfieldfrom=&chfie
ldto=Now&chfieldvalue=&cmdtype=doit&namedcmd=component%3DACPI&newqueryna
me=&order=Reuse+same+sort+as+last+time&field0-0-0=noop&type0-0-0=noop&va
lue0-0-0=


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id\x1356&alloc_id438&op=click

Re: No network after S3

[Peter.Zeipelt-zqRNUXuvxA0b1SvskN2V4Q]

Yes, it's right.
The mentioned NIC's are working under normal condition.
Of course, the first ping took almost a second.
But that isn't such a great problem for me,  that I have to look for
the reason now.
But again.
Before S3 the NIC is working.
Unloading the module before and reloading it after  suspending 
brings no success.
Even if the module isn't loaded before S3 and loaded after the wakeup
makes no difference.
And the problem is only after S3.
If I send the system to S4 and wake it up, the NIC is working.
But for the things I want to do (i.e. Fax-Services and answering
machine) the boottime
after S4 is to long.
And as mentioned in my first posting.
It makes no difference, if the NIC-driver is a module or compiled into
the kernel.
And the problem isn't specific on the ABIT-board with via chipset.
The same problem appears on an ASUS board with nforce2.
Unfortunately if have no modern board with  Intel chipset.

I hope, these informations are useful.
Many thanks for your help

Peter

Brown, Len schrieb:
> >I tried pci=noacpi. No diference.
> >If I understand you right, I should wait a little bit
> before dealing
> >with this powersaving stuff.
> 
> If the NIC were now working before attempting
> suspend/resume, then we'd
> need to get that fixed first.  However, I think you're
> saying that the
> NIC _does_ work under normal conditions, and the failure
> is only after
> returning from S3. yes?
> 
> So this seems suspend/resume specific, and your use of
> pci=noacpi
> further suggests it has nothing to do with the initial
> interrupt
> assignments.
> 
> We've got a number of post-resume failures, and this
> issue is rising on
> the priority list -- so I expect we'll spend some time
> looking at it
> soon.
> 
> Here's a bugzilla query of component=ACPI with the word
> "suspend".  I
> apologize in advance if outlook wraps this url on you...
> 
>
http://bugzilla.kernel.org/buglist.cgi?short_desc_type=allwordssubstr&sh
>
ort_desc=&component=ACPI&long_desc_type=allwordssubstr&long_desc=suspend
>
&kernel_version_type=allwordssubstr&kernel_version=&bug_status=NEW&bug_s
>
tatus=ASSIGNED&bug_status=RESOLVED&emailassigned_to1=1&emailtype1=substr
>
ing&email1=&emailassigned_to2=1&emailreporter2=1&emailcc2=1&emailtype2=s
>
ubstring&email2=&bugidtype=include&bug_id=&changedin=&chfieldfrom=&chfie
>
ldto=Now&chfieldvalue=&cmdtype=doit&namedcmd=component%3DACPI&newqueryna
>
me=&order=Reuse+same+sort+as+last+time&field0-0-0=noop&type0-0-0=noop&va
> lue0-0-0=
> 
> 
> -------------------------------------------------------
> SF.Net is sponsored by: Speed Start Your Linux Apps Now.
> Build and deploy apps & Web services for Linux with
> a free DVD software kit from IBM. Click Now!
> http://ads.osdn.com/?ad_id\x1356&alloc_id438&oplick
> _______________________________________________
> Acpi-devel mailing list
> Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/acpi-devel
>


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Thorsten Zachmann]

Hello 
On Monday 16 February 2004 07:10, Brown, Len wrote:
> >I tried pci=noacpi. No diference.
> >If I understand you right, I should wait a little bit before dealing
> >with this powersaving stuff.
>
> If the NIC were now working before attempting suspend/resume, then we'd
> need to get that fixed first.  However, I think you're saying that the
> NIC _does_ work under normal conditions, and the failure is only after
> returning from S3. yes?
>

It looks like I have the same problem on my Tecra S1. After a suspend (S3) the 
network and sound are no longer working. 
A ethereal trace shows that the interface still sends arp requests but is 
unable to receive something. If I look at /proc/interrupts the counter for 
the interrupt where the network card is connected is no longer incremented.

> So this seems suspend/resume specific, and your use of pci=noacpi
> further suggests it has nothing to do with the initial interrupt
> assignments.

pci=noacpi also didn't change the situation.

> We've got a number of post-resume failures, and this issue is rising on
> the priority list -- so I expect we'll spend some time looking at it
> soon.

I tried out the patches from bug report 1409 and 1661 but nothing changed.

Please let me know if there is any information that I can provide to track the 
problem down?

Thorsten



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Len Brown]

On Mon, 2004-02-16 at 08:17, Thorsten Zachmann wrote:
> Hello 
> On Monday 16 February 2004 07:10, Brown, Len wrote:
> > >I tried pci=noacpi. No diference.
> > >If I understand you right, I should wait a little bit before dealing
> > >with this powersaving stuff.
> >
> > If the NIC were now working before attempting suspend/resume, then we'd
> > need to get that fixed first.  However, I think you're saying that the
> > NIC _does_ work under normal conditions, and the failure is only after
> > returning from S3. yes?
> >
> 
> It looks like I have the same problem on my Tecra S1. After a suspend (S3) the 
> network and sound are no longer working. 
> A ethereal trace shows that the interface still sends arp requests but is 
> unable to receive something. If I look at /proc/interrupts the counter for 
> the interrupt where the network card is connected is no longer incremented.
> 
> > So this seems suspend/resume specific, and your use of pci=noacpi
> > further suggests it has nothing to do with the initial interrupt
> > assignments.
> 
> pci=noacpi also didn't change the situation.
> 
> > We've got a number of post-resume failures, and this issue is rising on
> > the priority list -- so I expect we'll spend some time looking at it
> > soon.
> 
> I tried out the patches from bug report 1409 and 1661 but nothing changed.
> 
> Please let me know if there is any information that I can provide to track the 
> problem down?

need /proc/interrupts -- PIC or APIC mode?  network interrupt shared
with ACPI, other devices, or dedicated interrupt?  We'll probably have
to dump out the state of the interrupt controller after resume.  for
IOAPIC we can do that with print_IO_APIC.  for PIC mode we'll need
something else.

-Len




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Thorsten Zachmann]

[-- Attachment #1: Type: text/plain, Size: 844 bytes --]

Hello Len,

On Tuesday 17 February 2004 07:53, Len Brown wrote:
> >
> > Please let me know if there is any information that I can provide to
> > track the problem down?
>
> need /proc/interrupts -- PIC or APIC mode?  network interrupt shared
> with ACPI, other devices, or dedicated interrupt?  

I think I use APIC mode. The interrupts for the network are not shared with 
another device as far as I can tell. /proc/interrupts only show eth0 on 
interrupt 6.

I have attached a file containing the output of /proc/interrupts, dmesg 
and /var/log/boot.msg.

> We'll probably have 
> to dump out the state of the interrupt controller after resume.  for
> IOAPIC we can do that with print_IO_APIC.  for PIC mode we'll need
> something else.

just tell me what I have to do. Where do I have to call print_IO_APIC?

Thanks for your help.

Thorsten

[-- Attachment #2: files.tar.gz --]
[-- Type: application/x-tgz, Size: 7958 bytes --]

Re: No network after S3

[Peter.Zeipelt-zqRNUXuvxA0b1SvskN2V4Q]

Hi Len

My  first mails concerned interrupts shared with ACPI
So I did some further testing with manually controlled
resources. 
Same result.
An interesting thing is, that the cheapest NIC an AT2500 with
realtek 8139b, did some pings after S3.
At first nothing happens after pressing enter.
Suddenly 4lines with a reply from the other host appeared.
Then I had to wait some seconds and several new lines 
with a reply appeared.
But each time, these lines appered "at once"  and not in the usual
"ping rythm".
After that , in the usual "ping rythm" ping told me
ping : sendmsg: No buffer space available
Here's my /proc/interrupts

           CPU0       
0:      1569738          XT-PIC  timer
1:          527          XT-PIC  i8042
2:            0          XT-PIC  cascade
5:            0          XT-PIC  acpi
6:           37          XT-PIC  floppy
8:            2          XT-PIC  rtc
10:           4          XT-PIC  HiSax
11:          19          XT-PIC  eth0
14:       23376          XT-PIC  ide0
15:          42          XT-PIC  ide1
NMI:          0
LOC:    1569739
ERR:       1528
MIS:          0

I don't know, if it's useful for you, but here are
some lines from /var/log/messages

Feb 17 22:01:26 penti kernel: PM: Preparing system for suspend
Feb 17 22:01:26 penti kernel: Stopping tasks:
==============================|
Feb 17 22:01:26 penti kernel: hdd: start_power_step(step: 0)
Feb 17 22:01:26 penti kernel: hdd: start_power_step(step: 1)
Feb 17 22:01:26 penti kernel: hdd: complete_power_step(step: 1, stat:
50, err: 0)
Feb 17 22:01:26 penti kernel: hdd: completing PM request, suspend
Feb 17 22:01:26 penti kernel: hdc: start_power_step(step: 0)
Feb 17 22:01:26 penti kernel: hdc: completing PM request, suspend
Feb 17 22:01:26 penti kernel: hda: start_power_step(step: 0)
Feb 17 22:01:26 penti kernel: hda: start_power_step(step: 1)
Feb 17 22:01:26 penti kernel: hda: complete_power_step(step: 1, stat:
50, err: 0)
Feb 17 22:01:26 penti kernel: hda: completing PM request, suspend
Feb 17 22:01:26 penti kernel: PM: Entering state.
Feb 17 22:01:26 penti kernel: Back to C!
Feb 17 22:01:26 penti kernel: PM: Finishing up.
Feb 17 22:01:26 penti kernel: eth0: link up, 100Mbps, full-duplex, lpa
0x45E1
Feb 17 22:01:26 penti kernel: hda: Wakeup request inited, waiting for
!BSY...
Feb 17 22:01:26 penti kernel: hda: start_power_step(step: 1000)
Feb 17 22:01:26 penti kernel: blk: queue c139ca00, I/O limit 4095Mb
(mask 0xffffffff)
Feb 17 22:01:26 penti kernel: hda: completing PM request, resume
Feb 17 22:01:26 penti kernel: hdc: Wakeup request inited, waiting for
!BSY...
Feb 17 22:01:26 penti kernel: hdc: start_power_step(step: 1000)
Feb 17 22:01:26 penti kernel: hdc: completing PM request, resume
Feb 17 22:01:26 penti kernel: hdd: Wakeup request inited, waiting for
!BSY...
Feb 17 22:01:26 penti kernel: hdd: start_power_step(step: 1000)
Feb 17 22:01:26 penti kernel: blk: queue c1396e00, I/O limit 4095Mb
(mask 0xffffffff)
Feb 17 22:01:26 penti kernel: hdd: completing PM request, resume
Feb 17 22:01:26 penti kernel: Restarting tasks... done
Feb 17 22:01:44 penti kernel: NETDEV WATCHDOG: eth0: transmit timed
out
Feb 17 22:01:44 penti kernel: eth0: Tx queue start entry 4  dirty entry
0.
Feb 17 22:01:44 penti kernel: eth0:  Tx descriptor 0 is 00002000.
(queue head)
Feb 17 22:01:44 penti kernel: eth0:  Tx descriptor 1 is 00002000.
Feb 17 22:01:44 penti kernel: eth0:  Tx descriptor 2 is 00002000.
Feb 17 22:01:44 penti kernel: eth0:  Tx descriptor 3 is 00002000.
Feb 17 22:01:44 penti kernel: eth0: link up, 100Mbps, full-duplex, lpa
0x45E1
Feb 17 22:01:56 penti kernel: NETDEV WATCHDOG: eth0: transmit timed
out
Feb 17 22:01:56 penti kernel: eth0: Tx queue start entry 4  dirty entry
0.
Feb 17 22:01:56 penti kernel: eth0:  Tx descriptor 0 is 00002000.
(queue head)
Feb 17 22:01:56 penti kernel: eth0:  Tx descriptor 1 is 00002000.
Feb 17 22:01:56 penti kernel: eth0:  Tx descriptor 2 is 00002000.
Feb 17 22:01:56 penti kernel: eth0:  Tx descriptor 3 is 00002000.
Feb 17 22:01:56 penti kernel: eth0: link up, 100Mbps, full-duplex, lpa
0x45E1

All these are books with seven seals for me.

I hope you can use this information.
What I haven't said before.
I tested the system with a well known commercial OS.
No problems with S3 or S4.
Btw: Are there linux-systems without problems?

Many thanks for your help and
good luck

Peter


Len Brown schrieb:
> On Mon, 2004-02-16 at 08:17, Thorsten Zachmann wrote:
> > Hello 
> > On Monday 16 February 2004 07:10, Brown, Len wrote:
> > > >I tried pci=noacpi. No diference.
> > > >If I understand you right, I should wait a little
> bit before dealing
> > > >with this powersaving stuff.
> > >
> > > If the NIC were now working before attempting
> suspend/resume, then we'd
> > > need to get that fixed first.  However, I think
> you're saying that the
> > > NIC _does_ work under normal conditions, and the
> failure is only after
> > > returning from S3. yes?
> > >
> > 
> > It looks like I have the same problem on my Tecra S1.
> After a suspend (S3) the 
> > network and sound are no longer working. 
> > A ethereal trace shows that the interface still sends
> arp requests but is 
> > unable to receive something. If I look at
> /proc/interrupts the counter for 
> > the interrupt where the network card is connected is no
> longer incremented.
> > 
> > > So this seems suspend/resume specific, and your use
> of pci=noacpi
> > > further suggests it has nothing to do with the
> initial interrupt
> > > assignments.
> > 
> > pci=noacpi also didn't change the situation.
> > 
> > > We've got a number of post-resume failures, and this
> issue is rising on
> > > the priority list -- so I expect we'll spend some
> time looking at it
> > > soon.
> > 
> > I tried out the patches from bug report 1409 and 1661
> but nothing changed.
> > 
> > Please let me know if there is any information that I
> can provide to track the 
> > problem down?
> 
> need /proc/interrupts -- PIC or APIC mode?  network
> interrupt shared
> with ACPI, other devices, or dedicated interrupt?  We'll
> probably have
> to dump out the state of the interrupt controller after
> resume.  for
> IOAPIC we can do that with print_IO_APIC.  for PIC mode
> we'll need
> something else.
> 
> -Len
> 
> 
> 
> 
> -------------------------------------------------------
> SF.Net is sponsored by: Speed Start Your Linux Apps Now.
> Build and deploy apps & Web services for Linux with
> a free DVD software kit from IBM. Click Now!
> http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
> _______________________________________________
> Acpi-devel mailing list
> Acpi-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
> https://lists.sourceforge.net/lists/listinfo/acpi-devel
>


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Thorsten Zachmann]

Hello Len,

> On Tuesday 17 February 2004 07:53, Len Brown wrote:
> > > Please let me know if there is any information that I can provide to
> > > track the problem down?
> >
> > need /proc/interrupts -- PIC or APIC mode?  network interrupt shared
> > with ACPI, other devices, or dedicated interrupt?
>
> I think I use APIC mode. The interrupts for the network are not shared with
> another device as far as I can tell. /proc/interrupts only show eth0 on
> interrupt 6.
>
> I have attached a file containing the output of /proc/interrupts, dmesg
> and /var/log/boot.msg.
>
> > We'll probably have
> > to dump out the state of the interrupt controller after resume.  for
> > IOAPIC we can do that with print_IO_APIC.  for PIC mode we'll need
> > something else.
>
> just tell me what I have to do. Where do I have to call print_IO_APIC?

I had a look in my kernel config and IO-APIC is enabled. Can you tell me where 
I have to call print_IO_APIC.  

is there any infomation I can send you, or any tests I can make to locate the 
source of the problem. I really like this bug fixed, as this is one of the 
features I need most on my notebook.

Thanks for you  help, have a nice day

Thorsten 


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: No network after S3

[Peter Meier]

<snip>

I'm having the same problem that my network is dead after I resume from
S3, but I'm getting some kind of (maybe?) helpfull error message:

b44: eth0: BUG!  Timeout waiting for bit 80000000 of register 428 to
clear.
b44: eth0: BUG!  Timeout waiting for bit 80000000 of register 428 to
clear.

I can rmmod b44 then, and modprobe it again, which leads to the
following:

b44.c:v0.92 (Nov 4, 2003)
PCI: Enabling device 0000:02:02.0 (0000 -> 0002)
PCI: Setting latency timer of device 0000:02:02.0 to 64
eh0: Broadcom 4400 10/100BaseT Ethernet ff:ff:ff:ff:ff:ff
eth0: PHY Reset would not complete.
b44: eth0: BUG!  Timeout waiting for bit 80000000 of register 428 to
clear. (several times)

So, somehow, the MAC address is incorrect. I can ping my local IP
address, but nothing else. There is no way to get the network card back
to live, except rebooting. If I rmmod b44 *before* going to sleep, and
modprobing it after I resume, I'm getting the same faulty MAC address
and no network.


Kernel is 2.6.3-mm3; I have IO-APIC enabled. cat /proc/interrupts gives
me the idea that the network card uses PIC.

On Tue, 2004-02-24 at 07:41, Thorsten Zachmann wrote:
> Hello Len,




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

b44 -- Re: No network after S3

[Len Brown]

On Tue, 2004-02-24 at 11:41, Peter Meier wrote:
> <snip>
> 
> I'm having the same problem that my network is dead after I resume
> from
> S3, but I'm getting some kind of (maybe?) helpfull error message:
> 
> b44: eth0: BUG!  Timeout waiting for bit 80000000 of register 428 to
> clear.
> b44: eth0: BUG!  Timeout waiting for bit 80000000 of register 428 to
> clear.

from b44.h:

#define B44_CAM_CTRL    0x0428UL /* EMAC CAM Control */
#define  CAM_CTRL_BUSY          0x80000000 /* CAM Busy */

looks like this bit is polled after writing to the CAM.
I have no idea what would cause the CAM busy bit to stay set forever.

> I can rmmod b44 then, and modprobe it again, which leads to the
> following:
> 
> b44.c:v0.92 (Nov 4, 2003)
> PCI: Enabling device 0000:02:02.0 (0000 -> 0002)
> PCI: Setting latency timer of device 0000:02:02.0 to 64
> eh0: Broadcom 4400 10/100BaseT Ethernet ff:ff:ff:ff:ff:ff
> eth0: PHY Reset would not complete.
> b44: eth0: BUG!  Timeout waiting for bit 80000000 of register 428 to
> clear. (several times)
> 
> So, somehow, the MAC address is incorrect. I can ping my local IP
> address, but nothing else. There is no way to get the network card
> back
> to live, except rebooting. If I rmmod b44 *before* going to sleep, and
> modprobing it after I resume, I'm getting the same faulty MAC address
> and no network.

apparently, the hardware is in a broken state, and re-loading the driver
doesn't seem to clear it.

It would be interesting to know if you can unload the driver before
going to S3, and then load it after resuming from S3.  ie. can the b44
hardware survive the S3 transition if it is quiescent?

> Kernel is 2.6.3-mm3; I have IO-APIC enabled. cat /proc/interrupts
> gives
> me the idea that the network card uses PIC.

system will be in XT-PIC mode or IO-APIC mode -- can't be both.  I
assume that /proc/interrupts shows this device receiving no interrupts
after the S3 resume?

thanks,
-Len





-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: b44 -- Re: No network after S3

[Pekka Pietikainen]

On Tue, Feb 24, 2004 at 05:00:24PM -0500, Len Brown wrote:
> > So, somehow, the MAC address is incorrect. I can ping my local IP
> > address, but nothing else. There is no way to get the network card
> > back
> > to live, except rebooting. If I rmmod b44 *before* going to sleep, and
> > modprobing it after I resume, I'm getting the same faulty MAC address
> > and no network.
I was recently able to reproduce a bug that sounds related, 
http://bugme.osdl.org/show_bug.cgi?id=2030 has the details.
Would be interesting to know whether the "fix" I found (running 
_only_ the initialization part of the bcm4400 driver) helps in this case too.

Broadcoms changelog provides some clues about what kind of access patters
make the chip break.

Current guesses are ssb_* (maybe ssb_is_core_up returning 1 
even when it isn't leading to ssb_pci_setup not getting called)
and b44_setup_phy (where the register access pattern differs slightly).
It didn't work even after altering the register access patterns to be ~= 
identical, making sure ssb_pci_setup gets run etc., alas, but I 
probably missed something.

Anyway, I'm working on it, but no guarantees on when I'll find the time to
nail it down.


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: b44 -- Re: No network after S3

[Peter Meier]

Hi,

that fix didnt work for me. Here is what I did:
I downloaded the broadcom driver 3.0.7 and installed the modules. I
rebootet, put the machine to sleep, and got follwing error message after
wake-up:

ASSERT failed: b44_LM_sb_coreid(pDevice) == SB_PCI
ASSERT failed: b44_LM_sb_coreid(pDevice) == SB_PCI
SBTMH_SERR; clearing...
ASSERT failed: 0
ASSERT failed: 0

Then, I rebootet, rmmod the driver before going to sleep, and modprobed
it after wake-up. Leads to follwing error message:

Broadcom 4401 Ethernet Driver bcm4400 ver. 3.0.7 (10/31/03)
PCI: Enabling device 0000:02:02.0 (0000 -> 0002)
PCI: Setting latency timer of device 0000:02:02.0 to 64
Get Adapter info failed

I then changed the line 417 in b44lm.c as suggested in bug id 2030.
Still, nothing changed. After playing around, I finally decied to get
back to my "normal" b44 - and had that bug described. Booting into XP
didnt help, and your "fix" described didnt work, too. Only thing that
solved the problem was disconnecting power and shutdown the system.

On Tue, 2004-02-24 at 23:34, Pekka Pietikainen wrote:
> I was recently able to reproduce a bug that sounds related, 
> http://bugme.osdl.org/show_bug.cgi?id=2030 has the details.
> Would be interesting to know whether the "fix" I found (running 
> _only_ the initialization part of the bcm4400 driver) helps in this case too.
> 
> Broadcoms changelog provides some clues about what kind of access patters
> make the chip break.
> 
> Current guesses are ssb_* (maybe ssb_is_core_up returning 1 
> even when it isn't leading to ssb_pci_setup not getting called)
> and b44_setup_phy (where the register access pattern differs slightly).
> It didn't work even after altering the register access patterns to be ~= 
> identical, making sure ssb_pci_setup gets run etc., alas, but I 
> probably missed something.
> 
> Anyway, I'm working on it, but no guarantees on when I'll find the time to
> nail it down.




-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click

Re: b44 -- Re: No network after S3

[Peter Meier]

> It would be interesting to know if you can unload the driver before
> going to S3, and then load it after resuming from S3.  ie. can the b44
> hardware survive the S3 transition if it is quiescent?

No, that does not help. The same error message.

> > Kernel is 2.6.3-mm3; I have IO-APIC enabled. cat /proc/interrupts
> > gives
> > me the idea that the network card uses PIC.
> 
> system will be in XT-PIC mode or IO-APIC mode -- can't be both.  I
> assume that /proc/interrupts shows this device receiving no interrupts
> after the S3 resume?

Yeah, /proc/interrupts shows XT-PIC mode.



-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click