[BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[mhf]

HW info

Using vga=0xf07, default8x16 font, display has 30 lines

On powerup from S3 console has only 25 lines but still scrolls 
at 30 lines. Setfont historically fixes it. 

Tested with 2.6.10, 2.6.11-rc1: OK

Tested with 2.6.11-rc2-Vanilla and 2.6.11-rc[234]+swsusp2.
When using setfont, screen goes blank. Power up after S3
returns console in 25 lines mode with 30 lines scroll. 
Several attempts - same result.

Another bug I see only on this HW and only with 2.6 is that
when - and only when - using gentoo emerge --usepackage in
text console, scroll area resets to _25_ when portage 
"dumps" the (binary) package contents which scrolls pretty
fast. I was unable to reproduce this in any other way. 
Tried also echo loop in bash but perhaps it is too slow
or not random enough. Note that 2.4.2[789] no problem.

Regards
Michael

[BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[mhf]

HW info

Hardware Celeron 433 with i810 chipset:

0000:00:00.0 Host bridge: Intel Corp. 82810E DC-133 GMCH [Graphics Memory C=
ontroller Hub] (rev 03)
=A0 =A0 =A0 =A0 Subsystem: Intel Corp. 82810E DC-133 GMCH [Graphics Memory =
Controller Hub]
=A0 =A0 =A0 =A0 Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-=
 ParErr- Stepping- SERR+ FastB2B-
=A0 =A0 =A0 =A0 Status: Cap- 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=3Dfast >TA=
bort- <TAbort-<MAbort+ >SERR- <PERR-
=A0 =A0 =A0 =A0 Latency: 0

0000:00:01.0 VGA compatible controller: Intel Corp. 82810E DC-133 CGC [Chip=
set Graphics Controller] (rev 03) (prog-if 00 [VGA])
=A0 =A0 =A0 =A0 Subsystem: FIRST INTERNATIONAL Computer Inc: Unknown device=
 9980
=A0 =A0 =A0 =A0 Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-=
 ParErr- Stepping- SERR- FastB2B-
=A0 =A0 =A0 =A0 Status: Cap+ 66Mhz+ UDF- FastB2B+ ParErr- DEVSEL=3Dmedium >=
TAbort- <TAbort- <MAbort- >SERR- <PERR-
=A0 =A0 =A0 =A0 Latency: 0
=A0 =A0 =A0 =A0 Interrupt: pin A routed to IRQ 9
=A0 =A0 =A0 =A0 Region 0: Memory at e8000000 (32-bit, prefetchable) [size=
=3D64M]
=A0 =A0 =A0 =A0 Region 1: Memory at eff80000 (32-bit, non-prefetchable) [si=
ze=3D512K]
=A0 =A0 =A0 =A0 Capabilities: [dc] Power Management version 1
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 Flags: PMEClk- DSI+ D1- D2- AuxCurrent=3D0m=
A PME(D0-,D1-,D2-,D3hot-,D3cold-)
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 Status: D0 PME-Enable- DSel=3D0 DScale=3D0 =
PME-

Software:

Gentoo current.

Using vga=3D0xf07, default8x16 font, display has 30 lines

On powerup from S3 console has only 25 lines but still scrolls=20
at 30 lines. Setfont historically fixes it.=20

Tested with 2.6.10, 2.6.11-rc1: OK

Tested with 2.6.11-rc2-Vanilla and 2.6.11-rc[234]+swsusp2.
When using setfont, screen goes blank. Power up after S3
returns console in 25 lines mode with 30 lines scroll.=20
Several attempts - same result.

Another bug I see only on this HW and only with 2.6 is that
when - and only when - using gentoo emerge --usepackage in
text console, scroll area resets to _25_ when portage=20
"dumps" the (binary) package contents which scrolls pretty
fast. I was unable to reproduce this in any other way.=20
Tried also echo loop in bash but perhaps it is too slow
or not random enough. Note that 2.4.2[789] no problem.

Regards
Michael

Re: [BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[Pavel Machek]

Hi!

> Any thoughts on this one?  We should come back from resume in 30-row mode,
> shouldn't we?

Well, current state of video resume is "we are happy to see anything
at all". 

> HW info
> 
> Using vga=0xf07, default8x16 font, display has 30 lines
> 
> On powerup from S3 console has only 25 lines but still scrolls 
> at 30 lines. Setfont historically fixes it. 
> 
> Tested with 2.6.10, 2.6.11-rc1: OK
> 
> Tested with 2.6.11-rc2-Vanilla and 2.6.11-rc[234]+swsusp2.
> When using setfont, screen goes blank. Power up after S3
> returns console in 25 lines mode with 30 lines scroll. 
> Several attempts - same result.

So... screen goes blank even when suspend is not involved, right?
Sounds like a bug to me ;-).

> Another bug I see only on this HW and only with 2.6 is that
> when - and only when - using gentoo emerge --usepackage in
> text console, scroll area resets to _25_ when portage 
> "dumps" the (binary) package contents which scrolls pretty
> fast. I was unable to reproduce this in any other way. 
> Tried also echo loop in bash but perhaps it is too slow
> or not random enough. Note that 2.4.2[789] no problem.

Well, dumping random stuff to console can produce funny results. I'd
call that normal. Try cat /dev/urandom, that should be "enough
random".

								Pavel
-- 
People were complaining that M$ turns users into beta-testers...
...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!

Re: [BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[James Courtier-Dutton]

Pavel Machek wrote:
> Hi!
> 
> 
>>Any thoughts on this one?  We should come back from resume in 30-row mode,
>>shouldn't we?
> 
> 
> Well, current state of video resume is "we are happy to see anything
> at all". 
> 
> 
>>HW info
>>
>>Using vga=0xf07, default8x16 font, display has 30 lines
>>
>>On powerup from S3 console has only 25 lines but still scrolls 
>>at 30 lines. Setfont historically fixes it. 
>>
>>Tested with 2.6.10, 2.6.11-rc1: OK
>>
>>Tested with 2.6.11-rc2-Vanilla and 2.6.11-rc[234]+swsusp2.
>>When using setfont, screen goes blank. Power up after S3
>>returns console in 25 lines mode with 30 lines scroll. 
>>Several attempts - same result.
> 
> 
> So... screen goes blank even when suspend is not involved, right?
> Sounds like a bug to me ;-).
> 
> 
>>Another bug I see only on this HW and only with 2.6 is that
>>when - and only when - using gentoo emerge --usepackage in
>>text console, scroll area resets to _25_ when portage 
>>"dumps" the (binary) package contents which scrolls pretty
>>fast. I was unable to reproduce this in any other way. 
>>Tried also echo loop in bash but perhaps it is too slow
>>or not random enough. Note that 2.4.2[789] no problem.
> 
> 
> Well, dumping random stuff to console can produce funny results. I'd
> call that normal. Try cat /dev/urandom, that should be "enough
> random".
> 
> 								Pavel

I am also getting strange effects. I boot into  2.6.11-rc4 and the 
console fonts looks fine. Come back a day later and the console font has 
  corrupt characters. E.g. Displays a "D" instead of an "L" and stuff 
like that. It is mostly readable, except for a few characters.
It is only the local console that is corrupted. ssh into the box 
displays correct characters, so all I can assume is that the VGA console 
is being programmed with different characters. The bad characters also 
survive a soft reboot( During BIOS boot up), until the linux kernel 
starts booting, and then it switches to a good font.

Re: [BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[James Courtier-Dutton]

Pavel Machek wrote:
> Hi!
> 
> 
>>Any thoughts on this one?  We should come back from resume in 30-row mode,
>>shouldn't we?
> 
> 
> Well, current state of video resume is "we are happy to see anything
> at all". 
> 
> 
>>HW info
>>
>>Using vga=0xf07, default8x16 font, display has 30 lines
>>
>>On powerup from S3 console has only 25 lines but still scrolls 
>>at 30 lines. Setfont historically fixes it. 
>>
>>Tested with 2.6.10, 2.6.11-rc1: OK
>>
>>Tested with 2.6.11-rc2-Vanilla and 2.6.11-rc[234]+swsusp2.
>>When using setfont, screen goes blank. Power up after S3
>>returns console in 25 lines mode with 30 lines scroll. 
>>Several attempts - same result.
> 
> 
> So... screen goes blank even when suspend is not involved, right?
> Sounds like a bug to me ;-).
> 
> 
>>Another bug I see only on this HW and only with 2.6 is that
>>when - and only when - using gentoo emerge --usepackage in
>>text console, scroll area resets to _25_ when portage 
>>"dumps" the (binary) package contents which scrolls pretty
>>fast. I was unable to reproduce this in any other way. 
>>Tried also echo loop in bash but perhaps it is too slow
>>or not random enough. Note that 2.4.2[789] no problem.
> 
> 
> Well, dumping random stuff to console can produce funny results. I'd
> call that normal. Try cat /dev/urandom, that should be "enough
> random".
> 
> 								Pavel

I am also getting strange effects. I boot into  2.6.11-rc4 and the
console fonts looks fine. Come back a day later and the console font has
  corrupt characters. E.g. Displays a "D" instead of an "L" and stuff
like that. It is mostly readable, except for a few characters.
It is only the local console that is corrupted. ssh into the box
displays correct characters, so all I can assume is that the VGA console
is being programmed with different characters. The bad characters also
survive a soft reboot( During BIOS boot up), until the linux kernel
starts booting, and then it switches to a good font.

Re: [BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[Pavel Machek]

Hi!

> >Well, dumping random stuff to console can produce funny results. I'd
> >call that normal. Try cat /dev/urandom, that should be "enough
> >random".
> 
> I am also getting strange effects. I boot into  2.6.11-rc4 and the
> console fonts looks fine. Come back a day later and the console font has
>  corrupt characters. E.g. Displays a "D" instead of an "L" and stuff
> like that. It is mostly readable, except for a few characters.
> It is only the local console that is corrupted. ssh into the box
> displays correct characters, so all I can assume is that the VGA console
> is being programmed with different characters. The bad characters also
> survive a soft reboot( During BIOS boot up), until the linux kernel
> starts booting, and then it switches to a good font.

I have seen something similar on S3 cards with bad video ram (we had
>3 of them). If it survives soft reboot... well, that looks like
hardware problem to me. [We may do something bad in linux, too, but
strange effects should not survive reboot, that's hw bug. I'd suggest
memtest on video ram, but someone would need to write that tool, first].

								Pavel
-- 
People were complaining that M$ turns users into beta-testers...
...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!

Re: [BUG] 2.6.11-rc[234] setfont fails on i810 after resume from ACPI-S3

[mhf]

On Sunday 27 February 2005 17:59, Pavel Machek wrote:
> Hi!
>
> > >Well, dumping random stuff to console can produce
> > > funny results. I'd call that normal. Try cat
> > > /dev/urandom, that should be "enough random".
> >
> > I am also getting strange effects. I boot into 
> > 2.6.11-rc4 and the console fonts looks fine. Come back
> > a day later and the console font has corrupt
> > characters. E.g. Displays a "D" instead of an "L" and
> > stuff like that. It is mostly readable, except for a
> > few characters. It is only the local console that is
> > corrupted. ssh into the box displays correct
> > characters, so all I can assume is that the VGA console
> > is being programmed with different characters. The bad
> > characters also survive a soft reboot( During BIOS boot
> > up), until the linux kernel starts booting, and then it
> > switches to a good font.
>
> I have seen something similar on S3 cards with bad video
> ram (we had
>
> >3 of them). If it survives soft reboot... well, that
> > looks like
>
> hardware problem to me. [We may do something bad in
> linux, too, but strange effects should not survive
> reboot, that's hw bug. I'd suggest memtest on video ram,
> but someone would need to write that tool, first].
>
>         Pavel

i810 uses system RAM (which was tested with memtest)

In my case, it also does work fine with 2.4.29, thus problem 
2.6 specific.


 Michael