Bug 11305

Created attachment 10367 [details]
X.org log

Note: I'm using a 8x8 font on the console, hence 80x50 text mode.

I am using a:
00:02.0 VGA compatible controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)
I am running:
xserver-xorg-video-intel_2.1.0-1 (2007-07-02)

I also happen to see the text cursor on the text console.

And if I log in xdm and change to text console while GNOME is loading it switches to a black screen with the hardware cursor on it.

I'm not able to reproduce this problem with upstream driver.

I am still having the problem with git version of this morning. Some more information that migh be important: I'm booting the kernel with vga=extended and use an 8x8 font. This is on a Dell Latitude X300 laptop with no external CRT, just the internal panel (if I plug an external CRT the problem doesn't appear).  00:02.0 VGA compatible controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02) 

How about if you don't use vga=extended?

Without vga=extended and loading a usual 8x16 fonts, the problem is still there. 

Created attachment 11258 [details]
xorg 7.3~rc1 intel 855gm driver 2.1.1 mouse cursor garbled on text console

Created attachment 11259 [details]
xorg 7.3~rc1 intel 855gm driver 2.1.1 mouse cursor garbled on text console X.org log

we still can't reproduce this bug....:(   please feel free to reopen if it still bothers you.

I've just upgraded my server to 1.4.0, pulled the intel driver git, and... I still can reproduce this bug....:( 

(pulled git up to bbf69c7446aa9748dacecccfe0b63d803f865b00)

There have been some VT patches committed in git recently. Could you retest with the git tip?

Seems to be fixed, great!

Created attachment 14243 [details]
xorg 7.3 intel 855gm driver 2.2.0.90 mouse cursor garbled on text console X.org log

I can see the same glitch in the text console with the 2.2.0.90-3 driver using X.Org 1.4.1~git20080131-1 from Debian.

I am reopening this bug as it seems it is not fixed.
I also sometimes see minor text console font corruption.

Same behavior here with git 8c8ca16120927fc2d3cf8bcd5212a4227a280708, same symptoms as before.

Samuel, do you mean it's fixed with git around 2007-11-15 (in comment#14), but re-appears with the latest git?

Yes.  When I have the time, I'll try to bisect it.

Samuel, any update?

I still don't have time to do this :)

I'll probably have it next week.

(In reply to comment #16)
> I am reopening this bug as it seems it is not fixed.
> I also sometimes see minor text console font corruption.
> 

Daniel, are you able to help us to confirm the latest status on git repository on your machine? thanks.

bisect showed that this bug was reintroduced in f69b48fe24ef94dac44b8123884ca71df675be4b

Ok, interesting.  Thanks for bisecting.  I'll have to double check the LeaveVT code and make sure we're doing everything in the right order for 855 chips...

Samuel, can you capture some register dumps with the latest git tree (assuming it still fails for you).  You can use the 'intel_reg_dumper' tool from src/reg_dumper in the tree to grab them.  I'd like to see:
  - at startup before X starts
  - after X starts
  - after VT switching back to a console from X
Also please attach the X log from all of this, with the "modedebug" option enabled in your intel driver section.

Thanks,
Jesse

Created attachment 15315 [details]
dump at startup before X starts

Created attachment 15316 [details]
dump after X starts

Created attachment 15317 [details]
dump after VT switching back to a console from X

Created attachment 15318 [details]
X log up to VT switch

Thanks.  I don't suppose this patch helps by itself?

--- a/src/i830_driver.c
+++ b/src/i830_driver.c
@@ -2024,7 +2024,7 @@ SaveHWState(ScrnInfoPtr pScrn)
    }

    vgaHWUnlock(hwp);
-   vgaHWSave(pScrn, vgaReg, VGA_SR_FONTS);
+   vgaHWSave(pScrn, vgaReg, VGA_SR_ALL);

    return TRUE;
 }
@@ -2233,7 +2233,7 @@ RestoreHWState(ScrnInfoPtr pScrn)
        OUTREG(FBC_CONTROL, pI830->saveFBC_CONTROL);
    }

-   vgaHWRestore(pScrn, vgaReg, VGA_SR_FONTS);
+   vgaHWRestore(pScrn, vgaReg, VGA_SR_ALL);
    vgaHWLock(hwp);

    return TRUE;

Created attachment 15319 [details]
dump after patched X starts

Created attachment 15320 [details]
dump after VT switching back to a console from patched X

Created attachment 15321 [details]
patched X log up to VT switch

Still the same symptoms yes.

Ok, thanks.  I went through our restore code again and it seems like we're following the rules, but I'm obviously missing something.  I'll check again.  Btw, do you have a picture of what your text console looks like when you switch to it?

Created attachment 15323 [details]
screenshot of the artifact

Here is a screenshot. Notice the cursor-shaped artifact at the bottom right of the screen, covering some of the 'a's I have typed.

I did see same screen like in comment #8, but can't reproduce it now.

Could you try current master tip or 2.2.99.901 driver?

Still there with c02ab432dd7058c700c35eecf6215daf5f262c51

(In reply to comment #36)
> Here is a screenshot. Notice the cursor-shaped artifact at the bottom right of
> the screen, covering some of the 'a's I have typed.

I also have the exact same artifact at the bottom right of the screen with current driver.
I also think this bug may be related to the random lock up of my laptop when I switch to the text console.
I would like to have this bug fixed. How may I help?

I don't have any new test patches yet, but once I do (hopefully next week) you can help out by testing.  Thanks.

Created attachment 16368 [details] [review]
Wait for vblank after disabling cursor

Can you give this patch a shot?  It fixes what I think are a couple of bugs in our VT switch code, and adds a vblank wait to the cursor disable path, which I think is necessary.

Created attachment 16371 [details]
Log with patch 16368

Still no luck :/

Still exactly the same behavior

Created attachment 16412 [details] [review]
Save/restore cursor state

Ok, I think we need to enable/disable both cursor planes on your machine for some reason.  Can you give this patch a try?  It should at least return things to the way we found them at VT switch/exit time...

Yes, it does work!

(while git tip a7188b1f2dd9a69fa7daefe478d283735226f9f3 does not)

Well that only took, what, about a year to fix? :)  Thanks for your patience and testing, Samuel, I pushed the fix as 33f033cbf346c13a687e469e8879579fcd5bb2fb.

The patch seems to fix the mouse pointer on text console problem. Great!

But I still see occasional text console minor font corruption.

Samuel, Could you confirm you are not seeing any text console minor font corruption?

It is quite rare and can be fixed by reloading the font, but I indeed
sometimes see that (it impacts a few letters only for instance).

(In reply to comment #48)
> It is quite rare and can be fixed by reloading the font, but I indeed
> sometimes see that (it impacts a few letters only for instance).

Since you can confirm it, should we reopen or file a new bug report? (I do not know if it is related to the original cause of this bug)

It looks to me like the font issue is not related to the cursor issue,
so a new bug should be opened.