Re: matroxfb: cannot determine memory size

[Helge Deller <deller@gmx.de>]

On 12/19/22 12:58, Paul Menzel wrote:
>
> Dear Helge,
>
>
> Am 16.12.22 um 22:21 schrieb Helge Deller:
>> * Paul Menzel <pmenzel@molgen.mpg.de>:
>>> [Cc: +regressions@, +stable@]
>>>
>>> #regzbot ^introduced: 62d89a7d49afe46e6b9bbe9e23b004ad848dbde4
>
>>> Am 16.12.22 um 00:02 schrieb Helge Deller:
>>>> On 12/15/22 17:39, Paul Menzel wrote:
>>>
>>>>> Between Linux 5.10.103 and 5.10.110/5.15.77, matrixfb fails to load.
>
> […]
>
>>>>> ### 5.15.77
>>>>>
>>>>>       [    0.000000] Linux version 5.15.77.mx64.440 (root@theinternet.molgen.mpg.de) (gcc (GCC) 10.4.0, GNU ld (GNU Binutils) 2.37) #1 SMP Tue Nov 8 15:42:33 CET 2022
>>>>>       [    0.000000] Command line: root=LABEL=root ro crashkernel=64G-:256M console=ttyS0,115200n8 console=tty0 init=/bin/systemd audit=0 random.trust_cpu=on systemd.unified_cgroup_hierarchy
>>>>>       […]
>>>>>       [    0.000000] DMI: Dell Inc. PowerEdge R715/0G2DP3, BIOS 1.5.2 04/19/2011
>>>>>       […]
>>>>>       [    9.436420] matroxfb: Matrox MGA-G200eW (PCI) detected
>>>>>       [    9.444502] matroxfb: cannot determine memory size
>>>>>       [    9.449316] matroxfb: probe of 0000:0a:03.0 failed with error -1
>>>>>
>>>>> We see it on several systems:
>>>>>
>>>>>       $ lspci -nn -s 0a:03.0 # Dell PowerEdge R715
>>>>>       0a:03.0 VGA compatible controller [0300]: Matrox Electronics Systems Ltd. MGA G200eW WPCM450 [102b:0532] (rev 0a)
>>>>>
>>>>>       $ lspci -nn -s 09:03.0 # Dell PowerEdge R910
>>>>>       09:03.0 VGA compatible controller [0300]: Matrox Electronics Systems Ltd. MGA G200eW WPCM450 [102b:0532] (rev 0a)
>
> Also Dell PowerEdge R815.
>
> […]
>
>>> I tested Linus’ master with commit 84e57d292203 (Merge tag
>>> 'exfat-for-6.2-rc1' of
>>> git://git.kernel.org/pub/scm/linux/kernel/git/linkinjeon/exfat) and the
>>> error is still there. Reverting commit fixes the issue.
>>>
>>> Tested on:
>>>
>>>      DMI: Dell Inc. PowerEdge R910/0KYD3D, BIOS 2.10.0 08/29/2013
>>>
>>> Current master:
>>>
>>>      [   36.221595] matroxfb 0000:09:03.0: vgaarb: deactivate vga console
>>>      [   36.228355] Console: switching to colour dummy device 80x25
>>>      [   36.234069] matroxfb: Matrox MGA-G200eW (PCI) detected
>>>      [   36.239316] PInS memtype = 7
>>>      [   36.242198] matroxfb: cannot determine memory size
>>>      [   36.242209] matroxfb: probe of 0000:09:03.0 failed with error -1
>>>
>>> After reverting 62d89a7d49af (video: fbdev: matroxfb: set maxvram of
>>> vbG200eW to the same as vbG200 to avoid black screen):
>>>
>>>      [   38.140763] matroxfb 0000:09:03.0: vgaarb: deactivate vga console
>>>      [   38.148057] Console: switching to colour dummy device 80x25
>>>      [   38.153789] matroxfb: Matrox MGA-G200eW (PCI) detected
>>>      [   38.159042] PInS memtype = 7
>>>      [   38.161953] matroxfb: 640x480x8bpp (virtual: 640x13107)
>>>      [   38.167175] matroxfb: framebuffer at 0xC5000000, mapped to 0x000000006f41c38c, size 8388608
>>>
>>>>> The master commit 62d89a7d49a was added to v5.18-rc1, and was also
>>>>> backported to the Linux 5.15 series in 5.15.33.
>>
>> Good.
>>
>> Could you test if the patch below works for you as well (on top of
>> git master) ? I believe the commit f8bf19f7f311 (video: fbdev:
>> matroxfb: set maxvram of vbG200eW to the same as vbG200 to avoid
>> black screen) changed the wrong value...
>
>> diff --git a/drivers/video/fbdev/matrox/matroxfb_base.c b/drivers/video/fbdev/matrox/matroxfb_base.c
>> index 0d3cee7ae726..5192c7ac459a 100644
>> --- a/drivers/video/fbdev/matrox/matroxfb_base.c
>> +++ b/drivers/video/fbdev/matrox/matroxfb_base.c
>> @@ -1378,8 +1378,8 @@ static struct video_board vbG200 = {
>>       .lowlevel = &matrox_G100
>>   };
>>   static struct video_board vbG200eW = {
>> -    .maxvram = 0x100000,
>> -    .maxdisplayable = 0x800000,
>> +    .maxvram = 0x800000,
>> +    .maxdisplayable = 0x100000,
>>       .accelID = FB_ACCEL_MATROX_MGAG200,
>>       .lowlevel = &matrox_G100
>>   };
>
> Thank you. That worked.
>
>      $ dmesg | grep -e matroxfb -e "Linux version" -e "DMI:"
>      [    0.000000] Linux version 6.1.0.mx64.440-13147-gfa99506bedb1 (pmenzel@dontpanic.molgen.mpg.de) (gcc (GCC) 11.1.0, GNU ld (GNU Binutils) 2.37) #1 SMP PREEMPT_DYNAMIC Mon Dec 19 12:13:21 CET 2022
>      [    0.000000] DMI: Dell Inc. PowerEdge R815/04Y8PT, BIOS 3.4.0 03/23/2018
>      [   29.033666] matroxfb 0000:0a:03.0: vgaarb: deactivate vga console
>      [   29.046608] matroxfb: Matrox MGA-G200eW (PCI) detected
>      [   29.054769] matroxfb: 640x480x8bpp (virtual: 640x1638)
>      [   29.059901] matroxfb: framebuffer at 0xE4000000, mapped to 0x00000000d36c9776, size 8388608
>      [   34.917829] matroxfb: Pixel PLL not locked after 5 secs
>      $ lspci -nn -s 0a:03.0
>      0a:03.0 VGA compatible controller [0300]: Matrox Electronics Systems Ltd. MGA G200eW WPCM450 [102b:0532] (rev 0a)
>
>> If it works, can you send a patch?
>
> Will do.

Great.

> If you have some explanation though, I could add to the commit message, that’d be great.

Look at the comment in the code, a few lines further down, for the vbG400 card:

/* from doc it looks like that accelerator can draw only to low 16MB :-( Direct accesses & displaying are OK for
    whole 32MB */
static struct video_board vbG400 = {
         .maxvram = 0x2000000,
         .maxdisplayable = 0x1000000,

it makes sense to have maxdisplayable smaller than maxvram.
But Z Liu made maxvram lower than maxdisplayable.

Helge