Re: [Linux-fbdev-devel] atyfb rmmod & vram crash, 2.6.24, 2.6.25-rc8

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

On Thu, 3 Apr 2008 08:22:19 +0300 "Denys Fedoryshchenko" <de...@vi...> wrote:

> Well, additionally i am not able to manage this FB driver to work on my VGA,
> while uvesafb works fine (but probably slower). But it is probably sync
> problems with my TV-Out, i am still investigating it. But again, uvesafb works
> great.

OK.

> Btw just a bit offtopic question, do i need to report here about such issues,
> while i can find for myself another solution? Maybe better to not take time of
> developers for such not important things? Or better to "bugtest" kernel and to
> make it better together?

No, you did fine.  You forgot to cc the fbdev mailing list, but I fixed
that.  Thanks for helping.

> On Wed, 2 Apr 2008 14:55:54 -0700, Andrew Morton wrote
> > On Wed, 2 Apr 2008 15:54:12 +0300
> > "Denys Fedoryshchenko" <de...@vi...> wrote:
> > 
> > > VGA card i dont know how much RAM there actually, but lspci here shows looks
> > > like 16MB:
> > > 
> > > 03:09.0 VGA compatible controller: ATI Technologies Inc 3D Rage Pro 215GP (rev
> > > 5c) (prog-if 00 [VGA controller])
> > >         Subsystem: ATI Technologies Inc Rage Pro Turbo
> > >         Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr-
> > > Stepping+ SERR- FastB2B- DisINTx-
> > >         Status: Cap- 66MHz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort-
> > > <TAbort- <MAbort- >SERR- <PERR- INTx-
> > >         Latency: 66 (2000ns min), Cache Line Size: 32 bytes
> > >         Interrupt: pin A routed to IRQ 21
> > >         Region 0: Memory at f0000000 (32-bit, prefetchable) [size=16M]
> > >         Region 1: I/O ports at 2000 [size=256]
> > >         Region 2: Memory at d0121000 (32-bit, non-prefetchable) [size=4K]
> > >         Kernel driver in use: atyfb
> > > 
> > > 
> > > 
> > > Loading without parameters
> > > 
> > > atyfb: using auxiliary register aperture
> > > atyfb: 3D RAGE PRO (Mach64 GP, PQFP, PCI) [0x4750 rev 0x7c]
> > > atyfb: Mach64 BIOS is located at c0000, mapped at c00c0000.
> > > atyfb: BIOS frequency table:
> > > atyfb: PCLK_min_freq 984, PCLK_max_freq 23600, ref_freq 2950, ref_divider 64
> > > atyfb: MCLK_pwd 4200, MCLK_max_freq 7500, XCLK_max_freq 10000, SCLK_freq 5000
> > > atyfb: 8M SGRAM (1:1), 29.498928 MHz XTAL, 236 MHz PLL, 75 Mhz MCLK, 100
> MHz XCLK
> > > atyfb: fb0: ATY Mach64 frame buffer device on PCI
> > > 
> > > It detects 8M.
> > > If i put parameter vram=8
> > > atyfb: 3D RAGE PRO (Mach64 GP, PQFP, PCI) [0x4750 rev 0x7c]
> > > atyfb: Mach64 BIOS is located at c0000, mapped at c00c0000.
> > > atyfb: BIOS frequency table:
> > > atyfb: PCLK_min_freq 984, PCLK_max_freq 23600, ref_freq 2950, ref_divider 64
> > > atyfb: MCLK_pwd 4200, MCLK_max_freq 7500, XCLK_max_freq 10000, SCLK_freq 5000
> > > atyfb: 0M SGRAM (1:1), 29.498928 MHz XTAL, 236 MHz PLL, 75 Mhz MCLK, 100
> MHz XCLK
> > > atyfb: not enough video RAM
> > > atyfb: can't set default video mode
> > 
> > Why did this happen?  It seems from reading the code that the vram 
> > arg is in kilobytes, yes?
> Also from various sources, like Documentation/fb/intel810.txt
> <<< SNIP >>>
>    e. "vram:<value>"
>         select amount of system RAM in MB to allocate for the video memory
> 
>         Recommendation: 1 - 4 MB.
>         (default = 4)
> <<< SNIP >>>
> 
> I tried also 8388608, sure it doesn't work too. This parameter seems doesn't
> work at all. When i didn't specify it - it shows amount or RAM (which looks
> invalid), when i specify it any value (i try 2 or 8 or 16) - it shows 0MB and
> crashing.

OK.

> > 
> > > And most interesting when i do rmmod
> > > tv1 ~ # rmmod atyfb
> > > Segmentation fault
> > > 
> > > and dmesg:
> > > BUG: unable to handle kernel NULL pointer dereference at 00000448
> > > IP: [<ce8fd0f2>] :atyfb:aty_set_crtc+0x9/0x19d
> > > *pde = 00000000
> > > Oops: 0000 [#1] SMP
> > > Modules linked in: atyfb(-) fb backlight cfbcopyarea cfbimgblt cfbfillrect
> > > rtc_cmos rtc_core rtc_lib e100 mii snd_intel8x0 snd_ac97_codec ac97_bus
> > > snd_pcm snd_timer snd soundcore snd_page_alloc i2c_i801 i2c_core rng_core
> > > iTCO_wdt uhci_hcd usbcore [last unloaded: atyfb]
> > > 
> > > Pid: 5198, comm: rmmod Not tainted (2.6.25-rc8-video #1)
> > > EIP: 0060:[<ce8fd0f2>] EFLAGS: 00010286 CPU: 1
> > > EIP is at aty_set_crtc+0x9/0x19d [atyfb]
> > > EAX: 00000000 EBX: 00000000 ECX: cd8c6854 EDX: ce9064a0
> > > ESI: ce9064a0 EDI: ce905830 EBP: cc9a1ed0 ESP: cc9a1ec8
> > >  DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
> > > Process rmmod (pid: 5198, ti=cc9a1000 task=cc9da000 task.ti=cc9a1000)
> > > Stack: 00000000 ccacd000 cc9a1ee0 ce900d20 cd8c6800 ce905830 cc9a1eec c01c8376
> > >        cd8c6854 cc9a1efc c0208c92 cd8c3854 cd8c6854 cc9a1f10 c02090c1 ce905830
> > >        00000000 c0369060 cc9a1f24 c02086bc 00000000 ce905830 00000880 cc9a1f34
> > > Call Trace:
> > >  [<ce900d20>] ? atyfb_pci_remove+0x1d/0xb3 [atyfb]
> > >  [<c01c8376>] ? pci_device_remove+0x19/0x39
> > >  [<c0208c92>] ? __device_release_driver+0x60/0x7d
> > >  [<c02090c1>] ? driver_detach+0x8e/0xcd
> > >  [<c02086bc>] ? bus_remove_driver+0x63/0x7f
> > >  [<c020914c>] ? driver_unregister+0x2a/0x2e
> > >  [<c01c852e>] ? pci_unregister_driver+0x1e/0x64
> > >  [<ce900d01>] ? atyfb_exit+0xd/0xf [atyfb]
> > >  [<c013be85>] ? sys_delete_module+0x195/0x1d0
> > >  [<c0110b8c>] ? do_page_fault+0x251/0x505
> > >  [<c012e5cd>] ? up_read+0x16/0x29
> > >  [<c0110b8c>] ? do_page_fault+0x251/0x505
> > >  [<c0103892>] ? sysenter_past_esp+0x5f/0x91
> > >  =======================
> > > Code: 01 00 00 00 e8 84 ff ff ff a8 02 74 0e 8b 07 89 83 dc 03 00 00 8b b3 50
> > > 04 00 00 89 f0 5b 5e 5f 5d c3 55 89 e5 56 53 89 c3 89 d6 <83> b8 48 04 00 00
> > > 00 74 5f 8b 52 2c 81 e2 ff ff ff fc 8b 80 48
> > > EIP: [<ce8fd0f2>] aty_set_crtc+0x9/0x19d [atyfb] SS:ESP 0068:cc9a1ec8
> > > ---[ end trace 1c998179e35e2f76 ]---
> > > 
> > > I will try to enable more debug, but probably this information is enough?
> > >
> > 
> > Looks like incoming arg `par' is NULL in aty_set_crtc().
> > 
> > I suspect the problem is that the module shouldn't have successfully 
> > loaded at all, given that its fb_ops.fb_set_par() failed?
> Maybe, at least it must not crash such ugly way. I am not enough programmer to
> give comments about code.

Alas, I don't think we have an atyfb maintainer, and our official fbdev
maintainer (Tony Daplas) has been out of contact for a couple of months. 
(This has happened before and I expect he'll return).  Meanwhile a few
other people are helping out.

But it may take us some time to get onto this.  If nothing happens, please
do send a repeat bug report in a month or so.