Re: [Linux-fbdev-devel] 2.6.18-rc6-mm1

[Jean D. <kh...@li...>]

Hi Andrew, Jurriaan, Antonino,

> > > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.18-rc6/2.6.18-rc6-mm1/
> >
> > This throws an oops on my IBM Thinkpad T23 notebook. Some parts scroll
> > off the screen, but the visible stack trace goes like this:
> > (...) 
> > hub 3-0:1.0: 2 ports detected
> > savagefb: mapped io at f8980000
> > savagefb: probed videoram:  16384k
> > savagefb: Detected current MCLK value of 71591 kHz
> > savagefb: 1024x768 TFT LCD panel detected and active
> > savagefb: Limiting video mode to 1024x768
> > savagefb: mapped framebuffer at f8a80000, pbase == e8000000
> > BUG: unable to handle kernel NULL pointer dereference at virtual address
> > 00000000
> >  printing eip:
> >  f8831013
> >  *pde = 00000000
> >  Oops: 0000 [#1]
> >  4K_STACKS PREEMPT 
> >  last sysfs file: /devices/system/cpu/cpu0/cpufreq/scaling_governor
> >  Modules linked in: savagefb fb_ddc cfbimgblt uhci_hcd usbcore
> >  CPU:    0
> >  EIP:    0060:[<f8831013>]    Not tainted VLI
> >  EFLAGS: 00010286   (2.6.18-rc6-mm1 #6) 
> >  EIP is at fb_ddc_read+0x13/0x1c4 [fb_ddc]
> >  eax: f7faf250   ebx: 00000000   ecx: f7faf244   edx: f7faf244
> >  esi: 00000000   edi: f7faf008   ebp: f7fbed98   esp: f7fbed68
> >  ds: 007b   es: 007b   ss: 0068
> >  Process modprobe (pid: 1776, ti=f7fbe000 task=f7fbdab0
> >  task.ti=f7fbe000)
> >  Stack: 00000001 f7faf244 00004988 01000000 f7faf000 f9a6d9e0 f7fbede8 00000000 
> >         f7faf244 f7faf208 f7faf000 f7faf008 f7fbedb0 f8846af9 f7faf250 f7faf208 
> >         f7faf000 f7faf008 f7fbede8 f884667c f7faf000 f7faf6b0 f7fbedd0 c01ebe91 
> >  Call Trace:
> >  [<c01039eb>] show_trace_log_lvl+0x15/0x28
> >  [<c0103a8a>] show_stack_log_lvl+0x8c/0x97
> >  [<c0103e48>] show_registers+0x188/0x21c
> >  [<c0104085>] die+0x1a9/0x283
> >  [<c0113bb5>] do_page_fault+0x3ed/0x4bf
> >  [<c03421df>] error_code+0x3f/0x44
> >  [<f8846af9>] savagefb_probe_i2c_connector+0x18/0x66 [savagefb]
> >  [<f884667c>] savagefb_probe+0x484/0x672 [savagefb]
> >  [<c01f6722>] pci_device_probe+0x3a/0x61
> >  [<c0269ecb>] really_probe+0x37/0xb0
> >  [<c0269fbc>] driver_probe_device+0x78/0x84
> >  [<c026a06a>] __driver_attach+0x38/0x60
> >  [<c026992a>] bus_for_each_dev+0x42/0x69
> >  [<c0269df7>] driver_attach+0x16/0x18
> >  [<c0269491>] bus_add_driver+0x66/0x179
> >  [<c026a2f7>] driver_register+0x77/0x7c
> >  [<c01f68bf>] __pci_register_driver+0x5e/0x7e
> >  [<f882d02e>] savagefb_init+0x2e/0x36 [savagefb]
> >  [<c0132158>] sys_init_module+0x1274/0x1443
> >  [<c0102dd8>] syscall_call+0x7/0xb
> >  =======================
> >  Code: <ff> 33 ff 53 08 6a 00 ff 33 ff 53 08 c7 45 d4 00 00 00 00 83 c4 10 6a 00 31 ff ff 33 ff 53 04 6a 
> >  EIP: [<f8831013>] fb_ddc_read+0x13/0x1c4 [fb_ddc] SS:ESP 0068:f7fbed68
> > 
> > Hope this helps,
> 
> Does, thanks.  I guess adapter->algo_data is NULL.
> 
> savagefb-use-generic-ddc-reading.patch, perhaps...

I indeed think this patch is causing the oops. The old code did call
savage_do_probe_i2c_edid even if no i2c bus had been created for the
board, and savage_do_probe_i2c_edid bailed out quickly in that case
("chan->par = NULL;" at the end of savage_setup_i2c_bus, and "if
(chan->par) {" at the beginning of savage_do_probe_i2c_edid.) The new
fb_ddc_read function has no such quick exit, it pretty much assumes
that an i2c bus was successfully created beforehand.

So my guess is that Jurriaan's graphics adapter is supported by the
savagefb driver, but the driver doesn't create an i2c bus for it
(either because the hardware doesn't have it, or we simply have no
support.) Jurriaan, please confirm that your adapter is not one of
Savage4, Savage2000, ProSavagePM, ProSavage8.

If my analysis is correct, the following patch should fix the problem.
It can probably be optimized/cleaned up, but I'll leave that to
Antonino. Jurriaan, can you please apply this patch on top of
2.6.18-rc6-mm1 and report success or failure?

* * * * *

Hot fix to savagefb-use-generic-ddc-reading.patch.

Signed-off-by: Jean Delvare <kh...@li...>
Cc: "Antonino A. Daplas" <ad...@po...>
---
 drivers/video/savage/savagefb-i2c.c |    5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

--- linux-2.6.18-rc6-mm1.orig/drivers/video/savage/savagefb-i2c.c	2006-09-09 23:21:24.000000000 +0200
+++ linux-2.6.18-rc6-mm1/drivers/video/savage/savagefb-i2c.c	2006-09-09 23:52:56.000000000 +0200
@@ -218,7 +218,10 @@
 	struct savagefb_par *par = info->par;
 	u8 *edid;
 
-	edid = fb_ddc_read(&par->chan.adapter);
+	if (par->chan.par)
+		edid = fb_ddc_read(&par->chan.adapter);
+	else
+		edid = NULL;
 
 	if (!edid) {
 		/* try to get from firmware */


-- 
Jean Delvare