Skip to contentSkip to conversation

crash on CONFIG_CFAG12864B=y in 2.6.20-rc3-mm1

DISCUSSION Daniel Walker 2 participants 9 messages Jan 7, 2007 – Feb 1, 2007

lore
Share & tools
lore
#0 lore raw

(forgot to CC LKML)

The options,

CONFIG_CFAG12864B=y
CONFIG_CFAG12864B_RATE=20

causes a crash at boot in 2.6.20-rc3-mm1. I don't have the hardware associated with the options. It looks like it just doesn't have guards to detect if the hardware doesn't exists.

Here is the crash,

ks0108: ERROR: parport didn't find 888 port

BUG: unable to handle kernel NULL pointer dereference at virtual address 0000004 printing eip:
c02dbff9
*pde = 00000000
Oops: 0000 [#1]
PREEMPT SMP
last sysfs file:
Modules linked in:
CPU:    3
EIP:    0060:[<c02dbff9>]    Not tainted VLI
EFLAGS: 00010246   (2.6.20-rc3-mm1 #11)

EIP is at ks0108_writecontrol+0x79/0xc0
eax: 00001008 ebx: 0000000a ecx: 673e2eb8 edx: 00000001
esi: 0000000a edi: 00000000 ebp: f7c3ff6c esp: f7c3ff50
ds: 007b es: 007b fs: 00d8 gs: 0000 ss: 0068
Process swapper (pid: 1, ti=f7c3e000 task=f7c26a90 task.ti=f7c3e000)

Stack: 00000001 f7552c40 f7c3ff60 c0120e3f 00000000 c049f450 00000000 f7c3ff74
       c02dc159 f7c3ff80 c02dc177 00000000 f7c3ff98 c048feda 00000378 c02d74db
       00000000 00000000 f7c3ffe0 c0478610 c03d9d35 00000004 f7c26a90 c0473fc4
Call Trace:
 [<c01053da>] show_trace_log_lvl+0x1a/0x30
 [<c0105499>] show_stack_log_lvl+0xa9/0xd0
 [<c01056c7>] show_registers+0x207/0x370
 [<c0105949>] die+0x119/0x250
 [<c011d267>] do_page_fault+0x277/0x610
 [<c038e9d4>] error_code+0x7c/0x84
 [<c02dc159>] cfag12864b_e+0x19/0x20
 [<c02dc177>] cfag12864b_page+0x17/0x30
 [<c048feda>] cfag12864b_init+0x8a/0x130
 [<c0478610>] init+0x110/0x250
 [<c0104fd3>] kernel_thread_helper+0x7/0x14
 =======================
Code: 8b 98 ec 00 00 00 0f b6 03 24 df 88 45 f3 80 75 f3 20 0f b6 43 01 20 45 f
EIP: [<c02dbff9>] ks0108_writecontrol+0x79/0xc0 SS:ESP 0068:f7c3ff50
 <0>Kernel panic - not syncing: Attempted to kill init!
#1 lore raw
Quoted 50 lines · Daniel Walker

(forgot to CC LKML)

The options,

CONFIG_CFAG12864B=y
CONFIG_CFAG12864B_RATE=20

causes a crash at boot in 2.6.20-rc3-mm1. I don't have the hardware associated with the options. It looks like it just doesn't have guards to detect if the hardware doesn't exists.

Here is the crash,

ks0108: ERROR: parport didn't find 888 port

BUG: unable to handle kernel NULL pointer dereference at virtual address
0000004 printing eip:
c02dbff9
*pde = 00000000
Oops: 0000 [#1]
PREEMPT SMP
last sysfs file:
Modules linked in:
CPU:    3
EIP:    0060:[<c02dbff9>]    Not tainted VLI
EFLAGS: 00010246   (2.6.20-rc3-mm1 #11)

EIP is at ks0108_writecontrol+0x79/0xc0
eax: 00001008 ebx: 0000000a ecx: 673e2eb8 edx: 00000001
esi: 0000000a edi: 00000000 ebp: f7c3ff6c esp: f7c3ff50
ds: 007b es: 007b fs: 00d8 gs: 0000 ss: 0068
Process swapper (pid: 1, ti=f7c3e000 task=f7c26a90 task.ti=f7c3e000)

Stack: 00000001 f7552c40 f7c3ff60 c0120e3f 00000000 c049f450 00000000
f7c3ff74
       c02dc159 f7c3ff80 c02dc177 00000000 f7c3ff98 c048feda 00000378
c02d74db
       00000000 00000000 f7c3ffe0 c0478610 c03d9d35 00000004 f7c26a90
c0473fc4
Call Trace:
 [<c01053da>] show_trace_log_lvl+0x1a/0x30
 [<c0105499>] show_stack_log_lvl+0xa9/0xd0
 [<c01056c7>] show_registers+0x207/0x370
 [<c0105949>] die+0x119/0x250
 [<c011d267>] do_page_fault+0x277/0x610
 [<c038e9d4>] error_code+0x7c/0x84
 [<c02dc159>] cfag12864b_e+0x19/0x20
 [<c02dc177>] cfag12864b_page+0x17/0x30
 [<c048feda>] cfag12864b_init+0x8a/0x130
 [<c0478610>] init+0x110/0x250
 [<c0104fd3>] kernel_thread_helper+0x7/0x14
 =======================
Code: 8b 98 ec 00 00 00 0f b6 03 24 df 88 45 f3 80 75 f3 20 0f b6 43 01 20
Quoted 7 lines

45 f
EIP: [<c02dbff9>] ks0108_writecontrol+0x79/0xc0 SS:ESP 0068:f7c3ff50
<0>Kernel panic - not syncing: Attempted to kill init!

As Daniel Walker pointed out, the driver doesn't probe for the hardware because it just uses the parallel port for output (there isn't any kind of input).

The driver shouldn't continue execution and using ks0108_writecontrol() (which writes to the parallel port) after the "ks0108: ERROR: parport didn't find 888 port" message.

Will check.

Thanks for the warning,
Miguel

(forgot to reply to all).

Signature · 2 lines
Miguel Ojeda
http://maxextreme.googlepages.com/index.htm
#2 lore raw

The problem is easy:

ks0108_init() prints the error message and exit with -EINVAL, so the module isn't loaded properly.

However, cfag12864b_init() is called, although ks0108 failed. It returns 0 and after a while cfag12864b calls ks0108_writecontrol() which dereferences the uninitialized pointer ks0108_parport:

	parport_write_control(ks0108_parport, byte ^ (bit(0) | bit(1) | bit(3)));

Why is cfag12864b_init() called if ks0108 module didn't load properly? Is that normal? If so, how a module should alarm other modules about it failed loading?

An easy solution woule be to export a function at ks0108.c like ks0108_inited() that would return if the module was properly inited or not. Is there any better solution?

Regards,
Miguel

Quoted 78 lines · Miguel Ojeda
Quoted 50 lines · Daniel Walker

(forgot to CC LKML)

The options,

CONFIG_CFAG12864B=y
CONFIG_CFAG12864B_RATE=20

causes a crash at boot in 2.6.20-rc3-mm1. I don't have the hardware associated with the options. It looks like it just doesn't have guards to detect if the hardware doesn't exists.

Here is the crash,

ks0108: ERROR: parport didn't find 888 port

BUG: unable to handle kernel NULL pointer dereference at virtual address
0000004 printing eip:
c02dbff9
*pde = 00000000
Oops: 0000 [#1]
PREEMPT SMP
last sysfs file:
Modules linked in:
CPU:    3
EIP:    0060:[<c02dbff9>]    Not tainted VLI
EFLAGS: 00010246   (2.6.20-rc3-mm1 #11)

EIP is at ks0108_writecontrol+0x79/0xc0
eax: 00001008 ebx: 0000000a ecx: 673e2eb8 edx: 00000001
esi: 0000000a edi: 00000000 ebp: f7c3ff6c esp: f7c3ff50
ds: 007b es: 007b fs: 00d8 gs: 0000 ss: 0068
Process swapper (pid: 1, ti=f7c3e000 task=f7c26a90 task.ti=f7c3e000)

Stack: 00000001 f7552c40 f7c3ff60 c0120e3f 00000000 c049f450 00000000
f7c3ff74
       c02dc159 f7c3ff80 c02dc177 00000000 f7c3ff98 c048feda 00000378
c02d74db
       00000000 00000000 f7c3ffe0 c0478610 c03d9d35 00000004 f7c26a90
c0473fc4
Call Trace:
 [<c01053da>] show_trace_log_lvl+0x1a/0x30
 [<c0105499>] show_stack_log_lvl+0xa9/0xd0
 [<c01056c7>] show_registers+0x207/0x370
 [<c0105949>] die+0x119/0x250
 [<c011d267>] do_page_fault+0x277/0x610
 [<c038e9d4>] error_code+0x7c/0x84
 [<c02dc159>] cfag12864b_e+0x19/0x20
 [<c02dc177>] cfag12864b_page+0x17/0x30
 [<c048feda>] cfag12864b_init+0x8a/0x130
 [<c0478610>] init+0x110/0x250
 [<c0104fd3>] kernel_thread_helper+0x7/0x14
 =======================
Code: 8b 98 ec 00 00 00 0f b6 03 24 df 88 45 f3 80 75 f3 20 0f b6 43 01 20
Quoted 7 lines

45 f
EIP: [<c02dbff9>] ks0108_writecontrol+0x79/0xc0 SS:ESP 0068:f7c3ff50
<0>Kernel panic - not syncing: Attempted to kill init!

As Daniel Walker pointed out, the driver doesn't probe for the hardware because it just uses the parallel port for output (there isn't any kind of input).

The driver shouldn't continue execution and using ks0108_writecontrol() (which writes to the parallel port) after the "ks0108: ERROR: parport didn't find 888 port" message.

Will check.

Thanks for the warning,
Miguel

(forgot to reply to all).

--
Miguel Ojeda
http://maxextreme.googlepages.com/index.htm

Signature · 2 lines
Miguel Ojeda
http://maxextreme.googlepages.com/index.htm
#3 lore raw
Quoted 9 lines · Miguel Ojeda

The problem is easy:

ks0108_init() prints the error message and exit with -EINVAL, so the module isn't loaded properly.

However, cfag12864b_init() is called, although ks0108 failed. It returns 0 and after a while cfag12864b calls ks0108_writecontrol() which dereferences the uninitialized pointer ks0108_parport:

Quoted 5 lines
	parport_write_control(ks0108_parport, byte ^ (bit(0) | bit(1) | bit(3)));

Why is cfag12864b_init() called if ks0108 module didn't load properly? Is that normal? If so, how a module should alarm other modules about it failed loading?

I don't know if this matters, but I had this driver built-in , and not a module ..

Daniel

#4 lore raw
Quoted 12 lines · Daniel Walker
Quoted 11 lines · Miguel Ojeda

The problem is easy:

ks0108_init() prints the error message and exit with -EINVAL, so the module isn't loaded properly.

However, cfag12864b_init() is called, although ks0108 failed. It returns 0 and after a while cfag12864b calls ks0108_writecontrol() which dereferences the uninitialized pointer ks0108_parport:

	parport_write_control(ks0108_parport, byte ^ (bit(0) | bit(1) | bit(3)));
Quoted 10 lines
Quoted 3 lines

Why is cfag12864b_init() called if ks0108 module didn't load properly? Is that normal? If so, how a module should alarm other modules about it failed loading?

I don't know if this matters, but I had this driver built-in , and not a module ..

Daniel

Well, I use the word "module" for both cases: When I modprobe cfag12864b, if ks0108 fails, it doesn't get linked. So I thought the same happen for built-in drivers (in other words, I didn't think cfag12864b would be linked if ks0108 failed).

So I'm waiting until someone tell me what is the right way to avoid drivers like cfag12864b been inited if their dependencies failed.

Anyway, thanks for discovering the bug (I wouldn't have discovered it as my motherboard has parallel port ;).

Miguel

Signature · 2 lines
Miguel Ojeda
http://maxextreme.googlepages.com/index.htm
#5 lore raw
Quoted 8 lines · Miguel Ojeda
Quoted 2 lines

Well, I use the word "module" for both cases: When I modprobe cfag12864b, if ks0108 fails, it doesn't get linked. So I thought the same happen for built-in drivers (in other words, I didn't think cfag12864b would be linked if ks0108 failed).

Quoted 5 lines

So I'm waiting until someone tell me what is the right way to avoid drivers like cfag12864b been inited if their dependencies failed.

Anyway, thanks for discovering the bug (I wouldn't have discovered it as my motherboard has parallel port ;).

I have a parallel port, but I don't have the cfag12864b or ks0108 .. I think the simplest way to fix it would be to add a flag to ks0108 that signals successful initialization , then force cfag12864b to check for that ..

I found this by booting a make allyesconfig kernel ..

Daniel

#6 lore raw
Quoted 10 lines · Daniel Walker
Quoted 8 lines · Miguel Ojeda
Quoted 2 lines

Well, I use the word "module" for both cases: When I modprobe cfag12864b, if ks0108 fails, it doesn't get linked. So I thought the same happen for built-in drivers (in other words, I didn't think cfag12864b would be linked if ks0108 failed).

Quoted 7 lines
Quoted 5 lines

So I'm waiting until someone tell me what is the right way to avoid drivers like cfag12864b been inited if their dependencies failed.

Anyway, thanks for discovering the bug (I wouldn't have discovered it as my motherboard has parallel port ;).

I have a parallel port, but I don't have the cfag12864b or ks0108 .. I

In fact, you don't need the cfag12864b display, as the driver just needs the parallel port for output (you can put some LEDs attached to the parallel port to check the driver's operations and it will be fine).

BTW, why ks0108 didn't find the 888 parallel port? Haven't you got such number?

Quoted 3 lines

think the simplest way to fix it would be to add a flag to ks0108 that signals successful initialization , then force cfag12864b to check for that ..

Sure, that won't be a problem; however, maybe there is other better way (?).

Quoted 6 lines

I found this by booting a make allyesconfig kernel ..

Daniel

Miguel

Signature · 2 lines
Miguel Ojeda
http://maxextreme.googlepages.com/index.htm
#7 lore raw
Quoted 17 lines · Miguel Ojeda
Quoted 16 lines · Daniel Walker
Quoted 13 lines · Miguel Ojeda
Quoted 2 lines

Well, I use the word "module" for both cases: When I modprobe cfag12864b, if ks0108 fails, it doesn't get linked. So I thought the same happen for built-in drivers (in other words, I didn't think cfag12864b would be linked if ks0108 failed).

So I'm waiting until someone tell me what is the right way to avoid drivers like cfag12864b been inited if their dependencies failed.

Anyway, thanks for discovering the bug (I wouldn't have discovered it as my motherboard has parallel port ;).

Quoted 8 lines
Quoted 1 line

I have a parallel port, but I don't have the cfag12864b or ks0108 .. I

In fact, you don't need the cfag12864b display, as the driver just needs the parallel port for output (you can put some LEDs attached to the parallel port to check the driver's operations and it will be fine).

BTW, why ks0108 didn't find the 888 parallel port? Haven't you got such number?

You know I think your right, it may not have a parallel port.. It's so common I tend to take it for granted .

Quoted 5 lines
Quoted 3 lines

think the simplest way to fix it would be to add a flag to ks0108 that signals successful initialization , then force cfag12864b to check for that ..

Sure, that won't be a problem; however, maybe there is other better way (?).

When you build the driver as a module the module loader will only load what is needed, and in order .. When the same driver is built-in the module_init() becomes a device_initcall() , and all device_initcall's get executed .. So you have to expect that cfag12864b drivers initcalls will run eventually and will need to be careful about it's initialization.

Daniel

#8 lore raw
Quoted 25 lines · Daniel Walker
Quoted 24 lines · Miguel Ojeda
Quoted 10 lines · Daniel Walker
Quoted 8 lines · Miguel Ojeda
Quoted 2 lines

Well, I use the word "module" for both cases: When I modprobe cfag12864b, if ks0108 fails, it doesn't get linked. So I thought the same happen for built-in drivers (in other words, I didn't think cfag12864b would be linked if ks0108 failed).

Quoted 7 lines
Quoted 5 lines

So I'm waiting until someone tell me what is the right way to avoid drivers like cfag12864b been inited if their dependencies failed.

Anyway, thanks for discovering the bug (I wouldn't have discovered it as my motherboard has parallel port ;).

I have a parallel port, but I don't have the cfag12864b or ks0108 .. I

In fact, you don't need the cfag12864b display, as the driver just needs the parallel port for output (you can put some LEDs attached to the parallel port to check the driver's operations and it will be fine).

Quoted 6 lines
Quoted 1 line

BTW, why ks0108 didn't find the 888 parallel port? Haven't you got such

number?

You know I think your right, it may not have a parallel port.. It's so common I tend to take it for granted .

Sadly, parallel ports are dying :(

Quoted 5 lines
Quoted 5 lines
Quoted 3 lines

think the simplest way to fix it would be to add a flag to ks0108 that signals successful initialization , then force cfag12864b to check for that ..

Sure, that won't be a problem; however, maybe there is other better way

Quoted 8 lines

(?).

When you build the driver as a module the module loader will only load what is needed, and in order .. When the same driver is built-in the module_init() becomes a device_initcall() , and all device_initcall's get executed .. So you have to expect that cfag12864b drivers initcalls will run eventually and will need to be careful about it's initialization.

That is the real problem, indeed.

Quoted 4 lines

Daniel

All right, if there is no other better way to avoid that, I will send the patch soon.

Miguel

Signature · 2 lines
Miguel Ojeda
http://maxextreme.googlepages.com/index.htm