Skip to contentSkip to conversation

2.5.4 Compile Error

DISCUSSION John Weber 6 participants 11 messages Feb 11, 2002

lore
Share & tools
lore

John Weber

author
#0 lore raw

Will try to specify the problem more precisely after I play around for a bit, but here goes the error:

[root@boolean linux-2.5.4]# make bzImage
gcc -D__KERNEL__ -I/usr/src/linux-2.5.4/include -Wall -Wstrict-prototypes -Wno-trigraphs -O2 -fomit-frame-pointer -fno-strict-aliasing -fno-common -pipe -mpreferred-stack-boundary=2 -march=i686 -DKBUILD_BASENAME=main -c -o init/main.o init/main.c
In file included from /usr/src/linux-2.5.4/include/asm/thread_info.h:13,

                 from /usr/src/linux-2.5.4/include/linux/thread_info.h:10,
                 from /usr/src/linux-2.5.4/include/linux/spinlock.h:7,
                 from /usr/src/linux-2.5.4/include/linux/mmzone.h:8,
                 from /usr/src/linux-2.5.4/include/linux/gfp.h:4,
                 from /usr/src/linux-2.5.4/include/linux/slab.h:14,
                 from /usr/src/linux-2.5.4/include/linux/proc_fs.h:5,
                 from init/main.c:15:

/usr/src/linux-2.5.4/include/asm/processor.h: In function `thread_saved_pc':
/usr/src/linux-2.5.4/include/asm/processor.h:444: dereferencing pointer to incomplete type
make: *** [init/main.o] Error 1

#1 lore raw
Quoted 4 lines · John Weber

/usr/src/linux-2.5.4/include/asm/processor.h: In function `thread_saved_pc':
/usr/src/linux-2.5.4/include/asm/processor.h:444: dereferencing pointer to incomplete type
make: *** [init/main.o] Error 1

since it's just for /usr/bin/ps, ie. not a fast path, I just un-inlined it in my alpha hacking. Same approach might work for here, too.

The basic problem, I'm guessing, is that asm/processor.h wants to know about the internals of task struct, but it can't yet.

Jeff

Signature · 3 lines
Jeff Garzik      | "I went through my candy like hot oatmeal
Building 1024    |  through an internally-buttered weasel."
MandrakeSoft     |             - goats.com
author
#2 lore raw
In reply to John Weber
Quoted 8 lines · Jeff Garzik
Quoted 5 lines · John Weber

/usr/src/linux-2.5.4/include/asm/processor.h: In function `thread_saved_pc':
/usr/src/linux-2.5.4/include/asm/processor.h:444: dereferencing pointer to incomplete type
make: *** [init/main.o] Error 1

Quoted 8 lines

since it's just for /usr/bin/ps, ie. not a fast path, I just un-inlined it in my alpha hacking. Same approach might work for here, too.

The basic problem, I'm guessing, is that asm/processor.h wants to know about the internals of task struct, but it can't yet.

Jeff

I don't know what the problem is, but un-inlining this function isn't correcting it.

The function thread_saved_pc() is a mystery to me. It is declared with a return type of unsigned long, and yet return this:

((unsigned long *)tsk->thread->esp)[3]

This is confusing to me in many ways:
- the "thread" member of task struct is not a pointer
- esp is of type unsigned long, so I don't understand the cast, and I certainly don't understand the [3] here.

Can anyone explain this code to me?

I'm a kernelnewbie, so I'm inclined to return:
return (tsk->thread).esp
What is this function trying to do?

#3 lore raw
Quoted 11 lines · John Weber

The function thread_saved_pc() is a mystery to me. It is declared with a return type of unsigned long, and yet return this:

((unsigned long *)tsk->thread->esp)[3]

This is confusing to me in many ways:
- the "thread" member of task struct is not a pointer
- esp is of type unsigned long, so I don't understand the cast, and I certainly don't understand the [3] here.

Can anyone explain this code to me?

The problem is an interdependency between processor.h and sched.h.

The old code was the same, except it did

	t->esp

where t was a thread_struct, instead of what we do now

	t->thread->esp

where t is a task_struct. And thus whereby before we passed

	p->thread

as the argument, now you pass just `p'. I.e., its the same net-affect. The error is because the function needs access to both task_struct (in sched.h) and thread_struct (in processor.h) but the two are interrelated so we can't include them in each other.

The contents of esp is a memory address, so typecasting it to (unsigned long *) is OK.

As for the [3], p[3] is the same as

	*(p+3)
ie,
	*(p+sizeof(p))

so that is legal.

So the fix, aside from reverting this change and the kernel/sched.c change and the sparc64 changes ... would be to solve the dependency issue.

Robert Love

#4 lore raw
In reply to John Weber
Quoted 4 lines · John Weber

I don't know what the problem is, but un-inlining this function isn't correcting it.

Try this:

include/asm-i386/processor.h +1 -8
@@ -435,14 +435,7 @@ extern int kernel_thread(int (*fn)(void 
 /* Copy and release all segment info associated with a VM */
 extern void copy_segments(struct task_struct *p, struct mm_struct * mm);
 extern void release_segments(struct mm_struct * mm);
-
-/*
- * Return saved PC of a blocked thread.
- */
-static inline unsigned long thread_saved_pc(struct task_struct *tsk)
-{
-	return ((unsigned long *)tsk->thread->esp)[3];
-}
+extern unsigned long thread_saved_pc(struct task_struct *tsk);
 
 unsigned long get_wchan(struct task_struct *p);
 #define KSTK_EIP(tsk)	(((unsigned long *)(4096+(unsigned long)(tsk)->thread_info))[1019])
arch/i386/kernel/process.c +8 -0
@@ -55,6 +55,14 @@ asmlinkage void ret_from_fork(void) __as
 int hlt_counter;
 
 /*
+ * Return saved PC of a blocked thread.
+ */
+unsigned long thread_saved_pc(struct task_struct *tsk)
+{
+	return ((unsigned long *)tsk->thread.esp)[3];
+}
+
+/*
  * Powermanagement idle function, if any..
  */
 void (*pm_idle)(void);

-

author
#8 lore raw
In reply to John Weber
Quoted 21 lines · Robert Love
Quoted 11 lines · John Weber

The function thread_saved_pc() is a mystery to me. It is declared with a return type of unsigned long, and yet return this:

((unsigned long *)tsk->thread->esp)[3]

This is confusing to me in many ways:
- the "thread" member of task struct is not a pointer
- esp is of type unsigned long, so I don't understand the cast, and I certainly don't understand the [3] here.

Can anyone explain this code to me?

The problem is an interdependency between processor.h and sched.h.

The old code was the same, except it did

	t->esp
Quoted 8 lines

where t was a thread_struct, instead of what we do now

	t->thread->esp

where t is a task_struct. And thus whereby before we passed

	p->thread

I understand all this, but thread is not a pointer. So shouldn't it be t->thread.esp ?

Quoted 3 lines

as the argument, now you pass just `p'. I.e., its the same net-affect. The error is because the function needs access to both task_struct (in sched.h) and thread_struct (in processor.h) but the two are interrelated
so we can't include them in each other.

Hmm... OK.

Quoted 8 lines

The contents of esp is a memory address, so typecasting it to (unsigned long *) is OK.

As for the [3], p[3] is the same as

	*(p+3)
ie,
	*(p+sizeof(p))

so that is legal.

*(p + (3*sizeof(p))) ?

I understand the syntax, but I don't understand why one would want to return the address of something 3 longs away. What is this function supposed to be doing?

#9 lore raw
Quoted 2 lines · John Weber

I understand all this, but thread is not a pointer. So shouldn't it be t->thread.esp ?

You are right, I missed this.

The fix is to change it to `((unsigned long *)tsk->thread.esp)[3]' but we also still have to fix the dependency problem, which means moving it out of processor.h. Andrew Morton just posted a patch to move it to process.c, I think he has it right. Give that a try.

Quoted 1 line

*(p + (3*sizeof(p))) ?

Right ;)

Robert Love

#5 lore raw
In reply to Andrew Morton
Quoted 4 lines · David S. Miller

From: Andrew Morton <[email protected]>
Date: Sun, 10 Feb 2002 22:29:58 -0800

Quoted 8 lines
Quoted 2 lines · John Weber

I don't know what the problem is, but un-inlining this function isn't correcting it.

Try this:

Not sufficient, you have to also add a dummy "struct task_struct;" declaration before the thread_saved_pc extern.

It's already there, line 425?

-

#6 lore raw

From: Andrew Morton <[email protected]>
Date: Sun, 10 Feb 2002 22:48:57 -0800

Quoted 2 lines · David S. Miller

Not sufficient, you have to also add a dummy "struct task_struct;" declaration before the thread_saved_pc extern.

It's already there, line 425?

Yep, my error.

#7 lore raw
In reply to Andrew Morton

From: Andrew Morton <[email protected]>
Date: Sun, 10 Feb 2002 22:29:58 -0800

Quoted 2 lines · John Weber

I don't know what the problem is, but un-inlining this function isn't correcting it.

Try this:

Not sufficient, you have to also add a dummy "struct task_struct;" declaration before the thread_saved_pc extern.

#10 lore raw
In reply to John Weber

On Mon, 11 Feb 2002 07:50:06 +0100, you wrote in linux.kernel:

Quoted 3 lines

I understand the syntax, but I don't understand why one would want to return the address of something 3 longs away. What is this function supposed to be doing?

Well, from the name thread_saved_pc() it tries to get some value of PC (the program counter ;) that's saved by that thread, right? My guess would be that the desired value is stored on the stack at esp[3].

Signature · 2 lines
Ciao,
Pascal