[patch] smaller kernels

[patch] smaller kernels

[Andrew Morton]

This is the result of a search-and-destroy mission against
oft-repeated strings in the kernel text.  These come about
due to the toolchain's inability to common up strings between
compilation units.

In particular, if there is a printk inside an inline function
then the string gets expanded into every object file which
included that inline function.

So the patch goes through and removes the offending strings 
in various ways.

Kernel size is reduced by another 110 kbytes in my build.

I haven't sent this to Linus.  gcc-3.0.x does the right thing,
and as 3.x will likely be the preferred compiler when 2.6 is
released, this change isn't needed for Linus kernels.  But it's
needed for 2.4.


--- 2.4.19-pre3/include/asm-i386/mmu_context.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/asm-i386/mmu_context.h	Tue Mar 19 21:56:15 2002
@@ -48,7 +48,7 @@ static inline void switch_mm(struct mm_s
 	else {
 		cpu_tlbstate[cpu].state = TLBSTATE_OK;
 		if(cpu_tlbstate[cpu].active_mm != next)
-			BUG();
+			out_of_line_bug();
 		if(!test_and_set_bit(cpu, &next->cpu_vm_mask)) {
 			/* We were in lazy tlb mode and leave_mm disabled 
 			 * tlb flush IPI delivery. We must flush our tlb.
--- 2.4.19-pre3/include/linux/kernel.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/kernel.h	Tue Mar 19 21:56:15 2002
@@ -181,4 +181,6 @@ struct sysinfo {
 	char _f[20-2*sizeof(long)-sizeof(int)];	/* Padding: libc5 uses this.. */
 };
 
+extern void out_of_line_bug(void) ATTRIB_NORET;
+
 #endif
--- 2.4.19-pre3/include/asm-i386/pci.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/asm-i386/pci.h	Tue Mar 19 21:56:15 2002
@@ -73,7 +73,7 @@ static inline dma_addr_t pci_map_single(
 					size_t size, int direction)
 {
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
 	flush_write_buffers();
 	return virt_to_bus(ptr);
 }
@@ -89,7 +89,7 @@ static inline void pci_unmap_single(stru
 				    size_t size, int direction)
 {
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
 	/* Nothing to do */
 }
 
@@ -101,7 +101,7 @@ static inline dma_addr_t pci_map_page(st
 				      unsigned long offset, size_t size, int direction)
 {
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
 
 	return (page - mem_map) * PAGE_SIZE + offset;
 }
@@ -110,7 +110,7 @@ static inline void pci_unmap_page(struct
 				  size_t size, int direction)
 {
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
 	/* Nothing to do */
 }
 
@@ -143,16 +143,16 @@ static inline int pci_map_sg(struct pci_
 	int i;
 
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
  
  	/*
  	 * temporary 2.4 hack
  	 */
  	for (i = 0; i < nents; i++ ) {
  		if (sg[i].address && sg[i].page)
- 			BUG();
+ 			out_of_line_bug();
  		else if (!sg[i].address && !sg[i].page)
- 			BUG();
+ 			out_of_line_bug();
  
  		if (sg[i].address)
  			sg[i].dma_address = virt_to_bus(sg[i].address);
@@ -172,7 +172,7 @@ static inline void pci_unmap_sg(struct p
 				int nents, int direction)
 {
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
 	/* Nothing to do */
 }
 
@@ -190,7 +190,7 @@ static inline void pci_dma_sync_single(s
 				       size_t size, int direction)
 {
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
 	flush_write_buffers();
 }
 
@@ -205,7 +205,7 @@ static inline void pci_dma_sync_sg(struc
 				   int nelems, int direction)
 {
 	if (direction == PCI_DMA_NONE)
-		BUG();
+		out_of_line_bug();
 	flush_write_buffers();
 }
 
--- 2.4.19-pre3/include/asm-i386/smplock.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/asm-i386/smplock.h	Tue Mar 19 21:56:15 2002
@@ -59,7 +59,7 @@ static __inline__ void lock_kernel(void)
 static __inline__ void unlock_kernel(void)
 {
 	if (current->lock_depth < 0)
-		BUG();
+		out_of_line_bug();
 #if 1
 	if (--current->lock_depth < 0)
 		spin_unlock(&kernel_flag);
--- 2.4.19-pre3/include/linux/dcache.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/dcache.h	Tue Mar 19 21:56:15 2002
@@ -242,11 +242,8 @@ extern char * __d_path(struct dentry *, 
  
 static __inline__ struct dentry * dget(struct dentry *dentry)
 {
-	if (dentry) {
-		if (!atomic_read(&dentry->d_count))
-			BUG();
+	if (dentry)
 		atomic_inc(&dentry->d_count);
-	}
 	return dentry;
 }
 
--- 2.4.19-pre3/include/linux/file.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/file.h	Tue Mar 19 21:56:15 2002
@@ -71,30 +71,7 @@ static inline void put_unused_fd(unsigne
 	write_unlock(&files->file_lock);
 }
 
-/*
- * Install a file pointer in the fd array.  
- *
- * The VFS is full of places where we drop the files lock between
- * setting the open_fds bitmap and installing the file in the file
- * array.  At any such point, we are vulnerable to a dup2() race
- * installing a file in the array before us.  We need to detect this and
- * fput() the struct file we are about to overwrite in this case.
- *
- * It should never happen - if we allow dup2() do it, _really_ bad things
- * will follow.
- */
-
-static inline void fd_install(unsigned int fd, struct file * file)
-{
-	struct files_struct *files = current->files;
-	
-	write_lock(&files->file_lock);
-	if (files->fd[fd])
-		BUG();
-	files->fd[fd] = file;
-	write_unlock(&files->file_lock);
-}
-
+void fd_install(unsigned int fd, struct file * file);
 void put_files_struct(struct files_struct *fs);
 
 #endif /* __LINUX_FILE_H */
--- 2.4.19-pre3/include/linux/nfs_fs.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/nfs_fs.h	Tue Mar 19 21:56:15 2002
@@ -168,7 +168,7 @@ nfs_file_cred(struct file *file)
 	struct rpc_cred *cred = (struct rpc_cred *)(file->private_data);
 #ifdef RPC_DEBUG
 	if (cred && cred->cr_magic != RPCAUTH_CRED_MAGIC)
-		BUG();
+		out_of_line_bug();
 #endif
 	return cred;
 }
--- 2.4.19-pre3/include/linux/quotaops.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/quotaops.h	Tue Mar 19 21:56:15 2002
@@ -40,8 +40,6 @@ extern int  dquot_transfer(struct inode 
 
 static __inline__ void DQUOT_INIT(struct inode *inode)
 {
-	if (!inode->i_sb)
-		BUG();
 	lock_kernel();
 	if (sb_any_quota_enabled(inode->i_sb) && !IS_NOQUOTA(inode))
 		inode->i_sb->dq_op->initialize(inode, -1);
@@ -51,11 +49,8 @@ static __inline__ void DQUOT_INIT(struct
 static __inline__ void DQUOT_DROP(struct inode *inode)
 {
 	lock_kernel();
-	if (IS_QUOTAINIT(inode)) {
-		if (!inode->i_sb)
-			BUG();
+	if (IS_QUOTAINIT(inode))
 		inode->i_sb->dq_op->drop(inode);	/* Ops must be set when there's any quota... */
-	}
 	unlock_kernel();
 }
 
--- 2.4.19-pre3/include/linux/sched.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/sched.h	Tue Mar 19 21:56:15 2002
@@ -896,7 +896,6 @@ static inline int task_on_runqueue(struc
 
 static inline void unhash_process(struct task_struct *p)
 {
-	if (task_on_runqueue(p)) BUG();
 	write_lock_irq(&tasklist_lock);
 	nr_threads--;
 	unhash_pid(p);
--- 2.4.19-pre3/include/linux/highmem.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/highmem.h	Tue Mar 19 21:56:15 2002
@@ -63,8 +63,6 @@ static inline void memclear_highpage_flu
 {
 	char *kaddr;
 
-	if (offset + size > PAGE_SIZE)
-		BUG();
 	kaddr = kmap(page);
 	memset(kaddr + offset, 0, size);
 	flush_dcache_page(page);
--- 2.4.19-pre3/include/linux/skbuff.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/linux/skbuff.h	Tue Mar 19 21:56:15 2002
@@ -756,9 +756,9 @@ static inline int skb_headlen(const stru
 	return skb->len - skb->data_len;
 }
 
-#define SKB_PAGE_ASSERT(skb) do { if (skb_shinfo(skb)->nr_frags) BUG(); } while (0)
-#define SKB_FRAG_ASSERT(skb) do { if (skb_shinfo(skb)->frag_list) BUG(); } while (0)
-#define SKB_LINEAR_ASSERT(skb) do { if (skb_is_nonlinear(skb)) BUG(); } while (0)
+#define SKB_PAGE_ASSERT(skb) do { if (skb_shinfo(skb)->nr_frags) out_of_line_bug(); } while (0)
+#define SKB_FRAG_ASSERT(skb) do { if (skb_shinfo(skb)->frag_list) out_of_line_bug(); } while (0)
+#define SKB_LINEAR_ASSERT(skb) do { if (skb_is_nonlinear(skb)) out_of_line_bug(); } while (0)
 
 /*
  *	Add data to an sk_buff
@@ -825,8 +825,6 @@ static inline unsigned char *skb_push(st
 static inline char *__skb_pull(struct sk_buff *skb, unsigned int len)
 {
 	skb->len-=len;
-	if (skb->len < skb->data_len)
-		BUG();
 	return 	skb->data+=len;
 }
 
@@ -1094,7 +1092,7 @@ static inline void *kmap_skb_frag(const 
 {
 #ifdef CONFIG_HIGHMEM
 	if (in_irq())
-		BUG();
+		out_of_line_bug();
 
 	local_bh_disable();
 #endif
--- 2.4.19-pre3/include/asm-i386/highmem.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/asm-i386/highmem.h	Tue Mar 19 21:56:15 2002
@@ -62,7 +62,7 @@ extern void FASTCALL(kunmap_high(struct 
 static inline void *kmap(struct page *page)
 {
 	if (in_interrupt())
-		BUG();
+		out_of_line_bug();
 	if (page < highmem_start_page)
 		return page_address(page);
 	return kmap_high(page);
@@ -71,7 +71,7 @@ static inline void *kmap(struct page *pa
 static inline void kunmap(struct page *page)
 {
 	if (in_interrupt())
-		BUG();
+		out_of_line_bug();
 	if (page < highmem_start_page)
 		return;
 	kunmap_high(page);
@@ -95,7 +95,7 @@ static inline void *kmap_atomic(struct p
 	vaddr = __fix_to_virt(FIX_KMAP_BEGIN + idx);
 #if HIGHMEM_DEBUG
 	if (!pte_none(*(kmap_pte-idx)))
-		BUG();
+		out_of_line_bug();
 #endif
 	set_pte(kmap_pte-idx, mk_pte(page, kmap_prot));
 	__flush_tlb_one(vaddr);
@@ -113,7 +113,7 @@ static inline void kunmap_atomic(void *k
 		return;
 
 	if (vaddr != __fix_to_virt(FIX_KMAP_BEGIN+idx))
-		BUG();
+		out_of_line_bug();
 
 	/*
 	 * force other mappings to Oops if they'll try to access
--- 2.4.19-pre3/include/net/tcp.h~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/include/net/tcp.h	Tue Mar 19 21:56:15 2002
@@ -1329,8 +1329,6 @@ static __inline__ int tcp_prequeue(struc
 		if (tp->ucopy.memory > sk->rcvbuf) {
 			struct sk_buff *skb1;
 
-			if (sk->lock.users) BUG();
-
 			while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
 				sk->backlog_rcv(sk, skb1);
 				NET_INC_STATS_BH(TCPPrequeueDropped);
--- 2.4.19-pre3/fs/open.c~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/fs/open.c	Tue Mar 19 21:56:15 2002
@@ -71,6 +71,30 @@ out:
 	return error;
 }
 
+/*
+ * Install a file pointer in the fd array.  
+ *
+ * The VFS is full of places where we drop the files lock between
+ * setting the open_fds bitmap and installing the file in the file
+ * array.  At any such point, we are vulnerable to a dup2() race
+ * installing a file in the array before us.  We need to detect this and
+ * fput() the struct file we are about to overwrite in this case.
+ *
+ * It should never happen - if we allow dup2() do it, _really_ bad things
+ * will follow.
+ */
+
+void fd_install(unsigned int fd, struct file * file)
+{
+	struct files_struct *files = current->files;
+	
+	write_lock(&files->file_lock);
+	if (files->fd[fd])
+		BUG();
+	files->fd[fd] = file;
+	write_unlock(&files->file_lock);
+}
+
 int do_truncate(struct dentry *dentry, loff_t length)
 {
 	struct inode *inode = dentry->d_inode;
--- 2.4.19-pre3/kernel/ksyms.c~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/kernel/ksyms.c	Tue Mar 19 21:56:15 2002
@@ -164,6 +164,7 @@ EXPORT_SYMBOL(mark_buffer_dirty);
 EXPORT_SYMBOL(set_buffer_async_io); /* for reiserfs_writepage */
 EXPORT_SYMBOL(__mark_buffer_dirty);
 EXPORT_SYMBOL(__mark_inode_dirty);
+EXPORT_SYMBOL(fd_install);
 EXPORT_SYMBOL(get_empty_filp);
 EXPORT_SYMBOL(init_private_file);
 EXPORT_SYMBOL(filp_open);
@@ -452,6 +453,7 @@ EXPORT_SYMBOL(nr_running);
 
 /* misc */
 EXPORT_SYMBOL(panic);
+EXPORT_SYMBOL(out_of_line_bug);
 EXPORT_SYMBOL(sprintf);
 EXPORT_SYMBOL(snprintf);
 EXPORT_SYMBOL(sscanf);
--- 2.4.19-pre3/kernel/panic.c~inline-BUG	Tue Mar 19 21:56:15 2002
+++ 2.4.19-pre3-akpm/kernel/panic.c	Tue Mar 19 21:56:15 2002
@@ -124,3 +124,21 @@ const char *print_tainted()
 }
 
 int tainted = 0;
+
+/*
+ * A BUG() call in an inline function in a header should be avoided,
+ * because it can seriously bloat the kernel.  So here we have
+ * helper functions.
+ * We lose the BUG()-time file-and-line info this way, but it's
+ * usually not very useful from an inline anyway.  The backtrace
+ * tells us what we want to know.
+ */
+
+void out_of_line_bug(void)
+{
+	BUG();
+
+	/* Satisfy __attribute__((noreturn)) */
+	for ( ; ; )
+		;
+}


-

Re: [patch] smaller kernels

[David S. Miller]

   From: Rusty Russell <rusty@rustcorp.com.au>
   Date: Mon, 25 Mar 2002 16:56:05 +1100

   And I'm not sure DaveM'll appreciate this:
   
   >  static inline char *__skb_pull(struct sk_buff *skb, unsigned int len)
   >  {
   >  	skb->len-=len;
   > -	if (skb->len < skb->data_len)
   > -		BUG();
   >  	return 	skb->data+=len;
   >  }

Rusty's right, I definitely won't take this, it catches problems
%99 of the time in the place that causes it.

Re: [patch] smaller kernels

[Rusty Russell]

On Tue, 19 Mar 2002 22:11:48 -0800
Andrew Morton <akpm@zip.com.au> wrote:

> This is the result of a search-and-destroy mission against
> oft-repeated strings in the kernel text.  These come about
> due to the toolchain's inability to common up strings between
> compilation units.

The name is horrible.  BUG() stands out: perhaps "BUG_LITE()"?

And I'm not sure DaveM'll appreciate this:

>  static inline char *__skb_pull(struct sk_buff *skb, unsigned int len)
>  {
>  	skb->len-=len;
> -	if (skb->len < skb->data_len)
> -		BUG();
>  	return 	skb->data+=len;
>  }

Rusty.
-- 
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.

Re: [patch] smaller kernels

[Andrew Morton]

Rusty Russell wrote:
> 
> On Tue, 19 Mar 2002 22:11:48 -0800
> Andrew Morton <akpm@zip.com.au> wrote:
> 
> > This is the result of a search-and-destroy mission against
> > oft-repeated strings in the kernel text.  These come about
> > due to the toolchain's inability to common up strings between
> > compilation units.
> 
> The name is horrible.  BUG() stands out: perhaps "BUG_LITE()"?

out_of_line_bug()? It's a bit ornate I guess, but it tells
you what it does when you look at it.

> And I'm not sure DaveM'll appreciate this:
> 
> >  static inline char *__skb_pull(struct sk_buff *skb, unsigned int len)
> >  {
> >       skb->len-=len;
> > -     if (skb->len < skb->data_len)
> > -             BUG();
> >       return  skb->data+=len;
> >  }

OK, we can out that back (out-of-line) for -pre6.

-

Re: [patch] smaller kernels

[Rusty Russell]

In message <3C9EBF13.4FEBB158@zip.com.au> you write:
> Rusty Russell wrote:
> > 
> > On Tue, 19 Mar 2002 22:11:48 -0800
> > Andrew Morton <akpm@zip.com.au> wrote:
> > 
> > > This is the result of a search-and-destroy mission against
> > > oft-repeated strings in the kernel text.  These come about
> > > due to the toolchain's inability to common up strings between
> > > compilation units.
> > 
> > The name is horrible.  BUG() stands out: perhaps "BUG_LITE()"?
> 
> out_of_line_bug()? It's a bit ornate I guess, but it tells
> you what it does when you look at it.

I like BUG() because it stands out in a fast code scan.  And the
shouting prevents people from sprinkling them everywhere like a
drunken soldier (although there are still some people who seem to want
to BUG before ops which would oops anyway...)

Rusty.
--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.

Re: [patch] smaller kernels

[Bill Davidsen]

On Sun, 24 Mar 2002, David S. Miller wrote:

>    From: Rusty Russell <rusty@rustcorp.com.au>
>    Date: Mon, 25 Mar 2002 16:56:05 +1100
> 
>    And I'm not sure DaveM'll appreciate this:
>    
>    >  static inline char *__skb_pull(struct sk_buff *skb, unsigned int len)
>    >  {
>    >  	skb->len-=len;
>    > -	if (skb->len < skb->data_len)
>    > -		BUG();
>    >  	return 	skb->data+=len;
>    >  }
> 
> Rusty's right, I definitely won't take this, it catches problems
> %99 of the time in the place that causes it.

  But, it makes the kernel smaller. And faster! ;-)

-- 
bill davidsen <davidsen@tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.

Re: [patch] smaller kernels

[Marcelo Tosatti]

On Mon, 25 Mar 2002, Rusty Russell wrote:

> On Tue, 19 Mar 2002 22:11:48 -0800
> Andrew Morton <akpm@zip.com.au> wrote:
> 
> > This is the result of a search-and-destroy mission against
> > oft-repeated strings in the kernel text.  These come about
> > due to the toolchain's inability to common up strings between
> > compilation units.
> 
> The name is horrible.  BUG() stands out: perhaps "BUG_LITE()"?
> 
> And I'm not sure DaveM'll appreciate this:
> 
> >  static inline char *__skb_pull(struct sk_buff *skb, unsigned int len)
> >  {
> >  	skb->len-=len;
> > -	if (skb->len < skb->data_len)
> > -		BUG();
> >  	return 	skb->data+=len;
> >  }

Andrew, 

I've just readded all asserts which you removed... if you really want to
remove any of those, please prove me that they are useless.

Re: [patch] smaller kernels

[Marcelo Tosatti]

On Mon, 25 Mar 2002, Andrew Morton wrote:

> Marcelo Tosatti wrote:
> > 
> > I've just readded all asserts which you removed... if you really want to
> > remove any of those, please prove me that they are useless.
> 
> I grepped a year's lkml traffic - nobody is hitting any
> of them...

Anyway, they may catch bugs introduced by new modifications...

> The quotaops.h checks were useless:
> 
> 	if (pointer == NULL)
> 		BUG();
> 	dereference(pointer);
> 
> The others can become calls to out_of_line_bug() if
> you want.

Please do, thanks.

Re: [patch] smaller kernels

[Andrew Morton]

Marcelo Tosatti wrote:
> 
> I've just readded all asserts which you removed... if you really want to
> remove any of those, please prove me that they are useless.

I grepped a year's lkml traffic - nobody is hitting any
of them...

The quotaops.h checks were useless:

	if (pointer == NULL)
		BUG();
	dereference(pointer);

The others can become calls to out_of_line_bug() if
you want.  That's dget(), unhash_process(),
memclear_highpage_flush(), __skb_pull() and tcp_prequeue()

-