[PATCH] 2.4: Fix steal_locks race

[PATCH] 2.4: Fix steal_locks race

[Herbert Xu]

[-- Attachment #1: Type: text/plain, Size: 492 bytes --]

Hi:

The steal_locks() call in binfmt_elf.c is buggy.  It steals locks from
a files entry whose reference was dropped much earlier.  This allows it
to steal other process's locks.

The following patch calls steal_locks() earlier so that this does not
happen.

Cheers,
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

[-- Attachment #2: p --]
[-- Type: text/plain, Size: 770 bytes --]

--- kernel-source-2.4/fs/binfmt_elf.c.orig	2003-08-08 20:46:56.000000000 +1000
+++ kernel-source-2.4/fs/binfmt_elf.c	2003-08-08 20:47:05.000000000 +1000
@@ -480,6 +480,7 @@
 	files = current->files;		/* Refcounted so ok */
 	if(unshare_files() < 0)
 		goto out_free_ph;
+	steal_locks(files, current->files);
 
 	/* exec will make our files private anyway, but for the a.out
 	   loader stuff we need to do it earlier */
@@ -797,7 +798,6 @@
 	if (current->ptrace & PT_PTRACED)
 		send_sig(SIGTRAP, current, 0);
 	retval = 0;
-	steal_locks(files, current->files);
 out:
 	return retval;
 
@@ -813,6 +813,7 @@
 out_free_fh:
 	ftmp = current->files;
 	current->files = files;
+	steal_locks(ftmp, current->files);
 	put_files_struct(ftmp);
 out_free_ph:
 	kfree(elf_phdata);

Re: [PATCH] 2.4: Fix steal_locks race

[Herbert Xu]

[-- Attachment #1: Type: text/plain, Size: 772 bytes --]

On Fri, Aug 08, 2003 at 08:53:21PM +1000, herbert wrote:
> 
> The steal_locks() call in binfmt_elf.c is buggy.  It steals locks from
> a files entry whose reference was dropped much earlier.  This allows it
> to steal other process's locks.
> 
> The following patch calls steal_locks() earlier so that this does not
> happen.

My patch is buggy too.  If a file is closed by another clone between
the two steal_locks calls the lock will again be lost.  Fortunately
this much harder to trigger than the previous bug.

The following patch fixes that.
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

[-- Attachment #2: p --]
[-- Type: text/plain, Size: 806 bytes --]

--- kernel-source-2.4/fs/binfmt_elf.c.orig	2003-08-09 10:43:56.000000000 +1000
+++ kernel-source-2.4/fs/binfmt_elf.c	2003-08-09 10:57:29.000000000 +1000
@@ -480,7 +480,6 @@
 	files = current->files;		/* Refcounted so ok */
 	if(unshare_files() < 0)
 		goto out_free_ph;
-	steal_locks(files, current->files);
 
 	/* exec will make our files private anyway, but for the a.out
 	   loader stuff we need to do it earlier */
@@ -603,6 +602,7 @@
 		goto out_free_dentry;
 
 	/* Discard our unneeded old files struct */
+	steal_locks(files, current->files);
 	put_files_struct(files);
 
 	/* OK, This is the point of no return */
@@ -813,7 +813,6 @@
 out_free_fh:
 	ftmp = current->files;
 	current->files = files;
-	steal_locks(ftmp, current->files);
 	put_files_struct(ftmp);
 out_free_ph:
 	kfree(elf_phdata);

[PATCH] 2.4: Restore current->files in flush_old_exec

[Herbert Xu]

[-- Attachment #1: Type: text/plain, Size: 614 bytes --]

Hi:

The unshare_files patch to flush_old_exec() did not restore the original
state when exec_mmap fails.  This patch fixes that.

At this point, I believe the unshare_files stuff should be fine from
a correctness point of view.  However, there is still a performance
problem as every ELF exec call ends up dupliating the files structure
as well as walking through all file locks.

Cheers,
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

[-- Attachment #2: q --]
[-- Type: text/plain, Size: 1206 bytes --]

Index: kernel-source-2.4/fs/exec.c
===================================================================
RCS file: /home/gondolin/herbert/src/CVS/debian/kernel-source-2.4/fs/exec.c,v
retrieving revision 1.5
diff -u -r1.5 exec.c
--- kernel-source-2.4/fs/exec.c	26 Jul 2003 02:54:45 -0000	1.5
+++ kernel-source-2.4/fs/exec.c	9 Aug 2003 01:00:28 -0000
@@ -557,7 +557,7 @@
 	char * name;
 	int i, ch, retval;
 	struct signal_struct * oldsig;
-	struct files_struct * files;
+	struct files_struct *files, *ftmp;
 
 	/*
 	 * Make sure we have a private signal table
@@ -576,8 +576,6 @@
 	retval = unshare_files();
 	if(retval)
 		goto flush_failed;
-	steal_locks(files, current->files);
-	put_files_struct(files);
 	
 	/* 
 	 * Release all of the old mmap stuff
@@ -586,6 +584,8 @@
 	if (retval) goto mmap_failed;
 
 	/* This is the point of no return */
+	steal_locks(files, current->files);
+	put_files_struct(files);
 	release_old_signals(oldsig);
 
 	current->sas_ss_sp = current->sas_ss_size = 0;
@@ -623,6 +623,9 @@
 	return 0;
 
 mmap_failed:
+	ftmp = current->files;
+	current->files = files;
+	put_files_struct(ftmp);
 flush_failed:
 	spin_lock_irq(&current->sigmask_lock);
 	if (current->sig != oldsig) {

Re: [PATCH] 2.4: Fix steal_locks race

[Andreas Gruenbacher]

Hello,

thanks a lot for analyzing this. Please see the patch I just posted unter
the subject:

  [PATCH] 2.4.22-rc2 steal_locks and load_elf_binary cleanups

That patch fixes the bug in a slightly cleaner way.

On Sat, 9 Aug 2003, Herbert Xu wrote:

> On Fri, Aug 08, 2003 at 08:53:21PM +1000, herbert wrote:
> >
> > The steal_locks() call in binfmt_elf.c is buggy.  It steals locks from
> > a files entry whose reference was dropped much earlier.  This allows it
> > to steal other process's locks.

This makes sense.

> > The following patch calls steal_locks() earlier so that this does not
> > happen.
>
> My patch is buggy too.  If a file is closed by another clone between
> the two steal_locks calls the lock will again be lost.  Fortunately
> this much harder to trigger than the previous bug.
>
> The following patch fixes that.
>

Re: [PATCH] 2.4: Restore current->files in flush_old_exec

[Andreas Gruenbacher]

On Sat, 9 Aug 2003, Herbert Xu wrote:

> Hi:
>
> The unshare_files patch to flush_old_exec() did not restore the original
> state when exec_mmap fails.  This patch fixes that.

Indeed. This is still needed.

> At this point, I believe the unshare_files stuff should be fine from
> a correctness point of view.  However, there is still a performance
> problem as every ELF exec call ends up dupliating the files structure
> as well as walking through all file locks.

Cheers,
Andreas.

Re: [PATCH] 2.4: Restore current->files in flush_old_exec

[Herbert Xu]

[-- Attachment #1: Type: text/plain, Size: 615 bytes --]

On Sat, Aug 09, 2003 at 11:11:16AM +1000, herbert wrote:
> 
> At this point, I believe the unshare_files stuff should be fine from
> a correctness point of view.  However, there is still a performance
> problem as every ELF exec call ends up dupliating the files structure
> as well as walking through all file locks.

Here is the patch that ensures files is only duplicated when necessary.
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

[-- Attachment #2: p --]
[-- Type: text/plain, Size: 1090 bytes --]

--- kernel-source-2.4/fs/binfmt_elf.c.orig	2003-08-09 11:28:18.000000000 +1000
+++ kernel-source-2.4/fs/binfmt_elf.c	2003-08-09 11:32:13.000000000 +1000
@@ -480,6 +480,10 @@
 	files = current->files;		/* Refcounted so ok */
 	if(unshare_files() < 0)
 		goto out_free_ph;
+	if (files == current->files) {
+		put_files_struct(files);
+		files = NULL;
+	}
 
 	/* exec will make our files private anyway, but for the a.out
 	   loader stuff we need to do it earlier */
@@ -602,8 +606,10 @@
 		goto out_free_dentry;
 
 	/* Discard our unneeded old files struct */
-	steal_locks(files, current->files);
-	put_files_struct(files);
+	if (files) {
+		steal_locks(files, current->files);
+		put_files_struct(files);
+	}
 
 	/* OK, This is the point of no return */
 	current->mm->start_data = 0;
@@ -811,9 +817,11 @@
 out_free_file:
 	sys_close(elf_exec_fileno);
 out_free_fh:
-	ftmp = current->files;
-	current->files = files;
-	put_files_struct(ftmp);
+	if (files) {
+		ftmp = current->files;
+		current->files = files;
+		put_files_struct(ftmp);
+	}
 out_free_ph:
 	kfree(elf_phdata);
 	goto out;

Re: [PATCH] 2.4: Fix steal_locks race

[Andreas Gruenbacher]

On Sat, 9 Aug 2003, Herbert Xu wrote:

> On Fri, Aug 08, 2003 at 08:53:21PM +1000, herbert wrote:
> >
> > The steal_locks() call in binfmt_elf.c is buggy.  It steals locks from
> > a files entry whose reference was dropped much earlier.  This allows it
> > to steal other process's locks.
> >
> > The following patch calls steal_locks() earlier so that this does not
> > happen.
>
> My patch is buggy too.  If a file is closed by another clone between
> the two steal_locks calls the lock will again be lost.  Fortunately
> this much harder to trigger than the previous bug.

I think this is not a strict bug---this scenario is not covered by POSIX
in the first place. Unless lock stealing is done atomically with
unshare_files there is a window of oportunity between unshare_files() and
steal_locks(), so locks can still get lost.


Cheers,
Andreas.

Re: [PATCH] 2.4: Restore current->files in flush_old_exec

[Andreas Gruenbacher]

On Sat, 9 Aug 2003, Herbert Xu wrote:

> On Sat, Aug 09, 2003 at 11:11:16AM +1000, herbert wrote:
> >
> > At this point, I believe the unshare_files stuff should be fine from
> > a correctness point of view.  However, there is still a performance
> > problem as every ELF exec call ends up dupliating the files structure
> > as well as walking through all file locks.
>
> Here is the patch that ensures files is only duplicated when necessary.

This patch is correct but unnecessary: steal_locks already tests for this
condition.


Cheers,
Andreas.

Re: [PATCH] 2.4: Fix steal_locks race

[Herbert Xu]

On Sat, Aug 09, 2003 at 04:04:53AM +0200, Andreas Gruenbacher wrote:
>
> > My patch is buggy too.  If a file is closed by another clone between
> > the two steal_locks calls the lock will again be lost.  Fortunately
> > this much harder to trigger than the previous bug.
> 
> I think this is not a strict bug---this scenario is not covered by POSIX
> in the first place. Unless lock stealing is done atomically with
> unshare_files there is a window of oportunity between unshare_files() and
> steal_locks(), so locks can still get lost.

It's not a standard compliance issue.  In this case the lock will never
be released and it will eventually lead to a crash when someone reads
/proc/locks.
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: [PATCH] 2.4: Restore current->files in flush_old_exec

[Herbert Xu]

On Sat, Aug 09, 2003 at 04:20:38AM +0200, Andreas Gruenbacher wrote:
> On Sat, 9 Aug 2003, Herbert Xu wrote:
> 
> > On Sat, Aug 09, 2003 at 11:11:16AM +1000, herbert wrote:
> > >
> > > At this point, I believe the unshare_files stuff should be fine from
> > > a correctness point of view.  However, there is still a performance
> > > problem as every ELF exec call ends up dupliating the files structure
> > > as well as walking through all file locks.
> >
> > Here is the patch that ensures files is only duplicated when necessary.
> 
> This patch is correct but unnecessary: steal_locks already tests for this
> condition.

Yes but when you call unshare_files twice one of them will have to
copy.
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: [PATCH] 2.4: Fix steal_locks race

[Andreas Gruenbacher]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 514 bytes --]

Here is an accumulate patch of Herbert's and my changes.

I could see no reason for ftmp in flush_old_exec and load_elf_binary, so I
removed that. Please correct me if that is wrong.

Cheers,
Andreas.


On Fri, 8 Aug 2003, Herbert Xu wrote:

> Hi:
>
> The steal_locks() call in binfmt_elf.c is buggy.  It steals locks from
> a files entry whose reference was dropped much earlier.  This allows it
> to steal other process's locks.
>
> The following patch calls steal_locks() earlier so that this does not
> happen.

[-- Attachment #2: Type: TEXT/PLAIN, Size: 4686 bytes --]

Index: linux-2.4.22-rc2.orig/fs/exec.c
===================================================================
--- linux-2.4.22-rc2.orig.orig/fs/exec.c	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/fs/exec.c	2003-08-09 04:47:53.000000000 +0200
@@ -582,8 +582,6 @@ int flush_old_exec(struct linux_binprm *
 	retval = unshare_files();
 	if(retval)
 		goto flush_failed;
-	steal_locks(files, current->files);
-	put_files_struct(files);
 	
 	/* 
 	 * Release all of the old mmap stuff
@@ -592,6 +590,8 @@ int flush_old_exec(struct linux_binprm *
 	if (retval) goto mmap_failed;
 
 	/* This is the point of no return */
+	steal_locks(files);
+	put_files_struct(files);
 	release_old_signals(oldsig);
 
 	current->sas_ss_sp = current->sas_ss_size = 0;
@@ -629,6 +629,8 @@ int flush_old_exec(struct linux_binprm *
 	return 0;
 
 mmap_failed:
+	put_files_struct(current->files);
+	current->files = files;
 flush_failed:
 	spin_lock_irq(&current->sigmask_lock);
 	if (current->sig != oldsig) {
Index: linux-2.4.22-rc2.orig/include/linux/fs.h
===================================================================
--- linux-2.4.22-rc2.orig.orig/include/linux/fs.h	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/include/linux/fs.h	2003-08-09 04:46:05.000000000 +0200
@@ -674,7 +674,7 @@ extern int __get_lease(struct inode *ino
 extern time_t lease_get_mtime(struct inode *);
 extern int lock_may_read(struct inode *, loff_t start, unsigned long count);
 extern int lock_may_write(struct inode *, loff_t start, unsigned long count);
-extern void steal_locks(fl_owner_t from, fl_owner_t to);
+extern void steal_locks(fl_owner_t from);
 
 struct fasync_struct {
 	int	magic;
Index: linux-2.4.22-rc2.orig/fs/binfmt_elf.c
===================================================================
--- linux-2.4.22-rc2.orig.orig/fs/binfmt_elf.c	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/fs/binfmt_elf.c	2003-08-09 04:48:27.000000000 +0200
@@ -444,7 +444,7 @@ static int load_elf_binary(struct linux_
 	struct elfhdr interp_elf_ex;
   	struct exec interp_ex;
 	char passed_fileno[6];
-	struct files_struct *files, *ftmp;
+	struct files_struct *files;
 	
 	/* Get the exec-header */
 	elf_ex = *((struct elfhdr *) bprm->buf);
@@ -480,7 +480,6 @@ static int load_elf_binary(struct linux_
 	files = current->files;		/* Refcounted so ok */
 	if(unshare_files() < 0)
 		goto out_free_ph;
-	steal_locks(files, current->files);
 
 	/* exec will make our files private anyway, but for the a.out
 	   loader stuff we need to do it earlier */
@@ -603,7 +602,9 @@ static int load_elf_binary(struct linux_
 		goto out_free_dentry;
 
 	/* Discard our unneeded old files struct */
+	steal_locks(files);
 	put_files_struct(files);
+	files = NULL;
 
 	/* OK, This is the point of no return */
 	current->mm->start_data = 0;
@@ -714,18 +715,16 @@ static int load_elf_binary(struct linux_
 			elf_entry = load_elf_interp(&interp_elf_ex,
 						    interpreter,
 						    &interp_load_addr);
-
-		allow_write_access(interpreter);
-		fput(interpreter);
-		kfree(elf_interpreter);
-
 		if (BAD_ADDR(elf_entry)) {
 			printk(KERN_ERR "Unable to load interpreter\n");
-			kfree(elf_phdata);
 			send_sig(SIGSEGV, current, 0);
 			retval = -ENOEXEC; /* Nobody gets to see this, but.. */
-			goto out;
+			goto out_free_dentry;
 		}
+
+		allow_write_access(interpreter);
+		fput(interpreter);
+		kfree(elf_interpreter);
 	}
 
 	kfree(elf_phdata);
@@ -811,10 +810,10 @@ out_free_interp:
 out_free_file:
 	sys_close(elf_exec_fileno);
 out_free_fh:
-	ftmp = current->files;
-	current->files = files;
-	steal_locks(ftmp, current->files);
-	put_files_struct(ftmp);
+	if (files) {
+		put_files_struct(current->files);
+		current->files = files;
+	}
 out_free_ph:
 	kfree(elf_phdata);
 	goto out;
Index: linux-2.4.22-rc2.orig/fs/locks.c
===================================================================
--- linux-2.4.22-rc2.orig.orig/fs/locks.c	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/fs/locks.c	2003-08-09 04:46:05.000000000 +0200
@@ -1937,11 +1937,11 @@ done:
 	return length;
 }
 
-void steal_locks(fl_owner_t from, fl_owner_t to)
+void steal_locks(fl_owner_t from)
 {
 	struct list_head *tmp;
 
-	if (from == to)
+	if (from == current->files)
 		return;
 
 	lock_kernel();
@@ -1949,7 +1949,7 @@ void steal_locks(fl_owner_t from, fl_own
 		struct file_lock *fl = list_entry(tmp, struct file_lock,
 						  fl_link);
 		if (fl->fl_owner == from)
-			fl->fl_owner = to;
+			fl->fl_owner = current->files;
 	}
 	unlock_kernel();
 }

Re: [PATCH] 2.4: Restore current->files in flush_old_exec

[Andreas Gruenbacher]

On Sat, 9 Aug 2003, Herbert Xu wrote:

> On Sat, Aug 09, 2003 at 04:20:38AM +0200, Andreas Gruenbacher wrote:
> > On Sat, 9 Aug 2003, Herbert Xu wrote:
> >
> > > On Sat, Aug 09, 2003 at 11:11:16AM +1000, herbert wrote:
> > > >
> > > > At this point, I believe the unshare_files stuff should be fine from
> > > > a correctness point of view.  However, there is still a performance
> > > > problem as every ELF exec call ends up dupliating the files structure
> > > > as well as walking through all file locks.
> > >
> > > Here is the patch that ensures files is only duplicated when necessary.
> >
> > This patch is correct but unnecessary: steal_locks already tests for this
> > condition.
>
> Yes but when you call unshare_files twice one of them will have to
> copy.

I see---that happens through flush_old_exec.

Re: [PATCH] 2.4: Fix steal_locks race

[Herbert Xu]

On Sat, Aug 09, 2003 at 04:59:03AM +0200, Andreas Gruenbacher wrote:
>
> @@ -714,18 +715,16 @@ static int load_elf_binary(struct linux_
>  			elf_entry = load_elf_interp(&interp_elf_ex,
>  						    interpreter,
>  						    &interp_load_addr);
> -
> -		allow_write_access(interpreter);
> -		fput(interpreter);
> -		kfree(elf_interpreter);
> -
>  		if (BAD_ADDR(elf_entry)) {
>  			printk(KERN_ERR "Unable to load interpreter\n");
> -			kfree(elf_phdata);
>  			send_sig(SIGSEGV, current, 0);
>  			retval = -ENOEXEC; /* Nobody gets to see this, but.. */
> -			goto out;
> +			goto out_free_dentry;
>  		}
> +
> +		allow_write_access(interpreter);
> +		fput(interpreter);
> +		kfree(elf_interpreter);
>  	}

This looks bad since you're past the point of no return.
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: [PATCH] 2.4: Fix steal_locks race

[Andreas Gruenbacher]

On Sat, 9 Aug 2003, Herbert Xu wrote:

> On Sat, Aug 09, 2003 at 04:04:53AM +0200, Andreas Gruenbacher wrote:
> >
> > > My patch is buggy too.  If a file is closed by another clone between
> > > the two steal_locks calls the lock will again be lost.  Fortunately
> > > this much harder to trigger than the previous bug.
> >
> > I think this is not a strict bug---this scenario is not covered by POSIX
> > in the first place. Unless lock stealing is done atomically with
> > unshare_files there is a window of oportunity between unshare_files() and
> > steal_locks(), so locks can still get lost.
>
> It's not a standard compliance issue.  In this case the lock will never
> be released and it will eventually lead to a crash when someone reads
> /proc/locks.

I don't see how this would happen. Could you please elaborate?

Re: [PATCH] 2.4: Fix steal_locks race

[Andreas Gruenbacher]

On Sat, 9 Aug 2003, Herbert Xu wrote:

> On Sat, Aug 09, 2003 at 04:59:03AM +0200, Andreas Gruenbacher wrote:
> >
> > @@ -714,18 +715,16 @@ static int load_elf_binary(struct linux_
> >  			elf_entry = load_elf_interp(&interp_elf_ex,
> >  						    interpreter,
> >  						    &interp_load_addr);
> > -
> > -		allow_write_access(interpreter);
> > -		fput(interpreter);
> > -		kfree(elf_interpreter);
> > -
> >  		if (BAD_ADDR(elf_entry)) {
> >  			printk(KERN_ERR "Unable to load interpreter\n");
> > -			kfree(elf_phdata);
> >  			send_sig(SIGSEGV, current, 0);
> >  			retval = -ENOEXEC; /* Nobody gets to see this, but.. */
> > -			goto out;
> > +			goto out_free_dentry;
> >  		}
> > +
> > +		allow_write_access(interpreter);
> > +		fput(interpreter);
> > +		kfree(elf_interpreter);
> >  	}
>
> This looks bad since you're past the point of no return.

This is an equivalence transformation except for the explicit
sys_close(elf_exec_fileno) in the unwind code, which would eventually
happen, anyway.

Re: [PATCH] 2.4: Restore current->files in flush_old_exec

[Andreas Gruenbacher]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 143 bytes --]

Here is an update ...

> On Sat, 9 Aug 2003, Herbert Xu wrote:
> > Yes but when you call unshare_files twice one of them will have to
> > copy.

[-- Attachment #2: Type: TEXT/PLAIN, Size: 4826 bytes --]

Index: linux-2.4.22-rc2.orig/fs/exec.c
===================================================================
--- linux-2.4.22-rc2.orig.orig/fs/exec.c	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/fs/exec.c	2003-08-09 05:04:35.000000000 +0200
@@ -582,8 +582,6 @@ int flush_old_exec(struct linux_binprm *
 	retval = unshare_files();
 	if(retval)
 		goto flush_failed;
-	steal_locks(files, current->files);
-	put_files_struct(files);
 	
 	/* 
 	 * Release all of the old mmap stuff
@@ -592,6 +590,8 @@ int flush_old_exec(struct linux_binprm *
 	if (retval) goto mmap_failed;
 
 	/* This is the point of no return */
+	steal_locks(files);
+	put_files_struct(files);
 	release_old_signals(oldsig);
 
 	current->sas_ss_sp = current->sas_ss_size = 0;
@@ -629,6 +629,8 @@ int flush_old_exec(struct linux_binprm *
 	return 0;
 
 mmap_failed:
+	put_files_struct(current->files);
+	current->files = files;
 flush_failed:
 	spin_lock_irq(&current->sigmask_lock);
 	if (current->sig != oldsig) {
Index: linux-2.4.22-rc2.orig/include/linux/fs.h
===================================================================
--- linux-2.4.22-rc2.orig.orig/include/linux/fs.h	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/include/linux/fs.h	2003-08-09 05:04:35.000000000 +0200
@@ -674,7 +674,7 @@ extern int __get_lease(struct inode *ino
 extern time_t lease_get_mtime(struct inode *);
 extern int lock_may_read(struct inode *, loff_t start, unsigned long count);
 extern int lock_may_write(struct inode *, loff_t start, unsigned long count);
-extern void steal_locks(fl_owner_t from, fl_owner_t to);
+extern void steal_locks(fl_owner_t from);
 
 struct fasync_struct {
 	int	magic;
Index: linux-2.4.22-rc2.orig/fs/binfmt_elf.c
===================================================================
--- linux-2.4.22-rc2.orig.orig/fs/binfmt_elf.c	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/fs/binfmt_elf.c	2003-08-09 05:06:05.000000000 +0200
@@ -444,7 +444,7 @@ static int load_elf_binary(struct linux_
 	struct elfhdr interp_elf_ex;
   	struct exec interp_ex;
 	char passed_fileno[6];
-	struct files_struct *files, *ftmp;
+	struct files_struct *files;
 	
 	/* Get the exec-header */
 	elf_ex = *((struct elfhdr *) bprm->buf);
@@ -480,7 +480,10 @@ static int load_elf_binary(struct linux_
 	files = current->files;		/* Refcounted so ok */
 	if(unshare_files() < 0)
 		goto out_free_ph;
-	steal_locks(files, current->files);
+	if (files == current->files) {
+		put_files_struct(files);
+		files = NULL;
+	}
 
 	/* exec will make our files private anyway, but for the a.out
 	   loader stuff we need to do it earlier */
@@ -603,7 +606,11 @@ static int load_elf_binary(struct linux_
 		goto out_free_dentry;
 
 	/* Discard our unneeded old files struct */
-	put_files_struct(files);
+	if (files) {
+		steal_locks(files);
+		put_files_struct(files);
+		files = NULL;
+	}
 
 	/* OK, This is the point of no return */
 	current->mm->start_data = 0;
@@ -714,18 +721,16 @@ static int load_elf_binary(struct linux_
 			elf_entry = load_elf_interp(&interp_elf_ex,
 						    interpreter,
 						    &interp_load_addr);
-
-		allow_write_access(interpreter);
-		fput(interpreter);
-		kfree(elf_interpreter);
-
 		if (BAD_ADDR(elf_entry)) {
 			printk(KERN_ERR "Unable to load interpreter\n");
-			kfree(elf_phdata);
 			send_sig(SIGSEGV, current, 0);
 			retval = -ENOEXEC; /* Nobody gets to see this, but.. */
-			goto out;
+			goto out_free_dentry;
 		}
+
+		allow_write_access(interpreter);
+		fput(interpreter);
+		kfree(elf_interpreter);
 	}
 
 	kfree(elf_phdata);
@@ -811,10 +816,10 @@ out_free_interp:
 out_free_file:
 	sys_close(elf_exec_fileno);
 out_free_fh:
-	ftmp = current->files;
-	current->files = files;
-	steal_locks(ftmp, current->files);
-	put_files_struct(ftmp);
+	if (files) {
+		put_files_struct(current->files);
+		current->files = files;
+	}
 out_free_ph:
 	kfree(elf_phdata);
 	goto out;
Index: linux-2.4.22-rc2.orig/fs/locks.c
===================================================================
--- linux-2.4.22-rc2.orig.orig/fs/locks.c	2003-08-09 04:44:07.000000000 +0200
+++ linux-2.4.22-rc2.orig/fs/locks.c	2003-08-09 05:04:35.000000000 +0200
@@ -1937,11 +1937,11 @@ done:
 	return length;
 }
 
-void steal_locks(fl_owner_t from, fl_owner_t to)
+void steal_locks(fl_owner_t from)
 {
 	struct list_head *tmp;
 
-	if (from == to)
+	if (from == current->files)
 		return;
 
 	lock_kernel();
@@ -1949,7 +1949,7 @@ void steal_locks(fl_owner_t from, fl_own
 		struct file_lock *fl = list_entry(tmp, struct file_lock,
 						  fl_link);
 		if (fl->fl_owner == from)
-			fl->fl_owner = to;
+			fl->fl_owner = current->files;
 	}
 	unlock_kernel();
 }

Re: [PATCH] 2.4: Fix steal_locks race

[Herbert Xu]

On Sat, Aug 09, 2003 at 05:13:52AM +0200, Andreas Gruenbacher wrote:
> On Sat, 9 Aug 2003, Herbert Xu wrote:
> 
> > On Sat, Aug 09, 2003 at 04:04:53AM +0200, Andreas Gruenbacher wrote:
> > >
> > > > My patch is buggy too.  If a file is closed by another clone between
> > > > the two steal_locks calls the lock will again be lost.  Fortunately
> > > > this much harder to trigger than the previous bug.
> > >
> > > I think this is not a strict bug---this scenario is not covered by POSIX
> > > in the first place. Unless lock stealing is done atomically with
> > > unshare_files there is a window of oportunity between unshare_files() and
> > > steal_locks(), so locks can still get lost.
> >
> > It's not a standard compliance issue.  In this case the lock will never
> > be released and it will eventually lead to a crash when someone reads
> > /proc/locks.
> 
> I don't see how this would happen. Could you please elaborate?

Suppose that A and B share current->files and fd has a POSIX lock on it.

A			B
unshare_files
steal_locks
			close(fd)
exec fails
steal_locks
put_files_struct

The close in B fails to release the lock as it has been stolen by the
new files structure.  The second steal_locks sets the fl_owner back to
the original files structure which no longer has fd in it and hence can
never release that lock.  The put_files_struct doesn't release the lock
either since it is now owned by the original file structure.
-- 
Debian GNU/Linux 3.0 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: [PATCH] 2.4: Fix steal_locks race

[Andreas Gruenbacher]

On Sat, 2003-08-09 at 05:19, Herbert Xu wrote:
> On Sat, Aug 09, 2003 at 05:13:52AM +0200, Andreas Gruenbacher wrote:
> > On Sat, 9 Aug 2003, Herbert Xu wrote:
> > 
> > > On Sat, Aug 09, 2003 at 04:04:53AM +0200, Andreas Gruenbacher wrote:
> > > >
> > > > > My patch is buggy too.  If a file is closed by another clone between
> > > > > the two steal_locks calls the lock will again be lost.  Fortunately
> > > > > this much harder to trigger than the previous bug.
> > > >
> > > > I think this is not a strict bug---this scenario is not covered by POSIX
> > > > in the first place. Unless lock stealing is done atomically with
> > > > unshare_files there is a window of oportunity between unshare_files() and
> > > > steal_locks(), so locks can still get lost.
> > >
> > > It's not a standard compliance issue.  In this case the lock will never
> > > be released and it will eventually lead to a crash when someone reads
> > > /proc/locks.
> > 
> > I don't see how this would happen. Could you please elaborate?
> 
> Suppose that A and B share current->files and fd has a POSIX lock on it.
> 
> A			B
> unshare_files
> steal_locks
> 			close(fd)
> exec fails
> steal_locks
> put_files_struct
> 
> The close in B fails to release the lock as it has been stolen by the
> new files structure.  The second steal_locks sets the fl_owner back to
> the original files structure which no longer has fd in it and hence can
> never release that lock.  The put_files_struct doesn't release the lock
> either since it is now owned by the original file structure.

In the patch I've sent there is no stealing back of locks, so that case
does not exist.


Cheers,
-- 
Andreas Gruenbacher <agruen@suse.de>
SuSE Labs, SuSE Linux AG <http://www.suse.de/>

Re: [PATCH] 2.4: Restore current->files in flush_old_exec

[Andreas Gruenbacher]

Hi,

On Sat, 2003-08-09 at 12:05, Herbert Xu wrote:
> On Sat, Aug 09, 2003 at 05:54:32AM +0200, Andreas Gruenbacher wrote:
> > On Sat, 9 Aug 2003, Andreas Gruenbacher wrote:
> > 
> > > Here is an update ...
> > Do you agree that this is correct?
> 
> It looks OK to me.  However, I still think the BAD_ADDR change is
> unnecessary.

Very good, thanks. The BAD_ADDR change is indeed not required. It saves
a funtion call so I think we should keep it, but I don't mind so much.


Cheers,
-- 
Andreas Gruenbacher <agruen@suse.de>
SuSE Labs, SuSE Linux AG <http://www.suse.de/>