lkml.org 
[lkml]   [2004]   [Nov]   [27]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
DateFri, 26 Nov 2004 00:13:14 +0100
FromAndrea Arcangeli <>
SubjectRe: Memory leak in 2.4.27 kernel, using mmap raw packet sockets
On Thu, Nov 25, 2004 at 03:12:42PM -0200, Marcelo Tosatti wrote:
> get_user_pages() bails out if ! VM_IO. 
> 
> Is that what you mean with "VM_IO enforcement" ? 

yes. It bails out if VM_IO is set (not ! clear ;)

> I thought about the BUG() to catch potential offenders, but I was
> not sure if it was possible for a PG_reserved page to be part of VMA's 
> which was being get_user_pages'd.

Exactly, it's much safer to go with the real fix of fixing it in
get_user_pages. If something we should put a bugcheck there.

> Now you tell me it is possible, and thats only the ZERO page. Fine. 

Yes, and the ZERO_PAGE is actually the _only_ reserved page we must
allow to go through. Every other reserved page must be discarded (or
kernel-crash with BUG_ON if Alan feels confortable with the VM_IO
enforcement).

> This is what you suggests plus some extra hopefully useful debugging 
> 
> 
> --- memory.c.orig	2004-11-25 14:51:00.074508952 -0200
> +++ memory.c	2004-11-25 15:08:38.026675776 -0200
> @@ -454,8 +454,9 @@
>  int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
>  		int len, int write, int force, struct page **pages, struct vm_area_struct **vmas)
>  {
> -	int i;
> +	int i, s;
>  	unsigned int flags;
> +	struct vm_area_struct *savevma = NULL;
> 
>  	/*
>  	 * Require read or write permissions.
> @@ -463,7 +464,7 @@
>  	 */
>  	flags = write ? (VM_WRITE | VM_MAYWRITE) : (VM_READ | VM_MAYREAD);
>  	flags &= force ? (VM_MAYREAD | VM_MAYWRITE) : (VM_READ | VM_WRITE);
> -	i = 0;
> +	i = s = 0;
> 
>  	do {
>  		struct vm_area_struct *	vma;
> @@ -499,9 +500,13 @@
>  				/* FIXME: call the correct function,
>  				 * depending on the type of the found page
>  				 */
> -				if (!pages[i] || PageReserved(pages[i]))
> -					goto bad_page;
> -				page_cache_get(pages[i]);
> +				if (!pages[i] || PageReserved(pages[i])) {
> +					if (pages[i] != ZERO_PAGE(start)) {
> +						savevma = vma;
> +						goto bad_page;
> +					}
> +				} else
> +					page_cache_get(pages[i]);
>  			}
>  			if (vmas)
>  				vmas[i] = vma;
> @@ -520,9 +525,15 @@
>  	 */
>  bad_page:
>  	spin_unlock(&mm->page_table_lock);
> +	s = i;
>  	while (i--)
>  		page_cache_release(pages[i]);
> -	i = -EFAULT;
> +	/* catch bad uses of PG_reserved on !VM_IO vma's */
> +	printk(KERN_ERR "get_user_pages PG_reserved page on"
> +			"vma:%p flags:%lx page:%d\n", savevma,
> +			savevma->flags, s);
> +	BUG();
> +	i = -EFAULT;
>  	goto out;
>  }

Yes, however I wouldn't turn on the debugging code just in case some
driver forgets to set VM_IO and it doesn't use remap_page_range. There's
nothing fundamentally fatal in having a reserved page in a non VM_IO
vma (I mean, after fixing the above bit ;).
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 14:08    [from the cache]
©2003-2008