lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <35e1582a-a64f-44c2-983f-bb1cde6dd98a@lucifer.local>
Date:   Wed, 22 Mar 2023 06:07:35 +0000
From:   Lorenzo Stoakes <lstoakes@...il.com>
To:     "Liam R. Howlett" <Liam.Howlett@...cle.com>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        David Hildenbrand <david@...hat.com>,
        Matthew Wilcox <willy@...radead.org>,
        Vlastimil Babka <vbabka@...e.cz>,
        maple-tree@...ts.infradead.org, Vernon Yang <vernon2gm@...il.com>
Subject: Re: [PATCH v2 4/4] mm/mmap/vma_merge: init cleanup, be explicit
 about the non-mergeable case

On Tue, Mar 21, 2023 at 10:08:58PM -0400, Liam R. Howlett wrote:
> * Lorenzo Stoakes <lstoakes@...il.com> [230321 16:46]:
> > Reorder the initial variables sensibly and set vma_start and vm_pgoff there
> 								^vma_pgoff
> 	Indicating it is used for the vm_area_struct *vma
>
> > rather than later so all initial values are set at the same time meaning we
> > don't have to set these later.
>
> I did these later to reduce the number of times we were checking prev.
> With this patch, we check prev 3 times, but before we were checking it
> once.  The compiler might do something clever here to reduce the
> checking?
>

Apologies for undoing your work! Which obviously wasn't my intention :)

I suspect the compiler would, but we probably shouldn't rely on it.

> I'm also not sure adding the conditional operator in the init helps your
> goal of cleaning it up.

The purpose was to group everything together, reduce indentation in the
prev case, have it resemble the next case more closely, reduce LoC and to
have the prev if block be solely concerned with merging the predecessor
rather than both setting these values and then checking to see if we can
merge the predecessor.

However, on second thoughts I think avoiding repeatedly checking it trumps
that so I will revert to the previous approach.

>
> >
> > Rather than setting err = -1 and only resetting if we hit merge cases,
> > explicitly check the non-mergeable case to make it abundantly clear that we
> > only proceed with the rest if something is mergeable, default err to 0 and
> > only update if an error might occur.
> >
> > Move the merge_prev, merge_next cases closer to the logic determining curr,
> > next.
> >
> > This has no functional impact.
> >
> > Signed-off-by: Lorenzo Stoakes <lstoakes@...il.com>
> > ---
> >  mm/mmap.c | 55 ++++++++++++++++++++++++++-----------------------------
> >  1 file changed, 26 insertions(+), 29 deletions(-)
> >
> > diff --git a/mm/mmap.c b/mm/mmap.c
> > index 7aec49c3bc74..d60cb0b7ae15 100644
> > --- a/mm/mmap.c
> > +++ b/mm/mmap.c
> > @@ -909,18 +909,18 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> >  			struct vm_userfaultfd_ctx vm_userfaultfd_ctx,
> >  			struct anon_vma_name *anon_name)
> >  {
> > -	pgoff_t pglen = (end - addr) >> PAGE_SHIFT;
> > -	pgoff_t vma_pgoff;
> >  	struct vm_area_struct *curr, *next, *res;
> >  	struct vm_area_struct *vma, *adjust, *remove, *remove2;
> > -	int err = -1;
> > +	struct vma_prepare vp;
> > +	int err = 0;
> >  	bool merge_prev = false;
> >  	bool merge_next = false;
> >  	bool vma_expanded = false;
> > -	struct vma_prepare vp;
> > +	unsigned long vma_start = prev ? prev->vm_start : addr;
> >  	unsigned long vma_end = end;
> > +	pgoff_t vma_pgoff = prev ? prev->vm_pgoff : 0;
> > +	pgoff_t pglen = (end - addr) >> PAGE_SHIFT;
> >  	long adj_start = 0;
> > -	unsigned long vma_start = addr;
> >
> >  	validate_mm(mm);
> >  	/*
> > @@ -940,6 +940,23 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> >  		/* Is there a VMA next to a hole (case 1 - 3) or prev (4)? */
> >  		next = vma_lookup(mm, end);
> >
> > +	/* Can we merge the predecessor? */
> > +	if (prev && addr == prev->vm_end && mpol_equal(vma_policy(prev), policy)
> > +	    && can_vma_merge_after(prev, vm_flags, anon_vma, file,
> > +				   pgoff, vm_userfaultfd_ctx, anon_name)) {
> > +		merge_prev = true;
> > +		vma_prev(vmi);
> > +	}
> > +
> > +	/* Can we merge the successor? */
> > +	merge_next = next && mpol_equal(policy, vma_policy(next)) &&
> > +		can_vma_merge_before(next, vm_flags,
> > +				     anon_vma, file, pgoff+pglen,
> > +				     vm_userfaultfd_ctx, anon_name);
> > +
> > +	if (!merge_prev && !merge_next)
> > +		return NULL; /* Not mergeable. */
> > +
> >  	/*
> >  	 * By default, we return prev. Cases 3, 4, 8 will instead return next
> >  	 * and cases 3, 8 will also update vma to point at next.
> > @@ -951,26 +968,6 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> >  	VM_WARN_ON(curr && (addr != curr->vm_start || end > curr->vm_end));
> >  	VM_WARN_ON(addr >= end);
> >
> > -	if (prev) {
> > -		vma_start = prev->vm_start;
> > -		vma_pgoff = prev->vm_pgoff;
> > -		/* Can we merge the predecessor? */
> > -		if (prev->vm_end == addr && mpol_equal(vma_policy(prev), policy)
> > -		    && can_vma_merge_after(prev, vm_flags, anon_vma, file,
> > -				   pgoff, vm_userfaultfd_ctx, anon_name)) {
> > -			merge_prev = true;
> > -			vma_prev(vmi);
> > -		}
> > -	}
> > -
> > -	/* Can we merge the successor? */
> > -	if (next && mpol_equal(policy, vma_policy(next)) &&
> > -	    can_vma_merge_before(next, vm_flags,
> > -				 anon_vma, file, pgoff+pglen,
> > -				 vm_userfaultfd_ctx, anon_name)) {
> > -		merge_next = true;
> > -	}
> > -
> >  	remove = remove2 = adjust = NULL;
> >  	/* Can we merge both the predecessor and the successor? */
> >  	if (merge_prev && merge_next &&
> > @@ -985,7 +982,7 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> >  				err = dup_anon_vma(prev, curr);
> >  		}
> >  	} else if (merge_prev) {
> > -		err = 0;				/* case 2 */
> > +							/* case 2 */
> >  		if (curr) {
> >  			err = dup_anon_vma(prev, curr);
> >  			if (end == curr->vm_end) {	/* case 7 */
> > @@ -995,7 +992,7 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> >  				adj_start = (end - curr->vm_start);
> >  			}
> >  		}
> > -	} else if (merge_next) {
> > +	} else { /* merge_next */
> >  		res = next;
> >  		if (prev && addr < prev->vm_end) {	/* case 4 */
> >  			vma_end = addr;
> > @@ -1011,7 +1008,7 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> >  			vma_start = addr;
> >  			vma_end = next->vm_end;
> >  			vma_pgoff = next->vm_pgoff;
> > -			err = 0;
> > +
>
> Was this blank line intentional?  I assume so, to give a gap for the
> comment below?  It's probably worth having.
>
> >  			if (curr) {			/* case 8 */
> >  				vma_pgoff = curr->vm_pgoff;
> >  				remove = curr;
> > @@ -1020,7 +1017,7 @@ struct vm_area_struct *vma_merge(struct vma_iterator *vmi, struct mm_struct *mm,
> >  		}
> >  	}
> >
> > -	/* Cannot merge or error in anon_vma clone */
> > +	/* Error in anon_vma clone. */
> >  	if (err)
> >  		return NULL;
> >
> > --
> > 2.39.2
> >
> >

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ