netfilter: arptables: use xt_entry_foreach() in copy_entries_to_user()

[PATCH net-next] netfilter: arptables: use xt_entry_foreach() in copy_entries_to_user()

Posted by Kshitiz Bartariya 2 weeks, 5 days ago

Replace the manual offset-based iteration with xt_entry_foreach(),
thereby removing FIXME. The byte offset semantics and user ABI
are preserved.

Signed-off-by: Kshitiz Bartariya <kshitiz.bartariya@zohomail.in>
---
 net/ipv4/netfilter/arp_tables.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
index 1cdd9c28ab2d..9f82ce0fcaa5 100644
--- a/net/ipv4/netfilter/arp_tables.c
+++ b/net/ipv4/netfilter/arp_tables.c
@@ -684,12 +684,11 @@ static int copy_entries_to_user(unsigned int total_size,
 
 	loc_cpu_entry = private->entries;
 
-	/* FIXME: use iterator macros --RR */
-	/* ... then go back and fix counters and names */
-	for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
+	num = 0;
+	xt_entry_foreach(e, loc_cpu_entry, total_size) {
 		const struct xt_entry_target *t;
 
-		e = loc_cpu_entry + off;
+		off = (unsigned char *)e - (unsigned char *)loc_cpu_entry;
 		if (copy_to_user(userptr + off, e, sizeof(*e))) {
 			ret = -EFAULT;
 			goto free_counters;
@@ -707,6 +706,7 @@ static int copy_entries_to_user(unsigned int total_size,
 			ret = -EFAULT;
 			goto free_counters;
 		}
+		num++;
 	}
 
  free_counters:
-- 
2.50.1 (Apple Git-155)

Re: [PATCH net-next] netfilter: arptables: use xt_entry_foreach() in copy_entries_to_user()

Posted by Simon Horman 2 weeks, 2 days ago

On Mon, Jan 19, 2026 at 12:07:04PM +0530, Kshitiz Bartariya wrote:
> Replace the manual offset-based iteration with xt_entry_foreach(),
> thereby removing FIXME. The byte offset semantics and user ABI
> are preserved.
> 
> Signed-off-by: Kshitiz Bartariya <kshitiz.bartariya@zohomail.in>
> ---
>  net/ipv4/netfilter/arp_tables.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
> index 1cdd9c28ab2d..9f82ce0fcaa5 100644
> --- a/net/ipv4/netfilter/arp_tables.c
> +++ b/net/ipv4/netfilter/arp_tables.c
> @@ -684,12 +684,11 @@ static int copy_entries_to_user(unsigned int total_size,
>  
>  	loc_cpu_entry = private->entries;
>  
> -	/* FIXME: use iterator macros --RR */
> -	/* ... then go back and fix counters and names */
> -	for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
> +	num = 0;
> +	xt_entry_foreach(e, loc_cpu_entry, total_size) {
>  		const struct xt_entry_target *t;
>  
> -		e = loc_cpu_entry + off;
> +		off = (unsigned char *)e - (unsigned char *)loc_cpu_entry;

This offset calculation makes me feel queasy.

Can the code start with off = 0 and increment it by e->next_offset
as the loop iterates, as was the case before this patch?
It would be similar to how num is handled.

>  		if (copy_to_user(userptr + off, e, sizeof(*e))) {
>  			ret = -EFAULT;
>  			goto free_counters;
> @@ -707,6 +706,7 @@ static int copy_entries_to_user(unsigned int total_size,
>  			ret = -EFAULT;
>  			goto free_counters;
>  		}
> +		num++;
>  	}
>  
>   free_counters:
> -- 
> 2.50.1 (Apple Git-155)
>

Re: [PATCH net-next] netfilter: arptables: use xt_entry_foreach() in copy_entries_to_user()

Posted by Florian Westphal 2 weeks, 2 days ago

Simon Horman <horms@kernel.org> wrote:
> On Mon, Jan 19, 2026 at 12:07:04PM +0530, Kshitiz Bartariya wrote:
> > Replace the manual offset-based iteration with xt_entry_foreach(),
> > thereby removing FIXME. The byte offset semantics and user ABI
> > are preserved.
> > 
> > Signed-off-by: Kshitiz Bartariya <kshitiz.bartariya@zohomail.in>
> > ---
> >  net/ipv4/netfilter/arp_tables.c | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> > 
> > diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
> > index 1cdd9c28ab2d..9f82ce0fcaa5 100644
> > --- a/net/ipv4/netfilter/arp_tables.c
> > +++ b/net/ipv4/netfilter/arp_tables.c
> > @@ -684,12 +684,11 @@ static int copy_entries_to_user(unsigned int total_size,
> >  
> >  	loc_cpu_entry = private->entries;
> >  
> > -	/* FIXME: use iterator macros --RR */
> > -	/* ... then go back and fix counters and names */
> > -	for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
> > +	num = 0;
> > +	xt_entry_foreach(e, loc_cpu_entry, total_size) {
> >  		const struct xt_entry_target *t;
> >  
> > -		e = loc_cpu_entry + off;
> > +		off = (unsigned char *)e - (unsigned char *)loc_cpu_entry;
> 
> This offset calculation makes me feel queasy.
> 
> Can the code start with off = 0 and increment it by e->next_offset
> as the loop iterates, as was the case before this patch?
> It would be similar to how num is handled.

I think this file should just be left alone resp. should be restricted
to bug fixes only.  Same for ip/ip6/ebtables.

Its too early to remove this file from the tree, but I don't see
value in making cosmetic improvements, sorry.

Re: [PATCH net-next] netfilter: arptables: use xt_entry_foreach() in copy_entries_to_user()

Posted by Simon Horman 2 weeks, 2 days ago

On Wed, Jan 21, 2026 at 11:05:21PM +0100, Florian Westphal wrote:
> Simon Horman <horms@kernel.org> wrote:
> > On Mon, Jan 19, 2026 at 12:07:04PM +0530, Kshitiz Bartariya wrote:
> > > Replace the manual offset-based iteration with xt_entry_foreach(),
> > > thereby removing FIXME. The byte offset semantics and user ABI
> > > are preserved.
> > > 
> > > Signed-off-by: Kshitiz Bartariya <kshitiz.bartariya@zohomail.in>
> > > ---
> > >  net/ipv4/netfilter/arp_tables.c | 8 ++++----
> > >  1 file changed, 4 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/net/ipv4/netfilter/arp_tables.c b/net/ipv4/netfilter/arp_tables.c
> > > index 1cdd9c28ab2d..9f82ce0fcaa5 100644
> > > --- a/net/ipv4/netfilter/arp_tables.c
> > > +++ b/net/ipv4/netfilter/arp_tables.c
> > > @@ -684,12 +684,11 @@ static int copy_entries_to_user(unsigned int total_size,
> > >  
> > >  	loc_cpu_entry = private->entries;
> > >  
> > > -	/* FIXME: use iterator macros --RR */
> > > -	/* ... then go back and fix counters and names */
> > > -	for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
> > > +	num = 0;
> > > +	xt_entry_foreach(e, loc_cpu_entry, total_size) {
> > >  		const struct xt_entry_target *t;
> > >  
> > > -		e = loc_cpu_entry + off;
> > > +		off = (unsigned char *)e - (unsigned char *)loc_cpu_entry;
> > 
> > This offset calculation makes me feel queasy.
> > 
> > Can the code start with off = 0 and increment it by e->next_offset
> > as the loop iterates, as was the case before this patch?
> > It would be similar to how num is handled.
> 
> I think this file should just be left alone resp. should be restricted
> to bug fixes only.  Same for ip/ip6/ebtables.
> 
> Its too early to remove this file from the tree, but I don't see
> value in making cosmetic improvements, sorry.

Thanks Florian,

No objection to that approach from my side.