Re: multiprocessor kernel problem

[Roger Crandell <rwc@lanl.gov>]

Rusty,

Excellent!  I applied the patch and netfilter is happy with 4 processors.  Thank you and
thanks to everyone else who sent suggestions.


Roger



Rusty Russell wrote:

> In message <m2elzp3uiq.fsf@euler.axel.nom> you write:
> > yes, but is it a dual machine or is it an N-way SMP with N > 2?  the
> > other guy with iptables/SMP problems also has a quad box.  could this
> > perhaps be a problem only when you have more than two processors?
>
> Yes, hacked my machine to think it had 4 cpus, and boom.
>
> There are two problems:
> (1) initialization of multiple tables was wrong, and
> (2) iterating through tables should not use cpu_number_map (doesn't
>     matter on X86 though).
>
> Please try attached patch.
>
> Thanks,
> Rusty,
> --
> Hacking time.
> --- working-2.4.0-test11-5/net/ipv4/netfilter/ip_tables.c.~1~   Sat Aug 12 00:23:40 2000
> +++ working-2.4.0-test11-5/net/ipv4/netfilter/ip_tables.c       Mon Dec  4 16:12:44 2000
> @@ -89,10 +89,8 @@
>         unsigned int hook_entry[NF_IP_NUMHOOKS];
>         unsigned int underflow[NF_IP_NUMHOOKS];
>
> -       char padding[SMP_ALIGN((NF_IP_NUMHOOKS*2+2)*sizeof(unsigned int))];
> -
>         /* ipt_entry tables: one per CPU */
> -       char entries[0];
> +       char entries[0] __attribute__((aligned(SMP_CACHE_BYTES)));
>  };
>
>  static LIST_HEAD(ipt_target);
> @@ -101,7 +99,7 @@
>  #define ADD_COUNTER(c,b,p) do { (c).bcnt += (b); (c).pcnt += (p); } while(0)
>
>  #ifdef CONFIG_SMP
> -#define TABLE_OFFSET(t,p) (SMP_ALIGN((t)->size)*cpu_number_map(p))
> +#define TABLE_OFFSET(t,p) (SMP_ALIGN((t)->size)*(p))
>  #else
>  #define TABLE_OFFSET(t,p) 0
>  #endif
> @@ -283,7 +281,8 @@
>         read_lock_bh(&table->lock);
>         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
>         table_base = (void *)table->private->entries
> -               + TABLE_OFFSET(table->private, smp_processor_id());
> +               + TABLE_OFFSET(table->private,
> +                              cpu_number_map(smp_processor_id()));
>         e = get_entry(table_base, table->private->hook_entry[hook]);
>
>  #ifdef CONFIG_NETFILTER_DEBUG
> @@ -860,7 +859,7 @@
>
>         /* And one copy for every other CPU */
>         for (i = 1; i < smp_num_cpus; i++) {
> -               memcpy(newinfo->entries + SMP_ALIGN(newinfo->size*i),
> +               memcpy(newinfo->entries + SMP_ALIGN(newinfo->size)*i,
>                        newinfo->entries,
>                        SMP_ALIGN(newinfo->size));
>         }
> @@ -1359,7 +1358,7 @@
>         int ret;
>         struct ipt_table_info *newinfo;
>         static struct ipt_table_info bootstrap
> -               = { 0, 0, { 0 }, { 0 }, { }, { } };
> +               = { 0, 0, { 0 }, { 0 }, { } };
>
>         MOD_INC_USE_COUNT;
>         newinfo = vmalloc(sizeof(struct ipt_table_info)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/