Re: [patch 3/3] radix-tree: RCU lockless readside

[Nick Piggin]

Andrew Morton wrote:
> On Thu, 22 Jun 2006 20:11:12 +0200
> Nick Piggin <npiggin@xxxxxxx> wrote:
>
>
> > On Thu, Jun 22, 2006 at 10:40:57AM -0700, Paul E. McKenney wrote:
> >
> > > On Thu, Jun 22, 2006 at 06:55:51PM +0200, Nick Piggin wrote:
> > >
> > > > No problem, will change.
> > >
> > > Thank you!
> >
> > OK, and with that I believe I've covered all your concerns.
> >
> > Attached is the incremental patch (plus a little bit of fuzz
> > that's gone to Andrew). The big items are:
> >
> > - documentation, clarification of comments
> > - tag lookups are now RCU safe (tested with harness)
> > - cleanups of various misuses of rcu_ API that Paul spotted
> > - thought I might put in a copyright -- is this OK?
> >
> > Andrew, please apply.
>
>
> Freeing unused kernel memory: 316k freed
> Write protecting the kernel read-only data: 384k
> No module found in object
> No module found in object
> No module found in object
> No module found in object
> input: AT Translated Set 2 keyboard as /class/input/input0
> No module found in object
> EXT3-fs: INFO: recovery required on readonly filesystem.
> EXT3-fs: write access will be enabled during recovery.
> kjournald starting.  Commit interval 5 seconds
> EXT3-fs: recovery complete.
> EXT3-fs: mounted filesystem with ordered data mode.
> BUG: NMI Watchdog detected LOCKUP on CPU0, eip c0264345, registers:
> CPU:    0
> EIP is at radix_tree_gang_lookup_tag+0x105/0x1a0
> eax: ffffffff   ebx: 00000040   ecx: ffffffc0   edx: 00000007
> esi: e701e9d8   edi: 000001c0   ebp: e6fbddd8   esp: e6fbdda8
> ds: 007b   es: 007b   ss: 0068
> Process fsck.ext3 (pid: 1565, ti=e6fbc000 task=c1fbcb90 task.ti=e6fbc000)
> Stack: e77f2dc4 e701e9d8 e701e9d8 00000002 00000fff 00000000 e701e8c8 e6fbde60 
>        0000000e c1c6c52c e6fbde60 c1c6c538 e6fbde00 c014b68f c1c6c52c e6fbde60 
>        00000000 0000000e 00000000 e6fbde58 00000000 00000001 e6fbde20 c0155631 
> Call Trace:
> Code: 89 fa 8d 4c 09 fa d3 e3 d3 ea 89 d9 83 e2 3f f7 d9 eb 13 8d 76 00 89 f8 89 df 21 c8 01 c7 74 26 42 83 fa 40 74 95 0f a3 16 19 c0 <85> c0 74 e7 83 7d dc 01 74 3a 31 f6 89 75 f0 e9 6e ff ff ff c7 
> console shuts up ...
>
>
> Not sure why, either.  It all looks like an equivalent transformation to
> me.
>
> fwiw, here's what I tested:


Arggh, line 755 has an extra semicolon. I caught and fixed this in the rtth
tree, but obviously forgot to transfer it over.

> @@ -741,15 +740,19 @@ __lookup_tag(struct radix_tree_node *slo
>  			for ( ; j < RADIX_TREE_MAP_SIZE; j++) {
>  				index++;
>  				if (tag_get(slot, tag, j)) {
> -					BUG_ON(slot->slots[j] == NULL);
> -					results[nr_found++] = slot->slots[j];
> -					if (nr_found == max_items)
> -						goto out;
> +					struct radix_tree_node *node = slot->slots[j];
> +					if (node) {
> +						results[nr_found++] = rcu_dereference(node);
> +						if (nr_found == max_items)
> +							goto out;
> +					}
>  				}
>  			}
>  		}
>  		shift -= RADIX_TREE_MAP_SHIFT;
> -		slot = slot->slots[i];
> +		slot = rcu_dereference(slot->slots[i]);
> +		if (slot == NULL);
> +			break;
>  	}

                         ^^^^^^^^

Up there.

--
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/