Message ID | alpine.LSU.2.00.1209271324060.2184@eggly.anvils |
---|---|
State | Superseded, archived |
Headers | show |
On Thu, Sep 27, 2012 at 01:31:28PM -0700, Hugh Dickins wrote: > Kernel build with CONFIG_DEBUG_SLAB or CONFIG_SLUB_DEBUG slub_debug=FPZ > gives me kernel BUG at fs/ext4/extents_status.c:142! That's the > BUG_ON(es->start + es->len < es->start) in extent_status_end() called > from ext4_es_insert_extent(). tree->cache_es has been freed and poisoned. > > This comes from when ext4_es_try_to_merge_left() merges es into leftward > es1, but ext4_es_insert_extent()'s out then updates cache_es to the freed > extent_status. ext4_es_try_to_merge_right() does not pose a problem. > > Change ext4_es_try_to_merge_left() to return whichever extent_status > should be recorded in tree->cache_es. Remove cache_es update from > both of them, leaving that to ext4_es_insert_extent()'s out label. > > Signed-off-by: Hugh Dickins <hughd@google.com> Hugh, thanks for finding this bug! Zheng, you were going to send me an updated patch series; can you take care of merging this patch into your patch series (and crediting Hugh with a Signed-off-by as appropriate)? Thanks!! - Ted -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Thu, Sep 27, 2012 at 04:39:20PM -0400, Theodore Ts'o wrote: > On Thu, Sep 27, 2012 at 01:31:28PM -0700, Hugh Dickins wrote: > > Kernel build with CONFIG_DEBUG_SLAB or CONFIG_SLUB_DEBUG slub_debug=FPZ > > gives me kernel BUG at fs/ext4/extents_status.c:142! That's the > > BUG_ON(es->start + es->len < es->start) in extent_status_end() called > > from ext4_es_insert_extent(). tree->cache_es has been freed and poisoned. > > > > This comes from when ext4_es_try_to_merge_left() merges es into leftward > > es1, but ext4_es_insert_extent()'s out then updates cache_es to the freed > > extent_status. ext4_es_try_to_merge_right() does not pose a problem. > > > > Change ext4_es_try_to_merge_left() to return whichever extent_status > > should be recorded in tree->cache_es. Remove cache_es update from > > both of them, leaving that to ext4_es_insert_extent()'s out label. > > > > Signed-off-by: Hugh Dickins <hughd@google.com> > > Hugh, thanks for finding this bug! > > Zheng, you were going to send me an updated patch series; can you take > care of merging this patch into your patch series (and crediting Hugh > with a Signed-off-by as appropriate)? Yeah, thanks for fixing it, although I have fixed this problem in my new patch set. I will add a Signed-off-by into the patch. Regards, Zheng -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
--- mmotm/fs/ext4/extents_status.c 2012-09-26 10:15:29.340071552 -0700 +++ linux/fs/ext4/extents_status.c 2012-09-27 11:52:59.284937056 -0700 @@ -244,24 +244,24 @@ static void ext4_es_free_extent(struct e kmem_cache_free(ext4_es_cachep, es); } -static void ext4_es_try_to_merge_left(struct ext4_es_tree *tree, - struct extent_status *es) +static struct extent_status * +ext4_es_try_to_merge_left(struct ext4_es_tree *tree, struct extent_status *es) { struct extent_status *es1; struct rb_node *node; node = rb_prev(&es->rb_node); if (!node) - return; + return es; es1 = rb_entry(node, struct extent_status, rb_node); if (es->start == extent_status_end(es1) + 1) { es1->len += es->len; rb_erase(&es->rb_node, &tree->root); - if (es == tree->cache_es) - tree->cache_es = es1; ext4_es_free_extent(es); + es = es1; /* Caller will update tree->cache_es to this */ } + return es; } static void ext4_es_try_to_merge_right(struct ext4_es_tree *tree, @@ -278,9 +278,8 @@ static void ext4_es_try_to_merge_right(s if (es1->start == extent_status_end(es) + 1) { es->len += es1->len; rb_erase(node, &tree->root); - if (es1 == tree->cache_es) - tree->cache_es = es; ext4_es_free_extent(es1); + /* Caller will update tree->cache_es to es */ } } @@ -318,7 +317,7 @@ int ext4_es_insert_extent(struct inode * es_debug("cached by [%u/%u)\n", es->start, es->len); es->start = offset; es->len += len; - ext4_es_try_to_merge_left(tree, es); + es = ext4_es_try_to_merge_left(tree, es); goto out; } else if (es && in_range(offset, es->start, es->len)) { es_debug("cached by [%u/%u)\n", es->start, es->len);
Kernel build with CONFIG_DEBUG_SLAB or CONFIG_SLUB_DEBUG slub_debug=FPZ gives me kernel BUG at fs/ext4/extents_status.c:142! That's the BUG_ON(es->start + es->len < es->start) in extent_status_end() called from ext4_es_insert_extent(). tree->cache_es has been freed and poisoned. This comes from when ext4_es_try_to_merge_left() merges es into leftward es1, but ext4_es_insert_extent()'s out then updates cache_es to the freed extent_status. ext4_es_try_to_merge_right() does not pose a problem. Change ext4_es_try_to_merge_left() to return whichever extent_status should be recorded in tree->cache_es. Remove cache_es update from both of them, leaving that to ext4_es_insert_extent()'s out label. Signed-off-by: Hugh Dickins <hughd@google.com> --- fs/ext4/extents_status.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html