null_blk: Null pointer deference problem in alloc_page_buffers
authorRaghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Tue, 21 Jan 2014 11:29:59 +0000 (16:59 +0530)
committerJens Axboe <axboe@kernel.dk>
Wed, 22 Jan 2014 04:22:40 +0000 (20:22 -0800)
 If we load the null_blk module with bs=8k we get following oops:
[ 3819.812190] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
[ 3819.812387] IP: [<ffffffff81170aa5>] create_empty_buffers+0x28/0xaf
[ 3819.812527] PGD 219244067 PUD 215a06067 PMD 0
[ 3819.812640] Oops: 0000 [#1] SMP
[ 3819.812772] Modules linked in: null_blk(+)

 Fix that by resetting block size to PAGE_SIZE if it is greater than PAGE_SIZE

Reported-by: Sumanth <sumantk2@linux.vnet.ibm.com>
Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com>
Reviewed-by: Matias Bjorling <m@bjorling.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/block/null_blk.c

index a2e69d2..a01d7f6 100644 (file)
@@ -622,6 +622,11 @@ static int __init null_init(void)
                irqmode = NULL_IRQ_NONE;
        }
 #endif
+       if (bs > PAGE_SIZE) {
+               pr_warn("null_blk: invalid block size\n");
+               pr_warn("null_blk: defaults block size to %lu\n", PAGE_SIZE);
+               bs = PAGE_SIZE;
+       }
 
        if (queue_mode == NULL_Q_MQ && use_per_node_hctx) {
                if (submit_queues < nr_online_nodes) {