Hi Jens,
I caught the backtrace below with the libnvdimm unit test. It's hard
to hit, and as far as I can tell it's been present for a while.
Wanted to get a second opinion on the potential fix.
The failure address of 0000000000000540 corresponds offset 1344 in
struct gendisk according to pahole.
struct gendisk {
[..]
struct request_queue * queue; /* 1344 8 */
[..]
}
which means that bdev->bd_disk was NULL. I think this means that
blk_get_backing_dev_info() should not assume that it can just use
bdev_get_queue() to get to the bdi. It seems it needs to hold its own
blkdev_get() reference.
[ 44.513201] BUG: unable to handle kernel NULL pointer dereference
at 0000000000000540
[ 44.515051] IP: [<ffffffff81485a80>] blk_get_backing_dev_info+0x10/0x20
[..]
[ 44.556549] Call Trace:
[ 44.557260] [<ffffffff812c6b19>] __inode_attach_wb+0x319/0x5c0
[ 44.560203] [<ffffffff811ef752>] __filemap_fdatawrite_range+0xf2/0x100
[ 44.561399] [<ffffffff811ef7d0>] filemap_write_and_wait+0x40/0x90
[ 44.562551] [<ffffffff812d6c69>] fsync_bdev+0x49/0x50
[ 44.563570] [<ffffffff8149f474>] invalidate_partition+0x24/0x50
[ 44.564693] [<ffffffff814a0062>] del_gendisk+0xf2/0x210
[ 44.565740] [<ffffffffa003b162>] pmem_release_disk+0x12/0x20 [nd_pmem]
[ 44.566933] [<ffffffff81615e9f>] devm_action_release+0xf/0x20
[ 44.568033] [<ffffffff81616979>] release_nodes+0x129/0x230
[ 44.569109] [<ffffffff81616cec>] devres_release_all+0x3c/0x60
[ 44.570216] [<ffffffff81611c49>] __device_release_driver+0xa9/0x160
[ 44.571384] [<ffffffff81611d25>] device_release_driver+0x25/0x40
--
To unsubscribe from this list: send the line "unsubscribe linux-block" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
