| Message ID | 20211103122157.1215783-5-mcgrof@kernel.org (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | block: add_disk() error handling stragglers | expand |
On Wed, Nov 03, 2021 at 05:21:48AM -0700, Luis Chamberlain wrote: > If nd_integrity_init() fails we'd get del_gendisk() called, > but that's not correct as we should only call that if we're > done with device_add_disk(). Fix this by providing unwinding > prior to the devm call being registered and moving the devm > registration to the very end. > > This should fix calling del_gendisk() if nd_integrity_init() > fails. I only spotted this issue through code inspection. It > does not fix any real world bug. Looks good, Reviewed-by: Christoph Hellwig <hch@lst.de> Should this grow a Fixes tag for the commit adding the problem?
On Wed, Nov 03, 2021 at 05:05:47PM +0100, Christoph Hellwig wrote: > On Wed, Nov 03, 2021 at 05:21:48AM -0700, Luis Chamberlain wrote: > > If nd_integrity_init() fails we'd get del_gendisk() called, > > but that's not correct as we should only call that if we're > > done with device_add_disk(). Fix this by providing unwinding > > prior to the devm call being registered and moving the devm > > registration to the very end. > > > > This should fix calling del_gendisk() if nd_integrity_init() > > fails. I only spotted this issue through code inspection. It > > does not fix any real world bug. > > Looks good, > > Reviewed-by: Christoph Hellwig <hch@lst.de> > > Should this grow a Fixes tag for the commit adding the problem? Given this driver is going to be removed, do we care? The only reason I re-added the patch was Dan could not remove the driver on time. Luis
diff --git a/drivers/nvdimm/blk.c b/drivers/nvdimm/blk.c index b6c6866f9259..4eef67918a7e 100644 --- a/drivers/nvdimm/blk.c +++ b/drivers/nvdimm/blk.c @@ -239,6 +239,7 @@ static int nsblk_attach_disk(struct nd_namespace_blk *nsblk) resource_size_t available_disk_size; struct gendisk *disk; u64 internal_nlba; + int rc; internal_nlba = div_u64(nsblk->size, nsblk_internal_lbasize(nsblk)); available_disk_size = internal_nlba * nsblk_sector_size(nsblk); @@ -255,20 +256,26 @@ static int nsblk_attach_disk(struct nd_namespace_blk *nsblk) blk_queue_logical_block_size(disk->queue, nsblk_sector_size(nsblk)); blk_queue_flag_set(QUEUE_FLAG_NONROT, disk->queue); - if (devm_add_action_or_reset(dev, nd_blk_release_disk, disk)) - return -ENOMEM; - if (nsblk_meta_size(nsblk)) { - int rc = nd_integrity_init(disk, nsblk_meta_size(nsblk)); + rc = nd_integrity_init(disk, nsblk_meta_size(nsblk)); if (rc) - return rc; + goto out_before_devm_err; } set_capacity(disk, available_disk_size >> SECTOR_SHIFT); device_add_disk(dev, disk, NULL); + + /* nd_blk_release_disk() is called if this fails */ + if (devm_add_action_or_reset(dev, nd_blk_release_disk, disk)) + return -ENOMEM; + nvdimm_check_and_set_ro(disk); return 0; + +out_before_devm_err: + blk_cleanup_disk(disk); + return rc; } static int nd_blk_probe(struct device *dev)
If nd_integrity_init() fails we'd get del_gendisk() called, but that's not correct as we should only call that if we're done with device_add_disk(). Fix this by providing unwinding prior to the devm call being registered and moving the devm registration to the very end. This should fix calling del_gendisk() if nd_integrity_init() fails. I only spotted this issue through code inspection. It does not fix any real world bug. Signed-off-by: Luis Chamberlain <mcgrof@kernel.org> --- drivers/nvdimm/blk.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-)