| Message ID | 20210604195900.2096121-9-satyat@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | ensure bios aren't split in middle of crypto data unit | expand |
On Fri, Jun 04, 2021 at 07:58:58PM +0000, Satya Tangirala wrote: > Handle any error from blk_ksm_register() in the callers. Previously, > the callers ignored the return value because blk_ksm_register() wouldn't > fail as long as the request_queue didn't have integrity support too, but > as this is no longer the case, it's safer for the callers to just handle > the return value appropriately. > > Signed-off-by: Satya Tangirala <satyat@google.com> > --- > drivers/md/dm-table.c | 16 +++++++++++++++- > 1 file changed, 15 insertions(+), 1 deletion(-) > > diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c > index 29cbfc3e3c4b..e44f304b5c1a 100644 > --- a/drivers/md/dm-table.c > +++ b/drivers/md/dm-table.c > @@ -1354,7 +1354,21 @@ static void dm_update_keyslot_manager(struct request_queue *q, > > /* Make the ksm less restrictive */ > if (!q->ksm) { > - blk_ksm_register(t->ksm, q); > + /* > + * This WARN_ON should never trigger since t->ksm isn't be > + * "empty" (i.e. will support at least 1 crypto capability), the > + * request queue doesn't support integrity (since > + * dm_table_construct_keyslot_manager() checks that), and > + * it also satisfies all the block layer constraints > + * "sufficiently" (as in the constraints of the DM device's > + * request queue won't preclude any of the intersection of the > + * supported capabilities of the underlying devices, since if > + * some capability were precluded by the DM device's request > + * queue's constraints, that capability would also have been > + * precluded by one of the child device's request queues) > + */ > + if (WARN_ON(!blk_ksm_register(t->ksm, q))) > + dm_destroy_keyslot_manager(t->ksm); This comment seems to be in the wrong place, as dm_update_keyslot_manager() already assumes that the crypto capabilities are either staying the same or expanding. This isn't something new that this WARN_ON() introduces. I think this explanation should go in dm_table_construct_keyslot_manager() instead, as that is where the new set of crypto capabilities is built. I.e. in dm_table_construct_keyslot_manager() we should explain how we "know" that the new set will really be equal or greater. - Eric
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 29cbfc3e3c4b..e44f304b5c1a 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1354,7 +1354,21 @@ static void dm_update_keyslot_manager(struct request_queue *q, /* Make the ksm less restrictive */ if (!q->ksm) { - blk_ksm_register(t->ksm, q); + /* + * This WARN_ON should never trigger since t->ksm isn't be + * "empty" (i.e. will support at least 1 crypto capability), the + * request queue doesn't support integrity (since + * dm_table_construct_keyslot_manager() checks that), and + * it also satisfies all the block layer constraints + * "sufficiently" (as in the constraints of the DM device's + * request queue won't preclude any of the intersection of the + * supported capabilities of the underlying devices, since if + * some capability were precluded by the DM device's request + * queue's constraints, that capability would also have been + * precluded by one of the child device's request queues) + */ + if (WARN_ON(!blk_ksm_register(t->ksm, q))) + dm_destroy_keyslot_manager(t->ksm); } else { blk_ksm_update_capabilities(q->ksm, t->ksm); dm_destroy_keyslot_manager(t->ksm);
Handle any error from blk_ksm_register() in the callers. Previously, the callers ignored the return value because blk_ksm_register() wouldn't fail as long as the request_queue didn't have integrity support too, but as this is no longer the case, it's safer for the callers to just handle the return value appropriately. Signed-off-by: Satya Tangirala <satyat@google.com> --- drivers/md/dm-table.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-)