On Wed, Apr 19, 2023 at 06:37:05PM +0300, Dan Carpenter wrote: > On Wed, Apr 19, 2023 at 10:36:49AM -0400, William Breathitt Gray wrote: > > The lock is acquired by rz_mtu3_lock_if_counter_is_valid(), so that > > function needs a sparse __acquires(&priv->lock) annotation too. > > I found this bug using Smatch. It's a competing static checker which > uses Sparse as a parser. I am the author of Smatch so I am naturally > biased. > > I don't think it's as simple as that. I don't think Sparse has > annotations for mutexes, only for spinlocks? Also it's really > complicated to annotate something as taking the lock on the success path > but not on the failure path. You have to set up a wrapper and use > __cond_lock(). > > Every other feature in Sparse is awesome, but for locking, it's better > to just use Smatch. > > regards, > dan carpenter Ah that's a fair point, I can see how involved that would be to set up a wrapper and handle the various paths correctly. The marginal benefit just doesn't seem worth the effort afterall. William Breathitt Gray
Attachment:
signature.asc
Description: PGP signature
