search for: trasnlate_desc

Hi all, In an attempt to remove the remaining traces of [smp_]read_barrier_depends() following my previous patches to strengthen READ_ONCE() for Alpha [1], I ended up trying to decipher the read_barrier_depends() usage in the vhost driver: --->8 // drivers/vhost/vhost.c static int get_indirect(struct vhost_virtqueue *vq, struct iovec iov[], unsigned int iov_size, unsigned int *out_num,

Hi all, In an attempt to remove the remaining traces of [smp_]read_barrier_depends() following my previous patches to strengthen READ_ONCE() for Alpha [1], I ended up trying to decipher the read_barrier_depends() usage in the vhost driver: --->8 // drivers/vhost/vhost.c static int get_indirect(struct vhost_virtqueue *vq, struct iovec iov[], unsigned int iov_size, unsigned int *out_num,

...ybody know which accesses are being ordered here? Usually you'd need > > a READ_ONCE()/rcu_dereference() beginning the chain, but I haven't managed > > to find one... > > > > I guess it was because we will read from the address stored in the iov like: > > 1) trasnlate_desc() that stores the userspace buffer pointer in the iov > > 2) copy_from_iter() that reads from those pointers Isn't that exactly the same flow as vhost_copy_from_user(), which doesn't have the barrier? Staring at the code some more, my best bet at the moment is that the load of '...

...ybody know which accesses are being ordered here? Usually you'd need > > a READ_ONCE()/rcu_dereference() beginning the chain, but I haven't managed > > to find one... > > > > I guess it was because we will read from the address stored in the iov like: > > 1) trasnlate_desc() that stores the userspace buffer pointer in the iov > > 2) copy_from_iter() that reads from those pointers Isn't that exactly the same flow as vhost_copy_from_user(), which doesn't have the barrier? Staring at the code some more, my best bet at the moment is that the load of '...

...is). > > Does anybody know which accesses are being ordered here? Usually you'd need > a READ_ONCE()/rcu_dereference() beginning the chain, but I haven't managed > to find one... > > Thanks, I guess it was because we will read from the address stored in the iov like: 1) trasnlate_desc() that stores the userspace buffer pointer in the iov 2) copy_from_iter() that reads from those pointers So we need a data dependency barrier in the middle as explained in the memory-barriers.txt? (since READ_ONCE is not used in iov iterator). Thanks > > Will > > [1] c2bc66082e10...

...ich accesses are being ordered here? Usually you'd need >>> a READ_ONCE()/rcu_dereference() beginning the chain, but I haven't managed >>> to find one... >>> >> I guess it was because we will read from the address stored in the iov like: >> >> 1) trasnlate_desc() that stores the userspace buffer pointer in the iov >> >> 2) copy_from_iter() that reads from those pointers > Isn't that exactly the same flow as vhost_copy_from_user(), which doesn't > have the barrier? There's a rmb before calling vhost_copy_from_user() (vhost_g...