On Mon 19-01-26 07:22:45, Christoph Hellwig wrote:
> Use the kernel's resizable hash table to find the fsverity_info. This
> way file systems that want to support fsverity don't have to bloat
> every inode in the system with an extra pointer. The tradeoff is that
> looking up the fsverity_info is a bit more expensive now, but the main
> operations are still dominated by I/O and hashing overhead.
>
> Signed-off-by: Christoph Hellwig <[email protected]>
...
> diff --git a/fs/verity/enable.c b/fs/verity/enable.c
> index 95ec42b84797..91cada0d455c 100644
> --- a/fs/verity/enable.c
> +++ b/fs/verity/enable.c
> @@ -264,9 +264,24 @@ static int enable_verity(struct file *filp,
> goto rollback;
> }
>
> + /*
> + * Add the fsverity_info into the hash table before finishing the
> + * initialization. This ensures we don't have to undo the enabling when
> + * memory allocation for the hash table fails. This is safe because
> + * looking up the fsverity_info always first checks the S_VERITY flag on
> + * the inode, which will only be set at the very end of the
> + * ->end_enable_verity method.
> + */
> + err = fsverity_set_info(vi);
> + if (err)
> + goto rollback;
OK, but since __fsverity_get_info() is just rhashtable_lookup_fast() what
prevents the CPU from reordering the hash table reads before the S_VERITY
check? I think you need a barrier in fsverity_get_info() to enforce the
proper ordering. The matching ordering during setting of S_VERITY is
implied by cmpxchg used to manipulate i_flags so that part should be fine.
Honza
--
Jan Kara <[email protected]>
SUSE Labs, CR
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
