On Wed, Apr 02, 2008 at 04:25:12PM +1000, Barry Naujok wrote: > It also introduces a new type, xfs_name, which is similar to the > dentry cache's qstr type. It contains a pointer to a zone allocated > string (MAXNAMELEN sized) and the length of the actual name. This > string does not need to be NULL terminated (a counted string). > > xfs_name_t is only used in the lookup path for this patch, but may > be used in other locations too if desired. It maybe desirable not > to use xfs_name_t at all in the lookup functions but stick to > separate parameters (which will mean 7 instead of 5 arguments). wouldn't it be a lot cleaner to just pass the struct qstr all the way down instead of inventing an almost the same type? > To avoid polluting the dcache, we implement a new directory inode > operations for lookup. xfs_vn_ci_lookup() interacts directly with > the dcache and the code was derived from ntfs_lookup() in > fs/ntfs/namei.c. The dentry hash and compare overrides introduced > in the ASCII-CI patch has been removed. So maybe we should introduce the method in that patch already? Especially as the old one is buggy and the new one should be one ore more helpers in generic code it could easily be a patch of it's own. > + /* > + * Directory without a 'disconnected' dentry; we need to do > + * d_instantiate() by hand because it takes dcache_lock which > + * we already hold. > + */ > + list_add(&result->d_alias, &inode->i_dentry); > + result->d_inode = inode; > + spin_unlock(&dcache_lock); > + security_d_instantiate(result, inode); please add a documented __d_instanciated helper in common code that is used by this and d_instanciate. > +typedef struct qstr bhv_vstr_t; we try to get rid of the bhv_v* names, they're ugly amd their naming is outdated. Just use struct qstr until some port comes up with s conflicting one. > + return kmem_zone_zalloc(xfs_da_name_zone, KM_SLEEP); please don't introduce new KM_SLEEP allocations. > + kmem_zone_free(xfs_da_name_zone, (void *)name); this cast should not be needed. -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
