On Tue, 07 May 2013 23:38:50 +0900 (JST), Ryusuke Konishi wrote: > Hi Vyacheslav, > On Tue, 07 May 2013 16:49:37 +0400, Vyacheslav Dubeyko wrote: >> Hi Ryusuke, >> >> On Mon, 2013-05-06 at 23:07 +0900, Ryusuke Konishi wrote: >> >> [snip] >>> I agree to use count of descriptor blocks of ifile to >>> calculate the approximate value of the total file nodes. >>> >>> Here are my comments on the patch: >>> >>> (1) Consider using nilfs_bmap_last_key() instead of repeating read >>> of descriptor blocks to decide new desc_blocks count. >>> The current logic may incur many device read accesses >>> even if it is updated in a stepwise fashion with >>> NILFS_IFILE_I(ifile)->mi.mi_desc_blocks_count. >>> [snip] >> Do you mean that nilfs_bmap_last_key() can initiate lesser read requests >> as my way of using nilfs_palloc_get_desc_block() call? > > Yes, nilfs_bmap_last_key() does not involve read of leaf blocks of > nilfs btree (i.e. data blocks including descriptor block, etc) unlike > with nilfs_palloc_get_desc_block function, however it is negligible > anyway as you pointed out. Oops, I have missed the point. Most important difference in these two functions is that you do not need to repeat calling the function in the case of nilfs_bmap_last_key(). nilfs_bmap_last_key() returns the maximum index of allocated data blocks with one call. Regards, Ryusuke Konishi -- 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
