On Mon, Aug 31, 2009 at 08:24:48AM -0400, Theodore Tso wrote: > On Mon, Aug 31, 2009 at 12:00:06PM +0530, Aneesh Kumar K.V wrote: > > Below are the possibilities i looked at > > > > a) mmap with no parallel write to the same offset. That would mean > > we don't have attached buffer heads because nobody attach buffer > > heads to the page. > > > > b) mmap happening to the hole. The buffer heads are not mapped. > > > > c) mmap with parallel write to the same offset. The parallel write > > did attach mapped buffer heads to the same page. So we should find > > all buffer heads mapped in the above case. > > > > if we will find buffer heads already be mapped in many workloads then > > i guess it make sense to add page lock. It will also avoid the > > journal_start that we do in write_begin. I will redo the patch > > The usage case I was worried about is the one where we are mmap'ing an > existing file (say, like an Oracle or DB2 table space, or a berkdb > database file), and we are writing into already allocated blocks. In > that case (which does use these code paths, right?) the second time we > write a particular page, the buffer heads will already be mapped. If the database is not being updated via a write(2), then even though the blocks are already allocated, we won't find buffer_heads attached to the page. ie, page_buffers(page) will be NULL The page_mkwrite -> write_begin path would be allocating the buffer_heads and attaching them to the page. So even in the above case we will be doing write_begin -> write_end. That is, it is similar to the (a) i wrote above. > > For database applications where we aren't loading a table, but just > making changes to an already instantiated table, the buffer heads > would be mapped most of the time, would they not? > > - Ted -aneesh -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
