![]() | |||||||||||
Messages in this thread |
On Sat, Apr 30, 2005 at 11:07:53AM -0700, Mingming Cao wrote: > On Sat, 2005-04-30 at 22:41 +0530, Suparna Bhattacharya wrote: > > On Sat, Apr 30, 2005 at 09:40:43PM +0530, Suparna Bhattacharya wrote: > > > On Fri, Apr 29, 2005 at 04:22:59PM -0700, Mingming Cao wrote: > > > > On Fri, 2005-04-29 at 11:45 -0700, Badari Pulavarty wrote: > > > > > I touch tested your patch earlier and seems to work fine. Lets integrate > > > > > Mingming's getblocks() patches with this and see if we get any benifit > > > > > from the whole effort. > > > > > > > > > > > > > I tried Suparna's mpage_writepages_getblocks patch with my > > > > ext3_get_blocks patch, seems to work fine, except that still only one > > > > block is allocated at a time. I got a little confused.... > > > > > > > > I did not see any delayed allocation code in your patch, I assume you > > > > have to update ext3_prepare_write to not call ext3_get_block, so that > > > > block allocation will be defered at ext3_writepages time. So without the > > > > delayed allocation part, the get_blocks in mpage_writepages is doing > > > > multiple blocks look up only, right? > > > > > > That's right - so you could try it with mmapped writes (which don't > > > go through a prepare write) or with Badari's patch to not call > > > ext3_get_block in prepare write. > > > > BTW, I hope you are running with NOBH. > > No, it was not run with NOBH. Why we need to turn on nobh here? If the page has buffers, then get_blocks won't be invoked -- it either finds all the buffers in a page to be mapped contiguous, in which case it can directly issue the IO or enters the confused case where it goes through block_write_full_page. > > There are some issue running fsx tests with both patches, w/o direct io > option. I will spend more time on this next week. OK, I can take a look at the fsx logs, just in case it is similar to something I've seen before. Regards Suparna -- Suparna Bhattacharya (suparna@in.ibm.com) Linux Technology Center IBM Software Lab, India - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ | ||||||||||
| Last update: 2005-05-02 06:41 [from the cache] ©2003-2008 | |||||||||||