On Thu, 27 May 2010 17:10:43 -0300 Daniel Henrique Debonzi <debonzi at linux.vnet.ibm.com> wrote: > FUJITA Tomonori wrote: > > On Tue, 18 May 2010 16:21:00 -0300 > > Daniel Henrique Debonzi <debonzi at linux.vnet.ibm.com> wrote: > > > >> One idea to solve it and don't make necessary to touch all backing > >> stores to avoid this issue is to create a new template like bs_rdrw_sync > >> O_SYNC and O_DIRECT flags on open, and don't have the options as this > >> patch does. This way I got want I am interested in and don't need to > >> involve all the other stuff. What are your thoughts about it? > > > > Or a new option to touch only bs_rdrw. > > > > However, I want to know why you want to use O_SYNC and O_DIRECT. > > > > - what's the advantage of O_SYNC compared with the current approach > > (sync_file after write)? notably faster? > > We would like to be able to control those flags because for our case it > has better performance. Some tests done recently showed that we can get > good better results when using O_DIRECT than sync_file. I see. Thanks for the explanation. As we discussed, "--bsoflags" can't be applied to all the bs modes equally (e.g. bs_aio can use only O_DIRECT). But I guess that creating multiple options such as --bsoflags_rw, --bsoflags_aio, etc is too overdoing. I'll take the approach add only "--bsoflags". --bsoflags="async direct" is not consistent with the other options. Let's use "," like --bsoflags="async,direct". I don't think that we need "none" or "async". Let's support "sync" and "direct". Can you please send the updated patch? > > - why do you want to use O_DIRECT with bd_rdrw? bd_aio does aio with > > O_DIRECT? It's not enough for you? > > AIO does not work with the kernel version we are using at moment. Understood. -- To unsubscribe from this list: send the line "unsubscribe stgt" in the body of a message to majordomo at vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html |