Hello Jeff-san, > "Yang, Xiao/杨 晓" <yangx.jy@xxxxxxxxxxx> writes: > > > Hi Jeff, > > > > Thanks a lot for your reply. > > > > On 2022/9/27 3:52, Jeff Moyer wrote: > >> "Yang, Xiao/杨 晓" <yangx.jy@xxxxxxxxxxx> writes: > >> > >>> Hi Ross, Toshi, Jeff and others > >>> > >>> After reading the kernel patches[1][2][3] about dm, I think only > >>> three types(linear, stripe and log-writes) of dm can support DAX now, > right? > >> > >> Right. > >> > >>> I wonder why only three types of dm need to support DAX? Is there > >>> any reason/history and use case? > >> > >> It was easy to add DAX support to those targets. Note that if a > >> target needs to intercept I/O, then it is fundamentally incompatible with > DAX. I would like to know its use case. Though my team has been made effort for a long time to remove experimental of FS-DAX, we still need one more time to solve thin-pool (or other DM feature) v.s. DAX problem even if it is easy. To be honest, I'm under criticism for taking such a long time and still not be able to remove experimental. In addition, there are a few persons who insist Fujitsu should stop development of FS-DAX due to too much time for development I hope that my team can continue to make effort for FS-DAX. If there is a use-case of thin-pool, then I'll be able to persuade such persons by it. So I would appreciate it if you could tell me about the use case. Thanks, --- Yasunori Goto -- dm-devel mailing list dm-devel@xxxxxxxxxx https://listman.redhat.com/mailman/listinfo/dm-devel
