On 3/11/21 3:43 AM, Stefan Metzmacher wrote: > Hi Jens, > >> I'm sure we're going to find little things to patch up after this >> series, but testing has been pretty thorough, from the usual regression >> suite to production. Any issue that may crop up should be manageable. >> There's also a nice series of further reductions we can do on top of >> this, but I wanted to get the meat of it out sooner rather than later. >> The general worry here isn't that it's fundamentally broken. Most of the >> little issues we've found over the last week have been related to just >> changes in how thread startup/exit is done, since that's the main >> difference between using kthreads and these kinds of threads. In fact, >> if all goes according to plan, I want to get this into the 5.10 and 5.11 >> stable branches as well. > > That would mean that IORING_FEAT_SQPOLL_NONFIXED would be implicitly > be backported from 5.11 to 5.10, correct? Right, that would happen by default if we moved the new worker code back to 5.10 and 5.11. > I'm wondering if I can advice people to move to 5.10 (as it's an lts > release) in order to get a kernel that is most likely very useful to > use in combination with Samba's drafted usage of io_uring, where I'd > like to use IORING_FEAT_SQPOLL_NONFIXED and IORING_FEAT_NATIVE_WORKERS > in order to use SENDMSG/RECVMSG with msg_control buffers (where the > control buffers may reference file descriptors). Understandable! It's worth nothing that 5.10 would also need a backport of the TIF_NOTIFY_SIGNAL change - could be done without, but then we'd have diverging code bases to some degree, which would be something I'd love to avoid. -- Jens Axboe
