Al Viro wrote on Thu, Jul 02, 2015: > On Thu, Jul 02, 2015 at 07:56:29PM +0200, Dominique Martinet wrote: > > Using cache=none here so behavious is likely different with cache, but > > basically you can't get more than one tag per user thread accessing the > > 9P mount... > > Yes, and...? You can get a lot more than one user thread... Andrey is > using trinity(1) on client, and that's *definitely* not single-threaded - > the whole point is stressing the damn thing. I have run trinity quite a bit and it doesn't fork bomb as far as I can recall, with him running it with -C100 we're not quite at 2^16 yet? I do agree it's a problem, just don't think it's the one we're hitting -- I'll try again on a recent kernel to see if anything changed with rdma/tcp as well, but I'm starting to doubt I'll get any luck with anything other than virtio; which doesn't really help since it's not the same order of latencies. FWIW I don't *think* trinity can issue TFlush either without user interaction, that's a really special call. It can only happen in rpc() or zc_rpc() if it's interrupted by ERESTARTSYS which I understand as ^C? (I'll look into making the pools use IDA unless someone else steps up, sure. Thanks Jeff) -- Dominique -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
