From mboxrd@z Thu Jan 1 00:00:00 1970 From: hch@lst.de (Christoph Hellwig) Date: Fri, 8 Mar 2019 14:26:08 +0100 Subject: [PATCH] nvmet: disable direct I/O when unavailable In-Reply-To: <71982410-d373-a659-0c0f-8aadf06d1d96@grimberg.me> References: <20190221182234.22345-1-jthumshirn@suse.de> <6c3fa30a-d516-a453-a001-1014202ff9ce@suse.de> <71982410-d373-a659-0c0f-8aadf06d1d96@grimberg.me> Message-ID: <20190308132608.GB18218@lst.de> On Mon, Feb 25, 2019@01:18:10PM -0800, Sagi Grimberg wrote: > >> Frankly, we have a ton of testing related special cases in the kernel. >> >> This one is a) simple and small, only 10 LoC, b) far away from the fast >> path or any other place where it could have any impact on legitimate >> users and c) it prints an informal message showing you what happened. >> >> Sorry but this is a https://xkcd.com/386/ moment. > > How about we just fail it with a proper error message and let the > user set buffered_io instead of trying to get smart here... Yes, would be my preference.