From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Martin K. Petersen" Subject: Re: [PATCH] libata: Whitelist SSDs that are known to properly return zeroes after TRIM Date: Thu, 04 Dec 2014 20:53:22 -0500 Message-ID: References: <547FCECB.4090908@ubuntu.com> <547FD4F6.7090700@ubuntu.com> <547FE5BD.9070101@ubuntu.com> Mime-Version: 1.0 Content-Type: text/plain Return-path: Received: from userp1040.oracle.com ([156.151.31.81]:40430 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754792AbaLEBxa (ORCPT ); Thu, 4 Dec 2014 20:53:30 -0500 In-Reply-To: <547FE5BD.9070101@ubuntu.com> (Phillip Susi's message of "Wed, 03 Dec 2014 23:40:29 -0500") Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Phillip Susi Cc: "Martin K. Petersen" , Tejun Heo , linux-ide@vger.kernel.org >>>>> "Phillip" == Phillip Susi writes: Phillip> So your assertion is that you have seen it in writing, Actually I haven't. I have talked to a bunch of people that have done RAID qualification on SSDs. I also gleaned on a few vendor lists of supported drive configurations. The good news is that practice of each OEM having their own custom drive firmware build is not very common for SSDs. And most of the results I looked at was collected using off-the-shelf drives. Phillip> I'm not saying a hard hell no, but this certainly makes me Phillip> uncomfortable. You do realize that my patch *restricts* the drives we enable discard_zeroes_data on, right? Instead of relying solely on the drive's own reporting we now also require empirical evidence that they do the right thing. Phillip> I'd much rather see the manufacturers put it in writing that Phillip> yes, this make and model will perform this way even though it Phillip> is not strictly required by ATA8. Wishful thinking. The sad reality is that standards are deliberately written to be vague. And that any hard guarantees are part of product requirements documents. That's not specific to SSDs in any way. That's true for pretty much any piece of hardware. Phillip> What would be better still is if the bloody ATA standard got a Phillip> clue and said that if the drive claims that it does in fact Phillip> zero after TRIM, that the TRIM command becomes mandatory Phillip> instead of advisory. I agree that the standards efforts in this department are a total train wreck. -- Martin K. Petersen Oracle Linux Engineering