From mboxrd@z Thu Jan  1 00:00:00 1970
From: Brad Campbell <brad@fnarfbargle.com>
Subject: Re: 3.10: discard/trim support on md-raid1?
Date: Sat, 13 Jul 2013 18:47:46 +0800
Message-ID: <51E13052.30101@fnarfbargle.com>
References: <001701ce7fb4$889bdd40$99d397c0$@lucidpixels.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <001701ce7fb4$889bdd40$99d397c0$@lucidpixels.com>
Sender: linux-kernel-owner@vger.kernel.org
To: Justin Piszcz <jpiszcz@lucidpixels.com>
Cc: linux-kernel@vger.kernel.org, linux-raid@vger.kernel.org
List-Id: linux-raid.ids

On 13/07/13 18:34, Justin Piszcz wrote:
> And possibly:
> discard_zeroes_data: 1

Does it though?

Here's my 6 x SSD RAID10 that definitely discards.

brad@srv:~$ grep . /sys/block/md2/queue/*
/sys/block/md2/queue/add_random:0
/sys/block/md2/queue/discard_granularity:33553920
/sys/block/md2/queue/discard_max_bytes:65536
/sys/block/md2/queue/discard_zeroes_data:0
/sys/block/md2/queue/hw_sector_size:512
/sys/block/md2/queue/iostats:0
/sys/block/md2/queue/logical_block_size:512
/sys/block/md2/queue/max_hw_sectors_kb:16383
/sys/block/md2/queue/max_integrity_segments:0
/sys/block/md2/queue/max_sectors_kb:512
/sys/block/md2/queue/max_segments:128
/sys/block/md2/queue/max_segment_size:65536
/sys/block/md2/queue/minimum_io_size:65536
/sys/block/md2/queue/nomerges:0
/sys/block/md2/queue/nr_requests:128
/sys/block/md2/queue/optimal_io_size:196608
/sys/block/md2/queue/physical_block_size:512
/sys/block/md2/queue/read_ahead_kb:384
/sys/block/md2/queue/rotational:1
/sys/block/md2/queue/rq_affinity:0
/sys/block/md2/queue/scheduler:none
/sys/block/md2/queue/write_same_max_bytes:0

Mine does not have discard_zeroes_data == 1 which is correct as 3 of the drives are Samsung 830's 
and they don't return deterministic after trim.

I can verify that discard gets passed down the stack as I get a non-zero mismatch count after 
running a trim on the filesystem. Checking the mirrors manually sees 0 returned from the Intel 
330's, and apparently random data returned from the 830's.

Regards,
Brad