xfs
[Top] [All Lists]

Fixing XFS RAID - does XFS ever issue non-aligned discards?

To: xfs@xxxxxxxxxxx
Subject: Fixing XFS RAID - does XFS ever issue non-aligned discards?
From: MichaÅ Pecio <michal.pecio@xxxxxxxxx>
Date: Tue, 15 Dec 2015 14:16:05 +0100
Delivered-to: xfs@xxxxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=date:from:to:subject:message-id:mime-version:content-type :content-transfer-encoding; bh=YuxJzZZdJ9kOVaXfV6tgqt505MA+AQkYoWZxNcxZquI=; b=Om4gKYbdq/BYuy13qJEAEm2USjpJTY5se4Vz2EUaqgd0lo0pibyvbigqhfshQb+TZy xaXZ5Dlrvf5Rxpk08rU6Inz7bQ/xGlhQ4vayfdKvGXWpzN1nn67OaZy22ZR4iiIGqTp7 pQqHwFJAQAhOHK7cRKCv5rcIbWirQ1fuFUPwo2cCC/8rfpoZlinh9MpzvNS+O+C3kajw dJbFyk+maVL6srn70iFoA36P5XMffmQAjduEdo/XDbVeBqwedHKg6rCb3k/lfCANgbCp QcL0Af7wgxqqrkDcLA57KsHSgonSbXvZclNV9338BYIOM2UbaVM1Vv4+X+bxhonnP8gT vn2A==
Hi,

I'm using XFS with bsize=4096 on RAID0 of two Intel 320 SSDs under
Linux 3.18 LTS. I'm wondering if it's correct to assume that discards
issued by XFS will always be multiples of 4k and 4k-aligned?

I'm asking because I need to fix damage caused by the infamous RAID0
bug from May 2015, which, if I'm reading the code correctly, basically
permuted discard requests between component disks, i.e. sda discarded
offsets which were supposed to be discarded by sdb and vice-versa.

The only backup I have is very slow so I'm trying to minimize copying.
The list of files with 4k-aligned blocks of zeros is small and worth
reviewing manually, but there is much more files with smaller holes
and I'd like to know if I should be bothered by them. Fwiw, I checked
few and they've all been false positives so far.

Thanks for any help,
MP

<Prev in Thread] Current Thread [Next in Thread>