From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chris Mason Subject: Re: [BUG] can not allocate space for caching data Date: Mon, 20 Dec 2010 07:44:06 -0500 Message-ID: <1292848998-sup-5155@think> References: <4D0F4B26.7030406@cn.fujitsu.com> Content-Type: text/plain; charset=UTF-8 Cc: Linux Btrfs To: Miao Xie Return-path: In-reply-to: <4D0F4B26.7030406@cn.fujitsu.com> List-ID: Excerpts from Miao Xie's message of 2010-12-20 07:25:10 -0500: > Hi, Chris > > There is something wrong with this patch: > > commit 83a50de97fe96aca82389e061862ed760ece2283 > Author: Chris Mason > Date: Mon Dec 13 15:06:46 2010 -0500 > > Btrfs: prevent RAID level downgrades when space is low > > The extent allocator has code that allows us to fill > allocations from any available block group, even if it doesn't > match the raid level we've requested. > > This was put in because adding a new drive to a filesystem > made with the default mkfs options actually upgrades the metadata from > single spindle dup to full RAID1. > > But, the code also allows us to allocate from a raid0 chunk when we > really want a raid1 or raid10 chunk. This can cause big trouble because > mkfs creates a small (4MB) raid0 chunk for data and metadata which then > goes unused for raid1/raid10 installs. > > The allocator will happily wander in and allocate from that chunk when > things get tight, which is not correct. > > The fix here is to make sure that we provide duplication when the > caller has asked for it. It does all the dups to be any raid level, > which preserves the dup->raid1 upgrade abilities. > > Signed-off-by: Chris Mason > > Btrfs has added the space of single chunks and raid0 chunks into the space > information, so when we use btrfs_check_data_free_space() to check if there > is some space for storing file data, this function may return true. So we > write the data into the cache successfully. But, the extent allocator can > not allocate any space to store that cached data, and then the file system > panic. > > I think we subtract that space from the space information, or split the space > information into two types, one is used to manage the chunks with duplication, > the other manages the other chunks. Ok, do you have a test case that triggers this? I'll work out a patch. Yan Zheng's original idea of 'the chunks should be readonly' should help us deduct them from the total. -chris