From: Matthew Wilcox <willy@infradead.org>
To: Dave Chinner <david@fromorbit.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>,
Ross Zwisler <ross.zwisler@linux.intel.com>,
Jens Axboe <axboe@kernel.dk>, Rehas Sachdeva <aquannie@gmail.com>,
linux-mm@kvack.org, linux-fsdevel@vger.kernel.org,
linux-f2fs-devel@lists.sourceforge.net,
linux-nilfs@vger.kernel.org, linux-btrfs@vger.kernel.org,
linux-xfs@vger.kernel.org, linux-usb@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4 72/73] xfs: Convert mru cache to XArray
Date: Tue, 5 Dec 2017 18:02:08 -0800 [thread overview]
Message-ID: <20171206020208.GK26021@bombadil.infradead.org> (raw)
In-Reply-To: <20171206012901.GZ4094@dastard>
On Wed, Dec 06, 2017 at 12:36:48PM +1100, Dave Chinner wrote:
> > - if (radix_tree_preload(GFP_NOFS))
> > - return -ENOMEM;
> > -
> > INIT_LIST_HEAD(&elem->list_node);
> > elem->key = key;
> >
> > - spin_lock(&mru->lock);
> > - error = radix_tree_insert(&mru->store, key, elem);
> > - radix_tree_preload_end();
> > - if (!error)
> > - _xfs_mru_cache_list_insert(mru, elem);
> > - spin_unlock(&mru->lock);
> > + do {
> > + xas_lock(&xas);
> > + xas_store(&xas, elem);
> > + error = xas_error(&xas);
> > + if (!error)
> > + _xfs_mru_cache_list_insert(mru, elem);
> > + xas_unlock(&xas);
> > + } while (xas_nomem(&xas, GFP_NOFS));
>
> Ok, so why does this have a retry loop on ENOMEM despite the
> existing code handling that error? And why put such a loop in this
> code and not any of the other XFS code that used
> radix_tree_preload() and is arguably much more important to avoid
> ENOMEM on insert (e.g. the inode cache)?
If we need more nodes in the tree, xas_store() will try to allocate them
with GFP_NOWAIT | __GFP_NOWARN. If that fails, it signals it in xas_error().
xas_nomem() will notice that we're in an ENOMEM situation, and allocate
a node using your preferred GFP flags (NOIO in your case). Then we retry,
guaranteeing forward progress. [1]
The other conversions use the normal API instead of the advanced API, so
all of this gets hidden away. For example, the inode cache does this:
+ curr = xa_cmpxchg(&pag->pag_ici_xa, agino, NULL, ip, GFP_NOFS);
and xa_cmpxchg internally does:
do {
xa_lock_irqsave(xa, flags);
curr = xas_create(&xas);
if (curr == old)
xas_store(&xas, entry);
xa_unlock_irqrestore(xa, flags);
} while (xas_nomem(&xas, gfp));
> Also, I really don't like the pattern of using xa_lock()/xa_unlock()
> to protect access to an external structure. i.e. the mru->lock
> context is protecting multiple fields and operations in the MRU
> structure, not just the radix tree operations. Turning that around
> so that a larger XFS structure and algorithm is now protected by an
> opaque internal lock from generic storage structure the forms part
> of the larger structure seems like a bad design pattern to me...
It's the design pattern I've always intended to use. Naturally, the
xfs radix trees weren't my initial target; it was the page cache, and
the page cache does the same thing; uses the tree_lock to protect both
the radix tree and several other fields in that same data structure.
I'm open to argument on this though ... particularly if you have a better
design pattern in mind!
[1] I actually have this documented! It's in the xas_nomem() kernel-doc:
* If we need to add new nodes to the XArray, we try to allocate memory
* with GFP_NOWAIT while holding the lock, which will usually succeed.
* If it fails, @xas is flagged as needing memory to continue. The caller
* should drop the lock and call xas_nomem(). If xas_nomem() succeeds,
* the caller should retry the operation.
*
* Forward progress is guaranteed as one node is allocated here and
* stored in the xa_state where it will be found by xas_alloc(). More
* nodes will likely be found in the slab allocator, but we do not tie
* them up here.
*
* Return: true if memory was needed, and was successfully allocated.
next prev parent reply other threads:[~2017-12-06 2:02 UTC|newest]
Thread overview: 127+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-12-06 0:40 [PATCH v4 00/73] XArray version 4 Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 01/73] xfs: Rename xa_ elements to ail_ Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 02/73] xarray: Add the xa_lock to the radix_tree_root Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 03/73] page cache: Use xa_lock Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 04/73] xarray: Replace exceptional entries Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 05/73] xarray: Change definition of sibling entries Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 06/73] xarray: Add definition of struct xarray Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 07/73] xarray: Define struct xa_node Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 08/73] xarray: Add documentation Matthew Wilcox
2017-12-11 23:10 ` Randy Dunlap
2017-12-15 4:22 ` Matthew Wilcox
2017-12-15 12:34 ` Naming of tag operations in the XArray Matthew Wilcox
2017-12-19 0:16 ` Randy Dunlap
2017-12-15 17:10 ` Storing errors " Matthew Wilcox
2017-12-19 0:27 ` Randy Dunlap
2017-12-06 0:40 ` [PATCH v4 09/73] xarray: Add xa_load Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 10/73] xarray: Add xa_get_tag, xa_set_tag and xa_clear_tag Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 11/73] xarray: Add xa_store Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 12/73] xarray: Add xa_cmpxchg Matthew Wilcox
2017-12-06 0:40 ` [PATCH v4 13/73] xarray: Add xa_for_each Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 14/73] xarray: Add xas_for_each_tag Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 15/73] xarray: Add xa_get_entries, xa_get_tagged and xa_get_maybe_tag Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 16/73] xarray: Add xa_destroy Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 17/73] xarray: Add xas_next and xas_prev Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 18/73] xarray: Add xas_create_range Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 19/73] xarray: Add MAINTAINERS entry Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 20/73] idr: Convert to XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 21/73] ida: " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 22/73] page cache: Convert hole search " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 23/73] page cache: Add page_cache_range_empty function Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 24/73] page cache: Add and replace pages using the XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 25/73] page cache: Convert page deletion to XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 26/73] page cache: Convert page cache lookups " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 27/73] page cache: Convert delete_batch " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 28/73] page cache: Remove stray radix comment Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 29/73] mm: Convert page-writeback to XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 30/73] mm: Convert workingset " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 31/73] mm: Convert truncate " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 32/73] mm: Convert add_to_swap_cache " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 33/73] mm: Convert delete_from_swap_cache " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 34/73] mm: Convert cgroup writeback " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 35/73] mm: Convert __do_page_cache_readahead " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 36/73] mm: Convert page migration " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 37/73] mm: Convert huge_memory " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 38/73] mm: Convert collapse_shmem " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 39/73] mm: Convert khugepaged_scan_shmem " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 40/73] pagevec: Use xa_tag_t Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 41/73] shmem: Convert replace to XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 42/73] shmem: Convert shmem_confirm_swap " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 43/73] shmem: Convert find_swap_entry " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 44/73] shmem: Convert shmem_tag_pins " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 45/73] shmem: Convert shmem_wait_for_pins " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 46/73] shmem: Convert shmem_add_to_page_cache " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 47/73] shmem: Convert shmem_alloc_hugepage " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 48/73] shmem: Convert shmem_free_swap " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 49/73] shmem: Convert shmem_partial_swap_usage " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 50/73] shmem: Comment fixups Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 51/73] btrfs: Convert page cache to XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 52/73] fs: Convert buffer " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 53/73] fs: Convert writeback " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 54/73] nilfs2: Convert " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 55/73] f2fs: " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 56/73] lustre: " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 57/73] dax: Convert dax_unlock_mapping_entry " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 58/73] dax: Convert lock_slot " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 59/73] dax: More XArray conversion Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 60/73] dax: Convert __dax_invalidate_mapping_entry to XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 61/73] dax: Convert dax_writeback_one " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 62/73] dax: Convert dax_insert_pfn_mkwrite " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 63/73] dax: Convert dax_insert_mapping_entry " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 64/73] dax: Convert grab_mapping_entry " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 65/73] dax: Fix sparse warning Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 66/73] page cache: Finish XArray conversion Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 67/73] vmalloc: Convert to XArray Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 68/73] brd: " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 69/73] xfs: Convert m_perag_tree " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 70/73] xfs: Convert pag_ici_root " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 71/73] xfs: Convert xfs dquot " Matthew Wilcox
2017-12-06 0:41 ` [PATCH v4 72/73] xfs: Convert mru cache " Matthew Wilcox
2017-12-06 1:36 ` Dave Chinner
2017-12-06 2:02 ` Matthew Wilcox [this message]
2017-12-06 3:14 ` Dave Chinner
2017-12-06 4:45 ` Matthew Wilcox
2017-12-06 4:52 ` Matthew Wilcox
2017-12-06 8:44 ` Dave Chinner
2017-12-06 14:06 ` Matthew Wilcox
2017-12-07 0:38 ` Dave Chinner
2017-12-08 23:01 ` Matthew Wilcox
2017-12-10 23:57 ` Dave Chinner
2017-12-11 4:23 ` Matthew Wilcox
2017-12-11 21:55 ` Dave Chinner
2017-12-07 16:06 ` Theodore Ts'o
2017-12-07 22:22 ` Dave Chinner
2017-12-08 4:45 ` Byungchul Park
2017-12-08 7:25 ` Dave Chinner
2017-12-08 9:27 ` Byungchul Park
2017-12-08 17:35 ` Alan Stern
2017-12-08 22:36 ` Dave Chinner
2017-12-09 17:00 ` Joe Perches
2017-12-11 21:43 ` Dave Chinner
2017-12-11 22:12 ` Joe Perches
2017-12-11 22:43 ` Matthew Wilcox
2017-12-11 23:46 ` Joe Perches
2017-12-12 15:51 ` Alan Stern
2017-12-14 18:23 ` Joe Perches
2017-12-17 1:26 ` [RFC patch] checkpatch: Add a test for long function definitions (>200 lines) Joe Perches
2017-12-17 21:46 ` Linus Torvalds
2017-12-17 22:22 ` Joe Perches
2017-12-17 22:33 ` Luc Van Oostenryck
2017-12-11 23:38 ` [PATCH v4 72/73] xfs: Convert mru cache to XArray Dave Chinner
2017-12-21 12:05 ` Knut Omang
2017-12-07 22:38 ` Lockdep is less useful than it was Matthew Wilcox
2017-12-07 22:39 ` Matthew Wilcox
2017-12-08 0:14 ` Dave Chinner
2017-12-08 15:27 ` Theodore Ts'o
2017-12-08 18:14 ` Matthew Wilcox
2017-12-08 22:47 ` Dave Chinner
2017-12-06 0:41 ` [PATCH v4 73/73] usb: Convert xhci-mem to XArray Matthew Wilcox
2017-12-06 1:45 ` [PATCH v4 00/73] XArray version 4 Dave Chinner
2017-12-06 1:51 ` Dave Chinner
2017-12-06 1:53 ` Matthew Wilcox
2017-12-06 2:17 ` Dave Chinner
2017-12-06 2:27 ` Matthew Wilcox
2017-12-06 2:05 ` Matthew Wilcox
2017-12-06 2:38 ` Dave Chinner
2017-12-06 23:58 ` Ross Zwisler
2017-12-07 0:13 ` Matthew Wilcox
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20171206020208.GK26021@bombadil.infradead.org \
--to=willy@infradead.org \
--cc=aquannie@gmail.com \
--cc=axboe@kernel.dk \
--cc=david@fromorbit.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=linux-f2fs-devel@lists.sourceforge.net \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux-nilfs@vger.kernel.org \
--cc=linux-usb@vger.kernel.org \
--cc=linux-xfs@vger.kernel.org \
--cc=mawilcox@microsoft.com \
--cc=ross.zwisler@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).