linux-ext4 - [RFC PATCH 01/11] ext4: data=journal: introduce struct/kmem_cache ext4_journalled_wb_page/

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20200423233705.5878-2-mfo@canonical.com>
Date:   Thu, 23 Apr 2020 20:36:55 -0300
From:   Mauricio Faria de Oliveira <mfo@...onical.com>
To:     linux-ext4@...r.kernel.org, "Theodore Y. Ts'o" <tytso@....edu>
Cc:     dann frazier <dann.frazier@...onical.com>,
        Andreas Dilger <adilger@...ger.ca>, Jan Kara <jack@...e.com>
Subject: [RFC PATCH 01/11] ext4: data=journal: introduce struct/kmem_cache ext4_journalled_wb_page/_cachep

The 'struct ext4_journalled_wb_page' is a journal callback entry;
use 'kmem_cache ext4_journalled_wb_page_cachep' to alloc/free it.

It will be used to end page writeback on transaction commit,
only for writeback of mmaped pagecache in data=journal mode
(i.e., ext4_writepage() -> __ext4_journalled_writepage()).

The next patch will add the associated set page writeback
and add the callback entry to the list in the transaction;
and process that list in the ext4 journal commit callback.

This builds on top of the existing journal commit callback
functions (not used anymore), using t_private_list in jbd2.
(introduced with commit 18aadd47f884 ("ext4: expand commit
callback and"), and not used in commit a015434480dc ("ext4:
send parallel discards on commit completions").)

P.S.: ext4_journal_commit_callback() does have spinlocking
with s_md_lock that is not needed for this particular case
(which seems to be the only one), as it has no concurrency
on transaction commit time/only during transaction running
time (currently only done in __ext4_journalled_writepage()).

But no changes for now, just keep it as it is/has been, so
it remains general in case other usages arise that need it.

Signed-off-by: Mauricio Faria de Oliveira <mfo@...onical.com>
---
 fs/ext4/ext4_jbd2.h | 16 ++++++++++++++++
 fs/ext4/page-io.c   | 11 +++++++++++
 2 files changed, 27 insertions(+)

diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index 4b9002f0e84c..9ea8ee583931 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -209,6 +209,22 @@ static inline bool ext4_journal_callback_try_del(handle_t *handle,
 	return deleted;
 }

+/* The struct ext4_journalled_wb_page is a journal callback entry to
+ * end page writeback on transaction commit in the data=journal mode.
+ *
+ * This is used for writeback of mmaped pagecache in data=journal mode
+ * (see ext4_writepage() -> __ext4_journalled_writepage() for details.)
+ */
+struct ext4_journalled_wb_page {
+	/* First member must be the journal callback entry */
+	struct ext4_journal_cb_entry ejwp_jce;
+
+	/* Pointer to page marked for writeback */
+	struct page *ejwp_page;
+};
+
+extern struct kmem_cache *ext4_journalled_wb_page_cachep;
+
 int
 ext4_mark_iloc_dirty(handle_t *handle,
 		     struct inode *inode,
diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
index de6fe969f773..67fabeef6bf2 100644
--- a/fs/ext4/page-io.c
+++ b/fs/ext4/page-io.c
@@ -32,6 +32,7 @@

 static struct kmem_cache *io_end_cachep;
 static struct kmem_cache *io_end_vec_cachep;
+struct kmem_cache *ext4_journalled_wb_page_cachep;

 int __init ext4_init_pageio(void)
 {
@@ -44,6 +45,15 @@ int __init ext4_init_pageio(void)
 		kmem_cache_destroy(io_end_cachep);
 		return -ENOMEM;
 	}
+
+	ext4_journalled_wb_page_cachep = KMEM_CACHE(ext4_journalled_wb_page,
+						    SLAB_RECLAIM_ACCOUNT);
+	if (ext4_journalled_wb_page_cachep == NULL) {
+		kmem_cache_destroy(io_end_cachep);
+		kmem_cache_destroy(io_end_vec_cachep);
+		return -ENOMEM;
+	}
+
 	return 0;
 }

@@ -51,6 +61,7 @@ void ext4_exit_pageio(void)
 {
 	kmem_cache_destroy(io_end_cachep);
 	kmem_cache_destroy(io_end_vec_cachep);
+	kmem_cache_destroy(ext4_journalled_wb_page_cachep);
 }

 struct ext4_io_end_vec *ext4_alloc_io_end_vec(ext4_io_end_t *io_end)
-- 
2.20.1