From: Jan Kara <jack@suse.cz>
To: Mauricio Faria de Oliveira <mfo@canonical.com>
Cc: Jan Kara <jack@suse.cz>,
linux-ext4@vger.kernel.org,
dann frazier <dann.frazier@canonical.com>,
Mauricio Faria de Oliveira <mauricio.foliveira@gmail.com>
Subject: Re: [RFC PATCH v3 3/3] ext4: data=journal: write-protect pages on j_submit_inode_data_buffers()
Date: Wed, 16 Sep 2020 18:59:45 +0200 [thread overview]
Message-ID: <20200916165945.GO3607@quack2.suse.cz> (raw)
In-Reply-To: <20200910193127.276214-4-mfo@canonical.com>
On Thu 10-09-20 16:31:27, Mauricio Faria de Oliveira wrote:
> This implements journal callbacks j_submit|finish_inode_data_buffers()
> with different behavior for data=journal: to write-protect pages under
> commit, preventing changes to buffers writeably mapped to userspace.
>
> If a buffer's content changes between commit's checksum calculation
> and write-out to disk, it can cause journal recovery/mount failures
> upon a kernel crash or power loss.
>
> [ 27.334874] EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!
> [ 27.339492] JBD2: Invalid checksum recovering data block 8705 in log
> [ 27.342716] JBD2: recovery failed
> [ 27.343316] EXT4-fs (loop0): error loading journal
> mount: /ext4: can't read superblock on /dev/loop0.
>
> In j_submit_inode_data_buffers() we write-protect the inode's pages
> with write_cache_pages() and redirty w/ writepage callback if needed.
>
> In j_finish_inode_data_buffers() there is nothing do to.
>
> And in order to use the callbacks, inodes are added to the inode list
> in transaction in __ext4_journalled_writepage() and ext4_page_mkwrite().
>
> In ext4_page_mkwrite() we must make sure that:
>
> 1) the inode is always added to the list;
> thus we skip the 'all buffers mapped' optimization on data=journal;
>
> 2) the buffers are attached to transaction as dirty;
> as already done in __ext4_journalled_writepage().
>
> Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
> Suggested-by: Jan Kara <jack@suse.cz>
> Reported-by: Dann Frazier <dann.frazier@canonical.com>
> ---
> fs/ext4/inode.c | 29 ++++++++++++++------
> fs/ext4/super.c | 72 +++++++++++++++++++++++++++++++++++++++++++++++--
> 2 files changed, 91 insertions(+), 10 deletions(-)
>
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index bf596467c234..fa4109da056c 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -1910,6 +1910,9 @@ static int __ext4_journalled_writepage(struct page *page,
> err = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
> write_end_fn);
> }
> + if (ret == 0)
> + ret = err;
> + err = ext4_jbd2_inode_add_write(handle, inode, 0, len);
> if (ret == 0)
> ret = err;
> EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;
> @@ -6004,9 +6007,12 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
> len = PAGE_SIZE;
> /*
> * Return if we have all the buffers mapped. This avoids the need to do
> - * journal_start/journal_stop which can block and take a long time
> + * journal_start/journal_stop which can block and take a long time.
> + *
> + * This cannot be done for data journalling, as we have to add the
> + * inode to the transaction's list to writeprotect pages on commit.
> */
> - if (page_has_buffers(page)) {
> + if (page_has_buffers(page) && !ext4_should_journal_data(inode)) {
> if (!ext4_walk_page_buffers(NULL, page_buffers(page),
> 0, len, NULL,
> ext4_bh_unmapped)) {
> @@ -6032,12 +6038,14 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
> err = block_page_mkwrite(vma, vmf, get_block);
> if (!err && ext4_should_journal_data(inode)) {
> if (ext4_walk_page_buffers(handle, page_buffers(page), 0,
> - PAGE_SIZE, NULL, do_journal_get_write_access)) {
> - unlock_page(page);
> - ret = VM_FAULT_SIGBUS;
> - ext4_journal_stop(handle);
> - goto out;
> - }
> + PAGE_SIZE, NULL, do_journal_get_write_access))
> + goto out_err;
> + /* Make sure buffers are attached to the transaction as dirty */
> + if (ext4_walk_page_buffers(handle, page_buffers(page), 0,
> + PAGE_SIZE, NULL, write_end_fn))
> + goto out_err;
I think here we need to be a bit more careful. As I wrote in my answer to
cover letter we cannot call block_page_mkwrite(). Instead we need to lock
the page, compute 'len' again from i_size, then call __block_write_begin()
to map (or allocate) and read blocks, and then ext4_walk_page_buffers()
which needs to walk from 0 to len. And then unlock the page.
> + if (ext4_jbd2_inode_add_write(handle, inode, 0, PAGE_SIZE))
> + goto out_err;
> ext4_set_inode_state(inode, EXT4_STATE_JDATA);
> }
> ext4_journal_stop(handle);
> @@ -6049,6 +6057,11 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
> up_read(&EXT4_I(inode)->i_mmap_sem);
> sb_end_pagefault(inode->i_sb);
> return ret;
> +out_err:
> + unlock_page(page);
> + ret = VM_FAULT_SIGBUS;
> + ext4_journal_stop(handle);
> + goto out;
> }
Otherwise the patch looks good to me!
Honza
--
Jan Kara <jack@suse.com>
SUSE Labs, CR
next prev parent reply other threads:[~2020-09-16 20:49 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-10 19:31 [RFC PATCH v3 0/3] ext4/jbd2: data=journal: write-protect pages on transaction commit Mauricio Faria de Oliveira
2020-09-10 19:31 ` [RFC PATCH v3 1/3] jbd2: introduce/export functions jbd2_journal_submit|finish_inode_data_buffers() Mauricio Faria de Oliveira
2020-09-16 16:19 ` Jan Kara
2020-09-10 19:31 ` [RFC PATCH v3 2/3] jbd2, ext4, ocfs2: introduce/use journal callbacks j_submit|finish_inode_data_buffers() Mauricio Faria de Oliveira
2020-09-16 16:22 ` Jan Kara
2020-09-16 16:52 ` Jan Kara
2020-09-10 19:31 ` [RFC PATCH v3 3/3] ext4: data=journal: write-protect pages on j_submit_inode_data_buffers() Mauricio Faria de Oliveira
2020-09-16 16:59 ` Jan Kara [this message]
2020-09-16 16:15 ` [RFC PATCH v3 0/3] ext4/jbd2: data=journal: write-protect pages on transaction commit Jan Kara
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=20200916165945.GO3607@quack2.suse.cz \
--to=jack@suse.cz \
--cc=dann.frazier@canonical.com \
--cc=linux-ext4@vger.kernel.org \
--cc=mauricio.foliveira@gmail.com \
--cc=mfo@canonical.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).