2010-03-26 16:21:13

by Surbhi Palande

Subject: [PATCH] ext4: Ensure writecache to disk in no journal mode

Ensure that in the no journal mode the write cache is flushed to the disk by
calling a blkdev_issue_flush() which issues a WRITE_BARRIER if necessary.

Signed-off-by: Surbhi Palande <[email protected]>
---
fs/ext4/fsync.c | 8 ++++++--
1 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/fs/ext4/fsync.c b/fs/ext4/fsync.c
index d6049e4..1d73a50 100644
--- a/fs/ext4/fsync.c
+++ b/fs/ext4/fsync.c
@@ -67,8 +67,12 @@ int ext4_sync_file(struct file *file, struct dentry *dentry, int datasync)
if (ret < 0)
return ret;

- if (!journal)
- return simple_fsync(file, dentry, datasync);
+ if (!journal) {
+ ret = simple_fsync(file, dentry, datasync);
+ if (test_opt(inode->i_sb, BARRIER))
+ blkdev_issue_flush(inode->i_sb->s_bdev, NULL);
+ return ret;
+ }

/*
* data=writeback,ordered:
--
1.6.3.3

2010-03-26 16:37:39

by Dmitry Monakhov

Subject: Re: [PATCH] ext4: Ensure writecache to disk in no journal mode

Surbhi Palande <[email protected]> writes:

> Ensure that in the no journal mode the write cache is flushed to the disk by
> calling a blkdev_issue_flush() which issues a WRITE_BARRIER if necessary.
As soon as i understand, nojournal mode is assumed to be used for
fail-free block devices(raid + UPS). So we don't have to worry about
blkdev's wcache vs persistent storage correctness.
>
> Signed-off-by: Surbhi Palande <[email protected]>
> ---
> fs/ext4/fsync.c | 8 ++++++--
> 1 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/fs/ext4/fsync.c b/fs/ext4/fsync.c
> index d6049e4..1d73a50 100644
> --- a/fs/ext4/fsync.c
> +++ b/fs/ext4/fsync.c
> @@ -67,8 +67,12 @@ int ext4_sync_file(struct file *file, struct dentry *dentry, int datasync)
> if (ret < 0)
> return ret;
>
> - if (!journal)
> - return simple_fsync(file, dentry, datasync);
> + if (!journal) {
> + ret = simple_fsync(file, dentry, datasync);
> + if (test_opt(inode->i_sb, BARRIER))
> + blkdev_issue_flush(inode->i_sb->s_bdev, NULL);
> + return ret;
> + }
>
> /*
> * data=writeback,ordered:

2010-03-26 17:15:54

by Ric Wheeler

Subject: Re: [PATCH] ext4: Ensure writecache to disk in no journal mode

On 03/26/2010 12:37 PM, Dmitry Monakhov wrote:
> Surbhi Palande<[email protected]> writes:
>
>
>> Ensure that in the no journal mode the write cache is flushed to the disk by
>> calling a blkdev_issue_flush() which issues a WRITE_BARRIER if necessary.
>>
> As soon as i understand, nojournal mode is assumed to be used for
> fail-free block devices(raid + UPS). So we don't have to worry about
> blkdev's wcache vs persistent storage correctness.
>

I don't think that is a safe assumption. If users want that behavior,
they can mount with fs without barriers...

ric

>> Signed-off-by: Surbhi Palande<[email protected]>
>> ---
>> fs/ext4/fsync.c | 8 ++++++--
>> 1 files changed, 6 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/ext4/fsync.c b/fs/ext4/fsync.c
>> index d6049e4..1d73a50 100644
>> --- a/fs/ext4/fsync.c
>> +++ b/fs/ext4/fsync.c
>> @@ -67,8 +67,12 @@ int ext4_sync_file(struct file *file, struct dentry *dentry, int datasync)
>> if (ret< 0)
>> return ret;
>>
>> - if (!journal)
>> - return simple_fsync(file, dentry, datasync);
>> + if (!journal) {
>> + ret = simple_fsync(file, dentry, datasync);
>> + if (test_opt(inode->i_sb, BARRIER))
>> + blkdev_issue_flush(inode->i_sb->s_bdev, NULL);
>> + return ret;
>> + }
>>
>> /*
>> * data=writeback,ordered:
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>

2010-03-26 17:40:10

by Eric Sandeen

Subject: Re: [PATCH] ext4: Ensure writecache to disk in no journal mode

On 03/26/2010 12:16 PM, Ric Wheeler wrote:
> On 03/26/2010 12:37 PM, Dmitry Monakhov wrote:
>> Surbhi Palande<[email protected]> writes:
>>
>>
>>> Ensure that in the no journal mode the write cache is flushed to the disk by
>>> calling a blkdev_issue_flush() which issues a WRITE_BARRIER if necessary.
>>>
>> As soon as i understand, nojournal mode is assumed to be used for
>> fail-free block devices(raid + UPS). So we don't have to worry about
>> blkdev's wcache vs persistent storage correctness.

No, I don't think so - even with "fail-free" storage, a system crash
still results in an inconsistent filesystem; with nojournalling you've
made the decision to either fsck or re-mkfs after that event.

> I don't think that is a safe assumption. If users want that behavior,
> they can mount with fs without barriers...

yes, I agree with Ric - even if you don't have journalling, the proper
sequence of sync calls should still result in data permanently on disk
by default. (though I think using mount -o nobarrier for this purpose,
in absence of journalling, overloads the option a little...)

-Eric