linux-kernel - Re: [PATCH RFC v5 10/10] iomap: Rename ATOMIC flags again

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: <Z9E0JqQfdL4nPBH-@infradead.org>
Date: Wed, 12 Mar 2025 00:13:42 -0700
From: Christoph Hellwig <hch@...radead.org>
To: John Garry <john.g.garry@...cle.com>
Cc: brauner@...nel.org, djwong@...nel.org, cem@...nel.org,
	linux-xfs@...r.kernel.org, linux-fsdevel@...r.kernel.org,
	linux-kernel@...r.kernel.org, ojaswin@...ux.ibm.com,
	ritesh.list@...il.com, martin.petersen@...cle.com
Subject: Re: [PATCH RFC v5 10/10] iomap: Rename ATOMIC flags again

On Mon, Mar 10, 2025 at 06:39:46PM +0000, John Garry wrote:
> Dave Chinner thought that names IOMAP_ATOMIC_HW and IOMAP_ATOMIC_SW were
> not appropopiate. Specifically because IOMAP_ATOMIC_HW could actually be
> realised with a SW-based method in the block or md/dm layers.
> 
> So rename to IOMAP_ATOMIC_BIO and IOMAP_ATOMIC_FS.

Looking over the entire series and the already merged iomap code:
there should be no reason at all for having IOMAP_ATOMIC_FS.
The only thing it does is to branch out to
xfs_atomic_write_sw_iomap_begin from xfs_atomic_write_iomap_begin.

You can do that in a much simpler and nicer way by just having
different iomap_ops for the bio based vs file system based atomics.

I agree with dave that bio is a better term for the bio based
atomic, but please use the IOMAP_ATOMIC_BIO name above instead
of the IOMAP_BIO_ATOMIC actually used in the code if you change
it.

>   */
>  static inline blk_opf_t iomap_dio_bio_opflags(struct iomap_dio *dio,
> -		const struct iomap *iomap, bool use_fua, bool atomic_hw)
> +		const struct iomap *iomap, bool use_fua, bool bio_atomic)

Not new here, but these two bools are pretty ugly.

I'd rather have a

    blk_opf_t extra_flags;

in the caller that gets REQ_FUA and REQ_ATOMIC assigned as needed,
and then just clear 

>  
> -	if (atomic_hw && length != iter->len)
> +	if (bio_atomic && length != iter->len)
>  		return -EINVAL;

This could really use a comment explaining what the check is for.

> -		if (WARN_ON_ONCE(atomic_hw && n != length)) {
> +		if (WARN_ON_ONCE(bio_atomic && n != length)) {

Same.

> -#define IOMAP_ATOMIC_HW		(1 << 9) /* HW-based torn-write protection */
> -#define IOMAP_DONTCACHE		(1 << 10)
> -#define IOMAP_ATOMIC_SW		(1 << 11)/* SW-based torn-write protection */
> +#define IOMAP_DONTCACHE		(1 << 9)
> +#define IOMAP_BIO_ATOMIC	(1 << 10) /* Use REQ_ATOMIC on single bio */
> +#define IOMAP_FS_ATOMIC		(1 << 11) /* FS-based torn-write protection */

Please also fix the overly long lines here by moving the comments
above the definitions.  That should also give you enough space to
expand the comment into a full sentence describing the flag fully.