Re: [RFC PATCH 4/5] fs: propagate write stream

[Christoph Hellwig <hch@lst.de>]

On Thu, Oct 16, 2025 at 03:17:50PM +0530, Kanchan Joshi wrote:
> On 8/12/2025 1:54 PM, Christoph Hellwig wrote:
> > On Tue, Jul 29, 2025 at 08:21:34PM +0530, Kanchan Joshi wrote:
> >> bio->bi_write_stream is not set by the filesystem code.
> >> Use inode's write stream value to do that.
> > Just passing it through is going to create problems.  i.e. when
> > the file system does it's own placement or reserves ids.  We'll need
> > an explicit intercept point between the user write stream and what
> > does into the bio.
> > 
> 
> For that intercept point - will you prefer a generic helper, say 
> fs_resolve_write_stream(), that will call a new inode operation that 
> filesystem will implement?

I don't remember what the patch was doing, but basically the file system
should be explicitly set bio->bi_write_stream instead of automatically
inheriting it.  Where the file system uses helpers we'll have to find
a way to propagate it.  For iomap one option could be to make the file
system return it in struct iomap.