On 2025/11/19 17:12, Christoph Hellwig wrote:
On Wed, Nov 19, 2025 at 02:17:07PM +0800, Gao Xiang wrote:
Hongbo didn't Cc you on this thread (I think he just added
recipients according to MAINTAINERS), but I know you played
a key role in iomap development, so I think you should be
in the loop about the iomap change too.
Could you give some comments (maybe review) on this patch
if possible? My own opinion is that if the first two
patches can be applied in the next cycle (6.19) (I understand
it will be too late for the whole feature into 6.19) , it
would be very helpful to us so at least the vfs iomap branch
won't be coupled anymore if the first two patch can be landed
in advance.
The patch itself looks fine. But as Darrick said we really need
to get our house in order for the iomap branch so that it actually
works this close to the merge window.
Sigh.. I'm sorry to hear about that.
Anyway, personally I think patch 1 makes no change to iomap logic
(so I think it definitely does no harm to iomap stability), but
opens a chance for iomap users to control iter->private and pass
fs-specific contexts from iomap_begin to end (and patch 2 uses
this to get rid of kmap_to_page()). So honestly I'm eager to get
patches 1 and 2 merged.
However, it's really up to the iomap maintainers. Yet, if delayed
to the next development cycle, it might still need to resolve
cross-branch conflicts, and it could still causes some churn,
anyway...
Thanks,
Gao Xiang
