On Fri, Jun 11, 2021 at 3:46 AM Al Viro [off-list ref] wrote:

On Fri, Jun 11, 2021 at 11:12:31AM +1000, Stephen Rothwell wrote:

quoted

Hi all,

Today's linux-next merge of the gfs2 tree got conflicts in:

  Documentation/filesystems/porting.rst
  include/linux/uio.h
  lib/iov_iter.c

between various commits from the vfs tree and the same, older version,
of the commits from the gfs2 tree.

I fixed it up (I used the vfs tree versions) and can carry the fix as
necessary. This is now fixed as far as linux-next is concerned, but any
non trivial conflicts should be mentioned to your upstream maintainer
when your tree is submitted for merging.  You may also want to consider
cooperating with the maintainer of the conflicting tree to minimise any
particularly complex conflicts.

IMO iov_iter_fault_in_writeable() is a bloody bad idea to start with...

It seems that we need a mechanism for faulting in a page in for
writing. It's ugly that iov_iter_fault_in_writeable() writes to the
page for faulting it in, and if the operation that the page was
faulted in for then fails, those writes will be visible. Would an
implementation that reads and then writes back the same value be less
bloody bad?

Thanks,
Andreas