Re: [PATCH v3 02/14] dm: add ->copy_from_iter() dax operation support

[Kani, Toshimitsu]

On Fri, 2017-06-09 at 13:23 -0700, Dan Williams wrote:
> Allow device-mapper to route copy_from_iter operations to the
> per-target implementation. In order for the device stacking to work
> we
> need a dax_dev and a pgoff relative to that device. This gives each
> layer of the stack the information it needs to look up the operation
> pointer for the next level.
> 
> This conceptually allows for an array of mixed device drivers with
> varying copy_from_iter implementations.
> 
> Cc: Toshi Kani <toshi.kani@xxxxxxx>
> Reviewed-by: Mike Snitzer <snitzer@xxxxxxxxxx>
> Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>

I was worried about possible overhead with additional stub calls, but
it looks fine with a single thread fio write test with direct=1.

Â92.62%ÂÂ[kernel.kallsyms]ÂÂÂ[k] __copy_user_nocache
 0.04%ÂÂ[kernel.kallsyms]ÂÂÂ[k] entry_SYSCALL_64_fastpath
 0.08%ÂÂlibpthread-2.22.soÂÂ[.] __GI___libc_write
 0.01%ÂÂ[kernel.kallsyms]ÂÂÂ[k] sys_write
 0.02%ÂÂ[kernel.kallsyms]ÂÂÂ[k] vfs_write
 0.02%ÂÂ[kernel.kallsyms]ÂÂÂ[k] __vfs_write
 0.02%ÂÂ[kernel.kallsyms]ÂÂÂ[k] ext4_file_write_iter
 0.02%ÂÂ[kernel.kallsyms]ÂÂÂ[k] dax_iomap_rw
 0.03%ÂÂ[kernel.kallsyms]ÂÂÂ[k] iomap_apply
 0.04%ÂÂ[kernel.kallsyms]ÂÂÂ[k] dax_iomap_actor
 0.01%ÂÂ[kernel.kallsyms]ÂÂÂ[k] dax_copy_from_iter
 0.01%ÂÂ[kernel.kallsyms]ÂÂÂ[k] dm_dax_copy_from_iter
 0.01%ÂÂ[kernel.kallsyms]ÂÂÂ[k] linear_dax_copy_from_iter
 0.03%ÂÂ[kernel.kallsyms]ÂÂÂ[k] copy_from_iter_flushcache
 0.00%ÂÂ[kernel.kallsyms]ÂÂÂ[k] pmem_copy_from_iter

Multi-thread fio test hits hard in inode_lock(), no contention from the
dm-layer.

Reviewed-by: Toshi Kani <toshi.kani@xxxxxxx>

Thanks,
-Toshi