Re: [PATCH v2 23/35] ovl: copy-up on MAP_SHARED

From: Randy Dunlap
Date: Mon May 07 2018 - 15:28:58 EST


On 05/07/2018 01:37 AM, Miklos Szeredi wrote:
> A corner case of a corner case is when
>
> - file opened for O_RDONLY
> - which is then memory mapped SHARED
> - file opened for O_WRONLY
> - contents modified
> - contents read back though the shared mapping
>
> Unfortunately it looks very difficult to do anything about the established
> shared map after the file is copied up.
>
> Instead, when a read-only file is mapped shared, copy up the file before
> actually doing the map. This may result in unnecessary copy-ups (but so
> may copy-up on open(O_RDWR) for exampe).

for example).

>
> We can revisit this later if it turns out to be a performance problem in
> real life.
>
> Signed-off-by: Miklos Szeredi <mszeredi@xxxxxxxxxx>
> ---
> fs/overlayfs/Kconfig | 21 +++++++++++++++++++++
> fs/overlayfs/file.c | 22 ++++++++++++++++++++++
> fs/overlayfs/overlayfs.h | 7 +++++++
> fs/overlayfs/ovl_entry.h | 1 +
> fs/overlayfs/super.c | 22 ++++++++++++++++++++++
> 5 files changed, 73 insertions(+)
>
> diff --git a/fs/overlayfs/Kconfig b/fs/overlayfs/Kconfig
> index 17032631c5cf..991c0a5a0e00 100644
> --- a/fs/overlayfs/Kconfig
> +++ b/fs/overlayfs/Kconfig
> @@ -103,3 +103,24 @@ config OVERLAY_FS_XINO_AUTO
> For more information, see Documentation/filesystems/overlayfs.txt
>
> If unsure, say N.
> +
> +config OVERLAY_FS_COPY_UP_SHARED
> + bool "Overlayfs: copy up when mapping a file shared"

a shared file" ??

> + default n
> + depends on OVERLAY_FS
> + help
> + If this option is enabled then on mapping a file with MAP_SHARED
> + overlayfs copies up the file in anticipation of it being modified (just
> + like we copy up the file on O_WRONLY and O_RDWR in anticipation of
> + modification). This does not interfere with shared library loading, as
> + that uses MAP_PRIVATE. But there might be use cases out there where
> + this impacts performance and disk usage.
> +
> + This just selects the default, the feature can also be enabled or
> + disabled in the running kernel or individually on each overlay mount.
> +
> + To get maximally standard compliant behavior, enable this option.
> +
> + To get a maximally backward compatible kernel, disable this option.
> +
> + If unsure, say N.

For Kconfig (coding-style.rst):
Lines under a ``config`` definition are indented with one tab, while help text
is indented an additional two spaces.


--
~Randy