Hi,

On Wed, Jul 08, 2026 at 11:04:52PM +0200, Salvatore Bonaccorso wrote:
> Control: tags -1 + moreinfo
> 
> On Wed, Jul 08, 2026 at 11:50:10AM +0800, Gao Xiang wrote:
> > Package: src:linux
> > Version: 7.2~rc2-1~exp1
> > Severity: wishlist
> > 
> > Dear Debian kernel maintainers,
> > 
> > Currently EROFS already ships CONFIG_EROFS_FS_PAGE_CACHE_SHARE in
> > linux 7.0 as an optional feature.
> > 
> > since it's already shipped and users can choose to use it at runtime
> > or not, I hope this feature could be enabled in Debian.
> > 
> > In short, I hope the debian kernels can enabled
> > CONFIG_EROFS_FS_PAGE_CACHE_SHARE=y
> > 
> > to match the latest erofs-utils 1.9.2.
> 
> Looking at the option
> 
> | config EROFS_FS_PAGE_CACHE_SHARE
> |         bool "EROFS page cache share support (experimental)"
> |         depends on EROFS_FS && EROFS_FS_XATTR
> |         help
> |           This enables page cache sharing among inodes with identical
> |           content fingerprints on the same machine.
> |
> |           If unsure, say N.
> 
> it is yet marked as experimental. What are reasons for which we would
> already want to enable it? I'm aware I'm asking the maintainer, but
> maybe we should have it documented for the request. 

>From the EROFS policy perspective, almost every EROFS new feature will
be marked as EXPERIMENTAL initially (e.g. EROFS LZMA, DEFLATE, ZSTD were
all marked as EXPERIMENTAL): since experimental features are new, they
could have some bugs so users need to know they are using a new feature
(as the other new features, this feature also shows "EXPERIMENTAL EROFS
page cache share support in use. Use at your own risk!" when mounting.)

However, we always encourage users to enable, use and give feedback on
them if possible (unless on some embedded systems), so that new features
can drop experimental tag after a given period.

On the functionality side, as the option describes, this feature
enables inode sharing between filesystems, and it also needs to be
explicitly enabled with a mount option "inode_share".

Thanks,
Gao Xiang

> 
> Regards,
> Salvatore

Reply via email to