On 2/10/25 11:48, Luis Henriques wrote: > [re-sending -- for some reason I did a simple 'reply', not a 'reply-all'.] > > On Mon, Feb 10 2025, Bernd Schubert wrote: > >> On 2/10/25 10:48, Luis Henriques wrote: >>> Currently userspace is able to notify the kernel to invalidate the cache for >>> an inode. This means that, if all the inodes in a filesystem need to be >>> invalidated, then userspace needs to iterate through all of them and do this >>> kernel notification separately. >>> >>> This patch adds a new option that allows userspace to invalidate all the >>> inodes with a single notification operation. In addition to invalidate all >>> the inodes, it also shrinks the sb dcache. >>> >>> Signed-off-by: Luis Henriques <luis@xxxxxxxxxx> >>> --- >>> Hi! >>> >>> As suggested by Bernd, this patch v2 simply adds an helper function that >>> will make it easier to replace most of it's code by a call to function >>> super_iter_inodes() when Dave Chinner's patch[1] eventually gets merged. >>> >>> [1] https://lore.kernel.org/r/20241002014017.3801899-3-david@xxxxxxxxxxxxx >>> >>> fs/fuse/inode.c | 59 +++++++++++++++++++++++++++++++++++++++ >>> include/uapi/linux/fuse.h | 3 ++ >>> 2 files changed, 62 insertions(+) >>> >>> diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c >>> index e9db2cb8c150..be51b53006d8 100644 >>> --- a/fs/fuse/inode.c >>> +++ b/fs/fuse/inode.c >>> @@ -547,6 +547,62 @@ struct inode *fuse_ilookup(struct fuse_conn *fc, u64 nodeid, >>> return NULL; >>> } >>> >>> +static void inval_single_inode(struct inode *inode, struct fuse_conn *fc) >>> +{ >>> + struct fuse_inode *fi; >>> + >>> + fi = get_fuse_inode(inode); >>> + spin_lock(&fi->lock); >>> + fi->attr_version = atomic64_inc_return(&fc->attr_version); >>> + spin_unlock(&fi->lock); >>> + fuse_invalidate_attr(inode); >>> + forget_all_cached_acls(inode); >> >> >> Thank you, much easier to read. >> >> Could fuse_reverse_inval_inode() call into this? > > Yep, it could indeed. I'll do that in the next iteration, thanks! > >> What are the semantics >> for invalidate_inode_pages2_range() in this case? Totally invalidate? >> No page cache invalidation at all as right now? If so, why? > > So, if I change fuse_reverse_inval_inode() to use this help, it will still > need to keep the call to invalidate_inode_pages2_range(). But in the new > function fuse_reverse_inval_all(), I'm not doing it explicitly. Instead, > that function calls into shrink_dcache_sb(). I *think* that by doing so > the invalidation will eventually happen. Or am I wrong assuming that? I think it will drop it, if the dentry cache is the last user/reference of the inode. My issue is that it changes semantics a bit - without FUSE_INVAL_ALL_INODES the page cache is invalidated based on the given offset. Obviously we cannot give the offset for all inodes, but we at least document the different semantics in a comment above FUSE_INVAL_ALL_INODES? Sorry, should have asked earlier for it, just busy with multiple things in parallel... Thanks, Bernd
