public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Bernd Schubert <bschubert@ddn.com>
To: Joanne Koong <joannelkoong@gmail.com>,
	Luis Henriques <luis@igalia.com>,
	Miklos Szeredi <miklos@szeredi.hu>
Cc: "linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	Matt Harvey <mharvey@jumptrading.com>
Subject: Re: [RFC PATCH v3] fuse: add new function to invalidate cache for all inodes
Date: Mon, 10 Feb 2025 20:06:55 +0000	[thread overview]
Message-ID: <a5c3eb63-0fab-4751-af2f-8cb48c06b47f@ddn.com> (raw)
In-Reply-To: <2b65778e-7d26-4168-9346-6c1e01de350b@gmail.com>

On 2/10/25 20:33, Joanne Koong wrote:
> On 2/10/25 6:33 AM, 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@igalia.com>
>> ---
>> * Changes since v2
>> Use the new helper from fuse_reverse_inval_inode(), as suggested by
>> Bernd.
>>
>> Also updated patch description as per checkpatch.pl suggestion.
>>
>> * Changes since v1
>> 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@fromorbit.com
>>
>>   fs/fuse/inode.c           | 67 +++++++++++++++++++++++++++++++++++----
>>   include/uapi/linux/fuse.h |  3 ++
>>   2 files changed, 63 insertions(+), 7 deletions(-)
>>
>> diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
>> index e9db2cb8c150..45b9fbb54d42 100644
>> --- a/fs/fuse/inode.c
>> +++ b/fs/fuse/inode.c
>> @@ -547,25 +547,78 @@ 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);
>> +}
>> +
>> +static int fuse_reverse_inval_all(struct fuse_conn *fc)
>> +{
>> +    struct fuse_mount *fm;
>> +    struct super_block *sb;
>> +    struct inode *inode, *old_inode = NULL;
>> +
>> +    inode = fuse_ilookup(fc, FUSE_ROOT_ID, NULL);
>> +    if (!inode)
>> +        return -ENOENT;
>> +
>> +    fm = get_fuse_mount(inode);
> 
> I think if you pass in &fm as the 3rd arg to fuse_ilookup(), it'll pass
> back the fuse mount and we won't need get_fuse_mount().
> 
>> +    iput(inode);
>> +    if (!fm)
>> +        return -ENOENT;
>> +    sb = fm->sb;
>> +
>> +    spin_lock(&sb->s_inode_list_lock);
>> +    list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
>> +        spin_lock(&inode->i_lock);
>> +        if ((inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) ||
>> +            !atomic_read(&inode->i_count)) {
> 
> Will inode->i_count ever be 0? AFAIU, inode->i_count tracks the inode
> refcount, so if this is 0, doesn't this mean it wouldn't be on the sb-
>>s_inodes list?
> 
>> +            spin_unlock(&inode->i_lock);
>> +            continue;
>> +        }
>> +
>> +        __iget(inode);
>> +        spin_unlock(&inode->i_lock);
>> +        spin_unlock(&sb->s_inode_list_lock);
> 
> Maybe worth adding a comment here since there can be inodes added after
> the s_inode_list_lock is dropped and before it's acquired again that
> when inodes get added to the head of sb->s_inodes, it's always for I_NEW
> inodes.
> 
>> +        iput(old_inode);
> 
> Maybe a dumb question but why is old_inode needed? Why can't iput()just
> be called right after inval_single_inode()?

I had wondered the same in v1. Issue is that there is a list iteration
that releases the locks - if the put would be done immediately it could
not continue on "old_inode" as it might not exist anymore.



> 
>> +
>> +        inval_single_inode(inode, fc);
>> +
>> +        old_inode = inode;
>> +        cond_resched();
> 
> Could you explain why a cond_resched() is needed here?

Give other threads a chance to work? The list might be huge?



Thanks,
Bernd

  reply	other threads:[~2025-02-10 20:07 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-02-10 14:33 [RFC PATCH v3] fuse: add new function to invalidate cache for all inodes Luis Henriques
2025-02-10 19:33 ` Joanne Koong
2025-02-10 20:06   ` Bernd Schubert [this message]
2025-02-10 21:39     ` Luis Henriques

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=a5c3eb63-0fab-4751-af2f-8cb48c06b47f@ddn.com \
    --to=bschubert@ddn.com \
    --cc=joannelkoong@gmail.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=luis@igalia.com \
    --cc=mharvey@jumptrading.com \
    --cc=miklos@szeredi.hu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox