Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1735636 > unrolled thread

[PATCH 03/10] coda: honor AT_STATX_DONT_SYNC

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH 03/10] coda: honor AT_STATX_DONT_SYNC Miklos Szeredi <mszeredi@redhat.com> - 2017-09-20 10:50 +0200
    Re: [PATCH 03/10] coda: honor AT_STATX_DONT_SYNC Jan Harkes <jaharkes@cs.cmu.edu> - 2017-09-20 22:40 +0200
      Re: [PATCH 03/10] coda: honor AT_STATX_DONT_SYNC Miklos Szeredi <mszeredi@redhat.com> - 2017-09-21 14:50 +0200

#1735636 — [PATCH 03/10] coda: honor AT_STATX_DONT_SYNC

The description of this flag says "Don't sync attributes with the server".
In other words: always use the attributes cached in the kernel and don't
send network or local messages to refresh the attributes.

Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Cc: Jan Harkes <jaharkes@cs.cmu.edu>
---
 fs/coda/inode.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/coda/inode.c b/fs/coda/inode.c
index 6058df380cc0..734672b2cbdc 100644
--- a/fs/coda/inode.c
+++ b/fs/coda/inode.c
@@ -255,7 +255,10 @@ static void coda_evict_inode(struct inode *inode)
 int coda_getattr(const struct path *path, struct kstat *stat,
 		 u32 request_mask, unsigned int flags)
 {
-	int err = coda_revalidate_inode(d_inode(path->dentry));
+	int err = 0;
+
+	if (!(flags & AT_STATX_DONT_SYNC))
+		err = coda_revalidate_inode(d_inode(path->dentry));
 	if (!err)
 		generic_fillattr(d_inode(path->dentry), stat);
 	return err;
-- 
2.5.5

[toc] | [next] | [standalone]

#1736090

On Wed, Sep 20, 2017 at 10:39:58AM +0200, Miklos Szeredi wrote:
> The description of this flag says "Don't sync attributes with the server".
> In other words: always use the attributes cached in the kernel and don't
> send network or local messages to refresh the attributes.

What is the use case for this AT_STATX_DONT_SYNC flag?

I'm asking because the Coda userspace client potentially has attributes
that are not cached in the kernel but can be (re-)validated without
network communication. So if we just care about avoiding network
traffic we could propagate the flag up to userspace. If we want to avoid
context switches, disk I/O and only check on what happens to be cached
in the kernel the current approach works fine.

Jan

> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
> Cc: Jan Harkes <jaharkes@cs.cmu.edu>
> ---
>  fs/coda/inode.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/coda/inode.c b/fs/coda/inode.c
> index 6058df380cc0..734672b2cbdc 100644
> --- a/fs/coda/inode.c
> +++ b/fs/coda/inode.c
> @@ -255,7 +255,10 @@ static void coda_evict_inode(struct inode *inode)
>  int coda_getattr(const struct path *path, struct kstat *stat,
>  		 u32 request_mask, unsigned int flags)
>  {
> -	int err = coda_revalidate_inode(d_inode(path->dentry));
> +	int err = 0;
> +
> +	if (!(flags & AT_STATX_DONT_SYNC))
> +		err = coda_revalidate_inode(d_inode(path->dentry));
>  	if (!err)
>  		generic_fillattr(d_inode(path->dentry), stat);
>  	return err;
> -- 
> 2.5.5
> 
> 

[toc] | [prev] | [next] | [standalone]

#1736637

On Wed, Sep 20, 2017 at 10:04 PM, Jan Harkes <jaharkes@cs.cmu.edu> wrote:
> On Wed, Sep 20, 2017 at 10:39:58AM +0200, Miklos Szeredi wrote:
>> The description of this flag says "Don't sync attributes with the server".
>> In other words: always use the attributes cached in the kernel and don't
>> send network or local messages to refresh the attributes.
>
> What is the use case for this AT_STATX_DONT_SYNC flag?
>
> I'm asking because the Coda userspace client potentially has attributes
> that are not cached in the kernel but can be (re-)validated without
> network communication. So if we just care about avoiding network
> traffic we could propagate the flag up to userspace. If we want to avoid
> context switches, disk I/O and only check on what happens to be cached
> in the kernel the current approach works fine.

It's an interesting question, "server" could mean several different
things depending on the context.  I interpreted it as "any entity
outside the kernel the filesystem resides in", mainly because it's the
simplest interpretation.

Thanks,
Miklos


>> Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
>> Cc: Jan Harkes <jaharkes@cs.cmu.edu>
>> ---
>>  fs/coda/inode.c | 5 ++++-
>>  1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/fs/coda/inode.c b/fs/coda/inode.c
>> index 6058df380cc0..734672b2cbdc 100644
>> --- a/fs/coda/inode.c
>> +++ b/fs/coda/inode.c
>> @@ -255,7 +255,10 @@ static void coda_evict_inode(struct inode *inode)
>>  int coda_getattr(const struct path *path, struct kstat *stat,
>>                u32 request_mask, unsigned int flags)
>>  {
>> -     int err = coda_revalidate_inode(d_inode(path->dentry));
>> +     int err = 0;
>> +
>> +     if (!(flags & AT_STATX_DONT_SYNC))
>> +             err = coda_revalidate_inode(d_inode(path->dentry));
>>       if (!err)
>>               generic_fillattr(d_inode(path->dentry), stat);
>>       return err;
>> --
>> 2.5.5
>>
>>

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web