2013-03-05 20:03:04

by Tejun Heo

Subject: [PATCH 1/7] nfsd: remove unused get_new_stid()

get_new_stid() is no longer used since 3abdb607125 ("nfsd4: simplify
idr allocation"). Remove it.

Signed-off-by: Tejun Heo <[email protected]>
Cc: "J. Bruce Fields" <[email protected]>
Cc: [email protected]
---
fs/nfsd/nfs4state.c | 31 -------------------------------
1 file changed, 31 deletions(-)

diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index 16d39c6..d91d6db 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -230,37 +230,6 @@ static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
__nfs4_file_put_access(fp, oflag);
}

-static inline int get_new_stid(struct nfs4_stid *stid)
-{
- static int min_stateid = 0;
- struct idr *stateids = &stid->sc_client->cl_stateids;
- int new_stid;
- int error;
-
- error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
- /*
- * Note: the necessary preallocation was done in
- * nfs4_alloc_stateid(). The idr code caps the number of
- * preallocations that can exist at a time, but the state lock
- * prevents anyone from using ours before we get here:
- */
- WARN_ON_ONCE(error);
- /*
- * It shouldn't be a problem to reuse an opaque stateid value.
- * I don't think it is for 4.1. But with 4.0 I worry that, for
- * example, a stray write retransmission could be accepted by
- * the server when it should have been rejected. Therefore,
- * adopt a trick from the sctp code to attempt to maximize the
- * amount of time until an id is reused, by ensuring they always
- * "increase" (mod INT_MAX):
- */
-
- min_stateid = new_stid+1;
- if (min_stateid == INT_MAX)
- min_stateid = 0;
- return new_stid;
-}
-
static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
kmem_cache *slab)
{
--
1.8.1.4

2013-03-05 20:05:21

by J. Bruce Fields

Subject: Re: [PATCH 1/7] nfsd: remove unused get_new_stid()

On Tue, Mar 05, 2013 at 12:02:46PM -0800, Tejun Heo wrote:
> get_new_stid() is no longer used since 3abdb607125 ("nfsd4: simplify
> idr allocation"). Remove it.

Whoops, thanks for catching that.

Acked-by: J. Bruce Fields <[email protected]>

>
> Signed-off-by: Tejun Heo <[email protected]>
> Cc: "J. Bruce Fields" <[email protected]>
> Cc: [email protected]
> ---
> fs/nfsd/nfs4state.c | 31 -------------------------------
> 1 file changed, 31 deletions(-)
>
> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> index 16d39c6..d91d6db 100644
> --- a/fs/nfsd/nfs4state.c
> +++ b/fs/nfsd/nfs4state.c
> @@ -230,37 +230,6 @@ static void nfs4_file_put_access(struct nfs4_file *fp, int oflag)
> __nfs4_file_put_access(fp, oflag);
> }
>
> -static inline int get_new_stid(struct nfs4_stid *stid)
> -{
> - static int min_stateid = 0;
> - struct idr *stateids = &stid->sc_client->cl_stateids;
> - int new_stid;
> - int error;
> -
> - error = idr_get_new_above(stateids, stid, min_stateid, &new_stid);
> - /*
> - * Note: the necessary preallocation was done in
> - * nfs4_alloc_stateid(). The idr code caps the number of
> - * preallocations that can exist at a time, but the state lock
> - * prevents anyone from using ours before we get here:
> - */
> - WARN_ON_ONCE(error);
> - /*
> - * It shouldn't be a problem to reuse an opaque stateid value.
> - * I don't think it is for 4.1. But with 4.0 I worry that, for
> - * example, a stray write retransmission could be accepted by
> - * the server when it should have been rejected. Therefore,
> - * adopt a trick from the sctp code to attempt to maximize the
> - * amount of time until an id is reused, by ensuring they always
> - * "increase" (mod INT_MAX):
> - */
> -
> - min_stateid = new_stid+1;
> - if (min_stateid == INT_MAX)
> - min_stateid = 0;
> - return new_stid;
> -}
> -
> static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
> kmem_cache *slab)
> {
> --
> 1.8.1.4
>