nfsd: Avoid /* Fallthrough */
Subsystems (3)
[email protected]
maintainerPATCH 1/2 nfsd: Avoid /* Fallthrough */
From: Trond Myklebust <[email protected]>
The '/* Fallthrough */' comment is already deprecated. Avoid it.
--- fs/nfsd/nfsfh.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
fs/nfsd/nfsfh.c +3 -3
@@ -302,9 +302,9 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)switch (rqstp->rq_vers) {case 3 : -if (!(dentry->d_sb->s_export_op->flags &EXPORT_OP_NOWCC )) -break ; -/* Fallthrough */ +if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOWCC ) + fhp->fh_no_wcc = true; +break ;case 2 : fhp->fh_no_wcc = true; }
Signature · 1 line
2.28.0
[email protected]
maintainerPATCH 2/2 nfsd: Record NFSv4 pre/post-op attributes as non-atomic
From: Trond Myklebust <[email protected]>
For the case of NFSv4, specify to the client that the the pre/post-op attributes were not recorded atomically with the main operation.
--- fs/nfs/export.c | 3 ++- fs/nfsd/nfsfh.c | 4 ++++ fs/nfsd/nfsfh.h | 5 +++++ fs/nfsd/xdr4.h | 2 +- include/linux/exportfs.h | 3 +++ 5 files changed, 15 insertions(+), 2 deletions(-)
fs/nfs/export.c +2 -1
@@ -172,5 +172,6 @@ const struct export_operations nfs_export_ops = { .fh_to_dentry = nfs_fh_to_dentry, .get_parent = nfs_get_parent, .flags =EXPORT_OP_NOWCC |EXPORT_OP_NOSUBTREECHK | - EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS, +EXPORT_OP_CLOSE_BEFORE_UNLINK |EXPORT_OP_REMOTE_FS | + EXPORT_OP_NOATOMIC_ATTR, };
fs/nfsd/nfsfh.c +4 -0
@@ -301,6 +301,10 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) fhp->fh_export = exp;switch (rqstp->rq_vers) { +case 4 : +if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOATOMIC_ATTR ) + fhp->fh_no_atomic_attr = true; +break ;case 3 :if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOWCC ) fhp->fh_no_wcc = true;
fs/nfsd/nfsfh.h +5 -0
@@ -36,6 +36,11 @@ typedef struct svc_fh {bool fh_locked;/* inode locked by us */ bool fh_want_write;/* remount protection taken */ bool fh_no_wcc;/* no wcc data needed */ +bool fh_no_atomic_attr; +/* +* wcc data is not atomic with +* operation +*/ int fh_flags;/* FH flags */ #ifdef CONFIG_NFSD_V3 bool fh_post_saved;/* post-op attrs saved */
fs/nfsd/xdr4.h +1 -1
@@ -748,7 +748,7 @@ static inline void set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp) { BUG_ON(!fhp->fh_pre_saved); - cinfo->atomic = (u32 )fhp->fh_post_saved; + cinfo->atomic = (u32)(fhp->fh_post_saved && !fhp->fh_no_atomic_attr); cinfo->before_change = fhp->fh_pre_change; cinfo->after_change = fhp->fh_post_change;
include/linux/exportfs.h +3 -0
@@ -217,6 +217,9 @@ struct export_operations {#define EXPORT_OP_NOSUBTREECHK (0x2)/* no subtree checking */ #define EXPORT_OP_CLOSE_BEFORE_UNLINK (0x4)/* close files before unlink */ #define EXPORT_OP_REMOTE_FS (0x8)/* Filesystem is remote */ +#define EXPORT_OP_NOATOMIC_ATTR (0x10)/* Filesystem cannot supply +atomic attribute updates +*/ unsigned long flags; };
Signature · 1 line
2.28.0
Hi -
Quoted 7 lines · On Nov 30, 2020, at 11:14 PM, [email protected] wrote:
From: Trond Myklebust <[email protected]>
The '/* Fallthrough */' comment is already deprecated. Avoid it.
Signed-off-by Trond Myklebust <[email protected]>
I've squashed this into "nfsd: add a new EXPORT_OP_NOWCC flag to struct export_operations", in addition to addressing Jeff's comment about line in the patch description that states support for re-exporting NFS filesystems may never be supported.
These changes will appear next time I push into the cel-next branch.
Quoted 24 lines
---
fs/nfsd/nfsfh.c| 6 +++---1 file changed, 3 insertions(+), 3 deletions(-)fs/nfsd/nfsfh.c +3
-3@@ -302,9 +302,9 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp)switch (rqstp->rq_vers) {case 3 : -if (!(dentry->d_sb->s_export_op->flags &EXPORT_OP_NOWCC )) -break ; -/* Fallthrough */ +if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOWCC ) + fhp->fh_no_wcc = true; +break ;case 2 : fhp->fh_no_wcc = true; }Signature · 1 line
2.28.0
--
Chuck Lever
Re: [PATCH 2/2] nfsd: Record NFSv4 pre/post-op attributes as non-atomic
In reply to [email protected]Hi Trond-
Quoted 8 lines · On Nov 30, 2020, at 11:14 PM, [email protected] wrote:
From: Trond Myklebust <[email protected]>
For the case of NFSv4, specify to the client that the the pre/post-op attributes were not recorded atomically with the main operation.
Signed-off-by Trond Myklebust <[email protected]>
I've applied this one provisionally to my tree for the next merge window.
Quoted 81 lines
---
fs/nfs/export.c | 3 ++- fs/nfsd/nfsfh.c | 4 ++++ fs/nfsd/nfsfh.h | 5 +++++ fs/nfsd/xdr4.h | 2 +- include/linux/exportfs.h | 3 +++ 5 files changed, 15 insertions(+), 2 deletions(-)fs/nfs/export.c +2
-1@@ -172,5 +172,6 @@ const struct export_operations nfs_export_ops = { .fh_to_dentry = nfs_fh_to_dentry, .get_parent = nfs_get_parent, .flags =EXPORT_OP_NOWCC |EXPORT_OP_NOSUBTREECHK | - EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS, +EXPORT_OP_CLOSE_BEFORE_UNLINK |EXPORT_OP_REMOTE_FS | + EXPORT_OP_NOATOMIC_ATTR, };fs/nfsd/nfsfh.c +4
-0@@ -301,6 +301,10 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) fhp->fh_export = exp;switch (rqstp->rq_vers) { +case 4 : +if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOATOMIC_ATTR ) + fhp->fh_no_atomic_attr = true; +break ;case 3 :if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOWCC ) fhp->fh_no_wcc = true;fs/nfsd/nfsfh.h +5
-0@@ -36,6 +36,11 @@ typedef struct svc_fh {bool fh_locked;/* inode locked by us */ bool fh_want_write;/* remount protection taken */ bool fh_no_wcc;/* no wcc data needed */ +bool fh_no_atomic_attr; +/* +* wcc data is not atomic with +* operation +*/ int fh_flags;/* FH flags */ #ifdef CONFIG_NFSD_V3 bool fh_post_saved;/* post-op attrs saved */ fs/nfsd/xdr4.h +1
-1@@ -748,7 +748,7 @@ static inline void set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp) { BUG_ON(!fhp->fh_pre_saved); - cinfo->atomic = (u32 )fhp->fh_post_saved; + cinfo->atomic = (u32)(fhp->fh_post_saved && !fhp->fh_no_atomic_attr); cinfo->before_change = fhp->fh_pre_change; cinfo->after_change = fhp->fh_post_change;include/linux/exportfs.h +3
-0@@ -217,6 +217,9 @@ struct export_operations {#define EXPORT_OP_NOSUBTREECHK (0x2)/* no subtree checking */ #define EXPORT_OP_CLOSE_BEFORE_UNLINK (0x4)/* close files before unlink */ #define EXPORT_OP_REMOTE_FS (0x8)/* Filesystem is remote */ +#define EXPORT_OP_NOATOMIC_ATTR (0x10)/* Filesystem cannot supply +atomic attribute updates +*/ unsigned long flags; };Signature · 1 line
2.28.0
--
Chuck Lever
Re: [PATCH 2/2] nfsd: Record NFSv4 pre/post-op attributes as non-atomic
In reply to [email protected]Quoted 19 lines · On Mon, Nov 30, 2020 at 11:14:27PM -0500, [email protected] wrote:
From: Trond Myklebust <[email protected]>
For the case of NFSv4, specify to the client that the the pre/post-op attributes were not recorded atomically with the main operation.
Signed-off-by Trond Myklebust <[email protected]>--- fs/nfs/export.c | 3 ++- fs/nfsd/nfsfh.c | 4 ++++ fs/nfsd/nfsfh.h | 5 +++++ fs/nfsd/xdr4.h | 2 +- include/linux/exportfs.h | 3 +++ 5 files changed, 15 insertions(+), 2 deletions(-)fs/nfs/export.c +0
-0
fs/nfs/export.c const struct export_operations nfs_export_ops = { · line 176
Quoted 6 lines
fs/nfs/export.c +2
-1@@ -172,5 +172,6 @@ const struct export_operations nfs_export_ops = { .fh_to_dentry = nfs_fh_to_dentry, .get_parent = nfs_get_parent, .flags =EXPORT_OP_NOWCC |EXPORT_OP_NOSUBTREECHK | - EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS, +EXPORT_OP_CLOSE_BEFORE_UNLINK |EXPORT_OP_REMOTE_FS | + EXPORT_OP_NOATOMIC_ATTR,
So I still don't understand why we need a new flag for this.
Before you said it was because a filesystem might want to turn off the v4 atomic flag but still return v3 post-wcc attributes.
But it seems that a) we have no example of a filesystem that wants to do that, b) it would violate the protocol anyway.
Is that right?
If so, then let's just stick to one flag for both.
--b.
Quoted 61 lines
fs/nfs/export.c +0
-0};fs/nfsd/nfsfh.c +4
-0@@ -301,6 +301,10 @@ static __be32 nfsd_set_fh_dentry(struct svc_rqst *rqstp, struct svc_fh *fhp) fhp->fh_export = exp;switch (rqstp->rq_vers) { +case 4 : +if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOATOMIC_ATTR ) + fhp->fh_no_atomic_attr = true; +break ;case 3 :if (dentry->d_sb->s_export_op->flags &EXPORT_OP_NOWCC ) fhp->fh_no_wcc = true;fs/nfsd/nfsfh.h +5
-0@@ -36,6 +36,11 @@ typedef struct svc_fh {bool fh_locked;/* inode locked by us */ bool fh_want_write;/* remount protection taken */ bool fh_no_wcc;/* no wcc data needed */ +bool fh_no_atomic_attr; +/* +* wcc data is not atomic with +* operation +*/ int fh_flags;/* FH flags */ #ifdef CONFIG_NFSD_V3 bool fh_post_saved;/* post-op attrs saved */ fs/nfsd/xdr4.h +1
-1@@ -748,7 +748,7 @@ static inline void set_change_info(struct nfsd4_change_info *cinfo, struct svc_fh *fhp) { BUG_ON(!fhp->fh_pre_saved); - cinfo->atomic = (u32 )fhp->fh_post_saved; + cinfo->atomic = (u32)(fhp->fh_post_saved && !fhp->fh_no_atomic_attr); cinfo->before_change = fhp->fh_pre_change; cinfo->after_change = fhp->fh_post_change;include/linux/exportfs.h +3
-0@@ -217,6 +217,9 @@ struct export_operations {#define EXPORT_OP_NOSUBTREECHK (0x2)/* no subtree checking */ #define EXPORT_OP_CLOSE_BEFORE_UNLINK (0x4)/* close files before unlink */ #define EXPORT_OP_REMOTE_FS (0x8)/* Filesystem is remote */ +#define EXPORT_OP_NOATOMIC_ATTR (0x10)/* Filesystem cannot supply +atomic attribute updates +*/ unsigned long flags; };Signature · 1 line
2.28.0
Re: [PATCH 2/2] nfsd: Record NFSv4 pre/post-op attributes as non-atomic
Quoted 35 lines · J. Bruce Fields
Quoted 23 lines · On Mon, Nov 30, 2020 at 11:14:27PM -0500, [email protected] wrote:
From: Trond Myklebust <[email protected]>
For the case of NFSv4, specify to the client that the the pre/post- op
attributes were not recorded atomically with the main operation.Signed-off-by Trond Myklebust <[email protected]>---
fs/nfs/export.c | 3 ++- fs/nfsd/nfsfh.c | 4 ++++ fs/nfsd/nfsfh.h | 5 +++++ fs/nfsd/xdr4.h | 2 +- include/linux/exportfs.h | 3 +++ 5 files changed, 15 insertions(+), 2 deletions(-)fs/nfs/export.c +0
-0@@ -172,5 +172,6 @@ const struct export_operations nfs_export_ops = { .fh_to_dentry = nfs_fh_to_dentry, .get_parent = nfs_get_parent,Quoted 4 lines
.flags = EXPORT_OP_NOWCC|EXPORT_OP_NOSUBTREECHK| - EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS, + EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS| + EXPORT_OP_NOATOMIC_ATTR,
So I still don't understand why we need a new flag for this.
Before you said it was because a filesystem might want to turn off the
v4 atomic flag but still return v3 post-wcc attributes.
Quoted 7 lines
But it seems that a) we have no example of a filesystem that wants to do
that, b) it would violate the protocol anyway.Is that right?
If so, then let's just stick to one flag for both.
The "atomic" flag in NFSv4 is very specifically tied to a single attribute (the change attribute) and how it is collected in a very limited set of operations. The truth is that we probably _can_ eventually set it when re-exporting NFSv4 as NFSv4, assuming that the server actually supplies us with an atomic update.
The same is not true of WCC. We might be able to supply knfsd with atomic updates for some operations, but certainly not for something like READ or WRITE.
So I do think this needs to be separate flags. It is definitely describing completely different functionality.
Signature · 3 lines
Trond Myklebust Linux NFS client maintainer, Hammerspace [email protected]
Re: [PATCH 2/2] nfsd: Record NFSv4 pre/post-op attributes as non-atomic
Quoted 55 lines · Trond Myklebust
Quoted 39 lines · J. Bruce Fields
Quoted 23 lines · On Mon, Nov 30, 2020 at 11:14:27PM -0500, [email protected] wrote:
From: Trond Myklebust <[email protected]>
For the case of NFSv4, specify to the client that the the pre/post-
op
attributes were not recorded atomically with the main operation.Signed-off-by Trond Myklebust <[email protected]>---
fs/nfs/export.c | 3 ++- fs/nfsd/nfsfh.c | 4 ++++ fs/nfsd/nfsfh.h | 5 +++++ fs/nfsd/xdr4.h | 2 +- include/linux/exportfs.h | 3 +++ 5 files changed, 15 insertions(+), 2 deletions(-)fs/nfs/export.c +0
-0@@ -172,5 +172,6 @@ const struct export_operations nfs_export_ops=
{Quoted 8 lines
.fh_to_dentry =
nfs_fh_to_dentry,
.get_parent =nfs_get_parent,
.flags = EXPORT_OP_NOWCC|EXPORT_OP_NOSUBTREECHK| -
EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS, + EXPORT_OP_CLOSE_BEFORE_UNLINK|EXPORT_OP_REMOTE_FS |
+ EXPORT_OP_NOATOMIC_ATTR,So I still don't understand why we need a new flag for this.
Before you said it was because a filesystem might want to turn off the
v4 atomic flag but still return v3 post-wcc attributes.Quoted 8 lines
But it seems that a) we have no example of a filesystem that wants to
do
that, b) it would violate the protocol anyway.Is that right?
If so, then let's just stick to one flag for both.
The "atomic" flag in NFSv4 is very specifically tied to a single attribute (the change attribute) and how it is collected in a very limited set of operations. The truth is that we probably _can_ eventually set it when re-exporting NFSv4 as NFSv4, assuming that the server actually supplies us with an atomic update.
Quoted 7 lines
The same is not true of WCC. We might be able to supply knfsd with atomic updates for some operations, but certainly not for something like READ or WRITE.
So I do think this needs to be separate flags. It is definitely describing completely different functionality.
Actually. The above is a good reason to just drop the EXPORT_OP_NOATOMIC_ATTR altogether.
As far as I can tell, there is no need for it at all, since both the NFSv3 and NFSv4 client can supply atomic struct change_info4 in the cases where it is relevant (those cases being recording the changes to the parent directory/ies when doing CREATE, OPEN(O_CREAT), LINK, REMOVE and RENAME).
Signature · 3 lines
Trond Myklebust Linux NFS client maintainer, Hammerspace [email protected]
Re: [PATCH 2/2] nfsd: Record NFSv4 pre/post-op attributes as non-atomic
Quoted 5 lines · Trond Myklebust
As far as I can tell, there is no need for it at all, since both the NFSv3 and NFSv4 client can supply atomic struct
change_info4in the cases where it is relevant (those cases being recording the changes to the parent directory/ies when doing CREATE,OPEN(O_CREAT), LINK, REMOVE and RENAME).
I was wondering about that. We'd need some additional interface to allow nfs to supply that stuff to nfsd, right?
--b.
Re: [PATCH 2/2] nfsd: Record NFSv4 pre/post-op attributes as non-atomic
Quoted 10 lines · On Tue, 2020-12-01 at 15:53 -0500, [email protected] wrote:
Quoted 8 lines · Trond Myklebust
As far as I can tell, there is no need for it at all, since both the
NFSv3 and NFSv4 client can supply atomic structchange_info4in the cases where it is relevant (those cases being recording the changes to
the parent directory/ies when doing CREATE,OPEN(O_CREAT), LINK, REMOVE
and RENAME).
Quoted 2 lines
I was wondering about that. We'd need some additional interface to allow nfs to supply that stuff to nfsd, right?
The only problem is the pre-op attributes, since those are supplied by the server but never recorded anywhere. Otherwise, all you really need is to hold the inode lock on the directory to prevent the client from making updates after the operation is done.
Signature · 3 lines
Trond Myklebust Linux NFS client maintainer, Hammerspace [email protected]
Re: [PATCH 2/2] nfsd: Record NFSv4 pre/post-op attributes as non-atomic
Quoted 11 lines · Trond Myklebust
Quoted 10 lines · On Tue, 2020-12-01 at 15:53 -0500, [email protected] wrote:
Quoted 8 lines · Trond Myklebust
As far as I can tell, there is no need for it at all, since both the
NFSv3 and NFSv4 client can supply atomic structchange_info4in the cases where it is relevant (those cases being recording the changes to
the parent directory/ies when doing CREATE,OPEN(O_CREAT), LINK, REMOVE
and RENAME).
Quoted 7 lines
Quoted 2 lines
I was wondering about that. We'd need some additional interface to allow nfs to supply that stuff to nfsd, right?
The only problem is the pre-op attributes, since those are supplied by the server but never recorded anywhere. Otherwise, all you really need is to hold the inode lock on the directory to prevent the client from making updates after the operation is done.
We already have the lock over all those operations.
I was worried that wouldn't prevent a getattr in another thread from updating the change attribute. Is that not possible?
My only idea for returning extra information was a hack using kthread_data() as we're doing in nfsd_breaker_owns_lease(); nfs would call something like:
void nfsd_set_cinfo(struct nfs4_change_info *cinfo)
{
struct svc_rqst *rqst;
if (!i_am_nfsd())
return;
rqst = kthread_data(current);
rqst->cinfo = cinfo;
}Then nfsd fetches it afterwards. Maybe there's a better way.
--b.