Skip to contentSkip to conversation

nfsd: Remove incorrect check in nfsd4_validate_stateid

PATCH In review [email protected] 3 participants 7 messages Jul 18, 2023

Subsystems (2)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

PATCH 1/1 nfsd: Remove incorrect check in nfsd4_validate_stateid

From: Trond Myklebust <[email protected]>

If the client is calling TEST_STATEID, then it is because some event occurred that requires it to check all the stateids for validity and call FREE_STATEID on the ones that have been revoked. In this case, either the stateid exists in the list of stateids associated with that nfs4_client, in which case it should be tested, or it does not. There are no additional conditions to be considered.

Fixes 663e36f07666 ("nfsd4: kill warnings on testing stateids with mismatched clientids")
---
 fs/nfsd/nfs4state.c | 2 --
 1 file changed, 2 deletions(-)
fs/nfsd/nfs4state.c +0 -2
index 6e61fa3acaf1..3aefbad4cc09 100644
@@ -6341,8 +6341,6 @@ static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
 		CLOSE_STATEID(stateid))
 		return status;
-	if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid))
-		return status;
 	spin_lock(&cl->cl_lock);
 	s = find_stateid_locked(cl, stateid);
 	if (!s)
Signature · 1 line
2.41.0
maintainer
#1 lore raw
Quoted 22 lines · On Tue, 2023-07-18 at 08:38 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

If the client is calling TEST_STATEID, then it is because some event occurred that requires it to check all the stateids for validity and call FREE_STATEID on the ones that have been revoked. In this case, either the stateid exists in the list of stateids associated with that nfs4_client, in which case it should be tested, or it does not. There are no additional conditions to be considered.

Fixes 663e36f07666 ("nfsd4: kill warnings on testing stateids with mismatched clientids")
---
 fs/nfsd/nfs4state.c | 2 --
 1 file changed, 2 deletions(-)
fs/nfsd/nfs4state.c +0 -0
index 6e61fa3acaf1..3aefbad4cc09 100644

fs/nfsd/nfs4state.c static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid) · line 6346

Quoted 8 lines
fs/nfsd/nfs4state.c +0 -2
@@ -6341,8 +6341,6 @@ static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
 	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
 		CLOSE_STATEID(stateid))
 		return status;
-	if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid))
-		return status;
 	spin_lock(&cl->cl_lock);
 	s = find_stateid_locked(cl, stateid);
 	if (!s)

IDGI. Is this fixing an actual bug? Granted this code does seem unnecessary, but removing it doesn't seem like it will cause any user-visible change in behavior. Am I missing something?

Signature · 1 line
Jeff Layton <[email protected]>
author maintainer
#2 lore raw
Quoted 34 lines · Jeff Layton
Quoted 33 lines · On Tue, 2023-07-18 at 08:38 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

If the client is calling TEST_STATEID, then it is because some event
occurred that requires it to check all the stateids for validity and
call FREE_STATEID on the ones that have been revoked. In this case, either the stateid exists in the list of stateids associated with that
nfs4_client, in which case it should be tested, or it does not. There
are no additional conditions to be considered.

Fixes 663e36f07666 ("nfsd4: kill warnings on testing stateids with

mismatched clientids")

---
 fs/nfsd/nfs4state.c | 2 --

 1 file changed, 2 deletions(-)
fs/nfsd/nfs4state.c +0 -0
index 6e61fa3acaf1..3aefbad4cc09 100644
@@ -6341,8 +6341,6 @@ static __be32 nfsd4_validate_stateid(struct
 nfs4_client *cl, stateid_t *stateid)

        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) || CLOSE_STATEID(stateid))
                return status;
-       if (!same_clid(&stateid->si_opaque.so_clid, &cl-

Quoted 1 line

cl_clientid))

Quoted 8 lines
Quoted 4 lines

-               return status;
        spin_lock(&cl->cl_lock);
        s = find_stateid_locked(cl, stateid);
        if (!s)

IDGI. Is this fixing an actual bug? Granted this code does seem unnecessary, but removing it doesn't seem like it will cause any user-visible change in behavior. Am I missing something?

It was clearly triggering in
https://bugzilla.redhat.com/show_bug.cgi?id=2176575

Furthermore, if you look at commit 663e36f07666, you'll see that all it does is remove the log message because "it is expected". For some unknown reason, it did not register that "then the check is incorrect".

So yes, this is fixing a real bug.

Signature · 3 lines
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]
maintainer Reviewed-by
#3 lore raw
Quoted 44 lines · Trond Myklebust
Quoted 34 lines · Jeff Layton
Quoted 33 lines · On Tue, 2023-07-18 at 08:38 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

If the client is calling TEST_STATEID, then it is because some event
occurred that requires it to check all the stateids for validity and
call FREE_STATEID on the ones that have been revoked. In this case, either the stateid exists in the list of stateids associated with that
nfs4_client, in which case it should be tested, or it does not. There
are no additional conditions to be considered.

Fixes 663e36f07666 ("nfsd4: kill warnings on testing stateids with

mismatched clientids")

---
 fs/nfsd/nfs4state.c | 2 --

 1 file changed, 2 deletions(-)
fs/nfsd/nfs4state.c +0 -0
index 6e61fa3acaf1..3aefbad4cc09 100644
@@ -6341,8 +6341,6 @@ static __be32 nfsd4_validate_stateid(struct
 nfs4_client *cl, stateid_t *stateid)

        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) || CLOSE_STATEID(stateid))
                return status;
-       if (!same_clid(&stateid->si_opaque.so_clid, &cl-

Quoted 1 line

cl_clientid))

Quoted 8 lines
Quoted 4 lines

-               return status;
        spin_lock(&cl->cl_lock);
        s = find_stateid_locked(cl, stateid);
        if (!s)

IDGI. Is this fixing an actual bug? Granted this code does seem unnecessary, but removing it doesn't seem like it will cause any user-visible change in behavior. Am I missing something?

Quoted 7 lines

It was clearly triggering in
https://bugzilla.redhat.com/show_bug.cgi?id=2176575

Furthermore, if you look at commit 663e36f07666, you'll see that all it does is remove the log message because "it is expected". For some unknown reason, it did not register that "then the check is incorrect".

Yeah, that commit just removes the warning, AFAICT.

Quoted 2 lines

So yes, this is fixing a real bug.

My assumption was that for any stateid that the server hands out, the si_opaque.so_clid must match the clid. But...it looks like s2s copy might have changed that rule?

In any case, the patch looks fine, so I have no objection. I'm just trying to understand how this could happen.

#4 lore raw
In reply to Trond Myklebust
Quoted 48 lines · Trond Myklebust
Quoted 34 lines · Jeff Layton
Quoted 33 lines · On Tue, 2023-07-18 at 08:38 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

If the client is calling TEST_STATEID, then it is because some event
occurred that requires it to check all the stateids for validity and
call FREE_STATEID on the ones that have been revoked. In this case, either the stateid exists in the list of stateids associated with that
nfs4_client, in which case it should be tested, or it does not. There
are no additional conditions to be considered.

Fixes 663e36f07666 ("nfsd4: kill warnings on testing stateids with

mismatched clientids")

---
 fs/nfsd/nfs4state.c | 2 --
 1 file changed, 2 deletions(-)
fs/nfsd/nfs4state.c +0 -1
index 6e61fa3acaf1..3aefbad4cc09 100644
@@ -6341,8 +6341,6 @@ static __be32 nfsd4_validate_stateid(struct
 nfs4_client *cl, stateid_t *stateid)
        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
                CLOSE_STATEID(stateid))
                return status;
-       if (!same_clid(&stateid->si_opaque.so_clid, &cl-
 > cl_clientid))
Quoted 8 lines
Quoted 4 lines
fs/nfsd/nfs4state.c +0 -1
-               return status;
        spin_lock(&cl->cl_lock);
        s = find_stateid_locked(cl, stateid);
        if (!s)

IDGI. Is this fixing an actual bug? Granted this code does seem unnecessary, but removing it doesn't seem like it will cause any user-visible change in behavior. Am I missing something?

Quoted 6 lines

It was clearly triggering in
https://bugzilla.redhat.com/show_bug.cgi?id=2176575

Furthermore, if you look at commit 663e36f07666, you'll see that all it does is remove the log message because "it is expected". For some unknown reason, it did not register that "then the check is incorrect".

I don't think 663e36f altered this logic: it "returned status" when it emitted the warning, and it "returned status" after the warning was removed.

Quoted 1 line

So yes, this is fixing a real bug.

If there is a bug, wouldn't it have been introduced when the "!same_clid()" check was added?

Fixes 7df302f75ee2 ("NFSD: TEST_STATEID should not return NFS4ERR_STALE_STATEID")

--
Chuck Lever

author maintainer
#5 lore raw
Quoted 69 lines · Chuck Lever III
Quoted 55 lines

On Jul 18, 2023, at 9:51 AM, Trond Myklebust <[email protected]> wrote:

Quoted 40 lines · Jeff Layton
Quoted 39 lines · On Tue, 2023-07-18 at 08:38 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

If the client is calling TEST_STATEID, then it is because some event
occurred that requires it to check all the stateids for validity
and
call FREE_STATEID on the ones that have been revoked. In this case,
either the stateid exists in the list of stateids associated with
that
nfs4_client, in which case it should be tested, or it does not. There
are no additional conditions to be considered.

Fixes 663e36f07666 ("nfsd4: kill warnings on testing stateids

with
mismatched clientids")
Cc: [email protected]
Signed-off-by: Trond Myklebust
<[email protected]>
---
 fs/nfsd/nfs4state.c | 2 --

 1 file changed, 2 deletions(-)
fs/nfsd/nfs4state.c +0 -0
index 6e61fa3acaf1..3aefbad4cc09 100644
@@ -6341,8 +6341,6 @@ static __be32

nfsd4_validate_stateid(struct
nfs4_client *cl, stateid_t *stateid)
        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) || CLOSE_STATEID(stateid))
                return status;
-       if (!same_clid(&stateid->si_opaque.so_clid, &cl-

Quoted 1 line

cl_clientid))

Quoted 8 lines
Quoted 4 lines

-               return status;
        spin_lock(&cl->cl_lock);
        s = find_stateid_locked(cl, stateid);
        if (!s)

IDGI. Is this fixing an actual bug? Granted this code does seem unnecessary, but removing it doesn't seem like it will cause any user-visible change in behavior. Am I missing something?

Quoted 8 lines

It was clearly triggering in
https://bugzilla.redhat.com/show_bug.cgi?id=2176575

Furthermore, if you look at commit 663e36f07666, you'll see that all it
does is remove the log message because "it is expected". For some unknown reason, it did not register that "then the check is incorrect".

I don't think 663e36f altered this logic: it "returned status" when it emitted the warning, and it "returned status" after the warning was removed.

Quoted 5 lines
Quoted 1 line

So yes, this is fixing a real bug.

If there is a bug, wouldn't it have been introduced when the "!same_clid()" check was added?

Correct.

Quoted 3 lines
Fixes 7df302f75ee2 ("NFSD: TEST_STATEID should not return

NFS4ERR_STALE_STATEID")

It can't fix anything older than that patch, because it won't apply.

Signature · 3 lines
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]
#6 lore raw
Quoted 80 lines · Trond Myklebust
Quoted 69 lines · Chuck Lever III
Quoted 55 lines

On Jul 18, 2023, at 9:51 AM, Trond Myklebust <[email protected]> wrote:

Quoted 40 lines · Jeff Layton
Quoted 39 lines · On Tue, 2023-07-18 at 08:38 -0400, [email protected] wrote:

From: Trond Myklebust <[email protected]>

If the client is calling TEST_STATEID, then it is because some event
occurred that requires it to check all the stateids for validity
and
call FREE_STATEID on the ones that have been revoked. In this case,
either the stateid exists in the list of stateids associated with
that
nfs4_client, in which case it should be tested, or it does not. There
are no additional conditions to be considered.

Fixes 663e36f07666 ("nfsd4: kill warnings on testing stateids

with
mismatched clientids")
Cc: [email protected]
Signed-off-by: Trond Myklebust
<[email protected]>

---
 fs/nfsd/nfs4state.c | 2 --
 1 file changed, 2 deletions(-)
fs/nfsd/nfs4state.c +0 -0
index 6e61fa3acaf1..3aefbad4cc09 100644
@@ -6341,8 +6341,6 @@ static __be32

nfsd4_validate_stateid(struct
nfs4_client *cl, stateid_t *stateid)

        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
                CLOSE_STATEID(stateid))
                return status;

- if (!same_clid(&stateid->si_opaque.so_clid, &cl-

Quoted 1 line

cl_clientid))

Quoted 8 lines
Quoted 4 lines
-               return status;
        spin_lock(&cl->cl_lock);
        s = find_stateid_locked(cl, stateid);
        if (!s)

IDGI. Is this fixing an actual bug? Granted this code does seem unnecessary, but removing it doesn't seem like it will cause any user-visible change in behavior. Am I missing something?

Quoted 8 lines

It was clearly triggering in
https://bugzilla.redhat.com/show_bug.cgi?id=2176575

Furthermore, if you look at commit 663e36f07666, you'll see that all it
does is remove the log message because "it is expected". For some unknown reason, it did not register that "then the check is incorrect".

I don't think 663e36f altered this logic: it "returned status" when it emitted the warning, and it "returned status" after the warning was removed.

Quoted 5 lines
Quoted 1 line

So yes, this is fixing a real bug.

If there is a bug, wouldn't it have been introduced when the "!same_clid()" check was added?

Quoted 7 lines

Correct.

Quoted 3 lines
Fixes 7df302f75ee2 ("NFSD: TEST_STATEID should not return

NFS4ERR_STALE_STATEID")

It can't fix anything older than that patch, because it won't apply.

Testing now. I plan to apply it to nfsd-fixes (for 6.5-rc).

--
Chuck Lever