With clang 15, the following -Werror warnings is produced:
sys/fs/nfsclient/nfs_clkdtrace.c:544:19: error: a function declaration without a prototype is deprecated in all versions of C [-Werror,-Wstrict-prototypes]
dtnfsclient_unload()
^
void
This is because dtnfsclient_unload() is declared with a (void) argument
list, but defined with an empty argument list. Make the definition match
the declaration.
MFC after: 3 days
Commit 981ef32230 added optional use of the session
slots marked bad to recover a new session when all
slots are marked bad. The recovery worked against
a FreeBSD NFSv4.1/4.2 server, but not a Linux one.
It turns out that it was a bug in the FreeBSD client
and not the Linux server.
This patch fixes the client so that DeleteSession
followed by CreateSession after receiving a
NFSERR_BADSESSION error reply works against the
Linux server (and conforms to the RFC).
This also implies that the FreeBSD NFSv4.1/4.2
server needs to be fixed in a future commit.
Without the fix, the FreeBSD server does a full
recovery, including creation of a new ClientID,
but since "intr" mounts were broken, this does
not result in a regression.
This patch only affects the case where a CreateSession
is done for an already confirmed ClientID, which was
not being done prior to commit 981ef32230.
PR: 260011
MFC after: 2 weeks
Commit 326bcf9394 added a new "cred" argument to nfscl_reqstart().
Fsinfo is a NFSv3 RPC and since the "cred" argument is not
used for NFSv3, it does not matter what is passed in.
However, to be consistent with the rest of the patch, change the
argument to NULL.
This patch should not result in a semantics change.
PR: 260011
MFC after: 2 weeks
Commit 981ef32230 enabled marking of potentially bad
session slots when an RPC is interrupted if the "intr"
mount option is used. As such, it no longer makes
sense to call nfscl_hasexpired() for I/O operations that
reply NFSERR_BADSTATEID for NFSv4.1/4.2, which does a full
recovery of NFSv4 open state, destroying all byte range locks.
Recovery of open state should not be usually needed, since
the session slot has been marked potentially bad and,
although opens for the process that has been terminated via
a signal may be broken, locks for other processes will still
be valid.
This patch disables calls to nfscl_hasexpired for NFSv4.1/4.2
mounts, when I/O RPCs receive NFSERR_BADSTATEID replies.
It does not affect the behaviour of NFSv4.0 mounts nor
hard (non "intr") mounts.
PR: 260011
MFC after: 2 weeks
This patch moves nfsrpc_destroysession() into nfscommon.ko
and also modifies its arguments slightly. This will allow
the function to be called from nfsv4_sequencelookup() in
a future commit.
This patch should not result in a semantics change.
PR: 260011
MFC after: 2 weeks
To deal with broken session slots caused by the use of the
"soft" and/or "intr" mount options, nfsv4_sequencelookup()
will be modified to track the potentially broken session
slots. Then, when all session slots are potentially
broken, do a DeleteSession operation, so that the NFSv4
server will reply NFSERR_BADSESSION to uses of the session.
These changes will be done in future commits. However,
to do the DeleteSession RPC, a "cred" argument is needed
for nfscl_reqstart(). This patch adds this argument,
which is unused at this time. If the argument is NULL,
it indicates that DeleteSession should not be done
(usually because the RPC does not use sessions).
This patch should not cause any semantics change.
PR: 260011
MFC after: 2 weeks
Commit a7bb120f8b added a printf for the case where recovery
has not marked the session defunct by setting nfsess_defunct
to 1. It turns out that nfscl_hasexpired() calls
nfsrpc_setclient() directly, without setting nfsess_defunct.
This patch replaces the printf with code that sets
nfsess_defunct to 1 to handle this case.
If SIGTERM is issued to a process when it is doing I/O on
an "intr" mount, the NFSv4 server may reply NFSERR_BADSTATEID,
due to the Open being prematurely closed.
This can result in a call to nfscl_hasexpired() to do a
recovery.
This would explain at least one hang described in the PR.
PR: 260011
MFC after: 2 weeks
The vnode_vtype() macro was used to make the code compatible
with Mac OSX, for the Mac OSX port.
For FreeBSD, this macro just obscured the code, so
avoid using it to clean up the code.
This commit should not result in a semantics change.
The definition of "APPLE" was used by the Mac OSX port.
For FreeBSD, this definition is never used, so remove
the references to it to clean up the code.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c and
nfs_clstate.c.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c and
nfs_clvfsops.c. Future commits will do the same for other functions.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c.
Future commits will do the same for other functions.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c.
Future commits will do the same for other functions.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c.
Future commits will do the same for other functions.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c.
Future commits will do the same for other functions.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c.
Future commits will do the same for other functions.
This commit should not result in a semantics change.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
defined in nfs_clrpcops.c and called in nfs_clvnops.c.
Future commits will do the same for other functions.
This commit should not result in a semantics change.
When a NFSv4 byte range write lock is unlocked, all
data modifications need to be flushed to the server
to satisfy the coherency requirements for byte range
locking. However, if a write delegation for the
file is held by the client, flushing is not required,
since no other NFSv4 client can have the file NFSv4
Opened.
Found by inspection as suggested by a similar change
that was done to the Linux NFSv4 client.
Commits 3ad1e1c1ce and 57014f21e7 added a Lookup+Open
RPC for NFSv4.1/4.2, which can reduce the RPC count by
10-20% for some loads. This has now received a fair amount
of testing, so I think it is ok to enable it.
Note that the Lookup+Open RPC is only used when the
"oneopenown" mount option is specified. As such, this
change won't affect most NFSv4.1/4.2 mounts.
When a NFSv4.1/4.2 session to the NFS server (not a pNFS DS) is
replaced, the old session should always be marked defunct by
nfsess_defunct being set non-zero.
However, the hang reported by the PR suggests that this might
be the case.
This patch adds a printf() to indicate this has somehow happened.
PR: 260011
MFC after: 2 weeks
The NFSERR_BADSESSION reply from a NFSv4.1/4.2 server
is handled by newnfs_request(). It should not be handled
separately after newnfs_request() has returned.
These two cases were spotted during code inspection.
One of them should only redo what newnfs_request() already
did by the same "nfscl" thread. The other might have
resulted in recovery being done twice, but the code is
only used for "pnfs" mounts, so that would be rare.
Also, since NFSERR_BADSESSION should only be replied by
a server after the server reboots, this would be extremely
rare.
MFC after: 2 weeks
When the NFSv4.1/4.2 client is doing a pnfs mount to
mirrored DS(s), asynchronous threads are used to do the
RPCs against the DS(s) concurrently. If a DS is slow
to reply, it is possible for the "cred" to be free'd
before the asynchronous thread is done with it, causing
a panic/crash.
This patch fixes the problem by acquiring a refcount on
the "cred" while it is being used by the asynchronous thread
for a DS RPC. This bug was found during a recent IETF
NFSv4 testing event.
This bug only affects "pnfs" mounts to mirrored pNFS
servers.
MFC after: 2 weeks
For IO_APPEND VOP_WRITE()s, the code first does a
Getattr RPC to acquire the file's size, before it
can do the Write RPC.
Although NFS does not have an append write operation,
an NFSv4 compound can use a Verify operation to check
that the client's notion of the file's size is
correct, followed by the Write operation.
This patch modifies the NFSv4 client to use an Appendwrite
RPC, which does a Verify to check the file's size before
doing the Write. This avoids the need for a Getattr RPC
to preceed this RPC and reduces the RPC count by half for
IO_APPEND writes, so long as the client knows the file's
size.
The nfsd structure was moved from the stack to be malloc()'d,
since the kernel stack limit was being exceeded.
While here, fix the types of a few variables, although
there should not be any semantics change caused by these
type changes.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for assorted functions
local to nfs_clrpcops.c.
Future commits will do the same for other functions.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for nfscl_nget().
Future commits will do the same for other functions.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for nfscl_loadattrcache().
Future commits will do the same for other functions.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for nfscl_request().
Future commits will do the same for other functions.
The "void *stuff" (also called fstuff and dstuff) argument
was used by the Mac OSX port. For FreeBSD, this argument
is always NULL, so remove it to clean up the code.
This commit gets rid of "stuff" for nfscl_postop_attr().
Future commits will do the same for other functions.
For IO_APPEND VOP_WRITE()s, the code first does a
Getattr RPC to acquire the file's size, before it
can do the Write RPC.
Although NFS does not have an append write operation,
an NFSv4 compound can use a Verify operation to check
that the client's notion of the file's size is
correct, followed by the Write operation.
This patch modifies nfscl_wcc_data() to optionally
acquire the file's size, for use with an AppendWrite.
Although the "stuff" arguments are always NULL
(these were used for the Mac OSX port and should be
cleared out someday), make the argument to
nfscl_wcc_data() explicitly NULL for clarity.
This patch does not cause any semantics change until
the AppendWrite is added in a future commit.
Commit 867c27c23a modified the NFS client so that
it did IO_APPEND writes directly to the NFS server
bypassing the buffer cache, via a call to
nfs_directio_write(). Unfortunately, this (very old)
function assumed that the uio iov was for user space
addresses. As such, a IO_APPEND VOP_WRITE() that
was for system space, such as ktrace(1) does, would
write bogus data.
This patch fixes nfs_directio_write() so that it
handles kernel space uio iovs.
Reported by: bz
Tested by: bz
MFC after: 2 weeks
A NFSv4.1/4.2 pNFS mount needs to do a
separate Open+LayoutGet RPC, so do not do
a Lookup+Open RPC for these mounts.
The Lookup+Open RPCs are still disabled,
until further testing is done, so this patch
has no effect at this time.
Use of the Lookup+Open RPC is currently disabled,
due to a problem detected during testing. This
patch fixes this problem. The problem was that
nfscl_postop_attr() does not parse the attributes
if nd_repstat != 0. It also would parse the
return status for the operation, where the
Lookup+Open code had already parsed it.
The first change in the patch does not make any
semantics change, but makes the code identical
to what is done later in the function, so that
it is apparent that the semantics should be the
same in both places.
Lookup+Open remains disabled while further
testing is being done, so this patch has no
effect at this time.
ler@, markj@ reported a use after free in nfscl_cleanupkext().
They also provided two possible causes:
- In nfscl_cleanup_common(), "own" is the owner string
owp->nfsow_owner. If we free that particular
owner structure, than in subsequent comparisons
"own" will point to freed memory.
- nfscl_cleanup_common() can free more than one owner, so the use
of LIST_FOREACH_SAFE() in nfscl_cleanupkext() is not sufficient.
I also believe there is a 3rd:
- If nfscl_freeopenowner() or nfscl_freelockowner() is called
without the NFSCLSTATE mutex held, this could race with
nfscl_cleanupkext().
This could happen when the exclusive lock is held
on the client, such as when delegations are being returned
or when recovering from NFSERR_EXPIRED.
This patch fixes them as follows:
1 - Copy the owner string to a local variable before the
nfscl_cleanup_common() call.
2 - Modify nfscl_cleanup_common() so that it will never free more
than the first matching element. Normally there should only
be one element in each list with a matching open/lock owner
anyhow (but there might be a bug that results in a duplicate).
This should guarantee that the FOREACH_SAFE loops in
nfscl_cleanupkext() are adequate.
3 - Acquire the NFSCLSTATE mutex in nfscl_freeopenowner()
and nfscl_freelockowner(), if it is not already held.
This serializes all of these calls with the ones done in
nfscl_cleanup_common().
Reported by: ler
Reviewed by: markj
Tested by: cy
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D34334
ler@, markj@ reported a use after free in nfscl_cleanupkext().
They also provided two possible causes:
- In nfscl_cleanup_common(), "own" is the owner string
owp->nfsow_owner. If we free that particular
owner structure, than in subsequent comparisons
"own" will point to freed memory.
- nfscl_cleanup_common() can free more than one owner, so the use
of LIST_FOREACH_SAFE() in nfscl_cleanupkext() is not sufficient.
I also believe there is a 3rd:
- If nfscl_freeopenowner() or nfscl_freelockowner() is called
without the NFSCLSTATE mutex held, this could race with
nfscl_cleanupkext().
This could happen when the exclusive lock is held
on the client, such as when delegations are being returned.
This patch fixes them as follows:
1 - Copy the owner string to a local variable before the
nfscl_cleanup_common() call.
2 - Modify nfscl_cleanup_common() to return whether or not a
free was done.
When a free was done, do a goto to restart the loop, instead
of using FOREACH_SAFE, which was not safe in this case.
3 - Acquire the NFSCLSTATE mutex in nfscl_freeopenowner()
and nfscl_freelockowner(), if it not already held.
This serializes all of these calls with the ones done in
nfscl_cleanup_common().
Reported by: ler
Reviewed by: markj
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D34334
The function nfscl_getcookie(), which is essentially the
same as ncl_getcookie(), is never called, so delete it.
This is probably cruft left over from the port of the
NFSv4 code to FreeBSD several years ago.
Found while modifying the code to better use the
directory offset cookies.
MFC after: 2 weeks
The UFS and ZFS file systems only support Allow/Deny ACEs
in the NFSv4 ACLs. This patch does not allow the server
to parse Audit/Alarm ACEs. The NFSv4 client is still
allowed to pase Audit/Alarm ACEs, since non-FreeBSD NFSv4
servers may use them.
This patch should not have a significant effect, since the
UFS and ZFS file systems will not handle these ACEs anyhow.
It simply serves as an additional "safety belt" for the
NFSv4 server.
MFC after: 2 weeks
This reverts commit 0fa074b53e.
I now see that the implementation of the "dacl" operation
requires that the NFSv4 server to "automatic inheritance"
and I do not plan on doing this. As such, this patch is
harmless, but unneeded.
kib@ reported a problem which was resolved by
reverting commit 867c27c23a, which changed the NFS
client to use direct RPCs to the server for
IO_APPEND writes. He also spotted that the
code only invalidated buffer cache buffers
when they were marked NMODIFIED (had been
written into).
This patch modifies the NFS VOP_WRITE() to
always invalidate the buffer cache buffers
and pages for the file when IO_APPEND is
specified. It also includes some cleanup
suggested by kib@.
Reported by: kib
Tested by: kib
Reviewed by: kib
MFC after: 10 weeks
NFSv4.1/4.2 has an alternative to the acl attribute, called
dacl, that includes support for the ACL_ENTRY_INHERITED flag,
called NFSV4ACE_INHERITED in NFSv4.
This patch adds a dacl argument to nfsrv_buildacl(),
nfsrv_dissectacl() and nfsrv_dissectace(), so that they
will handle NFSV4ACE_INHERITED when dacl == true.
Since these functions are always called with dacl == false
for this patch, semantics should not have changed.
A future patch will add support for dacl.
MFC after: 2 weeks
Commit 867c27c23a modified the NFS client so that
it does IO_APPEND writes directly to the NFS server,
bypassing the buffer cache. However, this could result
in stale data in client pages when the file is mmap(2)'d.
As such, the NFS client needs to call vm_object_is_active()
to check if the file is mmap(2)'d and only do direct
output if the file is not mmap(2)'d.
This patch adds this check.
Although a simple patch, I have given it a long MFC,
since the related commit 867c27c23a made a significant
semantics change and, as such, has a long MFC.
MFC after: 3 months
Commit 867c27c23a enabled the n_directio_opens code
in open/close, which sets/clears NNONCACHE, for
IO_APPEND. This code should not be enabled unless
newnfs_directio_enable is non-zero.
This patch reverts that part of commit 867c27c23a.
A future patch that fixes the case where the
file that is being written IO_APPEND is mmap()'d.
MFC after: 3 months
The check for "not first operation" in CB_SEQUENCE
was done after the slot, etc. was updated. This patch
moves the check to the beginning of CB_SEQUENCE
processing.
While here, also fix the check for "no CB_SEQUENCE operation first"
by moving the check to the beginning of callback operation parsing,
since the check was in a couple of the other operations, but
not all of them.
Reported by: rtm@lcs.mit.edu
Tested by: rtm@lcs.mit.edu
PR: 260412
MFC after: 2 weeks
IO_APPEND writes have always been very slow over NFS, due to
the need to acquire an up to date file size after flushing
all writes to the NFS server.
This patch switches the IO_APPEND writes to use direct I/O,
bypassing the buffer cache. As such, flushing of writes
normally only occurs when the open(..O_APPEND..) is done.
It does imply that all writes must be done synchronously
and must be committed to stable storage on the file server
(NFSWRITE_FILESYNC).
For a simple test program that does 10,000 IO_APPEND writes
in a loop, performance improved significantly with this patch.
For a UFS exported file system, the test ran 12x faster.
This drops to 3x faster when the open(2)/close(2) are done
for each loop iteration.
For a ZFS exported file system, the test ran 40% faster.
The much smaller improvement may have been because the ZFS
file system I tested against does not have a ZIL log and
does have "sync" enabled.
Note that IO_APPEND write performance is still much slower
than when done on local file systems.
Although this is a simple patch, it does result in a
significant semantics change, so I have given it a
large MFC time.
Tested by: otis
MFC after: 3 months
As reported in PR#260343, nfs_allocate() did not check
the filesize rlimit. This patch adds that check.
PR: 260343
Reviewed by: asomers
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D33422
For pNFS mounts to mirrored Flexible File layout pNFS servers,
the "must_commit" component in the nfsclwritedsdorpc
structure must be checked and the "must_commit" argument passed
into nfscl_doiods() must be updated. Technically, only writes to
the DS with a writeverf change must be redone, but since this
occurrence will be rare, the must_commit argument to nfscl_doiosd()
is set to 1, so all writes to all DSs will be redone.
This bug would affect few, since use of mirrored pNFS servers
is rare and "writeverf" rarely changes. Normally "writeverf"
only changes when a NFS server reboots.
MFC after: 2 weeks
