When a lookup operation crosses into a new mountpoint, the mountpoint
must first be busied before the root vnode can be locked. When a
filesystem is unmounted, the vnode covered by the mountpoint must
first be locked, and then the busy count for the mountpoint drained.
Ordinarily, these two operations work fine if executed concurrently,
but with a stacked filesystem the root vnode may in fact use the
same lock as the covered vnode. By design, this will always be
the case for unionfs (with either the upper or lower root vnode
depending on mount options), and can also be the case for nullfs
if the target and mount point are the same (which admittedly is
very unlikely in practice).
In this case, we have LOR. The lookup path holds the mountpoint
busy while waiting on what is effectively the covered vnode lock,
while a concurrent unmount holds the covered vnode lock and waits
for the mountpoint's busy count to drain.
Attempt to resolve this LOR by allowing the stacked filesystem
to specify a new flag, VV_CROSSLOCK, on a covered vnode as necessary.
Upon observing this flag, the vfs_lookup() will leave the covered
vnode lock held while crossing into the mountpoint. Employ this flag
for unionfs with the caveat that it can't be used for '-o below' mounts
until other unionfs locking issues are resolved.
Reported by: pho
Tested by: pho
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D35054
providing the support for lseek(2) SEEK_DATA and SEEK_HOLE.
Reviewed by: markj
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D37024
When the user specifies SEEK_END, unlike SEEK_CUR, VOP_ADVLOCK must
adjust lock offsets itself.
Sort-of related to bug 266886.
MFC after: 2 weeks
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D37040
When the NFS server does the Setxattr or Rmxattr operation,
the Change attribute (va_filerev) needs to be updated.
Without this patch, that was not happening for the
pNFS server configuration. This patch does a Setattr
against the DS file to make the Change attribute
change.
This bug was discovered during a recent IETF NFSv4 testing
event, where the Change attribute wasn't changed in the
operation reply.
MFC after: 1 month
When the NFS server does Setxattr or Removexattr, the
operations must be done IO_SYNC. If a server
crashes/reboots immediately after replying it must
have the extended attribute changes.
Since UFS does extended attributes asynchronously
by default and there is no "ioflag" argument in
the VOP calls, follow the VOP calls with VOP_FSYNC(),
to ensure the operation has been done synchronously.
This was found by inspection while investigating a
bug discovered during a recent IETF NFSv4 testing
event, where the Change attribute wasn't changed
in the operation reply.
This bug will take further work for ZFS and the
pNFS server configuration, but is now fixed for
a non-pNFS UFS exported file system.
MFC after: 1 month
Commit efe58855f3 modifies IN_LOOPBACK() so that it uses a VNET
variable. Without this patch, nfscl_getmyip() uses IN_LOOPBACK()
when the VNET is not set and crashes the system.
nfscl_getmyip() is only called when a NFSv4.0 (not NFSv4.1/4.2)
mount is done.
This patch re-organizes nfscl_getmyip() so that IN_LOOPBACK()
is before the CURVENT_RESTORE() macro, to avoid the crashes.
Reviewed by: karels, zlei.huang_gmail.com
Differential Revision: https://reviews.freebsd.org/D37008
When doing small reads and writes use an intermediate buffer to store the
data to save locking the remote process to access data.
Reviewed by: imp @
Differential Revision: https://reviews.freebsd.org/D36633
MFC after: 1 week
Sponsored by: NVIDIA Networking
Commit ab17854f97 fixed access to v_mount when the
vnode is unlocked for nfs_copy_file_range().
This patch does the same for nfs_advlockasync().
MFC after: 1 week
Commit ab17854f97 fixed access to v_mount when the
vnode is unlocked for nfs_copy_file_range().
This patch does the same for nfs_ioctl().
Reviewed by: kib, markj
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D36846
and we checked that it is not reclaimed.
Reviewed by: markj, rmacklem
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D36722
Use 'td' as the local thread name.
Wrap long lines.
Remove unneeded blank lines.
Reviewed by: asomers, jah, markj
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D36625
This removes some of the complexity needed to maintain HASBUF and
allows for removing injecting SAVENAME by filesystems.
Reviewed by: kib (previous version)
Differential Revision: https://reviews.freebsd.org/D36542
Commit 40ada74ee1 modified the NFSv4.1/4.2 client so
that it would issue a DestroySession to the server when
all session slots are marked bad. This handles the
case where session slots get broken when "intr" or "soft"
NFSv4 fairly well.1/4.2 mounts are done.
There are two other cases where having an NFSv4.1/4.2
RPC attempt terminate without completion can leave
state in a non-determinate condition.
One is file locking RPCs. If the "nolockd" option is
used, this avoids file locking RPCs by doing locking
locally within the client.
The other is Open locks, but since all FreeBSD Open
locks are done with OPEN_SHARE_DENY_NONE, the locking
state for these should not be critical.
This patch enables use of "nolockd" for NFSv4 mounts,
so that it can be combined with "intr" and/or "soft",
making the latter more usable.
Use of "intr" or "soft" NFSv4 mounts are still not
recommended, but when combined with "nolockd" should
now work fairly well.
A man page update will be done as a separate commit.
MFC after: 2 weeks
Commit 40ada74ee1 modified the NFSv4.1/4.2 client so
that it would issue a DestroySession to the server when
all session slots are marked bad. Once this is done,
the Sequence operation should get a NFSERR_BADSESSION
reply from the server.
Without this patch, the code was setting ND_HASSLOTID
when, in fact, there was no slot marked in use by
nfsv4_sequencelookup(). This would result in the
code freeing a slot not in use. The effect of this
was minimal, since the session was already destroyed.
This patch fixes the code so that it does not set
ND_HASSLOTID for this case.
MFC after: 2 weeks
The NFSv4.1/4.2 client does recovery when it receives a
NFSERR_BADSESSION reply from the server. If the server has
not rebooted, this is often caused by multiple clients using
the same /etc/hostid and, as such, not being recognized as
different clients by the server.
This trivial patch adds a console message to suggest that
client's /etc/hostid's need to be checked for uniqueness.
MFC after: 2 weeks
The NFSv4.1/4.2 server generates a console message that indicates
that there is no session. I was until recently perplexed w.r.t. how
this could occur. It turns out that the common cause is multiple NFS
clients with the same /etc/hostid.
The host uuid is used by the FreeBSD NFSv4.1/4.2 client as a unique
identifier for the client. If multiple clients use the same host uuid,
this indicates to the NFSv4.1/4.2 server that they are the same client
and confusion occurs.
This trivial patch modifies the console message to suggest that the
client's /etc/hostid needs to be checked for uniqueness.
Reviewed by: asomers
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D36377
When the NFSv4.1/4.2 client is handling a server error
of NFSERR_BADSESSION, it retries RPCs with a new session.
Without this patch, the nd_slotid was not being updated
for the new session.
This would result in a bogus console message like
"Wrong session srvslot=X slot=Y" and then it would
free the incorrect slot, often generating a
"freeing free slot!!" console message as well.
This patch fixes the problem.
Note that FreeBSD NFSv4.1/4.2 servers only
generate a NFSERR_BADSESSION error after a reboot
or after a client does a DestroySession operation.
PR: 260011
MFC after: 1 week
When the NFSv4.1/4.2 client is handling a server error
of NFSERR_BADSESSION, it retries RPCs with a new session.
Without this patch, the nd_slotid was not being updated
for the new session.
This would result in a bogus console message like
"Wrong session srvslot=X slot=Y" and then it would
free the incorrect slot, often generating a
"freeing free slot!!" console message as well.
This patch fixes the problem.
Note that FreeBSD NFSv4.1/4.2 servers only
generate a NFSERR_BADSESSION error after a reboot
or after a client does a DestroySession operation.
PR: 260011
MFC after: 1 week
When a session has been marked defunct by the server
sending a NFSERR_BADSESSION reply to the NFSv4.1/4.2
client, nfsv4_sequencelookup() returns NFSERR_BADSESSION
without actually assigning a session slot.
Without this patch, newnfs_request() would erroneously
free slot 0.
This could result in the slot being reused prematurely,
but most likely just generated a "freeing free slot!!"
console message.
This patch fixes the code to not do the erroneous
freeing of the slot for this case.
PR: 260011
MFC after: 1 week
It is only ever xlocked in drain_dev_clone_events and the only consumer of
that routine does not need it -- eventhandler code already makes sure the
relevant callback is no longer running.
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D36268
Make most AST handlers dynamically registered. This allows to have
subsystem-specific handler source located in the subsystem files,
instead of making subr_trap.c aware of it. For instance, signal
delivery code on return to userspace is now moved to kern_sig.c.
Also, it allows to have some handlers designated as the cleanup (kclear)
type, which are called both at AST and on thread/process exit. For
instance, ast(), exit1(), and NFS server no longer need to be aware
about UFS softdep processing.
The dynamic registration also allows third-party modules to register AST
handlers if needed. There is one caveat with loadable modules: the
code does not make any effort to ensure that the module is not unloaded
before all threads processed through AST handler in it. In fact, this
is already present behavior for hwpmc.ko and ufs.ko. I do not think it
is worth the efforts and the runtime overhead to try to fix it.
Reviewed by: markj
Tested by: emaste (arm64), pho
Discussed with: jhb
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D35888
Currently the cuse(3) mmap(2) offset is split into 128 banks of 16 Mbytes.
Allow cuse(3) to make allocations that span multiple banks at the expense
of any fragmentation issues that may arise. Typically mmap(2) buffers are
well below 16 Mbytes. This allows 8K video resolution to work using webcamd.
Reviewed by: markj @
Differential Revision: https://reviews.freebsd.org/D35830
MFC after: 1 week
Sponsored by: NVIDIA Networking
