When coding the pNFS server, I added vn_start_write() calls in nfsrv_copymr()
done while the vnodes were locked, not realizing I had introduced LORs and
possible deadlock when an exported file system on the MDS is suspended.
This patch fixes the LORs by moving the vn_start_write() calls up to before
where the vnodes are locked. For "tvp", the vn_start_write() probaby isn't
necessary, because NFS mounts can't be suspended. However, I think doing
so is harmless.
Thanks go to kib@ for letting me know that I had introduced these LORs.
This patch only affects the behaviour of the pNFS server when pnfsdscopymr(8)
is used to recover a mirrored DS.
After a re-read of the appropriate section of RFC5661, I decided that a
few things should be changed related to LayoutRecall callback handling.
Here are the things fixed by this patch.
- For two of the three cases that LayoutRecall is done, I now think
setting the clora_changed argument false is correct.
- All errors other than NFSERR_DELAY returned by LayoutRecall appear
permanent, so don't retry for any of them. (NFSERR_DELAY is retried by
newnfs_request(), so it is not affected by this patch.)
- Instead of waiting "forever" (actually until the process is SIGTERM'd)
for Layouts to be returned during a mirror copy, fail and return
ENXIO after about 1minute.
Waiting for a <ctrl>C made sense when pnfsdscopymr() was done by itself,
but did not make sense when done via find(1).
This patch only affects the pNFS server.
At least on x86, fhandle_t is a packed structure, so I believe an
assignment will copy all the bits. However, for some current/future
architectures, there might be padding in the structure that doesn't get
copied via an assignment.
Since NFS assumes a file handle is an opaque blob of bits that can be
compared via memcmp()/bcmp(), all the bits including any padding must be
copied.
This patch replaces the assignments with a call to a byte copy function.
Spotted during code inspection.
I believe that a ReclaimComplete with rca_one_fs == TRUE is only
to be used after a file system has been transferred to a different
file server. However, RFC5661 is somewhat vague w.r.t. this and
the ESXi 6.7 client does both a ReclaimComplete with rca_one_fs == TRUE
and one with ReclaimComplete with rca_one_fs == FALSE.
Therefore, just ignore the rca_one_fs == TRUE operation and return
NFS_OK without doing anything instead of replying NFS4ERR_NOTSUPP.
This allows the ESXi 6.7 NFSv4.1 client to do a mount.
After discussion on the NFSv4 IETF working group mailing list, doing this
along with setting a flag to note that a ReclaimComplete with rca_one_fs TRUE
was an appropriate way to handle this.
The flag that indicates that a ReclaimComplete with rca_one_fs == TRUE was
done may be used to disable replies of NFS4ERR_GRACE for non-reclaim
state operations in a future commit.
This patch along with r332790, r334492 and r336357 allow ESXi 6.7 NFSv4.1 mounts
work ok. ESX 6.5 NFSv4.1 mounts do not work well, due to what I believe are
violations of RFC-5661 and should not be used.
Reported by: andreas.nagy@frequentis.com
Tested by: andreas.nagy@frequentis.com, daniel@ftml.net (earlier version)
MFC after: 2 weeks
Relnotes: yes
The ESXi NFSv4.1 client will generate warning messages when the reason for
not issuing a delegation is two. Two refers to a resource limit and I do
not see why it would be considered invalid. However it probably was not the
best choice of reason for not issuing a delegation.
This patch changes the reasons used to ones that the ESXi client doesn't
complain about. This change does not affect the FreeBSD client and does
not appear to affect behaviour of the Linux NFSv4.1 client.
RFC5661 defines these "reasons" but does not give any guidance w.r.t. which
ones are more appropriate to return to a client.
Tested by: andreas.nagy@frequentis.com
PR: 226650
MFC after: 2 weeks
The pnfsdskill(8) command will normally fail if there is no valid mirror
for the DS to be disabled. However, a system administrator may need to
disable a DS which does not have a valid mirror so that the nfsd threads
can be terminated. This patch adds the kernel code needed by pnfsdskill(8)
to implement this "forced" case of disabling a DS.
This patch only affects the pNFS server.
If a mirrored DS is being recovered that has a lot of large sparse files,
pnfsdscopymr(8) would use a lot of space on the recovered mirror since it
would write the "holes" in the file being mirrored.
This patch adds code to check for a "hole" and skip doing the write.
The check is done on a "per PNFSDS_COPYSIZ size block", which is currently 64K.
I think that most file server file systems will be using a blocksize at
least this large. If the file server is using a smaller blocksize and
smaller holes need to be preserved, PNFSDS_COPYSIZ could be decreased.
The block of 0s is malloc()d, since pnfsdcopymr(8) should be an infrequent
occurrence.
an NFSERR_STALE error reported via a LayoutReturn.
The current FreeBSD client can generate these errors for an operational
DS while doing a recovery of a mirror after a mirrored DS has been repaired.
I am not sure why these errors occur, but my best current guess is a race
between the Layout Recall issued by the kernel code run from pnfsdscopymr(8)
and a Read operation on the DS for the file bing copied.
The errors are not fatal, since the client falls back on doing I/O through
the MDS, which can do the I/O successfully as a proxy. (The fact that the
MDS can do this indicates that the file does still exist on the functioning
DS.)
This change only affects the pNFS server and only when a client does a
LayoutReturn with the NFSERR_STALE error report.
The recently added feature of the pNFS server will set an fsid for the
MDS file system to define the file system a DS should store files for.
For a case where a DS handling all file systems has failed, it was possible
for the code to check for a mirror with a specified fs, even though
nfsdev_mdsisset was 0, possibly causing a false successful check for a mirror.
This patch adds a check for nfsdev_mdsisset != 0 to avoid this.
It only affects the pNFS server for a rare case. Found via code inspection.
Without this patch, the pNFS server distributes the data storage files across
all of the specified DSs.
A tester noted that it would be nice if a system administrator could control
which DSs are used to store the file data for a given exported MDS file system.
This patch adds the kernel support to do this. It also makes a slight semantic
change to nfsv4_findmirror(), since some uses of it no longer require that
the DS being searched for have a current mirror.
A patch that will be committed in a few minutes will modify the nfsd daemon
to support this feature.
The patch should only affect sites using the pNFS server (specified via the
"-p" command line option for nfsd.
Suggested by: james.rose@framestore.com
This patch adds a counter that limits the number of disabled mirrored DSs
to mirror level - 1. It also makes a small change that keeps a Write that
has failed with EACCES when attempted by a client to a DS from disabling
the DS.
This patch only affects the pNFS server.
This code merge adds a pNFS service to the NFSv4.1 server. Although it is
a large commit it should not affect behaviour for a non-pNFS NFS server.
Some documentation on how this works can be found at:
http://people.freebsd.org/~rmacklem/pnfs-planb-setup.txt
and will hopefully be turned into a proper document soon.
This is a merge of the kernel code. Userland and man page changes will
come soon, once the dust settles on this merge.
It has passed a "make universe", so I hope it will not cause build problems.
It also adds NFSv4.1 server support for the "current stateid".
Here is a brief overview of the pNFS service:
A pNFS service separates the Read/Write oeprations from all the other NFSv4.1
Metadata operations. It is hoped that this separation allows a pNFS service
to be configured that exceeds the limits of a single NFS server for either
storage capacity and/or I/O bandwidth.
It is possible to configure mirroring within the data servers (DSs) so that
the data storage file for an MDS file will be mirrored on two or more of
the DSs.
When this is used, failure of a DS will not stop the pNFS service and a
failed DS can be recovered once repaired while the pNFS service continues
to operate. Although two way mirroring would be the norm, it is possible
to set a mirroring level of up to four or the number of DSs, whichever is
less.
The Metadata server will always be a single point of failure,
just as a single NFS server is.
A Plan B pNFS service consists of a single MetaData Server (MDS) and K
Data Servers (DS), all of which are recent FreeBSD systems.
Clients will mount the MDS as they would a single NFS server.
When files are created, the MDS creates a file tree identical to what a
single NFS server creates, except that all the regular (VREG) files will
be empty. As such, if you look at the exported tree on the MDS directly
on the MDS server (not via an NFS mount), the files will all be of size 0.
Each of these files will also have two extended attributes in the system
attribute name space:
pnfsd.dsfile - This extended attrbute stores the information that
the MDS needs to find the data storage file(s) on DS(s) for this file.
pnfsd.dsattr - This extended attribute stores the Size, AccessTime, ModifyTime
and Change attributes for the file, so that the MDS doesn't need to
acquire the attributes from the DS for every Getattr operation.
For each regular (VREG) file, the MDS creates a data storage file on one
(or more if mirroring is enabled) of the DSs in one of the "dsNN"
subdirectories. The name of this file is the file handle
of the file on the MDS in hexadecimal so that the name is unique.
The DSs use subdirectories named "ds0" to "dsN" so that no one directory
gets too large. The value of "N" is set via the sysctl vfs.nfsd.dsdirsize
on the MDS, with the default being 20.
For production servers that will store a lot of files, this value should
probably be much larger.
It can be increased when the "nfsd" daemon is not running on the MDS,
once the "dsK" directories are created.
For pNFS aware NFSv4.1 clients, the FreeBSD server will return two pieces
of information to the client that allows it to do I/O directly to the DS.
DeviceInfo - This is relatively static information that defines what a DS
is. The critical bits of information returned by the FreeBSD
server is the IP address of the DS and, for the Flexible
File layout, that NFSv4.1 is to be used and that it is
"tightly coupled".
There is a "deviceid" which identifies the DeviceInfo.
Layout - This is per file and can be recalled by the server when it
is no longer valid. For the FreeBSD server, there is support
for two types of layout, call File and Flexible File layout.
Both allow the client to do I/O on the DS via NFSv4.1 I/O
operations. The Flexible File layout is a more recent variant
that allows specification of mirrors, where the client is
expected to do writes to all mirrors to maintain them in a
consistent state. The Flexible File layout also allows the
client to report I/O errors for a DS back to the MDS.
The Flexible File layout supports two variants referred to as
"tightly coupled" vs "loosely coupled". The FreeBSD server always
uses the "tightly coupled" variant where the client uses the
same credentials to do I/O on the DS as it would on the MDS.
For the "loosely coupled" variant, the layout specifies a
synthetic user/group that the client uses to do I/O on the DS.
The FreeBSD server does not do striping and always returns
layouts for the entire file. The critical information in a layout
is Read vs Read/Writea and DeviceID(s) that identify which
DS(s) the data is stored on.
At this time, the MDS generates File Layout layouts to NFSv4.1 clients
that know how to do pNFS for the non-mirrored DS case unless the sysctl
vfs.nfsd.default_flexfile is set non-zero, in which case Flexible File
layouts are generated.
The mirrored DS configuration always generates Flexible File layouts.
For NFS clients that do not support NFSv4.1 pNFS, all I/O operations
are done against the MDS which acts as a proxy for the appropriate DS(s).
When the MDS receives an I/O RPC, it will do the RPC on the DS as a proxy.
If the DS is on the same machine, the MDS/DS will do the RPC on the DS as
a proxy and so on, until the machine runs out of some resource, such as
session slots or mbufs.
As such, DSs must be separate systems from the MDS.
Tested by: james.rose@framestore.com
Relnotes: yes
Under some fairly unusual circumstances, the Linux NFSv4.1 client is
doing a BindConnectiontoSession operation for TCP connections.
It is also used by the ESXi6.5 NFSv4.1 client.
This patch adds this operation to the NFSv4.1 server.
Reported by: andreas.nagy@frequentis.com
Tested by: andreas.nagy@frequentis.com
MFC after: 2 weeks
If a client did a DestroySession on a session while it was still in use,
the server might try to use the session structure after it is free'd.
I think the client has violated RFC5661 if it does this, but this patch
makes DestroySession block all other nfsd threads so no thread could
be using the session when it is free'd. After the DestroySession, nfsd
threads will not be able to find the session. The patch also adds a check
for nd_sessionid being set, although if that was not the case it would have
been all 0s and unlikely to have a false match.
This might fix the crashes described in PR#228497 for the FreeNAS server.
PR: 228497
MFC after: 1 week
Since NFSv4.1 clients normally create a single session which supports
both fore and back channels, it is unlikely that a callback will fail
due to a lack of a back channel.
However, if this failure occurred, the session wasn't being dereferenced
and would never be free'd.
Found by inspection during pNFS server development.
Tested by: andreas.nagy@frequentis.com
MFC after: 2 months
The NFSv4 protocol requires that the server only allow reclaim of state
and not issue any new open/lock state for a grace period after booting.
The NFSv4.0 protocol required this grace period to be greater than the
lease duration (over 2minutes). For NFSv4.1, the client tells the server
that it has done reclaiming state by doing a ReclaimComplete operation.
If all NFSv4 clients are NFSv4.1, the grace period can end once all the
clients have done ReclaimComplete, shortening the time period considerably.
This patch does this. If there are any NFSv4.0 mounts, the grace period
will still be over 2minutes.
This change is only an optimization and does not affect correct operation.
Tested by: andreas.nagy@frequentis.com
MFC after: 2 months
For a fairly rare case of a client doing an ExchangeID after a hard reboot,
the old confirmed clientid still exists, but some clients use a new
co_verifier. For this case, the server was not freeing up the sessions on
the old confirmed clientid.
This patch fixes this case. It also adds two LIST_INIT() macros, which are
actually no-ops, since the structure is malloc()d with M_ZERO so the pointer
is already set to NULL.
It should have minimal impact, since the only way I could exercise this
code path was by doing a hard power cycle (pulling the plus) on a machine
running Linux with a NFSv4.1 mount on the server.
Originally spotted during testing of the ESXi 6.5 client.
Tested by: andreas.nagy@frequentis.com
MFC after: 2 months
When an NFSv4.1 session is busy due to a callback being in progress,
nfsrv_freesession() should return NFSERR_BACKCHANBUSY instead of NFS_OK.
The only effect this has is that the DestroySession operation will report
the failure for this case and this probably has little or no effect on a
client. Spotted by inspection and no failures related to this have been
reported.
MFC after: 2 months
The Linux client now uses the TestStateID operation, so this patch adds
support for it to the NFSv4.1 server. The FreeBSD client never uses this
operation, so it should not be affected.
MFC after: 2 months
This patch adds two missing LIST_INIT()s. Found by inspection.
In practice, these are currently no-ops, since the structure they are
in is malloc'd with M_ZERO and all LIST_INIT does is set the pointer
in the list head to NULL. (In other words, the M_ZERO has already
correctly initialized it.)
MFC after: 2 months
Mechanically replace uses of MALLOC/FREE with appropriate invocations of
malloc(9) / free(9) (a series of sed expressions). Something like:
* MALLOC(a, b, ... -> a = malloc(...
* FREE( -> free(
* free((caddr_t) -> free(
No functional change.
For now, punt on modifying contrib ipfilter code, leaving a definition of
the macro in its KMALLOC().
Reported by: jhb
Reviewed by: cy, imp, markj, rmacklem
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D14035
This reduces noise when kernel is compiled by newer GCC versions,
such as one used by external toolchain ports.
Reviewed by: kib, andrew(sys/arm and sys/arm64), emaste(partial), erj(partial)
Reviewed by: jhb (sys/dev/pci/* sys/kern/vfs_aio.c and sys/kern/kern_synch.c)
Differential Revision: https://reviews.freebsd.org/D10385
there are no write delegations issued.
manu@ reported on the freebsd-current@ mailing list that there was
a significant performance hit in nfsrv_checkgetattr() caused by
the acquisition/release of a state lock, even when there were no
write delegations issued.
This patch add a count of outstanding issued write delegations to the
NFSv4 server. This count allows nfsrv_checkgetattr() to return without
acquiring any lock when the count is 0, avoiding the performance hit
for the case where no write delegations are issued.
Reported by: manu
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D13327
Mainly focus on files that use BSD 2-Clause license, however the tool I
was using misidentified many licenses so this was mostly a manual - error
prone - task.
The Software Package Data Exchange (SPDX) group provides a specification
to make it easier for automated tools to detect and summarize well known
opensource licenses. We are gradually adopting the specification, noting
that the tags are considered only advisory and do not, in any way,
superceed or replace the license texts.
The client IP address was not being reported for some NFSv4 mounts by
nfsdumpstate. Upon investigation, two problems were found for mounts
using IPv4. One was that the code (originally written and tested on i386)
assumed that a "u_long" was a "uint32_t" and would exactly store an
IPv4 host address. Not correct for 64bit arches.
Also, for NFSv4.1 mounts, the field was not being filled in. This was
basically correct, because NFSv4.1 does not use a callback address.
However, it meant that nfsdumpstate could not report the client IP addr.
This patch should fix both of these issues.
For IPv6, the address will still not be reported. The original NFSv4 RFC
only specified IPv4 callback addresses. I think this has changed and, if so,
a future commit to fix reporting of IPv6 addresses will be needed.
Reported by: manu
PR: 223036
MFC after: 2 weeks
The NFSv4 RFCs give a server the option of allowing the use of an open
stateid for write access to be used for a Read operation.
This patch enables this by default and adds a sysctl to disable it,
for anyone who does not want this capability.
Allowing this is particularily useful for a pNFS Data Server (DS), since
they are not permitted to allow the use of special stateids.
Discovered during recent testing of the pNFS server under development.
MFC after: 2 weeks
The NFSv4.1 server failed to update the nfs-stablerestart file for
a client when the client was issued its first Open. As such, recovery
of Opens after a server reboot failed with NFSERR_NOGRACE.
This patch fixes this.
It also changes the code so that it malloc()'s the 1024 byte array
instead of allocating it on the kernel stack for both NFSv4.0 and NFSv4.1.
Note that this bug only affected NFSv4.1 and only when clients attempted
to reclaim Opens after a server reboot.
MFC after: 2 weeks
delegations enabled and the Linux NFSv4.1 client was reported in
reviews.freebsd.org/D7891.
I believe that the FreeBSD server behaviour conforms to the RFC and that
the Linux client has a bug. Therefore, I do not think the proposed patch
is appropriate. When nfsrv_writedelegifpos is non-zero, the FreeBSD
server will issue a write delegation for a read open if possible.
The Linux client then erroneously assumes that the credentials used for
the read open can write the file.
This patch reverses the default value for nfsrv_writedelegifpos to 0 so
that the default behaviour is Linux compatible and adds a sysctl that can
be used to set nfsrv_writedelegifpos.
This change should only affect users that are mounting a FreeBSD server
with delegations enabled (they are not enabled by default) with a Linux
NFSv4.1 client mount.
Reported by: fatih.acar@gandi.net
Tested by: fatih.acar@gandi.net
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D7891
the patch in D1626 plus changes so that it includes counts for
NFSv4.1 (and the draft of NFSv4.2).
Also, make all the counts uint64_t and add a vers field at the
beginning, so that future revisions can easily be implemented.
There is code in place to handle the old vesion of the nfsstats
structure for backwards binary compatibility.
Subsequent commits will update nfsstat(8) to use the new fields.
Submitted by: will (earlier version)
Reviewed by: ken
MFC after: 1 month
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D1626
The ordering of acquisition of the state and session mutexes was
reversed in two cases executed when an NFSv4.1 client created/freed
a session. Since clients will typically do this only when mounting
and dismounting, the likelyhood of causing a deadlock was low but possible.
This can only occur for NFSv4.1 mounts, since the others do not
use sessions.
This was detected while testing the pNFS server/client where the
client crashed during dismounting.
The patch also reorders the unlocks, although that isn't necessary
for correct operation.
MFC after: 2 weeks
(opens, locks, etc) is retained, which I believe is correct behaviour.
However, for NFSv4.1, the server also retained a reference to the xprt
(RPC transport socket structure) for the backchannel. This caused
svcpool_destroy() to not call SVC_DESTROY() for the xprt and allowed
a socket upcall to occur after the mutexes in the svcpool were destroyed,
causing a crash.
This patch fixes the code so that the backchannel xprt structure is
dereferenced just before svcpool_destroy() is called, so the code
does do an SVC_DESTROY() on the xprt, which shuts down the socket upcall.
Tested by: g_amanakis@yahoo.com
PR: 204340
MFC after: 2 weeks
that already has a confirmed ClientID, the nfsrv_setclient() function would
not fill in the clientidp being returned. As such, the value of ClientID
returned would be whatever garbage was on the stack.
An NFSv4.1 client would not normally do this, but it appears that it can
happen for certain Linux clients. When it happens, the client persistently
retries the ExchangeID and Create_session after Create_session fails when
it uses the bogus clientid. With this patch, the correct clientid is replied.
This problem was identified in a packet trace supplied by
Ahmed Kamal via email.
Reported by: email.ahmedkamal@googlemail.com
MFC after: 2 weeks
unconfirmed clientid structure for the same client on the last hash list,
this old entry would not be removed/deleted. I do not think this bug would have
caused serious problems, since the new entry would have been before the old one
on the list. This old entry would have eventually been scavenged/removed.
Detected while reading the code looking for another bug.
MFC after: 3 days
No appreciable change in performance was observed after increasing
the sizes of these tables and then testing with a single client.
However, there was an email that indicated high CPU overheads for
a heavily loaded NFSv4 and it is hoped that increasing the sizes
of the hash tables via these tunables might help.
The tables remain the same size by default.
Differential Revision: https://reviews.freebsd.org/D2596
MFC after: 2 weeks
was reported via email. This was caused by a LOR between the
sleep lock used to serialize the local locking (nfsrv_locklf())
and locking the vnode. I believe this patch fixes the problem
by delaying relocking of the vnode until the sleep lock is
unlocked (nfsrv_unlocklf()). To avoid nfsvno_advlock() having the side
effect of unlocking the vnode, unlocking the vnode was moved to before
the functions that call nfsvno_advlock().
It shouldn't affect the execution of the default case where
vfs.nfsd.enable_locallocks=0.
Reported by: loic.blot@unix-experience.fr
Discussed with: kib
MFC after: 1 week
This fix addresses only issues with the pynfs reports, none of these
issues are know to create problems for extant real clients.
Submitted by: Bart Hsiao <bart.hsiao@gmail.com>
Reworked by: myself
Reviewed by: rmacklem
Approved by: rmacklem
Sponsored by: QNAP Systems Inc.
into head. The code is not believed to have any effect
on the semantics of non-NFSv4.1 server behaviour.
It is a rather large merge, but I am hoping that there will
not be any regressions for the NFS server.
MFC after: 1 month
This simplifies the code and should avoid the clang sparc
port from generating an abort() call.
Requested by: rdivacky
Submitted by: jhb
MFC after: 2 weeks
when a Getattr for a file is done by a client other than the one that
holds the file's delegation. This would only happen when delegations
are enabled and the problem is fixed by this patch.
MFC after: 1 week
- Use NFSD_MONOSEC (which maps to time_uptime) instead of the seconds
portion of wall-time stamps to manage timeouts on events.
- Remove unused nd_starttime from the per-request structure in the new
NFS server.
- Use nanotime() for the modification time on a delegation to get as
precise a time as possible.
- Use time_second instead of extracting the second from a call to
getmicrotime().
Submitted by: bde (3)
Reviewed by: bde, rmacklem
MFC after: 2 weeks
to head. I don't think the NFS client behaviour will change unless
the new "minorversion=1" mount option is used. It includes basic
NFSv4.1 support plus support for pNFS using the Files Layout only.
All problems detecting during an NFSv4.1 Bakeathon testing event
in June 2012 have been resolved in this code and it has been tested
against the NFSv4.1 server available to me.
Although not reviewed, I believe that kib@ has looked at it.
used, when the code should actually protect the tested
variable with a mutex. Since the tsleep()s had a 10sec
timeout, the race would have only delayed the allocation
of a new clientid for a client. The sleeps will also
rarely occur, since having a callback in progress when
a client acquires a new clientid, is unlikely.
in practice, since having a callback in progress when
a fresh clientid is being acquired by a client is unlikely.
MFC after: 1 month
