buffer for the last vnode on the mount back to the server, it
returns. At that point, the code continues with the unmount,
including freeing up the nfs specific part of the mount structure.
It is possible that an nfsiod thread will try to check for an
empty I/O queue in the nfs specific part of the mount structure
after it has been free'd by the unmount. This patch avoids this problem by
setting the iodmount entries for the mount back to NULL while holding the
mutex in the unmount and checking the appropriate entry is non-NULL after
acquiring the mutex in the nfsiod thread.
Reported and tested by: pho
Reviewed by: kib
MFC after: 2 weeks
respective functionality, allowing to synchronize TSC on APs to match BSP's
during boot. It may be unsafe in general case due to theoretical chance of
later drift if CPUs are using different clock rate or source, but it allows
to use TSC in some cases when difference caused by some initialization bug,
while TSCs are known to increment synchronously.
Reviewed by: jimharris, kib
MFC after: 1 month
option. This can occur when an nfsiod thread that already holds
a buffer lock attempts to acquire a vnode lock on an entry in
the directory (a LOR) when another thread holding the vnode lock
is waiting on an nfsiod thread. This patch avoids the deadlock by disabling
readahead for this case, so the nfsiod threads never do readdirplus.
Since readaheads for directories need the directory offset cookie
from the previous read, they cannot normally happen in parallel.
As such, testing by jhb@ and myself didn't find any performance
degredation when this patch is applied. If there is a case where
this results in a significant performance degradation, mounting
without the "rdirplus" option can be done to re-enable readahead
for directories.
Reported and tested by: jhb
Reviewed by: jhb
MFC after: 2 weeks
it will work with either the old or new server.
The FHA code keeps a cache of currently active file handles for
NFSv2 and v3 requests, so that read and write requests for the same
file are directed to the same group of threads (reads) or thread
(writes). It does not currently work for NFSv4 requests. They are
more complex, and will take more work to support.
This improves read-ahead performance, especially with ZFS, if the
FHA tuning parameters are configured appropriately. Without the
FHA code, concurrent reads that are part of a sequential read from
a file will be directed to separate NFS threads. This has the
effect of confusing the ZFS zfetch (prefetch) code and makes
sequential reads significantly slower with clients like Linux that
do a lot of prefetching.
The FHA code has also been updated to direct write requests to nearby
file offsets to the same thread in the same way it batches reads,
and the FHA code will now also send writes to multiple threads when
needed.
This improves sequential write performance in ZFS, because writes
to a file are now more ordered. Since NFS writes (generally
less than 64K) are smaller than the typical ZFS record size
(usually 128K), out of order NFS writes to the same block can
trigger a read in ZFS. Sending them down the same thread increases
the odds of their being in order.
In order for multiple write threads per file in the FHA code to be
useful, writes in the NFS server have been changed to use a LK_SHARED
vnode lock, and upgrade that to LK_EXCLUSIVE if the filesystem
doesn't allow multiple writers to a file at once. ZFS is currently
the only filesystem that allows multiple writers to a file, because
it has internal file range locking. This change does not affect the
NFSv4 code.
This improves random write performance to a single file in ZFS, since
we can now have multiple writers inside ZFS at one time.
I have changed the default tuning parameters to a 22 bit (4MB)
window size (from 256K) and unlimited commands per thread as a
result of my benchmarking with ZFS.
The FHA code has been updated to allow configuring the tuning
parameters from loader tunable variables in addition to sysctl
variables. The read offset window calculation has been slightly
modified as well. Instead of having separate bins, each file
handle has a rolling window of bin_shift size. This minimizes
glitches in throughput when shifting from one bin to another.
sys/conf/files:
Add nfs_fha_new.c and nfs_fha_old.c. Compile nfs_fha.c
when either the old or the new NFS server is built.
sys/fs/nfs/nfsport.h,
sys/fs/nfs/nfs_commonport.c:
Bring in changes from Rick Macklem to newnfs_realign that
allow it to operate in blocking (M_WAITOK) or non-blocking
(M_NOWAIT) mode.
sys/fs/nfs/nfs_commonsubs.c,
sys/fs/nfs/nfs_var.h:
Bring in a change from Rick Macklem to allow telling
nfsm_dissect() whether or not to wait for mallocs.
sys/fs/nfs/nfsm_subs.h:
Bring in changes from Rick Macklem to create a new
nfsm_dissect_nonblock() inline function and
NFSM_DISSECT_NONBLOCK() macro.
sys/fs/nfs/nfs_commonkrpc.c,
sys/fs/nfsclient/nfs_clkrpc.c:
Add the malloc wait flag to a newnfs_realign() call.
sys/fs/nfsserver/nfs_nfsdkrpc.c:
Setup the new NFS server's RPC thread pool so that it will
call the FHA code.
Add the malloc flag argument to newnfs_realign().
Unstaticize newnfs_nfsv3_procid[] so that we can use it in
the FHA code.
sys/fs/nfsserver/nfs_nfsdsocket.c:
In nfsrvd_dorpc(), add NFSPROC_WRITE to the list of RPC types
that use the LK_SHARED lock type.
sys/fs/nfsserver/nfs_nfsdport.c:
In nfsd_fhtovp(), if we're starting a write, check to see
whether the underlying filesystem supports shared writes.
If not, upgrade the lock type from LK_SHARED to LK_EXCLUSIVE.
sys/nfsserver/nfs_fha.c:
Remove all code that is specific to the NFS server
implementation. Anything that is server-specific is now
accessed through a callback supplied by that server's FHA
shim in the new softc.
There are now separate sysctls and tunables for the FHA
implementations for the old and new NFS servers. The new
NFS server has its tunables under vfs.nfsd.fha, the old
NFS server's tunables are under vfs.nfsrv.fha as before.
In fha_extract_info(), use callouts for all server-specific
code. Getting file handles and offsets is now done in the
individual server's shim module.
In fha_hash_entry_choose_thread(), change the way we decide
whether two reads are in proximity to each other.
Previously, the calculation was a simple shift operation to
see whether the offsets were in the same power of 2 bucket.
The issue was that there would be a bucket (and therefore
thread) transition, even if the reads were in close
proximity. When there is a thread transition, reads wind
up going somewhat out of order, and ZFS gets confused.
The new calculation simply tries to see whether the offsets
are within 1 << bin_shift of each other. If they are, the
reads will be sent to the same thread.
The effect of this change is that for sequential reads, if
the client doesn't exceed the max_reqs_per_nfsd parameter
and the bin_shift is set to a reasonable value (22, or
4MB works well in my tests), the reads in any sequential
stream will largely be confined to a single thread.
Change fha_assign() so that it takes a softc argument. It
is now called from the individual server's shim code, which
will pass in the softc.
Change fhe_stats_sysctl() so that it takes a softc
parameter. It is now called from the individual server's
shim code. Add the current offset to the list of things
printed out about each active thread.
Change the num_reads and num_writes counters in the
fha_hash_entry structure to 32-bit values, and rename them
num_rw and num_exclusive, respectively, to reflect their
changed usage.
Add an enable sysctl and tunable that allows the user to
disable the FHA code (when vfs.XXX.fha.enable = 0). This
is useful for before/after performance comparisons.
nfs_fha.h:
Move most structure definitions out of nfs_fha.c and into
the header file, so that the individual server shims can
see them.
Change the default bin_shift to 22 (4MB) instead of 18
(256K). Allow unlimited commands per thread.
sys/nfsserver/nfs_fha_old.c,
sys/nfsserver/nfs_fha_old.h,
sys/fs/nfsserver/nfs_fha_new.c,
sys/fs/nfsserver/nfs_fha_new.h:
Add shims for the old and new NFS servers to interface with
the FHA code, and callbacks for the
The shims contain all of the code and definitions that are
specific to the NFS servers.
They setup the server-specific callbacks and set the server
name for the sysctl and loader tunable variables.
sys/nfsserver/nfs_srvkrpc.c:
Configure the RPC code to call fhaold_assign() instead of
fha_assign().
sys/modules/nfsd/Makefile:
Add nfs_fha.c and nfs_fha_new.c.
sys/modules/nfsserver/Makefile:
Add nfs_fha_old.c.
Reviewed by: rmacklem
Sponsored by: Spectra Logic
MFC after: 2 weeks
prevents us from creating UMA_ZONE_PCPU zones earlier.
As bandaid shift initialization of counter(9) zone later.
Reviewed by: kib
Reported & tested by: Lytochkin Boris <lytboris gmail.com>
(Wasting 4k just as a temporary placeholder for a boot environment seems
a bit ridiculous, but hey.)
Tested: gxemul:
$ gxemul -e malta -d i:/home/adrian/work/freebsd/svn/mfsroot-rspro.img -C 4Kc /tftpboot/kernel.MALTA
* Add ah_ratesArray[] to the ar5416 HAL state - this stores the maximum
values permissable per rate.
* Since different chip EEPROM formats store this value in a different place,
store the HT40 power detector increment value in the ar5416 HAL state.
* Modify the target power setup code to store the maximum values in the
ar5416 HAL state rather than using a local variable.
* Add ar5416RateToRateTable() - to convert a hardware rate code to the
ratesArray enum / index.
* Add ar5416GetTxRatePower() - which goes through the gymnastics required
to correctly calculate the target TX power:
+ Add the power detector increment for ht40;
+ Take the power offset into account for AR9280 and later;
+ Offset the TX power correctly when doing open-loop TX power control;
+ Enforce the per-rate maximum value allowable.
Note - setting a TPC value of 0x0 in the TX descriptor on (at least)
the AR9160 resulted in the TX power being very high indeed. This didn't
happen on the AR9220. I'm guessing it's a chip bug that was fixed at
some point. So for now, just assume the AR5416/AR5418 and AR9130 are
also suspect and clamp the minimum value here at 1.
Tested:
* AR5416, AR9160, AR9220 hostap, verified using (2GHz) spectrum analyser
* Looked at target TX power in TX descriptor (using athalq) as well as TX
power on the spectrum analyser.
TODO:
* The TX descriptor code sets the target TX power to 0 for AR9285 chips.
I'm not yet sure why. Disable this for TPC and ensure that the TPC
TX power is set.
* AR9280, AR9285, AR9227, AR9287 testing!
* 5GHz testing!
Quirks:
* The per-packet TPC code is only exercised when the tpc sysctl is set
to 1. (dev.ath.X.tpc=1.) This needs to be done before you bring the
interface up.
* When TPC is enabled, setting the TX power doesn't end up with a call
through to the HAL to update the maximum TX power. So ensure that
you set the TPC sysctl before you bring the interface up and configure
a lower TX power or the hardware will be clamped by the lower TX
power (at least until the next channel change.)
Thanks to Qualcomm Atheros for all the hardware, and Sam Leffler for use
of his spectrum analyser to verify the TX channel power.
ath_tx_rate_fill_rcflags(). Include setting up the TX power cap in the
rate scenario setup code being passed to the HAL.
Other things:
* add a tx power cap field in ath_rc.
* Add a three-stream flag in ath_rc.
* Delete the LDPC flag from ath_rc - it's not a per-rate flag, it's a
global flag for the transmission.
The following change from illumos brought caused DTrace to
pause in an interactive environment:
3026 libdtrace should set LD_NOLAZYLOAD=1 to help the pid provider
This was not detected during testing because it doesn't
affect scripts.
We shouldn't be changing the environment, especially since the
LD_NOLAZYLOAD option doesn't apply to our (GNU) ld.
Unfortunately the change from upstream was made in such a way
that it is very difficult to separate this change from the
others so, at least for now, it's better to just revert
everything.
Reference:
https://www.illumos.org/issues/3026
Reported by: Navdeep Parhar and Mark Johnston
directly referencing ni->ni_txpower.
This provides the hardware with a slightly more accurate idea of
the maximum TX power to be using.
This is part of a series to get per-packet TPC to work (better).
Tested:
* AR5416, hostap mode
the given node.
This takes into account the per-node cap, the ic cap and the
per-channel regulatory caps.
This is designed to replace references to ni_txpower in various net80211
drivers - ni_txpower doesn't necessarily reflect the actual cap for
the given node (eg if the node has the default value of 50dBm (100) and
the administrator has manually configured a lower TX power.)
signal.
- Fix the old ksem implementation for POSIX semaphores to not restart
sem_wait() or sem_timedwait() if interrupted by a signal.
MFC after: 1 week
Pointy-hat to: me, for not realizing snprintf() is available in kernel.
Thanks to: jh, for bringing me the good news of snprintf(), Pawel Worach, for
noting that the panic can be provoked in i386 and not in amd64
The notes format is a header of sizeof(int), which stores the size of
the corresponding data structure to provide some versioning, and data
in the format as it is returned by a related sysctl call.
The userland tools (procstat(1)) will be taught to extract this data,
providing additional info for postmortem analysis.
PR: kern/173723
Suggested by: jhb
Discussed with: jhb, kib
Reviewed by: jhb (initial version), kib
MFC after: 1 month
Only look for FDT partitions if our potential parent is a DISK device.
Excluding direct recursion on the flashmap geoms was insufficient
because it did not prevent the underlying device from being retrieved if
flashmap geoms were further partitioned.
Reviewed by: imp
Sponsored by: DARPA, AFRL
LK_EXCLOTHER. LK_EXCLOTHER is only used to acquire a
usecount on a vnode during NFSv4 recovery from an
expired lease.
Reported and tested by: pho
MFC after: 2 weeks
