jamie [Sat, 20 Oct 2018 16:20:36 +0000 (16:20 +0000)]
MFC r339409, r339420:
Add a new jail permission, allow.read_msgbuf. When true, jailed processes
can see the dmesg buffer (this is the current behavior). When false (the
new default), dmesg will be unavailable to jailed users, whether root or
not.
The security.bsd.unprivileged_read_msgbuf sysctl still works as before,
controlling system-wide whether non-root users can see the buffer.
hselasky [Fri, 19 Oct 2018 08:38:34 +0000 (08:38 +0000)]
MFC r339388:
Fix for reception of large full speed isochronous frames via the transaction
translator, when using the DWC OTG USB controller driver. Make sure to re-try
getting the complete split packets until a DATA0 packet is received. Larger
isochronous frames may be split into multiple MDATA packets terminated
by a single DATA0 packet.
mav [Fri, 19 Oct 2018 04:37:27 +0000 (04:37 +0000)]
MFC r339372: Skip VDEV_IO_DONE stage only for ZIO_TYPE_FREE.
Device removal code uses zio_vdev_child_io() with ZIO_TYPE_NULL parent,
that never happened before. It confused FreeBSD-specific TRIM code,
which does not use VDEV_IO_DONE for logical ZIO_TYPE_FREE ZIOs. As
result of that stage being skipped device removal ZIOs leaked references
and memory that supposed to be freed by VDEV_IO_DONE, making it stuck.
It is a quick patch rather then a nice fix, but hopefully we'll be able
to drop it all together when alternative TRIM implementation finally get
landed.
mav [Fri, 19 Oct 2018 04:30:25 +0000 (04:30 +0000)]
MFC r339329: Add ZIO_TYPE_FREE support for indirect vdevs.
Upstream code expects only ZIO_TYPE_READ and some ZIO_TYPE_WRITE
requests to removed (indirect) vdevs, while on FreeBSD there is also
ZIO_TYPE_FREE (TRIM). ZIO_TYPE_FREE requests do not have the data
buffers, so don't need the pointer adjustment.
mav [Fri, 19 Oct 2018 04:28:30 +0000 (04:28 +0000)]
MFC r339335: Avoid zero-sized kmem_alloc() in vdev_compact_children().
The device evacuation code adds a dependency that
vdev_compact_children() be able to properly empty the vdev_child
array by setting it to NULL and zeroing vdev_children. Under Linux,
kmem_alloc() and related functions return a sentinel pointer rather
than NULL for zero-sized allocations.
This is a part of ZoL port of device removal patch:
This fixes a regression in lld that made it stop emitting STT_FILE
symbols, which ctfmerge relies upon to uniquify function table entries
that reference STB_LOCAL symbols. Consequently, ctfmerge stopped
emitting entries for static functions into the function table, and
dtrace no longer gets type info for them.
gjb [Thu, 18 Oct 2018 17:33:30 +0000 (17:33 +0000)]
Document the krpc module requirement in 11.x that was not present
in 10.x if the system has a custom kernel configuration that excludes
NFS and, for example, uses MODULES_OVERRIDE="zfs opensolaris".
Reported by: eugen (via -stable)
Sponsored by: The FreeBSD Foundation
kp [Thu, 18 Oct 2018 04:36:25 +0000 (04:36 +0000)]
MFC r334375, r334379:
pf: Replace rwlock on PF_RULES_LOCK with rmlock
Given that PF_RULES_LOCK is a mostly read lock, replace the rwlock with rmlock.
This change improves packet processing rate in high pps environments.
Benchmarking by olivier@ shows a 65% improvement in pps.
While here, also eliminate all appearances of "sys/rwlock.h" includes since it
is not used anymore.
jamie [Wed, 17 Oct 2018 16:17:57 +0000 (16:17 +0000)]
MFC r339211:
Fix the test prohibiting jails from sharing IP addresses.
It's not supposed to be legal for two jails to contain the same IP address,
unless both jails contain only that one address. This is the behavior
documented in jail(8), and is there to prevent confusion when multiple
jails are listening on IADDR_ANY.
VIMAGE jails (now the default for GENERIC kernels) test this correctly,
but non-VIMAGE jails have been performing an incomplete test when nested
jails are used.
imp [Wed, 17 Oct 2018 04:10:23 +0000 (04:10 +0000)]
Direct commit since these files have gone away in head
Move pc98's biosdisk.c to the new style disk access. This is missing
support in common/part.c, however, for pc98 partitions, so it's
unlikely to actually work. Lack of a pc98 machine that's in sevice
limits my ability to test, but this allows pc98 to compile again.
np [Wed, 17 Oct 2018 02:05:31 +0000 (02:05 +0000)]
MFC r335352:
cxgbe(4): Some mailbox commands require access to the Tx pipeline and
can time out if it's backed up due to a non-stop deluge of PAUSE frames
from a misbehaving peer. Detect this situation and toggle MPS TxEn
to allow forward progress.
np [Wed, 17 Oct 2018 01:59:45 +0000 (01:59 +0000)]
MFC r334987:
cxgbe(4): Remove homemade version of htobe32 from the driver.
It was needed only for ia64 where it was implemented as a call to
bswapXX, which was always a real function. htobeXX with a constant
argument is calculated at compile-time everywhere else.
np [Wed, 17 Oct 2018 01:20:18 +0000 (01:20 +0000)]
MFC r338924:
cxgbe(4): Link related changes.
- Switch to using 32b port/link capabilities in the driver. The 32b
format is used internally by firmwares > 1.16.45.0 and the driver will
now interact with the firmware in its native format, whether it's 16b
or 32b. Note that the 16b format doesn't have room for 50G, 200G, or
400G speeds.
- Add a bit in the pause_settings knobs to allow negotiated PAUSE
settings to override manual settings.
- Ensure that manual link settings persist across an administrative
down/up as well as transceiver unplug/replug.
np [Wed, 17 Oct 2018 00:57:28 +0000 (00:57 +0000)]
MFC r333139:
cxgbe(4): Destroy the cdev before disabling interrupts in driver detach.
Filter work requests are submitted in the nexus cdev's ioctl which then
blocks waiting for a reply. If driver detach runs in this state and
disables interrupts the ioctl will never complete and detach will hang
in destroy_cdev.
np [Wed, 17 Oct 2018 00:45:01 +0000 (00:45 +0000)]
MFC r325840, r327811, and r329701.
r325840:
CXGBE: fix big-endian behaviour
The setbit/clearbit pair casts the bitfield pointer
to uint8_t* which effectively treats its contents as
little-endian variable. The ffs() function accepts int as
the parameter, which is big-endian. Use uint8_t here to
avoid mismatch, as we have only 4 doorbells.
ken [Fri, 12 Oct 2018 19:44:19 +0000 (19:44 +0000)]
MFC r339076
This has been edited slightly from the version in head. In head, the probe
sections of dadone() were split out into separate functions. In stable/11,
dadone() is still a single function.
So, for stable/11, this describes the change:
sys/cam/scsi/scsi_da.c:
In the DA_CCB_PROBE_DONE case in dadone(), free the data pointer
before returning.
------------------------------------------------------------------------
r339076 | ken | 2018-10-01 13:00:46 -0600 (Mon, 01 Oct 2018) | 12 lines
Fix a da(4) driver memory leak for SCSI SMR devices.
In the probe case for SCSI SMR Host Aware or Most Managed drives, be sure
to free allocated memory.
sys/cam/scsi/scsi_da.c:
In dadone_probezone(), free the data pointer before returning.
jhb [Thu, 11 Oct 2018 19:06:54 +0000 (19:06 +0000)]
MFC 338055: Remove some vestiges of IPI_LAZYPMAP on i386.
The support for lazy pmap invalidations on i386 was removed in r281707.
This removes the constant for the IPI and stops accounting for it when
sizing the interrupt count arrays.
hselasky [Thu, 11 Oct 2018 07:34:56 +0000 (07:34 +0000)]
MFC r339235:
Add missing steering rules for virtual function, VF, in mlx4en(4) driver.
When acting as a VF it is required to add steering rules for all unicast
addresses. Even if promiscious mode is selected. Else incoming data packets
will be dropped.
jhb [Wed, 10 Oct 2018 21:28:04 +0000 (21:28 +0000)]
Disable the KASSERT for curcpu == 0 in netisr for EARLY_AP_STARTUP.
In the EARLY_AP_STARTUP case, thread0 can migrate to another CPU
before this SYSINIT is run. However, the only part of this SYSINIT
that assumes it runs on CPU 0 is in the !EARLY_AP_STARTUP case when it
creates the netisr for the boot CPU. In the EARLY_AP_STARTUP case we
start up the netisr's for the first N CPUs during the SYSINIT itself
and don't depend on running on the boot CPU for correct operation.
This is a direct comit to stable/11 as the assertion was removed as part
of a different change in r302595.
According to the Intel SDM (Volme 3, 9.11.7) the BIOS signature MSR
should be zeroed before executing cpuid (although in practice it does
not seem to matter).
PR: 192487
Submitted by: Dan Lukes
Reported by: Henrique de Moraes Holschuh
emaste [Wed, 10 Oct 2018 15:37:10 +0000 (15:37 +0000)]
MFC r339019: clang: allow ifunc resolvers to accept arguments
Previously Clang required ifunc resolution functions to take no
arguments, presumably because GCC documented ifunc resolvers as taking
no arguments. However, GCC accepts resolvers accepting arguments, and
our rtld passes CPU ID information (cpuid, hwcap, etc.) to ifunc
resolvers. Just remove the check from the in-tree compiler for our in-
tree compiler.
emaste [Wed, 10 Oct 2018 13:12:52 +0000 (13:12 +0000)]
MFC r334072, r334247 (eadler): Add the text '@generated' to src.conf.5
This is a cross-tool approach to identifying generated code. Some tools,
notably phabricator, handle this marker specially. See
https://reviews.freebsd.org/differential/diff/42870/ for such an
example.
jhb [Tue, 9 Oct 2018 22:35:43 +0000 (22:35 +0000)]
MFC 338976: Don't clear DR6 for debug exceptions from userland.
This reverts part of r333368. The attempt to clear DR6 was occuring
too soon as trapsignal() does not pause to let the debugger notice the
SIGTRAP and query DR6. The signal exchange does not occur until much
later during ast(). As a result, GDB was no longer recognizing
hardware breakpoints and watchpoints on x86.
In addition, any userland programs that want to inspect DR6 in a
SIGTRAP handler don't have a way to do this if we clear DR6 in the
exception handler.
Instead of relying on the kernel to clear DR6, debuggers will have to
explicitly clear it after a trace trap (which they needed to do on
older kernels anyway).
jhb [Mon, 8 Oct 2018 17:18:55 +0000 (17:18 +0000)]
MFC 337400: Remove spurious ABI tags from kdump output.
The abidump routine output an ABI tag when -A was specified for records
that were not displayed due to type or pid filtering. To fix, split
the code to lookup the ABI from the code to display the ABI, move the
code to display the ABI into dumpheader(), and move dumpheader() later
in the main loop as a simplification. Previously dumpheader() was
called under a condition that repeated conditions made later in the
main loop.
jhb [Fri, 5 Oct 2018 21:10:03 +0000 (21:10 +0000)]
MFC 338022: Fix casts between 64-bit physical addresses and pointers in EFI.
Compiling FreeBSD/i386 with modern GCC triggers warnings for various
places that convert 64-bit EFI_ADDRs to pointers and vice versa.
- Cast pointers to uintptr_t rather than to uint64_t when assigning
to a 64-bit integer.
- Cast 64-bit integers to uintptr_t before a cast to a pointer.
hselasky [Fri, 5 Oct 2018 07:49:01 +0000 (07:49 +0000)]
MFC r338993:
When multiple threads are involved receiving completion events in LibUSB
make sure there is always a master polling thread, by setting the "ctx_handler"
field in the context. Else the reception of completion events can stop.
This happens if event threads are created and destroyed during runtime.
kevans [Wed, 3 Oct 2018 17:21:45 +0000 (17:21 +0000)]
MFC r338040: diff(1): Refactor -B a little bit
Instead of doing a second pass to skip empty lines if we've specified -I, go
ahead and check both at once. Ignore critera has been split out into its own
function to try and keep the logic cleaner.
kevans [Wed, 3 Oct 2018 17:20:30 +0000 (17:20 +0000)]
MFC r338646: dd(1): Correct padding in status=progress
Output padding is specified via outlen, which is set using the return value
of fprintf. Because it's printing that padding plus a trailing byte, it
grows by one each iteration rather than reflecting actual length.
Additionally, iec was sized improperly for scaling up similarly to si.
Fixing this revealed that the humanize_number(3) call to populate persec
was using the wrong width.
kevans [Wed, 3 Oct 2018 17:17:38 +0000 (17:17 +0000)]
MFC r338219, r338250: FDT in Loader fixes
r338219:
fdt_fixups: relocate the /chosen node after applying fixups
As indicated by the comment, any fixups applied (which might include
overlays) can invalidate the previously located node by adding nodes or
setting/adding properties. The later fdt_setprop of fixup-applied property
would then fail because of the bad/wrong node offset.
This would have generally been harmless, but potentially caused multiple
applications of fixups and caused a little bit of bloat.
r338250:
efiloader: Setup FDT in autoload to fix overlays clobbering kenv
manu found in the noted PR that overlays seemed to be clobbering the kenv
and killing the boot. Further inspection revealed that one can `fdt ls` at
the loader prompt for a successful boot, but autoboot breaks it.
In the autoboot case, first setup of FDT is happening in the middle of
bi_load, which triggers loading of the DTBO from /boot.
This is bad, bad, bad. Files in the loader are loaded somewhere in the
middle of the address space one after another. bi_load starts building the
needed kernel bootinfo immediately after the highest-addr loaded file. File
loads in the middle of bi_load suddenly clobber bootinfo and everything goes
off the rails.
The solution to this is to use take advantage of arch_autoload to setup FDT
in efiloader compiled with LOADER_FDT_SUPPORT. This matches how it works in
ubldr land, and is how it should have worked when overlay support was added
to efiloader since fdt_setup_fdtp now has the potential to load files
(courtesy of fdt_platform_load_dtb).
As noted by cem in r338035, coccinelle invokes diff(1) with the -B flag.
This was not previously implemented here, so one was forced to create a link
for GNU diff to /usr/local/bin/diff
Implement the -B flag and add some primitive tests for it. It is implemented
in the same fashion that -I is implemented; each chunk's lines are scanned,
and if a non-blank line is encountered then the chunk will be output.
Otherwise, it's skipped.
Notable fixes:
- Overlays may now be generated properly without -@
- /__local_fixups__ were not including unit address in their structure
- The error reporting a magic token was misleading, reporting
"Bad magic token in header. Got d00dfeed expected 0xd00dfeed"
if the token was missing. This has been split out into a separate message.
r338232:
dtc(1): Update to 0892ec7; HACKING and implicit header fixes
Fixes courtesy of arichardson and jmg:
- HACKING was pointing to the wrong place
- Added headers were being relied on implicitly, but libstdc++ did not
comply with the unspoken wishes of dtc.
perf: 2.75x faster ddt_entry_compare()
First 256bits of ddt_key_t is a block checksum, which are expected
to be close to random data. Hence, on average, comparison only needs to
look at first few bytes of the keys. To reduce number of conditional
jump instructions, the result is computed as: sign(memcmp(k1, k2)).
Sign of an integer 'a' can be obtained as: `(0 < a) - (a < 0)` := {-1, 0, 1} ,
which is computed efficiently. Synthetic performance evaluation of
original and new algorithm over 1G random keys on 2.6GHz Intel(R) Xeon(R)
CPU E5-2660 v3:
old 6.85789 s
new 2.49089 s
perf: 2.8x faster vdev_queue_offset_compare() and vdev_queue_timestamp_compare()
Compute the result directly instead of using conditionals
perf: zfs_range_compare()
Speedup between 1.1x - 2.5x, depending on compiler version and
optimization level.
perf: spa_error_entry_compare()
`bcmp()` is not suitable for comparator use. Use `memcmp()` instead.
Reviewed by: Toomas Soome <tsoome@me.com>
Reviewed by: Sanjay Nadkarni <sanjay.nadkarni@nexenta.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Matthew Ahrens <mahrens@delphix.com>
Author: Jerry Jelinek <jerry.jelinek@joyent.com>
Relax allocation throttling for ditto blocks. Due to random imbalances
in allocation it tends to push block copies to one vdev, that looks
slightly better at the moment. Slightly less strict policy allows both
improve data security and surprisingly write performance, since we don't
need to touch extra metaslabs on each vdev to respect the min distance.
mav [Wed, 3 Oct 2018 15:34:49 +0000 (15:34 +0000)]
MFC r337970: 9738 Fix third block copy allocations, broken at 9112.
Use METASLAB_WEIGHT_CLAIM weight to allocate tertiary blocks.
Previous use of METASLAB_WEIGHT_SECONDARY for that caused errors
later on metaslab_activate_allocator() call, leading to massive
load of unneeded metaslabs and write freezes.
We were doing count_block() twice inside this function, once
unconditionally at the beginning (intended to catch the embedded block
case) and once near the end after processing the block.
The double-accounting caused the "zpool scrub" progress statistics in
"zpool status" to climb from 0% to 200% instead of 0% to 100%, and
showed double the I/O rate it was actually seeing.
This was apparently a regression introduced in commit 00c405b4b5e8,
which was an incorrect port of this OpenZFS commit:
mav [Wed, 3 Oct 2018 14:59:39 +0000 (14:59 +0000)]
MFC r337229: Reduce taskq and context-switch cost of zio pipe
When doing a read from disk, ZFS creates 3 ZIO's: a zio_null(), the
logical zio_read(), and then a physical zio. Currently, each of these
results in a separate taskq_dispatch(zio_execute).
On high-read-iops workloads, this causes a significant performance
impact. By processing all 3 ZIO's in a single taskq entry, we reduce the
overhead on taskq locking and context switching. We accomplish this by
allowing zio_done() to return a "next zio to execute" to zio_execute().
This results in a ~12% performance increase for random reads, from
96,000 iops to 108,000 iops (with recordsize=8k, on SSD's).
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: George Wilson <george.wilson@delphix.com> Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
External-issue: DLPX-59292
Closes #7736
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Prashanth Sreenivasa <pks@delphix.com>
Reviewed by: Robert Mustacchi <rm@joyent.com>
Approved by: Dan McDonald <danmcd@joyent.com>
Author: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Sebastien Roy <sebastien.roy@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Robert Mustacchi <rm@joyent.com>
Approved by: Dan McDonald <danmcd@omniti.com>
Author: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Robert Mustacchi <rm@joyent.com>
Approved by: Dan McDonald <danmcd@omniti.com>
Author: Matthew Ahrens <mahrens@delphix.com>
The zfs_dbuf_evict_key TSD (thread-specific data) is not necessary - we can
instead pass a flag down in a few places to prevent recursive dbuf eviction.
Making this change has 3 benefits:
1. The code semantics are easier to understand.
2. On Linux, performance is improved, because creating/removing TSD values
(by setting to NULL vs non-NULL) is expensive, and we do it very often.
3. According to Nexenta, the current semantics can cause a deadlock when
concurrently calling dmu_objset_evict_dbufs() (which is rare today, but they
are working on a "parallel unmount" change that triggers this more easily)
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Serapheim Dimitropoulos <serapheim.dimitro@delphix.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: Andy Stormont <astormont@racktopsystems.com>
Approved by: Richard Lowe <richlowe@richlowe.net>
Author: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com>
Author: Matthew Ahrens <mahrens@delphix.com>
mav [Wed, 3 Oct 2018 14:51:49 +0000 (14:51 +0000)]
MFC r337202: MFV r337200:
9438 Holes can lose birth time info if a block has a mix of birth times
Ultimately, the problem here is that when you truncate and write a file in
the same transaction group, the dbuf for the indirect block will be zeroed
out to deal with the truncation, and then written for the write. During
this process, we will lose hole birth time information for any holes in the
range. In the case where a dnode is being freed, we need to determine
whether the block should be converted to a higher-level hole in the zio
pipeline, and if so do it when the dnode is being synced out.
Reviewed by: Matt Ahrens <matt@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com>
Author: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com>
Author: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Sebastien Roy <sebastien.roy@delphix.com>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Don Brady <don.brady@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com>
Author: Matthew Ahrens <mahrens@delphix.com>
mav [Wed, 3 Oct 2018 14:48:55 +0000 (14:48 +0000)]
MFC r337191:
MFV r337190: 9486 reduce memory used by device removal on fragmented pools
In the most fragmented real-world cases, this reduces memory used by the
mapping from ~1GB to ~50MB of RAM per 1TB of storage removed. Less
fragmented cases will typically also see around 50-100MB of RAM per 1TB
of storage.
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Serapheim Dimitropoulos <serapheim.dimitro@delphix.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: Tim Chase <tim@chase2k.com>
Approved by: Robert Mustacchi <rm@joyent.com>
Author: Matthew Ahrens <mahrens@delphix.com>
mav [Wed, 3 Oct 2018 14:48:17 +0000 (14:48 +0000)]
MFC r337185:
MFV r337184: 9457 libzfs_import.c:add_config() has a memory leak
A memory leak occurs on lines 209 and 213 because the config is not freed
in the error case. The interface to add_config() seems less than ideal -
it would be better if it copied any data necessary from the config and the
caller freed it.
mav [Wed, 3 Oct 2018 14:47:29 +0000 (14:47 +0000)]
MFC r337183:
MFV r337182: 9330 stack overflow when creating a deeply nested dataset
Datasets that are deeply nested (~100 levels) are impractical. We just put
a limit of 50 levels to newly created datasets. Existing datasets should
work without a problem.
mav [Wed, 3 Oct 2018 14:46:25 +0000 (14:46 +0000)]
MFC r337181: 9539 Make zvol operations use _by_dnode routines
Continues what was started in 7801 add more by-dnode routines by fully
converting zvols to avoid unnecessary dnode_hold() calls. This saves a
small amount of CPU time and slightly improves latencies of operations
on zvols.
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: Rick McNeal <rick.mcneal@nexenta.com>
Approved by: Dan McDonald <danmcd@joyent.com>
Author: Richard Yao <richard.yao@prophetstor.com>
mav [Wed, 3 Oct 2018 14:44:16 +0000 (14:44 +0000)]
MFC r337177:
MFV r337175: 9487 Free objects when receiving full stream as clone
All objects after the last written or freed object are not supposed to
exist after receiving the stream. We should free them accordingly, as if
a freeobjects record for them had been included in the stream.
mav [Wed, 3 Oct 2018 14:43:17 +0000 (14:43 +0000)]
MFC r337172, MFV r337171:
9464 txg_kick() fails to see that we are quiescing, forcing transactions
to their next stages without leaving them accumulate changes
Ideally we would like txg_kick() to get triggered only when we are sure
that we are not syncing AND not quiescing any txg. This way we can kick
an open TXG to the quiescing state when we are sure that there is nothing
going on and we would benefit from the different states running
concurrently.