Remove incorrect (and inappropriate) comment in dprintf_dnode

This comment seems to misunderstand the ## preprocessor token, which
does token concatenation.  It is not needed here, since we are
concatenating string literals, which is performed by putting the
literals next to each other.

Additionally, the comment uses offensive language.

Reviewed-by: Igor Kozhukhov <igor@dilos.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
Closes #8698
Closes #8699

Use sigaction(2) instead of sigignore(3) for portability

sigignore(3) isn't portable.
This code fails to compile on platforms without sigignore(3).
Use sigaction(2).

--
zfs_main.c: In function 'zfs_do_diff':
zfs_main.c:7178:9: error: implicit declaration of function 'sigignore' [-Werror=implicit-function-declaration]
  (void) sigignore(SIGPIPE);
         ^~~~~~~~~

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8593

Use sigaction(2) instead of sigset(3) for portability

sigset(3) isn't portable.
This code fails to compile on platforms without sigset(3).
Use sigaction(2).

--
largest_file.c: In function 'main':
largest_file.c:75:9: error: implicit declaration of function 'sigset'; did you mean 'sigvec'? [-Werror=implicit-function-declaration]
  (void) sigset(SIGXFSZ, sigxfsz);
         ^~~~~~
         sigvec

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8593

Correct snprintf() size argument

The size argument of snprintf(3) in glibc and snprintf() in Linux
kernel includes trailing \0, as snprintf(3) man page explains it as
"write at most size bytes (including the trailing null byte ('\0'))",
i.e. snprintf() can just take buffer size.

e.g. For snprintf() in module/zfs/zfs_ctldir.c, a buffer size is
MAXPATHLEN, and a caller is passing MAXPATHLEN to snprintf(), so size
should just be `path_len` to do what the caller is trying to do.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8692

Clarify that deduped data is encrypted

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8691

Add CODE_OF_CONDUCT.md

Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8660

Linux 5.0 compat: Remove incorrect ASSERT

Not all block devices, notably scsi_debug, set a root_blkg on the
request queue.  Remove this assertion and allow the the existing
call to blkg_tryget() to gracefully handle the NULL (which it does).

Reviewed-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8678

Use NV_ENCODE_NATIVE for nvlist encoding variable

Use NV_ENCODE_NATIVE for nvlist encoding variable instead of 0.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Chris Dunlop <chris@onthe.net.au>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8653

Prevent `make distclean` removing config/config.rpath

`make distclean` removes an empty file config/config.rpath.
Avoid that by adding some text.

Also see e1245d83e9("Prevent `make distclean` removing 0 sized file").

--
 # find . -size 0
 ./config/config.rpath
 # ./autogen.sh && ./configure
 # git diff
 # make distclean
 # git diff
 diff --git a/config/config.rpath b/config/config.rpath
 deleted file mode 100644
 index e69de29bb..000000000

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8665

Use SEEK_{SET,CUR,END} for file seek "whence"

Use either SEEK_* or 0,1,2..., but not both.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8656

Fixes for the DMU free throttle

This patch fixes 2 issues with the DMU free throttle implemented
in dmu_free_long_range(). The first issue is that get_next_chunk()
was calculating the number of L1 blocks the free would dirty
incorrectly. In some cases involving extremely large files, this
code would greatly overestimate the number of effected L1 blocks,
causing excessive calls to txg_wait_open(). This patch corrects
the calculation.

The second issue is that the free throttle uses the total number
of free'd blocks in all (open, quiescing, and syncing) txgs to
determine whether to throttle. This causes large frees (such as
those created by the first issue) to cause 4 txg syncs before
any further frees were allowed to proceed. This patch ensures
that the accounting is done entirely in a per-txg fashion, so
that frees from a given txg don't affect those that immediately
follow it.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8655

Clarify and improve encryption documentation

- Remove the language that "all user data" is encrypted.  This is to
  avoid misunderstandings or arguments about what is "user data",
  especially in light of "user properties".
- Document that properties are unencrypted.
- Document that snapshot names are unencrypted.
- For consistency with the rest of the zfs.8 man page, use "ZFS" as the
  generic noun, not (bolded) "zfs".  The latter refers to the command.
  Likewise, use "ZFS" instead of "the kernel module".
- Give "a passphrase" as an example of a "user's key".

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8652

Duplicate the encryption copies=3 limitation

This adds the encryption copies=3 limitation language into the copies
property section.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8651

Deprecate dedupditto

This documents, in zpool.8, that dedupditto is deprecated and will be
made to have no effect in a future release.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8650

Eliminate useless double-bolding in man pages

As far as I know and can tell from testing, \fB\fB...\fR\fR is exactly
equivalent to \fB...\fR.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Alphabetize zpool-features.5 by short name

The features are sorted in the en_US locale, not the C locale.
Specifically, that means that bookmark_v2 comes _after_ bookmarks.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Standardize .RE placement in zpool-features.5

This command is being used to unindent, so it should be at the end of
each block.  This is consistent with the other man pages.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Add missing formatting to sha512 in zpool-features.5

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Correct GUID of large_blocks in zpool-features.5

It is org.open-zfs:large_blocks (plural).

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Change wording in zfs-module-parameters.5

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Document that hole_birth is effectively useless

The first sentence of this commit comes from the wiki, and was
originally written by:
Rich Ercolani <rincebrain@gmail.com>
with changes by:
Tom Caputi <tcaputi@datto.com>

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641
Closes #8642

Document send_holes_without_birth_time

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Correct bookmark_v2 dependencies

encryption depends on bookmark_v2.

bookmark_v2 depends on bookmarks.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Fix formatting for multi_vdev_crash_dump in zpool-features.5

This needs to use tabs instead of spaces to display correctly (i.e. with
things lined up).

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8641

Fix incorrect use of .Nm directive for ZPOOL_VDEV_NAME_GUID in zpool(8)

It should only affect "zpool".

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8644

config: libintl/libiconv for gettext() detection

Detect in autoconf whether `-lintl` and possibly `-liconv` are necessary
for translation functions like `gettext()`.

The actual autoconf code is just:

```
AM_ICONV
AM_GNU_GETTEXT([external])
LIBS="$LIBS $LTLIBINTL $LTLIBICONV"
```

References:

https://www.gnu.org/software/gettext/manual/html_node/AM_005fGNU_005fGETTEXT.html
https://www.gnu.org/software/gettext/manual/html_node/AM_005fICONV.html

The reason to check for `libiconv` and add it separately is that this is
sometimes necessary if users are linking statically.

The `config/*.m4` files were added by running `gettextize` and removing
everything else.

The empty file `config/config.rpath` is necessary to avoid an error with
some versions of autotools, see:

http://ramblingfoo.blogspot.com/2007/07/required-file-configrpath-not-found.html

The `config.rpath` copied by `gettextize` does not currently work, there
is some kind of missing interaction with `libtool` and it tries to apply
`libtool` flags to the compiler.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Signed-off-by: Rafael Kitover <rkitover@gmail.com>
Closes #8554

Drop unused ZNODE_STATS and ZNODE_STAT_ADD()

Unused since 5649246dd3("Remove znode move functionality"),
and ZNODE_STAT_ADD() will never be needed.

Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8636

Fix typo "/zbin/zpool" -> "/sbin/zpool"

Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8643

Fix incorrect "[UNUSED]" comments

These aren't unused.
`flag` in zfs_create() also isn't to indicate large file.

Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8635

Tag 0.8.0-rc4

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

Code improvement and bug fixes for QAT support

1. Support QAT when ZFS is root file-system:
   When ZFS module is loaded before QAT started, the QAT can
   be started again in post-process, e.g.:
   echo 0 > /sys/module/zfs/parameters/zfs_qat_compress_disable
   echo 0 > /sys/module/zfs/parameters/zfs_qat_encrypt_disable
   echo 0 > /sys/module/zfs/parameters/zfs_qat_checksum_disable
2. Verify alder checksum of the de-compress result
3. Allocate Digest, IV and AAD buffer in physical contiguous
   memory by QAT_PHYS_CONTIG_ALLOC.
4. Update the documentation for zfs_qat_compress_disable,
   zfs_qat_checksum_disable, zfs_qat_encrypt_disable.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Weigang Li <weigang.li@intel.com>
Signed-off-by: Chengfeix Zhu <chengfeix.zhu@intel.com>
Closes #8323
Closes #8610

Restructure vec_idx loops

This replaces empty for loops with while loops to make the code easier
to read.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reported-by: github.com/dcb314
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #6681
Closes #6682
Closes #6683
Closes #8623

Add option [-V|--version] to emit version string

Add the 'zfs version' and 'zpool version' subcommands to display
the version of the user space utilities and loaded zfs kernel
module.  For example:

$ zfs version
zfs-0.8.0-rc3_169_g67e0366b88
zfs-kmod-0.8.0-rc3_169_g67e0366b88

The '-V' and '--version' aliases were added to support the
common convention of using 'zfs --version` to obtain the version
information.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Signed-off-by: TerraTech <1118433+TerraTech@users.noreply.github.com>
Closes #2501
Closes #8567

zfs allow refreservation needed for zfs create -V

When creating a non-sparse volume, zfs create sets a refreservation.
Accordingly, one needs the "refreservation" ability in addition to the
"create" ability in order to create a non-sparse volume.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reported-by: github.com/homerlinux
Reported-by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8531
Closes #8624

Clarify GRUB's lack of support for sha512, skein, edonr

zfs.8 correctly said that GRUB did not support them, but
zpool-features.5 said that "Booting off pools...is supported."  Now,
zpool-features.5 discusses GRUB specifically and indicates its lack of
support for these features.  Also, I have clarified the wording in both
places to indicate that the pool feature cannot be used.  It's not a
filesystem dataset thing, but pool-wide.

I described this as "cannot be used".  I think technically the feature
can be enabled, just not active.  However, the effect is essentially the
same: you cannot enable those checksum algorithms on any dataset in the
pool, so you might as well not enable the feature (which is just
pointing a loaded gun at your foot).  In the past, an argument could be
made that having all the features enabled was useful for simplicity, as
long as you didn't activate the GRUB-incompatible features, but that's
getting less and less realistic over time.  A user can still do that,
but we should not encourage that.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626
Closes #8446

Update a comment to match the code

GRUB supports large_blocks.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Reference zfeature.c in a SPA_VERSION comment

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Remove zfs.h comments about GRUB

Nobody is going to be bumping SPA_VERSION again, as OpenZFS has moved on
to feature flags.  Also, there is no requirement to keep GRUB
up-to-date, nor has that been happening.

The ZPL_VERSION could be bumped, but that would likely be handled in a
similar way, by adding filesystem feature flags.  In any event, we do
not need this comment, and we certainly don't need a reference to the
GRUB 0.97 source code in a Solaris tree.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Reword the dedup limitation for Edon-R

The old wording was effectively "You can not use this (except you can)",
which just seems confusing.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Consistently captialize GUID for features

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Fix the spelling of deferred

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Update zdb.8's fsck reference

On Linux, this is in man section 8, not 1M.  Also, there is no fsdb on
Linux, so I removed that.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Refer to commands consistently in zpool-features.5

This had a mix of command vs subcommand, quoted vs not quoted, and
bolded vs. not bolded command names.

Also, fix man page sections from 1M (Solaris) to 8 (Linux).

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Eliminate most mentions of "special"

Previously, the "spare" vdev type was described as "A special
pseudo-vdev which...".  I wanted to eliminate the word "special" from
that, now that the allocation_classes feature exists and there is such a
thing as a "special vdev".  I ended up eliminating almost all instances
of the word "special" that are not referencing the allocation_classes
feature.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8626

Fix issues with truncated files in raw sends

When receiving a raw send stream only reallocated objects
whose contents were not freed by the standard indicators
should call dmu_free_long_range().

Furthermore, if calling dmu_free_long_range() is required
then the objects current block size must be used and not
the new block size.

Two additional test cases were added to provided realistic
test coverage for processing reallocated objects which are
part of a raw receive.

Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8528
Closes #8607

Fix hierarchy misspellings

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reported-by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: Richard Laager <rlaager@wiktel.com>
Closes #8563
Closes #8622

Don't hard-code number of ioctls for portability

Use (ZFS_IOC_LAST - ZFS_IOC_FIRST) instead of 256.
It seems 256 is just a number large enough to hold ioctls
at the moment.

Using 256 also causes compile-time warning or error
on platfoms whose enum zfs_ioc definition differs.

Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Igor Kozhukhov <igor@dilos.org>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8598

Sync reserved Illumos ioctl comment with actual number

It's 81 now.

Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Igor Kozhukhov <igor@dilos.org>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8598

compile with -fno-omit-frame-pointer

By default, depending on the version, gcc can reuse the frame pointer register.
This is a micro-optimization that might help on some very old x86 processors.
However, it also makes dynamic tracing less useful because the stacks cannot
be easily observed.

This rule change instructs gcc to use the -fno-omit-frame-pointer option
when compiling.

Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Elling <Richard.Elling@RichardElling.com>
Closes #8617

Cleanup nits from ab7615d92

This patch simply up cleans up a nit and corrects an error message
issue that were introduced in the Multiple DVA scrub patch.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8619

Fix issue in receive_object() during reallocation

When receiving an object to a previously allocated interior slot
the new object should be "allocated" by setting DMU_NEW_OBJECT,
not "reallocated" with dnode_reallocate().  For resilience verify
the slot is free as required in case the stream is malformed.

Add a test case to generate more realistic incremental send streams
that force reallocation to occur during the receive.

Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8067
Closes #8614

Fix TXG_MASK cstyle

Fix style issue for 'tx->tx_txg&TXG_MASK'.  There should be white
space around the '&' character.  Split the dnode_reallocate() ASSERT
to make it more readable to clearly separate the checks.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8606

ZTS: Make fault cleanup function more robust

The cleanup function of auto_online_001_pos does not account for the
possibility that the test may fail while a disk is still removed. If
the test run is using real disks, cleanup should involve restoring any
that are missing.

Reviewed-by: Giuseppe Di Natale <guss80@gmail.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: John Kennedy <john.kennedy@delphix.com>
Closes #8579

Allow zfs-tests to recover from hibernation

When a system sleeps during a zfs-test, the time spent
hibernating is counted against the test's runtime even
though the test can't and isn't running.
This patch tries to detect timeouts due to hibernation and
reruns tests that timed out due to system sleeping.
In this version of the patch, the existing behavior of returning
non-zero when a test was killed is preserved. With this patch applied
we still return nonzero and we also automatically rerun the test we
suspect of being killed due to system hibernation.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Signed-off-by: Alek Pinchuk <apinchuk@datto.com>
Closes #8575

Always call rw_init in zio_crypt_key_unwrap

The error path in zio_crypt_key_unwrap would call zio_crypt_key_destroy which
calls rw_destroy(&key->zk_salt_lock); which has not yet been initialized.

We move the rw_init() call to the start of zio_crypt_key_unwrap instead.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes #8604
Closes #8605

Avoid stack overwrite in zfs_setattr_dir()

The bulk[] array index, count, must be reset per-iteration in order to
not overwrite the stack.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Chris Dunlop <chris@onthe.net.au>
Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
Closes #8072
Closes #8597
Closes #8601

Don't assume pthread_t is uint_t for portability

POSIX doesn't define pthread_t as uint_t. It could be a pointer.
This code causes below compile error on a platform using pointer
for pthread_t.

--
kernel.c:815:25: error: cast from pointer to integer of different size [-Werror=pointer-to-int-cast]
    (void) printf("%u ", (uint_t)pthread_self());

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Igor Kozhukhov <igor@dilos.org>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8558

Unbreak build on Linux kernel < 3.10

d12614521a("Fixes for procfs files backed by linked lists")
uses PDE_DATA(), but since PDE_DATA() (public interface which
replaced old public interface PDE()) first appeared in upstream
kernel 3.10, it lacks visible local definition for kernel < 3.10.

Move the local PDE_DATA() definition to a ZoL header, to unbreak
build on kernel < 3.10.

--
module/spl/spl-procfs-list.c: In function 'procfs_list_open':
module/spl/spl-procfs-list.c:166: error: implicit declaration of function 'PDE_DATA'
module/spl/spl-procfs-list.c:166: warning: assignment makes pointer from integer without a cast

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: John Gallagher <john.gallagher@delphix.com>
Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@gmail.com>
Closes #8599

Fix 'zfs list -t snapshot' depth

Commit df583073 introduced the ability to list the snapshots for a
specified dataset.  This change inadvertently resulted in only the top-
level snapshots being listed when no dataset was specified.  Fix this
issue by adding an additional check to determine if a dataset was
provided to avoid incorrectly restricting the depth.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Alek Pinchuk <apinchuk@datto.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8591
Closes #8594

Fix buffer length in strlcpy()

The length used for the strlcpy() used the size of zv_value
when it should have used the size of zc_name.  Correct this
typo.

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Reviewed-by: Igor Kozhukhov <igor@dilos.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8595
Closes #8596

Revert "Fix issues with truncated files in raw sends"

This partially reverts commit 5dbf8b4ed.  This change resolved
the issues observed with truncated files in raw sends.  However,
the required changes to dnode_allocate() introduced a regression
for non-raw streams which needs to be understood.

The additional debugging improvements from the original patch
were not reverted.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #7378
Issue #8528
Issue #8540
Issue #8565
Close #8584

predictive prefetch disabled on new pools until export/reboot

When a pool is initially created (by `zpool create`), predictive
prefetch is inadvertently disabled, until the pool is export/import-ed,
or the machine is rebooted.

When device removal was introduced, we added some code to disable
predictive prefetching until indirect vdevs have been loaded.  This
resulted in the "default state" of prefetch being disabled, until we
proactively enable it after indirect vdevs are loaded.  Unfortunately
this resulted in a few bugs where in some code paths we neglect to
enable predictive prefetch.  The first of these was fixed by
https://github.com/zfsonlinux/zfs/commit/20507534d4ede14d4dd82c99fc8d461704ce7419

This commit fixes another case where we also need to explicitly enable
predictive prefetch, when the pool is initially created.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
Closes #8577

features.kernel layout should match features.pool

The features.kernel layout should match features.pool.

Reviewed-by: Sara Hartse <sara.hartse@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Don Brady <don.brady@delphix.com>
Closes #8566

Restrict kstats and print real pointers

There are several places where we use zfs_dbgmsg and %p to
print pointers. In the Linux kernel, these values obfuscated
to prevent information leaks which means the pointers aren't
very useful for debugging crash dumps. We decided to restrict
the permissions of dbgmsg (and some other kstats while we were
at it) and print pointers with %px in zfs_dbgmsg as well as
spl_dumpstack

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: John Gallagher <john.gallagher@delphix.com>
Signed-off-by: sara hartse <sara.hartse@delphix.com>
Closes #8467
Closes #8476

Hint about zpool free vs zfs available

Also describe free/allocated/fragmentation

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Signed-off-by: Josh Soref <jsoref@users.noreply.github.com>
Closes #7565
Closes #8483

Fix txg_wait_open() load average inflation

Callers of txg_wait_open() which set should_quiesce=B_TRUE should be
accounted for as iowait time.  Otherwise, the caller is understood
to be idle and cv_wait_sig() is used to prevent incorrectly inflating
the system load average.

Similarly txg_wait_wait() has been updated to use cv_wait_io() to
be accounted against iowait.

Reviewed-by: Tim Chase <tim@chase2k.com>
Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8550
Closes #8558

Move dracut specifics to dracut module

Dracut depends on the environment variable BOOTFS to be set after pool
import. This dracut specific systemd ExecStartPost command should not be
called for any non-dracut systems, so let's move it to a static systemd
unit that.

Reviewed-by: Manuel Amador (Rudd-O) <rudd-o@rudd-o.com>
Reviewed-by: Matthew Thode <prometheanfire@gentoo.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Michael Niewöhner <foss@mniewoehner.de>
Closes #8510

Fix man(1) warnings

The macOS man app strenuously objects to blank lines in man files.

mdoc warning: Empty input line #xyz

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: bunder2015 <omfgbunder@gmail.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Josh Soref <jsoref@users.noreply.github.com>
Closes #8559

Append snapshot name to "TIME SENT SNAPSHOT" output

Simply appends zhp->zfs_name to the "TIME SENT SNAPSHOT" output.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Signed-off-by: TerraTech <TerraTech@users.noreply.github.com>
Closes #8543

Do not iterate through filesystems unnecessarily

Currently, when attempting to list snapshots ZFS may do a lot of
extra work checking child datasets. This is because the code does
not realize that it will not be able to reach any snapshots
contained within snapshots that are at the depth limit since the
snapshots of those datasets are counted as an additional layer
deeper. This patch corrects this issue.

In addition, this patch adds the ability to do perform the commands:

$ zfs list -t snapshot <dataset>
$ zfs get -t snapshot <prop> <dataset>

as a convenient way to list out properties of all snapshots of a
given dataset without having to use the depth limit.

Reviewed-by: Alek Pinchuk <apinchuk@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8539

Fix systemd-import services

On debian, systemd complains about missing /bin/awk because it
actually is located at /usr/bin/awk. It is not a good idea to
hardcode binary paths because different linux distros use different
paths. According to systemd's man page it is absolutely safe to
miss paths for binaries located at standard locations (/bin,
/sbin, /usr/bin, ...).

Further, replace this more or less complicated awk command by
grep.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Michael Niewöhner <foss@mniewoehner.de>
Issue #8510

Remove hard dependency on bash

zfs-import-* services have a hard dependency on bash while not
everyone has bash installed. At this point /bin/sh is sufficient,
so use that.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Michael Niewöhner <foss@mniewoehner.de>
Issue #8510

Update raw send documentation

This patch simply clarifies some of the limitations related to
raw sends in the man page. No functional changes.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jason Cohen <jwittlincohen@gmail.com>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8503
Closes #8544

Add TRIM support

UNMAP/TRIM support is a frequently-requested feature to help
prevent performance from degrading on SSDs and on various other
SAN-like storage back-ends.  By issuing UNMAP/TRIM commands for
sectors which are no longer allocated the underlying device can
often more efficiently manage itself.

This TRIM implementation is modeled on the `zpool initialize`
feature which writes a pattern to all unallocated space in the
pool.  The new `zpool trim` command uses the same vdev_xlate()
code to calculate what sectors are unallocated, the same per-
vdev TRIM thread model and locking, and the same basic CLI for
a consistent user experience.  The core difference is that
instead of writing a pattern it will issue UNMAP/TRIM commands
for those extents.

The zio pipeline was updated to accommodate this by adding a new
ZIO_TYPE_TRIM type and associated spa taskq.  This new type makes
is straight forward to add the platform specific TRIM/UNMAP calls
to vdev_disk.c and vdev_file.c.  These new ZIO_TYPE_TRIM zios are
handled largely the same way as ZIO_TYPE_READs or ZIO_TYPE_WRITEs.
This makes it possible to largely avoid changing the pipieline,
one exception is that TRIM zio's may exceed the 16M block size
limit since they contain no data.

In addition to the manual `zpool trim` command, a background
automatic TRIM was added and is controlled by the 'autotrim'
property.  It relies on the exact same infrastructure as the
manual TRIM.  However, instead of relying on the extents in a
metaslab's ms_allocatable range tree, a ms_trim tree is kept
per metaslab.  When 'autotrim=on', ranges added back to the
ms_allocatable tree are also added to the ms_free tree.  The
ms_free tree is then periodically consumed by an autotrim
thread which systematically walks a top level vdev's metaslabs.

Since the automatic TRIM will skip ranges it considers too small
there is value in occasionally running a full `zpool trim`.  This
may occur when the freed blocks are small and not enough time
was allowed to aggregate them.  An automatic TRIM and a manual
`zpool trim` may be run concurrently, in which case the automatic
TRIM will yield to the manual TRIM.

Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Reviewed-by: Tim Chase <tim@chase2k.com>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: George Wilson <george.wilson@delphix.com>
Reviewed-by: Serapheim Dimitropoulos <serapheim@delphix.com>
Contributions-by: Saso Kiselkov <saso.kiselkov@nexenta.com>
Contributions-by: Tim Chase <tim@chase2k.com>
Contributions-by: Chunwei Chen <tuxoko@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8419
Closes #598

Send stream should only list included snaps

Currently, zfs send streams will include a list of all snapshots
on the source side if the '-p' option is provided. This can cause
performance problems on the receive side, especially if those
snapshots aren't present on the destination. These problems arise
because guid_to_name(), which is used for several receive side
functions, will search the entire receive-side pool if it can't
find a snapshot with a matching guid. This patch corrects the
issue by ensuring only streams that require this list of snapshots
include them.

Reviewed-by: Alek Pinchuk <apinchuk@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8533

Fix issues with truncated files in raw sends

This patch fixes a few issues with raw receives involving
truncated files:

* dnode_reallocate() now calls dnode_set_blksz() instead of
  dnode_setdblksz(). This ensures that any remaining dbufs with
  blkid 0 are resized along with their containing dnode upon
  reallocation.

* One of the calls to dmu_free_long_range() in receive_object()
  needs to check that the object it is about to free some contents
  or hasn't been completely removed already by a previous call to
  dmu_free_long_object() in the same function.

* The same call to dmu_free_long_range() in the previous point
  needs to ensure it uses the object's current block size and
  not the new block size. This ensures the blocks of the object
  that are supposed to be freed are completely removed and not
  simply partially zeroed out.

This patch also adds handling for DRR_OBJECT_RANGE records to
dprintf_drr() for debugging purposes.

Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #7378
Closes #8528

Update valid vdev types for get_disklist

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Signed-off-by: Richard Elling <Richard.Elling@RichardElling.com>
Closes #8532

ZTS: Detect e2fsprogs verity issue

The projectid_001_pos and projecttree_001_pos test cases use the lsattr
command to detect that the project quota bit is set correctly.  Due to
a bug in e2fsprogs-1.44.4 setting the Project 'P' bit also results in
the Verity 'V' bit being reported as set.  This will result in the test
case failing.

The issue has been resolved in e2fsprogs but in order to avoid testing
failures these two test cases are skipped when e2fsprogs-1.44.4 is
installed.

https://github.com/tytso/e2fsprogs/commit/7e5a95e3d

Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8534

Correct a very minor grammar issue

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Evan Allrich <evan@unguku.com>
Closes #8535

git ignore python 3.7 virtual environment directories

Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Elling <Richard.Elling@RichardElling.com>
Closes #8525

Fix vd_path and error in spa_vdev_remove()

Make a local copy of the vd_path and preserve the removal error
for use in spa_history_log_internal().  This is required because
after spa_vdev_exit() there is nothing preventing the vdev state
from changing.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Igor Kozhukhov <igor@dilos.org>
Closes #8522

Panic when running 'zpool split'

Added missing remove of detachable VDEV from txg's DTL list
to avoid use-after-free for the split VDEV

Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Roman Strashkin <roman.strashkin@nexenta.com>
Closes #5565
Closes #7856

ZFS Reads may result in unneccesary calls to zil_commit

ZFS supports O_RSYNC for read operations and when specified will ensure
the same level of data integrity that O_DSYNC and O_SYNC provides for
writes. O_RSYNC by itself has no effect so it must be combined with
either O_DSYNC or O_SYNC. However, many platforms don't support O_RSYNC
and have mapped O_SYNC to mean O_RSYNC within ZFS. This is incorrect
and causes unnecessary calls to zil_commit. Only platforms which
support O_RSYNC should implement the zil_commit functionality in the
read code path.

Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: George Wilson <george.wilson@delphix.com>
Closes #8523

MMP interval and fail_intervals in uberblock

When Multihost is enabled, and a pool is imported, uberblock writes
include ub_mmp_delay to allow an importing node to calculate the
duration of an activity test.  This value, is not enough information.

If zfs_multihost_fail_intervals > 0 on the node with the pool imported,
the safe minimum duration of the activity test is well defined, but does
not depend on ub_mmp_delay:

zfs_multihost_fail_intervals * zfs_multihost_interval

and if zfs_multihost_fail_intervals == 0 on that node, there is no such
well defined safe duration, but the importing host cannot tell whether
mmp_delay is high due to I/O delays, or due to a very large
zfs_multihost_interval setting on the host which last imported the pool.
As a result, it may use a far longer period for the activity test than
is necessary.

This patch renames ub_mmp_sequence to ub_mmp_config and uses it to
record the zfs_multihost_interval and zfs_multihost_fail_intervals
values, as well as the mmp sequence.  This allows a shorter activity
test duration to be calculated by the importing host in most situations.
These values are also added to the multihost_history kstat records.

It calculates the activity test duration differently depending on
whether the new fields are present or not; for importing pools with
only ub_mmp_delay, it uses

(zfs_multihost_interval + ub_mmp_delay) * zfs_multihost_import_intervals

Which results in an activity test duration less sensitive to the leaf
count.

In addition, it makes a few other improvements:
* It updates the "sequence" part of ub_mmp_config when MMP writes
  in between syncs occur.  This allows an importing host to detect MMP
  on the remote host sooner, when the pool is idle, as it is not limited
  to the granularity of ub_timestamp (1 second).
* It issues writes immediately when zfs_multihost_interval is changed
  so remote hosts see the updated value as soon as possible.
* It fixes a bug where setting zfs_multihost_fail_intervals = 1 results
  in immediate pool suspension.
* Update tests to verify activity check duration is based on recorded
  tunable values, not tunable values on importing host.
* Update tests to verify the expected number of uberblocks have valid
  MMP fields - fail_intervals, mmp_interval, mmp_seq (sequence number),
  that sequence number is incrementing, and that uberblock values match
  tunable settings.

Reviewed-by: Andreas Dilger <andreas.dilger@whamcloud.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Olaf Faaland <faaland1@llnl.gov>
Closes #7842

Mutex leak in dsl_dataset_hold_obj()

In addition to dsl_dataset_evict_async() releasing a hold, there is
an error case in dsl_dataset_hold_obj() which had missed 4 additional
release calls.  This was introduced in a1d477c24.

openzfsonosx-commit: https://github.com/openzfsonosx/zfs/commit/63ff7f1c

Authored by: Jorgen Lundman <lundman@lundman.net>
Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8517

QAT: Allocate digest_buffer using QAT_PHYS_CONTIG_ALLOC()

If the buffer 'digest_buffer' is allocated in the qat_checksum()
stack, it can't ensure that the address is physically contiguous,
and the DMA result of the buffer may be handled incorrectly.
Using QAT_PHYS_CONTIG_ALLOC() ensures a physically
contiguous allocation.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Chengfei, Zhu <chengfeix.zhu@intel.com>
Closes #8323
Closes #8521

Report holes when there are only metadata changes

Update the dirty check in dmu_offset_next() such that dnode's
are only considered dirty for the purpose or reporting holes
when there are pending data blocks or frees to be synced.  This
ensures that when there are only metadata updates to be synced
(atime) that holes are reported.

Reviewed-by: Debabrata Banerjee <dbanerje@akamai.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #6958
Closes #8505

Improve `zpool labelclear`

1) As implemented the `zpool labelclear` command overwrites
the calculated offsets of all four vdev labels even when only a
single valid label is found.  If the device as been re-purposed
but still contains a valid label this can result in space no
longer owned by ZFS being zeroed.  Prevent this by verifying
every label removed is intact before it's overwritten.

2) Address a small bug in zpool_do_labelclear() which prevented
labelclear from working on file vdevs.  Only block devices support
BLKFLSBUF, try the ioctl() but when it's reported as unsupported
this should not be fatal.

3) Fix `zpool labelclear` so it can be run on vdevs which were
removed from the pool with `zpool remove`.  Additionally, allow
intact but partial labels to be cleared as in the case of a failed
`zpool attach` or `zpool replace`.

4) Remove LABELCLEAR and LABELREAD variables for test cases.

Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Reviewed-by: Tim Chase <tim@chase2k.com>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8500
Closes #8373
Closes #6261

Add missing dmu_zfetch_fini() in dnode_move_impl()

As it turns out, on the Windows platform when rw_init() is called
(rather its bedrock call ExInitializeResourceLite) it is placed on
an active-list of locks, and is removed at rw_destroy() time.

dnode_move() has logic to copy over the old-dnode to new-dnode,
including calling dmu_zfetch_init(new-dnode). But due to the missing
dmu_zfetch_fini(old-dnode), kmem will call dnode_dest() to release the
memory (and in debug builds fill pattern 0xdeadbeef) over the Windows
active-lock's prev/next list pointers, making Windows sad.

But on other platforms, the contents of dmu_zfetch_fini() is one
call to list_destroy() and one to rw_destroy(), which is effectively
a no-op call and is not required. This commit is mostly for
"correctness" and can be skipped there.

Porting Notes:
* This leak exists on Linux but currently can never happen because
  the dnode_move() functionality is not supported.

openzfsonosx-commit: openzfsonosx/zfs@d95fe517

Authored by: Julian Heuking <JulianH@beckhoff.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes #8519

Add space in error message

This patch simply adds a missing space in the
ZFS_ERR_FROM_IVSET_GUID_MISSING error message.

Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Don Brady <don.brady@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8514

Fix l2arc_evict() destroy race

When destroying an arc_buf_hdr_t its identity cannot be discarded
until it is entirely undiscoverable.  This not only includes being
unhashed, but also being removed from the l2arc header list.
Discarding the header's identify prematurely renders the hash
lock useless because it will always hash to bucket zero.

This change resolves a race with l2arc_evict() by discarding the
identity after it has been removed from the l2arc header list.
This ensures either the header is not on the list or contains
the correct identify.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #7688
Closes #8144

Multiple DVA Scrubbing Fix

Currently, there is an issue in the sequential scrub code which
prevents self healing from working in some cases. The scrub code
will split up all DVA copies of a bp and issue each of them
separately. The problem is that, since each of the DVAs is no
longer associated with the others, the self healing code doesn't
have the opportunity to repair problems that show up in one of the
DVAs with the data from the others.

This patch fixes this issue by ensuring that all IOs issued by the
sequential scrub code include all DVAs. Initially, only the first
DVA of each is attempted. If an issue arises, the IO is retried
with all available copies, giving the self healing code a chance
to correct the issue.

To test this change, this patch also adds the ability for zinject
to specify individual DVAs to inject read errors into. We then
add a new test case that utilizes this functionality to ensure
scrubs and self-healing reads can handle and transparently fix
issues with individual copies of blocks.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matt Ahrens <mahrens@delphix.com>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8453

Make zpool status counters match error events count

The number of IO and checksum events should match the number of errors
seen in zpool status.  Previously there was a mismatch between the
two counts because zpool status would only count unrecovered errors,
while zpool events would get an event for *all* errors (recovered or
not).  This lead to situations where disks could be faulted for
"too many errors", while at the same time showing zero errors in zpool
status.

This fixes the zpool status error counters to increment at the same
times we post the error events.

Reviewed-by: Tom Caputi <tcaputi@datto.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Signed-off-by: Tony Hutter <hutter2@llnl.gov>
Closes #4851
Closes #7817

Fix memory leaks in zfsvfs_create_impl()

This patch simply fixes some small memory leaks that can happen
during error handling in zfsvfs_create_impl(). If the function
fails, it frees all the memory / references it created.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #8490

zfs-import: should be before swap

zfs-import must be done before swap in order for swap on zvol to work

Reviewed-by: Kash Pande <kash@tripleback.net>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Henrik Riomar <henrik.riomar@gmail.com>
Closes #8502

Better user experience for errata 4

This patch attempts to address some user concerns that have arisen
since errata 4 was introduced.

* The errata warning has been made less scary for users without
  any encrypted datasets.

* The errata warning now clears itself without a pool reimport if
  the bookmark_v2 feature is enabled and no encrypted datasets
  exist.

* It is no longer possible to create new encrypted datasets without
  enabling the bookmark_v2 feature, thus helping to ensure that the
  errata is resolved.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Issue ##8308
Closes #8504

Update commented zed.rc values to defaults

Update zed.rc values reflect their default value.  This helps
avoid confusion if a user expects functionality to be enabled.

Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Kash Pande <kash@tripleback.net>
Closes #8498

Use 'printf %s' instead of 'echo -n' for compatibility

The ksh 'echo -n' behavior on Illumos and Linux differs.  For
compatibility with others platforms switch to "printf '%s' ".

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Allan Jude <allanjude@freebsd.org>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Giuseppe Di Natale <guss80@gmail.com>
Signed-off-by: Igor Kozhukhov <igor@dilos.org>
Closes #8501

Add separate aggregation limit for non-rotating media

Before sequential scrub patches ZFS never aggregated I/Os above 128KB.
Sequential scrub bumped that to 1MB, supposedly to reduce number of
head seeks for spinning disks.  But for SSDs it makes little to no
sense, especially on FreeBSD, where due to MAXPHYS limitation device
will likely still see bunch of 128KB I/Os instead of one large.
Having more strict aggregation limit for SSDs allows to avoid
allocation of large memory buffer and copy to/from it, that is a
serious problem when throughput reaches gigabytes per second.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes #8494

Update CONTRIBUTING to point users to IRC as well as mailing list

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Kash Pande <kash@tripleback.net>
Closes #8466