Solve issues with L2ARC if compressed_arc_enabled=0

Maybe as a separate change, but is it time to stop shipping lib32 by default as well?

Update to latest -current for @bkidney_briankidney.ca

In D12396#257653, @seanc wrote:

avg@ added the loader tunable to make this adjustable in https://svnweb.freebsd.org/base?view=revision&revision=323797 however I have every reason to believe (and measurements from production) that 4K is a going to result in more waste than a 1K default. A 1K default may result in fragmentation with 4K slabs but there is less waste for everyone with 1K than 4K.

@mahrens , do you think OpenZFS will go back to a 1K default?

Newer version of ZSTD (1.3.2)

Switch from symlink to include files

Switch to Warner's fake-files concept

Update to latest -current and lots of cleanup suggested by imp

Thanks for cleaning up my mess

In D13088#274147, @nikita_elyzion.net wrote:

The best fix would have been to check the errno value inside zpool_read_label() just after the libc calls, like all errno checks should be done. But it necessitates to add one more return value to zpool_read_label():

• 0 == vdev label was read (continue the normal execution)
• -1 == ENOMEM, do the no_memory() call and fail.
• -2 == vdev label not found (don't enter the if block and continue the normal execution)

In D7538#273970, @karl_denninger.net wrote:

I'll try to get some time allocated to that - right now the machine that
I use for this effort is busy doing other things but I may be able to
free up some time on it in the next week or two.

Is there any intent to backport/MFC those changes into 11-STABLE?  That
I can probably test and adapt the code to more-quickly.

The needfree change is a good one, it's just that having it confined to
-HEAD causes problems for me right now due to "free computer of
appropriate configuration" constraints.

Thanks in advance.

• smime.p7s4 KBDownload

In D7538#273915, @karl_denninger.net wrote:

Well that's a bitch.

It's easily fixed but I'm not going to be doing much with -HEAD until it
stabilizes and gets closer to being -RELEASE.

• smime.p7s4 KBDownload

Note: the modern installer has an entire UI dedicated to you setting the date correctly during the install.

In D13088#273063, @nikita_elyzion.net wrote:

Thanks for your comments Allan and Andriy.
Concerning the errno set to 12 (ENOMEM), it's effectively set by the ioctl ZFS_IOC_POOL_STATS which is first called with a too small nvlist.
I'm attaching the conditional watchpoint where it happens.

Watchpoint 1 hit:
old value: 0
new value: 12
Process 11604 stopped
* thread #1, name = 'zpool', stop reason = watchpoint 1
    frame #0: 0x0000000800e951e9 libzfs.so.2`zcmd_ioctl(fd=3, request=5, zc=0x00007fffffffc640) at libzfs_compat.c:117
   114 	
   115 		if (ret == 0 && oldsize < zc->zc_nvlist_dst_size) {
   116 			ret = -1;
-> 117 			errno = ENOMEM;
   118 		}
   119 	
   120 		return (ret);
(lldb) bt
* thread #1, name = 'zpool', stop reason = watchpoint 1
  * frame #0: 0x0000000800e951e9 libzfs.so.2`zcmd_ioctl(fd=3, request=5, zc=0x00007fffffffc640) at libzfs_compat.c:117
    frame #1: 0x0000000800e87099 libzfs.so.2`zpool_refresh_stats(zhp=0x0000000802c30000, missing=0x00007fffffffd844) at libzfs_config.c:282
    frame #2: 0x0000000800e800e5 libzfs.so.2`zpool_open_canfail(hdl=0x0000000802c20000, pool="bld311") at libzfs_pool.c:995
    frame #3: 0x0000000800e802ef libzfs.so.2`zpool_open(hdl=0x0000000802c20000, pool=<unavailable>) at libzfs_pool.c:1051
    frame #4: 0x000000000040820c zpool`zpool_do_add(argc=3, argv=0x00007fffffffeb70) at zpool_main.c:540
    frame #5: 0x000000000040766f zpool`main(argc=<unavailable>, argv=0x00007fffffffeb60) at zpool_main.c:5642
    frame #6: 0x00000000004043ff zpool`_start + 399

Concerning your suggestion on how to fix it, I would rather suggest that to modify zpool_read_label() to return -1 in the case of a libc with errno == ENOMEM and return -2 in the case of other internal and logical ZFS "failure" ( no vdev label for exemple ).
In my opinion our real bug here is to check the value of errno after a call to an internal ZFS function and not after the libc call. Currently we are reading errno too late and we cannot easily differentiate a libc failure from a normal ZFS failure.

for example:

if (zpool_read_label(fd, &config) != 0 && errno == ENOMEM) {
        (void) no_memory(hdl);
        return (-1);
}

would be:

if (zpool_read_label(fd, &config) == -1) {
        (void) no_memory(hdl);
        return (-1);
}

I think since we are returning -1 (a failure), it makes more sense to set errno to 0 at the top of the function, so that we only zero it if it is from before this function was run.

Thanks for the feedback Warner. I'll work on that stuff later today or tomorrow.

Rebase to zstd 1.3.2

@yann.collet.73_gmail.com : Is there any tuning we can do to get the size of the compression context down, since we know the maximum size if the input will be 16MB?

I will test this on my macbookpro NVMe where FreeBSD currently cannot read the partition table.

Did you check for other man pages that .Xr this manpage?

I think an interface like mjg described is the best approach.

jail_attach() is not the correct approach, because it makes the actions of the administrator on the host visible within the jail.

bluetooth should not be unconditional.

Thank you very much

Approved

It might be useful to mention -DNO_CLEAN as a more supported shortcut

Approved

Approved By: allanjude

Cleanup debugging cruft

Sparc fixes

Update to support ZSTD 1.3.x

@imp I know you are working on a sys/boot cleanup, so I figured you might want to integrate this

Address Conrad's feedback re: M_NOWAIT

This now cleanly applies with 'arc patch' for me

Updated for newer version of zstd

Thank you

I agree with @mav about adding any other useful bits from GEOM that are available.

Approved

I think something else is going on. Can you try the with and without with 'truss -c' instead of 'time'

I am confused by your results:

In D12339#255993, @cem wrote:

Er, why is rename / renameat with AT_FDCWD not allowed in capability mode?

Resync

Looks good.

Thanks for keeping on top of this

In the future we should do away with the uuencode, and maybe switch to xz (doesn't save any space over bzip, but is one of the only remaining uses of bzip in buildworld)

Hopefully Colin can get a chance to look at this too, but looks good to me now.

If it is going to be on by default, it should be enabled even in the absence of -h when you do geli init. And we need a -H flag for init/label to be able to disable it.

I think it makes more sense to treat it like $NOTTYOK etc, and just make it 0 or 1