imp [Sat, 21 Mar 2020 00:01:06 +0000 (00:01 +0000)]
Redo r359164 now that it's baked : Eliminate misuse of $MACHINE for userland things.
Use TARGET_ARCH and/or MACHINE_ARCH exclusively. Change all __TT uses to __T
with appropriate translations. MACHINE/TARGET is to be used only for kernel
things, and this fixes the last few stragglers.
scottl [Fri, 20 Mar 2020 23:26:37 +0000 (23:26 +0000)]
When printing out the contents of the VSEC, include the contents of the
headers. Device documentation often times give offsets relative to the
start of the entire VSEC, not just the post-header data area, so this
change makes it easier to correlate offsets.
kib [Fri, 20 Mar 2020 21:06:58 +0000 (21:06 +0000)]
Introduce LINUXKPI_GENSRCS.
Centralize the list of generated files required by linuxkpi consumers,
into the common variable. This way, consumers that use the variable
are insulated from possible changes in the list.
emaste [Fri, 20 Mar 2020 19:28:55 +0000 (19:28 +0000)]
Use a separate copy of machdep.h in cp and ce drivers
Previously they included sys/dev/cx/machdep.h, but the cx driver was
retired in r359178. These drivers haven't had real development for
a decade or more so there's no real benefit in sharing this file; just
copy it to the ce and cp subdirs.
emaste [Fri, 20 Mar 2020 16:25:06 +0000 (16:25 +0000)]
ObsoleteFiles.inc: add blank line between some recent entries
We recently switched to including a blank line between ObsoleteFiles.inc
entries. I missed this when adding GCC 4.2.1 (because the change was
prepared months ago in a WIP tree); add a space for GCC 4.2.1 now, as
well as for the other entry after the switch.
markj [Fri, 20 Mar 2020 16:17:54 +0000 (16:17 +0000)]
Remove UNP_NASCENT, reverting r303855.
unp_connectat() no longer holds the link lock across calls to
sonewconn(), so the recursion described in r303855 can no longer occur.
No functional change intended.
Tested by: pho
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
emaste [Fri, 20 Mar 2020 16:02:45 +0000 (16:02 +0000)]
remove ancient pre-2000 ObsoleteFiles.inc entries
We support 10.3 as the minimum version to install from, which was
released in the mid-2010s. There's a lot of ancient ObsoleteFiles.inc
history that serves no purpose today; start by removing entries from
1999 and earlier.
imp [Fri, 20 Mar 2020 15:07:25 +0000 (15:07 +0000)]
Implement a workaround for kms-drm modules
pci_iov_if.h was added to pci.h, but none of the kms-drm branches have
that. Rather than play whack a mole with the branches, move its inclusion to
linux_pci.c which is the only part of the code that needs it now.
Longer term, other solutions will be needed, but this gives us time to get those
deployed on all the supported versions.
imp [Fri, 20 Mar 2020 15:07:21 +0000 (15:07 +0000)]
Eliminate misuse of $MACHINE for userland things.
Use TARGET_ARCH and/or MACHINE_ARCH exclusively. Change all __TT uses to __T
with appropriate translations. MACHINE/TARGET is to be used only for kernel
things.
np [Fri, 20 Mar 2020 05:12:16 +0000 (05:12 +0000)]
cxgbe(4): Split sge_nm_rxq into three cachelines.
This reduces the lines bouncing around between the driver rx ithread and
the netmap rxsync thread. There is no net change in the size of the
struct (it continues to waste a lot of space).
This kind of split was originally proposed in D17869 by Marc De La
Gueronniere @ Verisign, Inc.
markj [Thu, 19 Mar 2020 21:38:52 +0000 (21:38 +0000)]
Fix synchronization in the IPV6_2292PKTOPTIONS set handler.
The inpcb needs to be locked when we update output packet options.
Otherwise it is possible for the IPV6_2292PKTOPTIONS handler to free
packet option structures while another thread is reading or updating
them.
Note that the option handler is still kind of broken. For instance it
frees all options before performing privilege checks for individual
options. However, this can be fixed separately.
Reported by: syzbot+52eb0fd4ddc119787f9d@syzkaller.appspotmail.com
Reviewed by: bz, tuexen
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D24125
mav [Thu, 19 Mar 2020 17:20:50 +0000 (17:20 +0000)]
Fix SES on device slots without phys after r349321.
Broadcom 9400-8i8e HBAs report virtual SES device, where slots representing
external connectors are reported having no phys. Since sasdev_phys is NULL
there and proto_hdr is a union, ses_paths_iter() misinterpreted them as ATA.
Add explicit protocol check to properly differentiate them.
markj [Thu, 19 Mar 2020 15:40:05 +0000 (15:40 +0000)]
kern_dup(): Call filecaps_free_prep() in a write section.
filecaps_free_prep() bzeros the capabilities structure and we need to be
careful to synchronize with unlocked readers, which expect a consistent
rights structure.
Reviewed by: kib, mjg
Reported by: syzbot+5f30b507f91ddedded21@syzkaller.appspotmail.com
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D24120
markj [Thu, 19 Mar 2020 15:39:45 +0000 (15:39 +0000)]
Enter a write sequence when updating rights.
The Capsicum system calls modify file descriptor table entries. To
ensure that readers observe a consistent snapshot of descriptor writes,
the system calls need to signal to unlocked readers that an update is
pending.
Note that ioctl rights are always checked with the descriptor table lock
held, so it is not strictly necessary to signal unlocked readers.
However, we probably want to enable lockless ioctl checks eventually, so
use seqc_write_begin() in kern_cap_ioctls_limit() too.
Reviewed by: kib
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D24119
kp [Thu, 19 Mar 2020 12:54:43 +0000 (12:54 +0000)]
pfctl: improve rule load times with thousands of interfaces
r343287 / D18759 introduced ifa_add_groups_to_map() which is now run by
ifa_load/ifa_lookup/host_if. When loading an anchor or ruleset via pfctl that
does NOT contain ifnames as hosts, host() still ends up iterating all
interfaces twice, grabbing SIOCGIFGROUP ioctl twice for each. This adds an
unnecessary amount of time on systems with thousands or tens of thousands of
interfaces.
Prioritize the IPv4/6 check over the interface name lookup, which skips loading
the iftab and iterating all interfaces when the configuration does not contain
interface names.
Two arguments were reversed in calls to cam_strvis() in
nvme_da.c. This was found by a Coverity scan of this code within Dell
(Isilon). These are also marked in the FreeBSD Coverity scan as CIDs 1400526 & 1400531.
0mp [Thu, 19 Mar 2020 09:23:26 +0000 (09:23 +0000)]
Document geli(8) loader variables conventions
The geli(8) manual page has an example for preloading keyfiles during boot.
There is no detail though on how the lookup of these variables actually
works.
Let's document that the name of a device does not have to be a part
of the variable.
delphij [Thu, 19 Mar 2020 06:33:06 +0000 (06:33 +0000)]
Fix race condition in catopen(3).
The current code uses a rwlock to protect the cached list, which
in turn holds a list of catentry objects, and increments reference
count while holding only read lock.
Fix this by converting the reference counter to use atomic operations.
While I'm there, also perform some clean ups around memory operations.
PR: 202636
Reported by: Henry Hu <henry.hu.sh@gmail.com>
Reviewed by: markj
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D24095
mav [Thu, 19 Mar 2020 01:05:54 +0000 (01:05 +0000)]
MFOpenZFS: make zil max block size tunable
We've observed that on some highly fragmented pools, most metaslab
allocations are small (~2-8KB), but there are some large, 128K
allocations. The large allocations are for ZIL blocks. If there is a
lot of fragmentation, the large allocations can be hard to satisfy.
The most common impact of this is that we need to check (and thus load)
lots of metaslabs from the ZIL allocation code path, causing sync writes
to wait for metaslabs to load, which can take a second or more. In the
worst case, we may not be able to satisfy the allocation, in which case
the ZIL will resort to txg_wait_synced() to ensure the change is on
disk.
To provide a workaround for this, this change adds a tunable that can
reduce the size of ZIL blocks.
External-issue: DLPX-61719 Reviewed-by: George Wilson <george.wilson@delphix.com> Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
Closes #8865
openzfs/zfs@b8738257c2607c73c731ce8e0fd73282b266d6ef
kib [Wed, 18 Mar 2020 21:34:52 +0000 (21:34 +0000)]
Do not spuriously re-enable disabled io_apic pin on EOI for some configurations.
If EOI suppression is supported but reported ioapic version is so old
that it does not has EOI register (weird virtualization setup), fix
Intel trick of eoi-ing by flipping pin type (edge/level) to account
for the disabled pin.
Reported by: Juniper
Reviewed by: jhb
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D23965
kib [Wed, 18 Mar 2020 21:31:35 +0000 (21:31 +0000)]
Stop (trying to) renumber io apics.
It does not serve any purpose now, the io apic id is not seen by
software, and some Intel documents claim that the register is
implemented for FUD reasons. More, renumbering seems to not work on
new Intel machines which actually have mismatched MADT and hw IDs.
On older machines where separate APIC bus existed, unique numbering of
all APICs was required for bus arbitration to work, but it is no
longer true (that machines were SMP from pre-Pentium IV era).
When matching PCIe IOAPIC device against MADT-enumerated IOAPICs,
compare io_apic_id from BAR against io_apic_id read from the
MADT-pointed register page.
kib [Wed, 18 Mar 2020 21:24:34 +0000 (21:24 +0000)]
Widen the stored io_apic_id to 8 bits.
It seems that the newer Intel chipset did that, and Linux reads 8
bits. The only detail is that all seen datasheets, even under NDA,
claim that io apic id is 4 bits.
emaste [Wed, 18 Mar 2020 21:20:10 +0000 (21:20 +0000)]
Revert r359083, fixed properly by r359088
r359083 introduced a workaround for stale libomp dependencies during a
regular (no -DNO_CLEAN) buildworld. r359088 addressed the reason the
clean step missed libomp, so revert the workaround.
emaste [Wed, 18 Mar 2020 21:15:45 +0000 (21:15 +0000)]
src.opts.mk: drop C++11 compiler conditional from OPENMP
We no longer support older C++ compilers, so do not need to explicitly
test for C++11 support.
After r339946 we stopped running `cc --version` during cleandir/obj
stages, so stopped setting COMPILER_FEATURES. This in turn meant
lib/libomp was excluded from the clean stage in a normal buildworld
(i.e., one without -DNO_CLEAN), and this is what caused recent build
failures with errors about missing ittnotify_static.c.
This commit should obviate the need for the workaround committed in
r359083. Thanks to bdrewery for the insight and for pushing for a
correct fix. There are more cleanups to be done, but this change is
a simplification and an improvement over r359083.
Fix -Wdeprecated-copy-dtor and -Wdeprecated-dynamic-exception-spec
warnings.
Summary:
The former are like:
libcxx/include/typeinfo:322:11: warning: definition of implicit copy
constructor for 'bad_cast' is deprecated because it has a
user-declared destructor [-Wdeprecated-copy-dtor]
virtual ~bad_cast() _NOEXCEPT;
^
libcxx/include/typeinfo:344:11: note: in implicit copy constructor
for 'std::bad_cast' first required here
throw bad_cast();
^
Fix these by adding an explicitly defaulted copy constructor.
The latter are like:
libcxx/include/codecvt:105:37: warning: dynamic exception
specifications are deprecated [-Wdeprecated-dynamic-exception-spec]
virtual int do_encoding() const throw();
^~~~~~~
This is because we use -Wsystem-headers during buildworld, and the two
warnings above are now triggered by default with clang 10, preventing
most C++ code from compiling without NO_WERROR.
dim [Wed, 18 Mar 2020 20:44:40 +0000 (20:44 +0000)]
Merge commit b8ebc11f0 from llvm git (by Sanjay Patel):
[EarlyCSE] avoid crashing when detecting min/max/abs patterns (PR41083)
As discussed in PR41083:
https://bugs.llvm.org/show_bug.cgi?id=41083
...we can assert/crash in EarlyCSE using the current hashing scheme
and instructions with flags.
ValueTracking's matchSelectPattern() may rely on overflow (nsw, etc)
or other flags when detecting patterns such as min/max/abs composed
of compare+select. But the value numbering / hashing mechanism used
by EarlyCSE intersects those flags to allow more CSE.
Several alternatives to solve this are discussed in the bug report.
This patch avoids the issue by doing simple matching of min/max/abs
patterns that never requires instruction flags. We give up some CSE
power because of that, but that is not expected to result in much
actual performance difference because InstCombine will canonicalize
these patterns when possible. It even has this comment for abs/nabs:
/// Canonicalize all these variants to 1 pattern.
/// This makes CSE more likely.
(And this patch adds PhaseOrdering tests to verify that the expected
transforms are still happening in the standard optimization
pipelines.
I left this code to use ValueTracking's "flavor" enum values, so we
don't have to change the callers' code. If we decide to go back to
using the ValueTracking call (by changing the hashing algorithm
instead), it should be obvious how to replace this chunk.
emaste [Wed, 18 Mar 2020 20:12:46 +0000 (20:12 +0000)]
invoke _cleanobj_fast_depend_hack unconditionally
Apparently make ${CLEANDIR} is leaving stale entries in .depend files;
for now invoke the hacky cleanup in both the -DNO_CLEAN and normal
(no -DNO_CLEAN) cases.
In collaboration with: dim
Sponsored by: The FreeBSD Foundation
dim [Wed, 18 Mar 2020 18:26:53 +0000 (18:26 +0000)]
Merge llvm, clang, compiler-rt, libc++, libunwind, lld, lldb and openmp llvmorg-10.0.0-rc4-5-g52c365aa9ca. The actual release should follow Real
Soon Now.
freqlabs [Wed, 18 Mar 2020 16:21:11 +0000 (16:21 +0000)]
loader: Add a "kernel.loaded" hook
This hook can be useful, for example to run a local function to choose
different modules to load when a user has picked a different kernel
from the menu.
luporl [Wed, 18 Mar 2020 13:09:47 +0000 (13:09 +0000)]
Add fsck_flags config option
On UFS with SU+J, sometimes fsck's default recovery from journal marks the
filesystem as clean but some errors remain.
With SU only, default fsck in preen mode sometimes thinks all errors have
been fixed but some still remain.
To address the issues above, this change adds a new config option:
fsck_flags. By default it's set to -p, but the user may change it to -y
or -f -y, for instance, to force a full fsck after a system crash.
hrs [Wed, 18 Mar 2020 08:17:24 +0000 (08:17 +0000)]
Remove /etc/host.conf, which was automatically generated from
nsswitch.conf for backward compatibility. This file was used
over 19 years ago, before introducing nsdispatch() in the
name-service lookup APIs.
jhibbits [Wed, 18 Mar 2020 01:09:43 +0000 (01:09 +0000)]
powerpc: Axe PPC4xx support.
Summary:
The support was added almost a decade ago, and never completed. Just axe
it. It was also inadvertently broken 5 years ago, and nobody noticed.
np [Wed, 18 Mar 2020 01:09:40 +0000 (01:09 +0000)]
Remove spurious warning about invalid VPD data.
The warning used to be displayed for valid VPDs about 512B or above in
size. Fix the size check and add a break while here so that the routine
stops if if detects any problem.