imp [Thu, 1 Jan 2015 08:18:08 +0000 (08:18 +0000)]
Move more of the warnings into kern.mk, as well as centralize that as
the place where the C dialect is selected. Have a fairly long list
of newly requires warning suppression for clang 3.5.0, also
centralized in kern.mk. Survive the fallout of the removal of
bsd.sys.mk from bsd.kmod.mk.
ngie [Thu, 1 Jan 2015 04:10:56 +0000 (04:10 +0000)]
Unbreak the build for some modules that depend on -std flags to be passed
through via CFLAGS, as noted by Jenkins when building sparc64 by re-adding
bsd.sys.mk to bsd.kmod.mk:
https://jenkins.freebsd.org/job/FreeBSD_HEAD_sparc64/225/changes
Tested with:
% cd /sys/modules/bce; make obj; make depend; make all
imp [Thu, 1 Jan 2015 02:00:04 +0000 (02:00 +0000)]
Fix module builds on arm (and maybe others) by turning off a whole
raft of new warnings that appear to be on by default in clang 3.5.0.
Fix RPI-B build issues with new clang not liking the ability to pass
arbitrary flags to as, since some flags are more arbitrary (and thus
verboten) than others.
These warnings should be actually fixed in the code, but this is a
band-aide to get things (almost) building again.
np [Wed, 31 Dec 2014 23:19:16 +0000 (23:19 +0000)]
cxgbe(4): major tx rework.
a) Front load as much work as possible in if_transmit, before any driver
lock or software queue has to get involved.
b) Replace buf_ring with a brand new mp_ring (multiproducer ring). This
is specifically for the tx multiqueue model where one of the if_transmit
producer threads becomes the consumer and other producers carry on as
usual. mp_ring is implemented as standalone code and it should be
possible to use it in any driver with tx multiqueue. It also has:
- the ability to enqueue/dequeue multiple items. This might become
significant if packet batching is ever implemented.
- an abdication mechanism to allow a thread to give up writing tx
descriptors and have another if_transmit thread take over. A thread
that's writing tx descriptors can end up doing so for an unbounded
time period if a) there are other if_transmit threads continuously
feeding the sofware queue, and b) the chip keeps up with whatever the
thread is throwing at it.
- accurate statistics about interesting events even when the stats come
at the expense of additional branches/conditional code.
The NIC txq lock is uncontested on the fast path at this point. I've
left it there for synchronization with the control events (interface
up/down, modload/unload).
c) Add support for "type 1" coalescing work request in the normal NIC tx
path. This work request is optimized for frames with a single item in
the DMA gather list. These are very common when forwarding packets.
Note that netmap tx in cxgbe already uses these "type 1" work requests.
d) Do not request automatic cidx updates every 32 descriptors. Instead,
request updates via bits in individual work requests (still every 32
descriptors approximately). Also, request an automatic final update
when the queue idles after activity. This means NIC tx reclaim is still
performed lazily but it will catch up quickly as soon as the queue
idles. This seems to be the best middle ground and I'll probably do
something similar for netmap tx as well.
e) Implement a faster tx path for WRQs (used by TOE tx and control
queues, _not_ by the normal NIC tx). Allow work requests to be written
directly to the hardware descriptor ring if room is available. I will
convert t4_tom and iw_cxgbe modules to this faster style gradually.
adrian [Wed, 31 Dec 2014 22:52:43 +0000 (22:52 +0000)]
Migrate the RSS IPv6 hash code to use pointers to the v6 addresses
rather than passing them in by value.
The eventual aim is to do incremental hash construction rather than
all of the memcpy()'ing into a contiguous buffer for the hash
function, which does show up as taking quite a bit of CPU during
profiling.
Tested:
* a variety of laptops/desktop setups I have, with v6 connectivity
dim [Wed, 31 Dec 2014 20:31:32 +0000 (20:31 +0000)]
Upgrade our copy of clang, llvm and lldb to 3.5.0 release.
Please note that this version now requires C++11 support to build; see
UPDATING for more information.
Release notes for llvm and clang can be found here:
<http://llvm.org/releases/3.5.0/docs/ReleaseNotes.html>
<http://llvm.org/releases/3.5.0/tools/clang/docs/ReleaseNotes.html>
Thanks to Ed Maste, Roman Divacky, Andrew Turner, Justin Hibbits and
Antoine Brodin for their invaluable help with this import.
ian [Wed, 31 Dec 2014 16:15:43 +0000 (16:15 +0000)]
A couple small fixes to make clang 3.5 happy... Move END(sigcode) to the
end of the actual instruction sequence for the function but before some
misc data in the text segment. This eliminates a strange "size must be
constant" error from the integrated assembler. Also, the build_pagetables
function was missing an END(), but really the problem is that it shouldn't
have an ASENTRY() because it's not a function that needs to be a global
symbol with unwind info and all, it's just a little private subroutine
used in very early kernel init.
marius [Wed, 31 Dec 2014 16:06:26 +0000 (16:06 +0000)]
- Switching the mode of Ricoh R5CE823 to SD2.0 causes their PCI device ID
to change to 0xe822, which may be persistent across reboots and, thus,
confuse other OSes. Therefore, restore the original mode and frequency
setting on detach and shutdown.
- Report Ricoh R5CE822 as such.
- According to Linux, Ricoh R5CE822 also need SDHCI_QUIRK_LOWER_FREQUENCY.
- Nuke an unused softc member.
smh [Wed, 31 Dec 2014 04:54:48 +0000 (04:54 +0000)]
Correct zpool list displaying invalid EXPANDSZ for unavailable pool vdevs
When pools are unavailable their vdevs are also unavailable which means
that vdev_max_asize remains at the default zero. This default was being
used to calculate vs_esize resulting in a negative number as vdev_asize >
vdev_max_asize, which caused zpool list -v to display 16.0E for EXPANDSZ
of these vdevs.
smh [Wed, 31 Dec 2014 04:11:29 +0000 (04:11 +0000)]
Use the correct state name for unavailable pools in zpool list
This corrects inconsitencies between zpool list and zpool status which are
both described as displaying the pool <state> however zpool list would use
this hardcoded FAULTED instead of the correct UNAVAIL.
ian [Wed, 31 Dec 2014 02:38:23 +0000 (02:38 +0000)]
Change the order of operations for the initial cache setup. Turning off
the cache before clean/invalidate ensured that no new lines can come into
the cache or migrate between levels during the operation, but may not be
safe on some chips. Instead, if the cache was enabled on entry, do the
wbinv while it's still enabled, and then disable it and do a separate
invalidate pass. After the intitial writeback we know there are no
dirty lines left and no new dirty lines can be created as long as we
carefully avoid touching memory before turning the cache off. Add a
comment about that so no new code gets inserted between those points.
alc [Wed, 31 Dec 2014 00:54:38 +0000 (00:54 +0000)]
The physical memory allocator supports the use of distinct free lists for
managing pages from different address ranges. Generally speaking, this
feature is used to increase the likelihood that physical pages are
available that can meet special DMA requirements or can be accessed through
a limited-coverage direct mapping (e.g., MIPS). However, prior to this
change, the configuration of the free lists was static, i.e., it was
determined at compile time. Consequentally, free lists could be created
for address ranges that held no actual pages, for example, on 32-bit MIPS-
based systems with 512 MB or less of physical memory. This change makes
the creation of the free lists dynamic, i.e., it is based on the available
physical memory at boot time.
On 64-bit x86-based systems with 64 GB or more of physical memory, create
free lists for managing pages with physical addresses below 4 GB. This
change is to address reported problems with initializing devices that
require the allocation of physical pages below 4 GB on some systems with
128 GB or more of physical memory.
neel [Tue, 30 Dec 2014 23:38:31 +0000 (23:38 +0000)]
Initialize all fields of 'struct vm_exception exception' before passing it to
vm_inject_exception(). This fixes the issue that 'exception.cpuid' is
uninitialized when calling 'vm_inject_exception()'.
However, in practice this change is a no-op because vm_inject_exception()
does not use 'exception.cpuid' for anything.
Reported by: Coverity Scan
CID: 1261297
MFC after: 3 days
neel [Tue, 30 Dec 2014 22:19:34 +0000 (22:19 +0000)]
Replace bhyve's minimal RTC emulation with a fully featured one in vmm.ko.
The new RTC emulation supports all interrupt modes: periodic, update ended
and alarm. It is also capable of maintaining the date/time and NVRAM contents
across virtual machine reset. Also, the date/time fields can now be modified
by the guest.
Since bhyve now emulates both the PIT and the RTC there is no need for
"Legacy Replacement Routing" in the HPET so get rid of it.
The RTC device state can be inspected via bhyvectl as follows:
bhyvectl --vm=vm --get-rtc-time
bhyvectl --vm=vm --set-rtc-time=<unix_time_secs>
bhyvectl --vm=vm --rtc-nvram-offset=<offset> --get-rtc-nvram
bhyvectl --vm=vm --rtc-nvram-offset=<offset> --set-rtc-nvram=<value>
emaste [Tue, 30 Dec 2014 22:04:24 +0000 (22:04 +0000)]
Check for multiplication integer overflow in CHECK_EHDR
The initial fix in r276374 is valid only for 64-bit objects. Revert it
and return an error in CHECK_EHDR if the multiplication would overflow.
The original buffer overflow issue was found with the security/afl
fuzzer and has upstream elftoolchain ticket 462. The 32-bit object issue
with r276374 found by antoine@ during an i386 exp-run.
wblock [Tue, 30 Dec 2014 20:20:18 +0000 (20:20 +0000)]
Move the explanation for examples before the example itself. Add
numerous QEMU examples and explanations supplied by sbruno, with thanks
to Ingo Schwarze for help with the mdoc markup. Code, text, testing,
proofreading, cinematography, stunts, and the haunting theme song
supplied by sbruno.
nwhitehorn [Tue, 30 Dec 2014 16:55:53 +0000 (16:55 +0000)]
Fix loader's ability to read the 10.1 release PowerPC ISOs. There appears to
be some kind of problem with the version of makefs used for these disks.
There may be a better way to handle this problem, so I've set the MFC
timer for a fairly long time period.
hselasky [Tue, 30 Dec 2014 09:20:29 +0000 (09:20 +0000)]
Allow systems having a page size greater than 4K to use fewer
scatter-gather XHCI TRB entries for its payload data. The XHCI
controller can handle at least 65536 bytes per scatter-gather list
entry.
hselasky [Tue, 30 Dec 2014 08:33:51 +0000 (08:33 +0000)]
Add missed flushing of data which can happen when "xhci_configure_mask()"
is called from "xhci_configure_reset_endpoint()". Ensure the 3-strikes
error feature is always enabled except for ISOCHRONOUS transfers.
emaste [Tue, 30 Dec 2014 03:25:42 +0000 (03:25 +0000)]
Update elftoolchain to upstream rev 3136
This fixes two strip(1) issues found during ports exp-run and adds a
string hash implementation which significantly speeds up certain
operations on objects with large numbers of symbols.
This also improves libdwarf handling for stripped objects with
.eh_frame or .debug_frame (but not other debug) sections.
ian [Tue, 30 Dec 2014 03:19:46 +0000 (03:19 +0000)]
Add a new locore.S that #includes the right (v4 or v6) implementation.
The kernel build machinery really wants the entry point to be in a file
named locore.S so doing this avoids a bunch of changes to the build system
for relatively little benefit.
ian [Tue, 30 Dec 2014 03:17:53 +0000 (03:17 +0000)]
Rename locore.S to locore-v4.S and add a new locore-v6.S for starting up
armv6/7 systems. We need to use some new armv6/7 features at startup
and splitting the implemenations to separate files will be more maintainable
than adding even more #ifdef sections to locore.S.
Because of the standardized interfaces to cache and MMU operations in armv7,
we can tolerate the kernel being entered with caches enabled. This allows
running u-boot and loader(8) with caches enabled, and the performance
improvement can be dramatic (boot times can be cut from over a minute
to under 30 seconds). The new implementation also has more robust cache
and mmu sequences for launching AP cores, and it paves the way for
upcoming changes to the pmap code which will use the TEX remap feature.
Changes in mp_machdep.c work with the new behavior in locore-v6 mp_entry,
and also reuse the original boot-time page tables to get transitioned
from physical to virtual addressing before installing the normal tables.
Submitted by Svatopluk Kraus and Michal Meloun with some changes by me.
ian [Tue, 30 Dec 2014 02:56:31 +0000 (02:56 +0000)]
Add armv6 implementations of these cache operations to avoid duplicating
the #ifdef stuff at multiple points the functions are called from. Also
rework the armv7 implementations so that the invalidate operations work
from outermost to innermost cache level, and the writeback works from
inner to outer levels.
ian [Tue, 30 Dec 2014 02:51:04 +0000 (02:51 +0000)]
Add arm option ARM_NEW_PMAP, to allow us to begin adding the new pmap
code alongside the existing implementation and quickly toggle between
the two implementations when testing. Once the new code is past its
teething stage we can remove this option.
imp [Tue, 30 Dec 2014 02:39:44 +0000 (02:39 +0000)]
The sn driver doesn't actually include microcode at the current time,
although it may in the future. Or not since the ositech cards that
might need it have become perhaps too rare to support.
marius [Mon, 29 Dec 2014 20:33:34 +0000 (20:33 +0000)]
- No longer exclude malo(4) and mwl(4), they have been fixed in r275870
and r275871 respectively to build with PAE enabled.
- For the PAE kernel configuration file, no longer exclude devices that
are known to be 64-bit DMA clean from amd64.
emaste [Mon, 29 Dec 2014 19:38:12 +0000 (19:38 +0000)]
Rename DT_FEATURE_1 to DT_FEATURE
Track r270303:
This provides a minor cleanup in elfdump; there are otherwise no
consumers in the tree. Old SUN documentation can be found for either
variant, but GNU binutils switched to DT_FEATURE around 2000.
emaste [Mon, 29 Dec 2014 19:16:40 +0000 (19:16 +0000)]
Update elftoolchain to upstream rev 3130
This includes a number of libdwarf improvements (particularly DWARF4
related) and updates to elftoolchain tools such as strip(1). It also
includes a large number of miscellaneous fixes (memory leaks, sign and
cast warnings, integer overflow and underflow, etc.).
This is a merge of r276167,276170-276172 from the
projects/elftoolchain-update-r3130 branch.
dim [Mon, 29 Dec 2014 00:10:43 +0000 (00:10 +0000)]
In contrib/binutils/bfd/elf32-ppc.c, avoid warnings about case values
not being in the enumerated type 'enum elf_ppc_reloc_type', by casting
the switch argument to int.
marius [Sun, 28 Dec 2014 21:36:20 +0000 (21:36 +0000)]
Don't use a sub-device/-vendor wildcard for probing MCS9922 as other
chips with the same device and vendor IDs actually may provide different
functionality. While at it, canonicalize the description to match other
MosChip UARTs.
rmacklem [Sun, 28 Dec 2014 21:13:52 +0000 (21:13 +0000)]
r245508 modified the NFS client's Setattr RPC to
use VA_UTIMES_NULL to indicate whether it should
set the time to the current tod on the server.
This had the side effect of making the NFS client
use the client's timestamp for exclusive create,
starting with FreeBSD9.2.
Unfortunately a bug in some Solaris NFS servers
causes these servers to return NFS_OK to the
Setattr RPC done during exclusive create, but not
actually set the file's mode, leaving the file's
mode == 0.
This patch restores the NFS client's behaviour to
use the server's tod for the exclusive open's
Setattr RPC, to avoid the Solaris server bug and
to restore the pre-FreeBSD9.2 NFS behaviour.
Discussed on: freebsd-fs
PR: 186293
MFC after: 3 months
dim [Sun, 28 Dec 2014 19:55:44 +0000 (19:55 +0000)]
Fix the following -Werror warning from clang 3.5.0, while building
rtld-elf for powerpc 32 bit:
libexec/rtld-elf/powerpc/reloc.c:486:6: error: taking the absolute value of unsigned type 'Elf_Addr' (aka 'unsigned int') has no effect [-Werror,-Wabsolute-value]
if (abs(offset) < 32*1024*1024) { /* inside 32MB? */
^
libexec/rtld-elf/powerpc/reloc.c:486:6: note: remove the call to 'abs' since unsigned values cannot be negative
if (abs(offset) < 32*1024*1024) { /* inside 32MB? */
^~~
1 error generated.
Cast 'offset' to int, since that was intended, and should be safe to do
on architectures with 32-bit two's complement ints.
ian [Sun, 28 Dec 2014 18:38:25 +0000 (18:38 +0000)]
Add cache maintenance functions which will be used by startup code to
initially set up the MMU. Some day they may also be useful as part of
suspend/resume handling, when we get better at power management.
Submitted by: Svatopluk Kraus <onwahe@gmail.com>,
Michal Meloun <meloun@miracle.cz
ian [Sun, 28 Dec 2014 18:19:05 +0000 (18:19 +0000)]
Add new TLB and cache maintainence functions for armv6 and armv7. These
are inline functions that handle all the routine maintenance operations
except the flush-all and invalidate-all routines which are required only
during early kernel init.
These inline functions should be very much faster than the old mechanism
that involved jumping through the big cpufuncs table, especially for
common operations such as invalidating a single TLB entry. Note that
nothing is calling these yet, this just is just required infrastructure
for upcoming changes to the pmap-v6 code.
ian [Sun, 28 Dec 2014 18:12:56 +0000 (18:12 +0000)]
Add new code to read and parse cpu identification data using the new CPUID
mechanism defined for armv7 (and also present on some armv6 chips including
the arm1176 used on rpi). The information is parsed into a global cpuinfo
structure, which will be used by (upcoming) new cache and tlb maintenance
code to handle cpu-specific variations of the maintence sequences.
Submitted by: Svatopluk Kraus <onwahe@gmail.com>,
Michal Meloun <meloun@miracle.cz
jhibbits [Sun, 28 Dec 2014 06:41:13 +0000 (06:41 +0000)]
Add PowerPC64 function descriptor support for dt_link.c
Summary:
PowerPC64 uses function descriptors in a section .opd, exporting the descriptors
to the symbol table. This adds support for these into dt_link.c so that dtrace
USDT probes can be compiled.
Test Plan:
Tested only on powerpc64. No regression testing has been performed, so I want
someone with x86 hardware to regression test this.