- Move prototypes for various functions into out of C files and into
<machine/md_var.h>.
- Move some CPU-related variables out of i386/i386/identcpu.c to
initcpu.c to match amd64.
- Move the declaration of has_f00f_hack out of identcpu.c to machdep.c.
- Remove a misleading comment from i386/i386/initcpu.c (locore zeros
the BSS before it calls identify_cpu()) and remove explicit zero
assignments to reduce the diff with amd64.
Do not blindly announce 1000baseT half-duplex capability in
autonegotiation. Some controllers like cgem(4) do not support
half-duplex at gigabit speeds.
Create a /boot/dtb directory to house DTB blobs. The flattened device
tree support includes a device tree source compiler dtc(8) which
converts .dts files into .dtb files. /boot/loader will load dtb files
from this directory by default, allowing for fewer differences between
images for different SoCs. Compiled dts files will wind up here
eventually as an alternative to embedding them into the kernel.
Document this in hier(7), as well as add missing entries for
/boot/firmware and /boot/zfs, though the latter two should only be
considered place holders if someone wants to make them better.
ian [Wed, 3 Sep 2014 19:37:41 +0000 (19:37 +0000)]
Remove DIAGNOSTIC from the kernel config of low-end arm systems. Sanity
checks such as vmem_check() can make a low-end system go completely
unresponsive for as much as 3 seconds out of every 10.
Always seek back to the beginning of a regular directory, even if the
previous seek location was 0. Without this, readdir() would see
dd_loc of zero and call getdirentries() which would start reading
entries at the current seek location of the directory ignoring the
first batch of entries. Also, rewinddir() should always seek so that
it reads the directory from the beginning to get updated entries.
PR: 192935
Reported by: iron@mail.ua
MFC after: 3 days
Use of "find ." resulted in METALOG entries with an extra ./ -- e.g.,
./usr/share/zoneinfo/./America/Toronto. Avoid this by using globbing
via "find *" instead.
ian [Wed, 3 Sep 2014 17:51:03 +0000 (17:51 +0000)]
Fixes and enhancements for the if_cgem driver...
- miibus fixes as suggested by Yonghyeon Pyun.
- enable VLAN MTU support.
- fix a few WITNESS complaints in cgem_attach().
- have cgem_attach() properly init the ifnet struct before calling
mii_attach() to fix panic when using e1000phy.
- fix ethernet address changing.
- fix transmit queue overflow handling.
- tweak receive queue handling to reduce receive overflows.
- bring out MAC statistic counters to sysctls.
- add e1000phy to config file.
- implement receive hang work-around described in reference guide.
- change device name from if_cgem to cgem to be consistent with other
interfaces.
Import a hackish tool I use to examine the skew of the TSC across
CPUs in a system. The tool queries the kernel for its set of CPUs
and compares TSC values on each of the additional CPUs to the first
CPU in turn. It then outputs a table of simple statistics.
Do not direct commit to contrib/llvm. Make the change a patch file instead.
Reverts 271025 but still functionally patches it. Original intent is still
the same. Pointed out by rdivacky.
MFV: Only emit movw on ARMv6T2
Building for the FreeBSD default target ARMv6 was emitting movw ASM on certain
test cases (found building qmake4/5 for ARM). Don't do that, moreover, the AS
in base doesn't understand this instruction for this target. One would need
to use --integrated-as to get this to build if desired.
Building for the FreeBSD default target ARMv6 was emitting movw ASM on certain
test cases (found building qmake4/5 for ARM). Don't do that, moreover, the AS
in base doesn't understand this instruction for this target. One would need
to use --integrated-as to get this to build if desired.
* Unconditionally turn on SIOCGI2C probing for all interfaces
on "ifconfig -v". I've seen no measurable timing difference
for doing additional SIOCGI2C call for system with 4k vlans.
* Determine appropriate handler (SFP/QSFP) by reading identification
byte (which is the same for both SFF-8472 and SFF-8436) instead
of checking driver name.
Leave the C11 keywords alone when we have a recent version of GCC.
As GCC also gained support for the C11 keywords over time, we can patch
up <sys/cdefs.h> to not define these anymore. This has the advantage
that error messages for static assertions are printed natively and that
_Alignas() will work with even a type outside of C11 mode.
All C11 keywords are supported with GCC 4.7 and higher, with the
exception of _Thread_local and _Generic. These are only supported as of
GCC 4.9.
Right now, thread_single(SINGLE_EXIT) returns after the p_numthreads
reaches 1. The p_numthreads counter is decremented in thread_exit() by
a call to thread_unlink(). This means that the exiting threads may
still execute on other CPUs when thread_single(SINGLE_EXIT) returns.
As result, vmspace could be destroyed while paging structures are
still used on other CPUs by exiting threads.
Delay the return from thread_single(SINGLE_EXIT) until all threads are
really destroyed by thread_stash() after the last switch out. The
p_exitthreads counter already provides the required mechanism, move
the wait from the thread_wait() (which is called from wait(2) code)
into thread_single().
Reported by: many (as "panic: pmap active <addr>")
Reviewed by: alc, jhb
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Partially revert r270964. Don't test for C++11 to define _Thread_local.
In addition to Clang 3.3, it turns out that GCC 4.7 in Ports also does
not support the _Thread_local keyword. Let's document this in a bit more
detail.
Invoke make_dtb with MACHINE defined for enhanced cross building
friendliness. This should restore old-fashioned kernel building in a
cross environment, though this has only had limited testing.
Define _DTRACE_VERSION in sdt.h rather than setting it manually. This is
similar to what illumos does, and makes it easier to enable USDT probes
in third-party software that doesn't make use of the system makefiles.
Allow standalone debug for non-default ${PROG} targets
This allows WITH_DEBUG_FILES to produce standalone debug for the ELF
runtime linker.
We previously disabled standalone debug files for bsd.prog.mk consumers
that included a non-default ${PROG} target, but this is not required.
Consumers that do not support standalone debug are still handled by
disabling it for statically linked binaries, and for those that specify
a non-default binary format.
Fix for WITNESS warning while doing xpt_rescan.
This happen when converting any JBOD to RAID or creating
any new RAID from Unconfigured Drives.
Without this fix, user may see below call trace if WITNESS is enabled.
witness_warn() at witness_warn+0x4b5/frame 0xfffffe011f929a00
uma_zalloc_arg() at uma_zalloc_arg+0x3b/frame 0xfffffe011f929a70
malloc() at malloc+0x192/frame 0xfffffe011f929ac0
mrsas_bus_scan_sim() at mrsas_bus_scan_sim+0x32/frame 0xfffffe011f929af0
mrsas_aen_handler() at mrsas_aen_handler+0x11c/frame 0xfffffe011f929b20
taskqueue_run_locked() at taskqueue_run_locked+0xf0/frame 0xfffffe011f929b80
taskqueue_thread_loop() at taskqueue_thread_loop+0x9b/frame 0xfffffe011f929bb0
fork_exit() at fork_exit+0x84/frame 0xfffffe011f929bf0
fork_trampoline() at fork_trampoline+0xe/frame 0xfffffe011f929bf0
Submitted by: kadesai
Reviewed by: ambrisko
MFC after: 3 days
The proper compatibility string for the AHCI device is
allwinner,sun4i-ahci, so use that instead of the older string which
had become FreeBSD specific.
- Remove c++0x hack from <sys/cdefs.h> that was needed when Clang did
not fully implement C++11. We can now safely test against C++11 to
check whether thread_local is available, like we do for all other
C++11 keywords.
- Don't use __clang__ to test for thread safety annotation presence. It
turns out we have a proper attribute for this.
ian [Tue, 2 Sep 2014 03:23:05 +0000 (03:23 +0000)]
Create an interface for drivers to enable or disable their clocks as listed
in the clocks=<...> properties of their FDT data. The clock properties
consist of 2-cell tuples, each containing a clock device node reference and
a clock number. A clock device driver can register itself as providing
this interface, then other drivers can turn the FDT clock node reference
into the corresponding device_t so that they can use the interface to query
and manipulate their clocks.
This provides convenience functions to enable or disable all the clocks
listed in the properties for a device, so most drivers will be able to
manage their clocks with a single call to fdt_clock_enable_all(dev).
ian [Tue, 2 Sep 2014 02:56:43 +0000 (02:56 +0000)]
The anatop driver controls power and PLLs and needs to be available before
just about anything else (other than octop which provides it config info),
so make it an EARLY_DRIVER_MODULE().
ian [Tue, 2 Sep 2014 02:54:55 +0000 (02:54 +0000)]
The ocotp driver provides access to registers containing chip configuration
data that is needed by other drivers, so make it an EARLY_DRIVER_MODULE()
that loads before just about anything else.
Fix a bug where command line arguments could be misprocessed if getopts is
used prior to f_dialog_init() -- e.g., in a script that sets
DIALOG_SELF_INITIALIZE to NULL, preventing f_dialog_init() from being run
automaticaly when `dialog.subr' is included. Caused by sub-shell processing
of arguments inheriting prior value of $OPTIND, used by getopts. Solved by
unsetting OPTIND prior to [re-]processing of positional arguments.
ian [Mon, 1 Sep 2014 22:40:31 +0000 (22:40 +0000)]
Create a mechanism for looking up a device_t associated with an ofw/fdt
xref handle, and for registering that association. Also use the same data
for faster translations between node and xref handles.
Now when fdt properties contain &othernode references, a driver can find
the device instance that corresponds to &othernode, and thus can use
interfaces provided by that instance.
Add bsearch_b to the libc map and the stdlib header.
bsearch_b is the Apple blocks enabled version of bsearch(3).
This was added to libc in Revision 264042 but the commit
missed the declaration required to make use of it.
While here move some other block-related functions to the
BSD_VISIBLE block as these are non-standard.
While there, add a NO_WTHREAD_SAFETY flag that can be used to disable
this specific warning flag. Disable it for auditdistd. We can easily
patch up auditdistd to have the right annotations to build, but as
auditdistd is intended to be portable across other operating systems,
it's not worth the effort.
According to the ISO C standard, lgamma(-integer) returns
inf and raises the divided-by-zero exception. Compilers
constant fold one/zero to inf but do not raise the exception.
Introduce a volatile vzero to prevent the constant folding.
Move the declaration of zero into the main declaration block.
Add lock annotations to the header files of our threading libraries.
This change extends all of the functions present in the <pthread.h> and
<threads.h> headers to have lock annotations. This will allow Clang to
warn about the following:
- Locking a function twice,
- Unlocking a function without a mutex being locked,
- Forgetting to unlock a mutex before returning,
- Destroying or reinitializing a mutex that is currenty locked,
- Using an unlocked mutex in combination with a condition variable.
Enabling these annotations already allowed me to catch a bug in one of
our userspace tools (r270749).
Clang has support for annotating mutexes and code that uses mutexes to
validate certain aspects of thread safety:
- Whether acquiring/releasing locks is done properly (e.g., whether you
unlock a mutex before leaving a function).
- Whether a lock is held while reading/writing data from/to memory.
Analysis is performed at the function level. Functions can be annotated
to indicate they:
- (try to) pick up a lock,
- release a lock,
- can only be called when (not) holding a lock,
- assert that a lock is held.
Variables and structure members can be annotated to indicate that they
are guarded by a certain lock. In C++, these annotations can refer
to both global variables, but also other class/structure members. In C,
it is only possible to refer to global variables.
This change adds wrappers for the annotations used by Clang to
<sys/cdefs.h>. They currently have no effect, but this is on purpose.
This change will be merged back to FreeBSD 9 and 10, which means we can
safely experiment with these annotations on HEAD without making it
harder to port changes back.
Reviewed by: announced on arch@ and toolchain@
MFC after: 3 weeks
Add vt(4) support to the console initialisation script, specifically:
- Identify the console driver used and print syscons or vt as appropriate.
- If vt is used and a keymap could not be loaded, then try to replace the
keymap name configured in rc.conf based on a replacement list in this
script. Warn about the fact, that a syscons keyname is configured and
report the replacement used under vt.
- If no replacement keymap is found, no keymap is loaded and a warning
is displayed, which points at the conversion script and allows the
conversion of keymaps not part of the official distribution.
This patch has been sent to the -hackers list for review, but no comment
has been received, yet. It is tested to work under syscons and vt on my
system (on vt with either the syscons or vt keymap file name in rc.conf).
ian [Mon, 1 Sep 2014 14:57:04 +0000 (14:57 +0000)]
Do not generate unwind info in asm functions if _STANDALONE is defined.
The .fnend op causes the assembler to emit RELOC references to unwind
support functions that don't exist in libstand.
Make SOCK_RAW sockets to be truly raw, not modifying received and sent
packets at all. Swapping byte order on SOCK_RAW was actually a bug, an
artifact from the BSD network stack, that used to convert a packet to
native byte order once it is received by kernel.
Other operating systems didn't follow this, and later other BSD
descendants fixed this, leaving us alone with the bug. Now it is
clear that we should fix the bug.
kargl [Sun, 31 Aug 2014 21:38:03 +0000 (21:38 +0000)]
Compute sin(pi*x) without actually doing the pi*x multiplication.
sin_pi(x) is only called for x < 0 and |x| < 2**(p-1) where p is
the precision of x. The new argument reduction is an optimization
compared to the old code, and it removes a chunk of dead code.
Accuracy tests in the intervals (-21,-20), (-20,-19), ... (-1,0)
show no differences between the old and new code.
marius [Sun, 31 Aug 2014 17:56:54 +0000 (17:56 +0000)]
- Nuke unused sdhci_softc.
- Static'ize sdhci_debug local to sdhci.c.
- Const'ify PCI device description strings.
- Nuke redundant resource ID members from sdhci_pci_softc.
- Nuke unused hw.sdhci_pci.debug tunable.
- Add support for using MSI instead of INTx, controllable via the tunable
hw.sdhci.enable_msi (defaulting to on) and tested with a RICOH R5CE823 SD
controller.
- Use NULL instead of 0 for pointers.
ian [Sun, 31 Aug 2014 17:21:51 +0000 (17:21 +0000)]
In ARM asm code, ensure that every ENTRY(foo) has a matching END(foo).
The EABI unwind info requires a .fnend for every .fnstart, and newer
binutils will complain about seeing two .fnstart in a row. This change
allows newer tools to compile our code.
glebius [Sun, 31 Aug 2014 06:46:21 +0000 (06:46 +0000)]
o Remove struct if_data from struct ifnet. Now it is merely API structure
for route(4) socket and ifmib(4) sysctl.
o Move fields from if_data to ifnet, but keep all statistic counters
separate, since they should disappear later.
o Provide function if_data_copy() to fill if_data, utilize it in routing
socket and ifmib handler.
o Provide overridable ifnet(9) method to fetch counters. If no provided,
if_get_counters_compat() would be used, that returns old counters.
glebius [Sun, 31 Aug 2014 06:23:54 +0000 (06:23 +0000)]
Remove ability to write to struct if_data residing in struct ifnet
via net.link.generic.IFMIB_IFDATA.*.IFDATA_GENERAL sysctl. Reasons
for removal are:
- No code in tree uses this possibility.
- The documentation ifmib(4) doesn't say that such possibility
exist. The example provided in manual page only reads data.
- On many interfaces the feature simply doesn't work, since they
do accounting in hardware, and overwrite if_data on tick.
ian [Sat, 30 Aug 2014 22:21:57 +0000 (22:21 +0000)]
Fix the handling of MMU type in the AP entry code. The ARM_MMU_V6/V7
symbols are always #defined to 0 or 1, so use #if SYM not #if defined(SYM).
Also, it helps if you include the header file that defines the symbols.
neel [Sat, 30 Aug 2014 19:59:42 +0000 (19:59 +0000)]
The "SUB" instruction used in getcc() actually does 'x -= y' so use the
proper constraint for 'x'. The "+r" constraint indicates that 'x' is an
input and output register operand.
While here generate code for different variants of getcc() using a macro
GETCC(sz) where 'sz' indicates the operand size.
Update the status bits in %rflags when emulating AND and OR opcodes.
neel [Sat, 30 Aug 2014 18:35:16 +0000 (18:35 +0000)]
Set the 'inst_length' to '0' early on before any error conditions are detected
in the emulation of the task switch. If any exceptions are triggered then the
guest %rip should point to instruction that caused the task switch as opposed
to the one after it.