ae [Fri, 15 May 2015 12:19:45 +0000 (12:19 +0000)]
Add an ability accept encapsulated packets from different sources by one
gif(4) interface. Add new option "ignore_source" for gif(4) interface.
When it is enabled, gif's encapcheck function requires match only for
packet's destination address.
kib [Fri, 15 May 2015 08:40:17 +0000 (08:40 +0000)]
Some third-party malloc(3) implementations use pthread_setspecific(3)
to handle per-thread information. Since our pthread_setspecific()
implementation calls calloc(3) to allocate per-thread specific data
storage, things get complicated.
Switch the allocator to use bare mmap(2). There is some loss of the
allocated page, since e.g. on amd64, PTHREAD_KEYS_MAX * sizeof(struct
pthread_specific_elem) is 3K (it actually spans whole page due to
padding), but I believe it is more acceptable than additional code for
specialized allocator().
The alternatives would either to make the specific data array be part of
the struct thread, or use internal bindings to call the libc malloc,
avoiding interposing.
Also do the style pass over the thr_spec.c, esp. simplify the
conditionals nesting by returning early when an error detected.
Remove trivial comments.
Found by: yuri@rawbw.com
PR: 200138
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
kib [Fri, 15 May 2015 08:30:29 +0000 (08:30 +0000)]
On amd64, make proc0 pmap initialization slightly more correct. In
particular, switch to the proc0 pmap to have expected %cr3 and PCID
for the thread0 during initialization, and the up to date pm_active
mask.
pmap_pinit0() should be done after proc0->p_vmspace is assigned so
that the amd64 pmap_activate() find the correct curproc pmap.
Sponsored by: The FreeBSD Foundation
MFC after: 3 weeks
kib [Fri, 15 May 2015 07:54:31 +0000 (07:54 +0000)]
Right now, the process' p_boundary_count counter is decremented by the
suspended thread itself, on the return path from
thread_suspend_check(). A consequence is that return from
thread_single_end(SINGLE_BOUNDARY) may leave p_boundary_count
non-zero, it might be even equal to the threads count.
Now, assume that we have two threads in the process, both calling
execve(2). Suppose that the first thread won the race to be the
suspension thread, and that afterward its exec failed for any reason.
After the first thread did thread_single_end(SINGLE_BOUNDARY), second
thread becomes the process suspension thread and checks
p_boundary_count. The non-zero value of the count allows the
suspension loop to finish without actually suspending some threads.
In other words, we enter exec code with some threads not suspended.
Fix this by decrementing p_boundary_count in the
thread_single_end()->thread_unsuspend_one() during marking the thread
as runnable. This way, a return from thread_single_end() guarantees
that the counter is cleared. We do not care whether the unsuspended
thread has a chance to run.
Add some asserts to ensure the state of the process when single
boundary suspension is lifted. Also make thread_unuspend_one()
static.
In collaboration with: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
arybchik [Fri, 15 May 2015 06:49:43 +0000 (06:49 +0000)]
sfxge: do not change CSUM_TSO when IFCAP_TSOx is changed
It is simply not required since the kernel checks corresponding
IFCAP_TSOx capability and CSUM_TSO in hw-assisted offloads.
Note that CSUM_TSO is two bits (CSUM_IP_TSO|CSUM_IP6_TSO) and both bits
are set in IPv4 and IPv6 mbufs.
Reviewed by: gnn
Sponsored by: Solarflare Communications, Inc.
MFC after: 2 days
Differential Revision: https://reviews.freebsd.org/D2546
ganbold [Fri, 15 May 2015 00:39:51 +0000 (00:39 +0000)]
It appears to be armv7_sleep is a duplication of armv7_cpu_sleep.
For consistency with the naming conventions used by the other
implementations kill armv7_sleep and keep armv7_cpu_sleep.
grehan [Thu, 14 May 2015 21:08:48 +0000 (21:08 +0000)]
Bump the size of the blockif scatter-gather list to 67.
The Windows virtio driver ignores the advertized seg_max
field and assumes the host can accept up to 67 segments
in indirect descriptors, triggering an assert in the bhyve
process.
No objection from: mav
Reviewed by: neel
Reported and tested by: Leon Dang (ldang@nahannisys.com)
MFC after: 2 weeks
emaste [Thu, 14 May 2015 19:48:15 +0000 (19:48 +0000)]
Update to ELF Tool Chain r3197
Highlights:
- Fix man page markup, whitespace, and typos
- Fix sh_info of SHT_GROUP section to point to the correct string
- Improve validation in readelf and elfcopy/strip
- Handle DWARF 4's DW_AT_high_pc in addr2line
royger [Thu, 14 May 2015 16:29:11 +0000 (16:29 +0000)]
netfront: wait for backend to connect before sending ARP
Netfront has to wait for the backend to switch to state XenbusStateConnected
before sending the ARP request, or else the backend might not be connected
and thus the packet will be lost.
Sponsored by: Citrix Systems R&D
MFC after: 1 week
pfg [Thu, 14 May 2015 15:49:48 +0000 (15:49 +0000)]
Add new __unreachable() builtin
This is one of the few post gcc4.2 builtins that has been implemented by
clang:
__builtin_unreachable is used to indicate that a specific point in the
program cannot be reached, even if the compiler might otherwise think it
can. This is useful to improve optimization and eliminates certain
warnings.
jonathan [Thu, 14 May 2015 15:14:03 +0000 (15:14 +0000)]
Allow sizeof(cpuset_t) to be queried in capability mode.
This allows functions that retrieve and inspect pthread_attr_t objects to
work correctly: querying the cpuset_t size is part of querying CPU
affinity information, which is part of creating a complete pthread_attr_t.
emaste [Thu, 14 May 2015 14:07:44 +0000 (14:07 +0000)]
Correct language code -- "Danish" is English
The menu entry "Danish ISO-8859-1 (macbook)" was first added to the
syscons(4) INDEX.keymaps in r241851 with no language code, and then in
r256367 incorrectly tagged with "da". It is a Danish keyboard map, but
the description is in English and therefore must be "en".
This error subsequently propagated into the vt(4) INDEX.keymaps.
PR: 146793, 193656
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
fanf [Thu, 14 May 2015 10:31:37 +0000 (10:31 +0000)]
whois: special case certain query suffixes
This extends the existing support for -NORID handles
to include -NICAT, -ARIN, and -RIPE handles.
The suffix machinery is also used to work around a problem with
lack of referrals from the Nominet server for .uk: names under
.ac.uk need to be queried directly at JANET's whois server.
mav [Thu, 14 May 2015 10:04:42 +0000 (10:04 +0000)]
Do not promote large async writes to sync.
Present implementation of large sync writes is too strict and so can be
quite slow. Instead of doing that, execute large async write in chunks,
syncing each chunk separately.
It would be good to fix large sync writes too, but I leave it to somebody
with more skills in this area.
zbb [Wed, 13 May 2015 18:57:03 +0000 (18:57 +0000)]
Add support for ARM GICv3 interrupt controller used in some ARM64 chips
GICv3 allows to distribute interrupts to more than 8 cores served by
the previous GIC revisions. GICv3 introduces additional logic in form
of Re-Distributors associated with particular CPUs to determine
the highest priority interrupts and manage PPIs and LPIs
(Locality-specific Peripheral Interrupts). Interrupts routing is
based on CPUs' affinity numbers. CPU interface was changed to be
accessible via CPU System Registers and this is the preferred
(and supported) method in this driver.
Obtained from: Semihalf
Reviewed by: andrew, emaste, ian, imp
Sponsored by: The FreeBSD Foundation
stas [Wed, 13 May 2015 18:52:18 +0000 (18:52 +0000)]
Fix pmcstat symbol resolution for userland processes.
When examining existing processes pmcstat fails to
correctly determine the locations of executable sections
of the process due to a miscalculated virtual load address.
This does not affect the newly launched processes as the
same value passed as a "start address" to the pmcstat_image_link()
thus nullifying the effect of it. The issue manifests itself
in processes not being reported in the pmcstat(8) output and
"dubious frames" being reported.
Fix it for now by ignoring all the sections except the executable
one. This won't fix the issue for objects with multiple
executable sections but helps in majority of real world usecases.
The real solution would be to modify the MAP-IN event to include
the appropriate load address so pmcstat(8) won't have to manually
parse object files to try to determine it.
pfg [Wed, 13 May 2015 15:26:44 +0000 (15:26 +0000)]
Adjust visibility macros.
The GCC visibility attributes were introduced in GCC 4.0.
Apparently the "protected" attribute was introduced only
until GCC 4.2, but we are not currently using it.
loos [Wed, 13 May 2015 02:25:54 +0000 (02:25 +0000)]
Fix the vmstat -i output on ARM.
The consumers of hw.intrnames expect a NULL byte at end of the string
containing the interrupt names.
On ARM all the interrupt name slots are initialized and this leave no room
for the terminating NULL byte, which makes vmstat read beyond the end of
intrnames.
melifaro [Tue, 12 May 2015 20:42:42 +0000 (20:42 +0000)]
Remove ptei->value check from ipfw_link_table_values():
even if there was non-zero number of restarts, we would unref/clear
all value references and start ipfw_link_table_values() once again
with (mostly) cleared "tei" buffer.
Additionally, ptei->ptv stores only to-be-added values, not existing ones.
This is a forgotten piece of previous value refconting implementation,
and now it is simply incorrect.
adrian [Tue, 12 May 2015 16:55:50 +0000 (16:55 +0000)]
Do not check sequence number for QoS Null frames; set it for generated QoS Null
frames to 0
From IEEE Std. 802.11-2012, 8.3.2.1 "Data frame format", p. 415 (513):
"The Sequence Control field for QoS (+)Null frames is ignored by the receiver
upon reception."
At this moment, any <mode>_input() function interprets them as regular QoS data
frames with TID = 0. As a result, stations, that use another TX sequence for
QoS Null frames (e.g. wpi(4), where (QoS) Null frames are generated by the
firmware), may experience significant packet loss with any other NIC in hostap
mode.
andrew [Tue, 12 May 2015 10:03:14 +0000 (10:03 +0000)]
Teach bits of libc about Thumb. This adds the if-then instructions needed
to handle the ARM conditional execution.
While here fix a bug found by this in the hard-float code, cc is the
opposite of cs. The former is used for 'less than' in floating-point code
and is executed when the C (carry) bit is clear, the latter is used when
greater than, equal, or unordered, and is executed when the C bit is set.
ganbold [Tue, 12 May 2015 08:53:54 +0000 (08:53 +0000)]
Disable WPI in case of aml8726-m3.
The aml8726-m3 SoC is identified as a Cortex A9-r2 rev 4 CPU and
it hangs sometimes during the boot when WFI is used by the kernel.
Differential Revision: https://reviews.freebsd.org/D2473
Submitted by: John Wehle
Suggested by: ian@
tuexen [Tue, 12 May 2015 08:08:16 +0000 (08:08 +0000)]
Ensure that the COOKIE-ACK can be sent over UDP if the COOKIE-ECHO was
received over UDP.
Thanks to Felix Weinrank for makeing me aware of the problem and to
Irene Ruengeler for providing the fix.
ae [Tue, 12 May 2015 07:37:27 +0000 (07:37 +0000)]
Add new socket ioctls SIOC[SG]TUNFIB to set FIB number of encapsulated
packets on tunnel interfaces. Add support of these ioctls to gre(4),
gif(4) and me(4) interfaces. For incoming packets M_SETFIB() should use
if_fib value from ifnet structure, use proper value in gre(4) and me(4).
hrs [Tue, 12 May 2015 03:31:57 +0000 (03:31 +0000)]
- Remove ND6_IFF_IGNORELOOP. This functionality was useless in practice
because a link where looped back NS messages are permanently observed
does not work with either NDP or ARP for IPv4.
neel [Tue, 12 May 2015 00:30:39 +0000 (00:30 +0000)]
Allow configuration of the sector size advertised to the guest.
The default behavior is to infer the logical and physical sector sizes from
the block device backend. However older versions of Windows only work with
specific logical/physical combinations:
- Vista and Windows 7: 512/512
- Windows 7 SP1: 512/512 or 512/4096
For this reason allow the sector size to be specified using the following
block device option: sectorsize=logical[/physical]
gjb [Tue, 12 May 2015 00:27:52 +0000 (00:27 +0000)]
Always use the 'make install' directory as the source for images.
For RE purposes, we use the default (/R within the chroot), so
this helps avoid copying files multiple times and xz(1)-compressing
additional times when not needed.
Again, this Makefile is not for general consumption.
gjb [Mon, 11 May 2015 22:14:03 +0000 (22:14 +0000)]
Add a new file, Makefile.mirrors, which is intended to replace
a 474-line kludge of a shell script to pre-create the directory
hierarchy on ftp-master.
This is not in any way connected to the build, and there is no
intention to do so. This only intent here is to try to make
things a little bit easier for me. But I've probably just made
things worse.
grehan [Mon, 11 May 2015 21:24:10 +0000 (21:24 +0000)]
Handling indirect descriptors is a capability of the host and
not one that needs to be negotiated. Use the host capabilities
field and not the negotiated field when verifying that indirect
descriptors are supported.
Found with the Redhat Windows viostor driver, which clears
the indirect capability in the negotiated caps and then starts
using them.
Reported and tested by: Leon Dang (ldang@nahannisys.com)
MFC after: 2 weeks
jhibbits [Mon, 11 May 2015 20:58:05 +0000 (20:58 +0000)]
Add a PCI bridge for the Freescale PCIe Root Complex
Summary:
The Freescale PCIe Root Complex shows up as a Processor class device, PowerPC
subclass, so the generic PCI code ignores it for a bridge. This adds support
for it.
As part of this, update the Freescale PCI hostbridge driver, to allow probing
beyond the root complex, instead of only allowing "proper" PCI-PCI bridges.
alc [Mon, 11 May 2015 19:55:01 +0000 (19:55 +0000)]
Retire pmap_lazyfix(). This function only existed in the new armv6 pmap
because the i386 pmap on which the new armv6 pmap is based had it, and in
r281707 pmap_lazyfix() was removed from the i386 pmap.
Discussed with: kib
Submitted by: Michal Meloun (via Svatopluk Kraus)
andrew [Mon, 11 May 2015 19:20:30 +0000 (19:20 +0000)]
Add the kernel support for Thumb-2. It is only supported on ARMv7 as the
main ARMv6 target, the Raspberry Pi, doesn't support Thumb-2.
This as been tested with a Thumb-2 userland, however building one is
currently unsupported as there are known toolchain issues breaking some
binaries. Further work will also be needed to decide on the method of
selecting which instruction set to build for, and to benchmark both to
find how building everything as Thumb-2 will affect performance.
thomas [Mon, 11 May 2015 16:45:33 +0000 (16:45 +0000)]
Unbreak MIPS build following rev. 282726
Introduce further adjustments to the renaming of libmd
symbols: make sure that we do not generate dangling weak
aliases, as this causes build failures on MIPS.
andrew [Mon, 11 May 2015 14:10:54 +0000 (14:10 +0000)]
Add ofw_bus_find_compatible to find a compatible ofw node. This will be
used on ARM to help find the correct node to use to start secondary CPUs
as this happens before device enumeration.