adrian [Sat, 7 Nov 2015 04:04:00 +0000 (04:04 +0000)]
Add a sched_yield() to work around low memory conditions in the current code.
Things seem to get stuck in low memory conditions where no bufs are available,
the reclamation path is called to wakeup the daemon, but no sleeping is done.
Because of this, we are stuck in a tight loop in the current process and
never run said reclamation path.
This was introduced in r289279 . This is only a temporary workaround
to restore system usefulness until the more permanent solutions can be
found.
bapt [Sat, 7 Nov 2015 02:18:19 +0000 (02:18 +0000)]
Protecting against rm -rf / is now POSIXLY_CORRECT per posix 1003.1
edition 2013. No need anymore to disable the protection if one set
the POXILY_CORRECT environment variable.
Reviewed by: imp
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D4092
cem [Sat, 7 Nov 2015 01:43:01 +0000 (01:43 +0000)]
Round out SYSCTL macros to the full set of fixed-width types
Add S8, S16, S32, and U32 types; add SYSCTL*() macros for them, as well
as for the existing 64-bit types. (While SYSCTL*QUAD and UQUAD macros
already exist, they do not take the same sort of 'val' parameter that
the other macros do.)
Clean up the documented "types" in the sysctl.9 document. (These are
macros and thus not real types, but the manual page documents intent.)
The sysctl_add_oid(9) arg2 has been bumped from intptr_t to intmax_t to
accommodate 64-bit types on 32-bit pointer architectures.
This is just the kernel support piece; the userspace sysctl(1) support
will follow in a later patch.
Submitted by: Ravi Pokala <rpokala@panasas.com>
Reviewed by: cem
Relnotes: no
Sponsored by: Panasas
Differential Revision: https://reviews.freebsd.org/D4091
adrian [Fri, 6 Nov 2015 23:07:43 +0000 (23:07 +0000)]
[netinet6]: Create a new IPv6 netisr which expects the frames to have been verified.
This is required for fragments and encapsulated data (eg tunneling) to be redistributed
to the RSS bucket based on the eventual IPv6 header and protocol (TCP, UDP, etc) header.
* Add an mbuf tag with the state of IPv6 options parsing before the frame is queued
into the direct dispatch handler;
* Continue processing and complete the frame reception in the correct RSS bucket /
netisr context.
smh [Fri, 6 Nov 2015 20:45:19 +0000 (20:45 +0000)]
Switch zfs_panic_recover to panic for bad DVA
As reported by Coverity a null pointer de-reference panic would be triggered
when zfs_recover was set so switch to straight panic as it can never be
recovered.
skra [Fri, 6 Nov 2015 17:12:33 +0000 (17:12 +0000)]
Make interrupt dispatching MP safe. Use GPU interrupt bit in per-core
interrupt status register to process shared interrupts only if the bit
is active and only on core to which they are routed.
andrew [Fri, 6 Nov 2015 14:36:21 +0000 (14:36 +0000)]
Mark the arm64 nexus devices to be attached early. This allows multipass
to work correctly. Without it the pass quickly moves to the final pass
before the nexus device attaches.
tuexen [Fri, 6 Nov 2015 13:08:16 +0000 (13:08 +0000)]
Fix typos in field names of struct sctp_extrcvinfo.
Provide defines to allow applications to compile.
Thanks to Jens Hoelscher for making me aware of the typos.
hselasky [Fri, 6 Nov 2015 12:54:27 +0000 (12:54 +0000)]
Fix for unaligned IP-header.
The mbuf length fields must be set before m_adj() is called else
m_adj() will not always adjust the mbuf and an unaligned read
exception can trigger inside the network stack. This can happen on
platforms where unaligned reads are not supported. Adjust a length
check to include the 2-byte ethernet alignment while at it.
avos [Fri, 6 Nov 2015 11:29:24 +0000 (11:29 +0000)]
urtwn(4): simplify urtwn_tsf_sync_enable().
- Drop TSF initialization; device can discover it without our help.
- Do not touch R92C_BCN_CTRL_EN_BCN bit in STA mode.
- Add 'static' keyword for function definition.
Tested with RTL8188EU, STA mode.
Reviewed by: kevlo
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D3801
jhibbits [Fri, 6 Nov 2015 04:56:52 +0000 (04:56 +0000)]
Write 2- and 4-byte aligned values as single writes in ddb(4)
On the mpc85xx SoC family, writes to any part of a word in the CCSR affect the
full word. This prevents single-byte writes from taking the desired effect.
bdrewery [Fri, 6 Nov 2015 04:45:29 +0000 (04:45 +0000)]
Add a FAST_DEPEND option, off by default, which speeds up the build significantly.
This speeds up buildworld by 16% on my system and buildkernel by 35%.
Rather than calling mkdep(1), which is just a wrapper around 'cc -E',
use the modern -MD -MT -MF flags to gather and generate dependencies during
compilation. This flag was introduced in GCC "a long time ago", in GCC 3.0,
and is also supported by Clang. (It appears that ICC also supports this but I
do not have access to test it). This avoids running the preprocessor *twice*
for every build, in both 'make depend' and 'make all'. This is especially
noticeable when using ccache since it does not cache preprocessor results from
mkdep(1) / 'cc -E', but still speeds up compilation with the -MD flags.
For 'make depend' a tree-walk is still done to ensure that all DPSRCS
are generated when expected, and that beforedepend/afterdepend and
_EXTRADEPEND are all still respected. In time this may change but for now
I've been conservative. The time for a tree-walk with -j combined with
SUBDIR_PARALLEL is not significant. For example, it takes about 9 seconds
with -j15 to walk all of src/ for 'make depend' now on my system.
A .depend file is still generated with the various rules that apply to
the final target, or custom rules. Otherwise there are now
per-built-object-file .depend files, such as .depend.filename.o. These
are included directly by make rather than populating .depend with a loop
and .depend lines, which only added overhead to the now almost-NOP 'make
depend' phase.
Before this I experimented with having mkdep(1) called in parallel per-file.
While this improved the kernel and lib/libc 'make depend' phase, it resulted
in slower build times overall.
The -M flags are removed from CFLAGS when linking since they have no effect.
Enabling this by default, for src or out-of-src, can be done once more testing
has been done, such as a ports exp-run, and with more compilers.
The system I used for testing was:
WITNESS
Build options: -j20 WITH_LLDB=yes WITH_DEBUG_FILES=yes WITH_FAST_DEPEND=yes
DISK: ZFS 3-way mirror with very slow disks using SSD l2arc/log.
The arc was fully populated with src tree files.
RAM: 76GiB
CPU: Intel(R) Xeon(R) CPU L5520 @2.27GHz
2 package(s) x 4 core(s) x 2 SMT threads = hw.ncpu=16
buildworld:
x buildworld-before
+ buildworld-fastdep
+-------------------------------------------------------------------------------+
|+ |
|+ |
|+ xx x|
| |_MA___||
|A |
+-------------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 3 3744.13 3794.31 3752.25 3763.5633 26.935139
+ 3 3153.34 3155.16 3154.2 3154.2333 0.91045776
Difference at 95.0% confidence
-609.33 +/- 43.1943
-16.1902% +/- 1.1477%
(Student's t, pooled s = 19.0569)
buildkernel:
x buildkernel-before
+ buildkernel-fastdep
+-------------------------------------------------------------------------------+
|+ x |
|++ xx|
| A||
|A| |
+-------------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 3 571.57 573.94 571.79 572.43333 1.3094401
+ 3 369.12 370.57 369.3 369.66333 0.79033748
Difference at 95.0% confidence
-202.77 +/- 2.45131
-35.4225% +/- 0.428227%
(Student's t, pooled s = 1.0815)
avos [Thu, 5 Nov 2015 22:44:36 +0000 (22:44 +0000)]
iwn(4): various simple fixes
- Fix mbuf leaks in iwn_raw_xmit() and iwn_xmit_task()
(regression since r288178).
- Check IWN_FLAG_RUNNING flag under lock.
- Remove m->m_pkthdr.rcvif initialization (fixed in r283994).
- Enclose some values in return statements into parentheses.
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D4069
bdrewery [Thu, 5 Nov 2015 22:09:00 +0000 (22:09 +0000)]
Allow 'make buildenv' to default to the caller's shell by using SHELL.
Also pass BUILDENV=1 into the sub-shell to allow modifying PS1 in .profile such
as:
if [ -n "${BUILDENV}" ]; then
PS1="(buildenv) ${PS1}"
fi
SHELL defaults to 'sh' in share/mk/sys.mk, but is typically passed down by
the shell invoking make as well. Rather than forcing all 'buildenv' users
to use plain /bin/sh, let them use their favorite shell.
imp [Thu, 5 Nov 2015 21:48:12 +0000 (21:48 +0000)]
Fix CC being wrong during install* targets.
Move CROSS_TOOLS stuff to top of file (before bsd.compiler.mk) so that
decisions made by bsd.compiler.mk can properly affect the defaults in
src.opts.mk. Move that to after bsd.compiler.mk. Add a comment about
why we include bsd.compiler.mk here despite the fact that src.opts.mk
currently does too. Also remove bsd.arch.inc.mk that's been OBE.
jhb [Thu, 5 Nov 2015 21:33:15 +0000 (21:33 +0000)]
Chelsio T5 chips do not properly echo the No Snoop and Relaxed Ordering
attributes when replying to a TLP from a Root Port. As a workaround,
disable No Snoop and Relaxed Ordering in the Root Port of each T5 adapter
during attach so that CPU-initiated requests do not contain these flags.
Note that this affects CPU-initiated requests to all devices under this
root port.
jhb [Thu, 5 Nov 2015 21:26:06 +0000 (21:26 +0000)]
Add helper routines for PCI device drivers to read, write, and modify
PCI-Express capability registers (that is, PCI config registers in the
standard PCI config space belonging to the PCI-Express capability
register set).
Note that all of the current PCI-e registers are either 16 or 32-bits,
so only widths of 2 or 4 bytes are supported.
jhb [Thu, 5 Nov 2015 19:52:18 +0000 (19:52 +0000)]
Start a cleanup of the MAINTAINERS file. This adds an "OLD" line at the
top of the list. Any entries in the list that are still below this line
after December 5th will be removed.
lidl [Thu, 5 Nov 2015 17:37:14 +0000 (17:37 +0000)]
Restrict 520.pfdenied to only list rules that blocked traffic.
Before this change, the 520.pfdenied script listed all rules that
matched /^block/ in the rule. Restrict the printed output to only
those rules that result in packets being dropped.
andrew [Thu, 5 Nov 2015 16:55:27 +0000 (16:55 +0000)]
Fix the open solaris atomic functions on arm64. Without this we may use the
wrong value in the comparison, leading to incorrectly setting the new
value.
This has been observed in the ZFS code. Without this we can lose track of
the reference count in a zrlock object.
We should move to use the generic atomic functions, however as this has
been observed I would prefer to have this working, then move to the generic
functions.
royger [Thu, 5 Nov 2015 14:33:46 +0000 (14:33 +0000)]
xen/intr: fix the event channel enabled per-cpu mask
Fix two issues with the current event channel code, first ENABLED_SETSIZE is
not correctly defined and then using a BITSET to store the per-cpu masks is
not portable to other arches, since on arm32 the event channel arrays shared
with the hypervisor are of type uint64_t and not long. Partially restore the
previous code but switch the bit operations to use the recently introduced
xen_{set/clear/test}_bit versions.
gnn [Thu, 5 Nov 2015 07:26:32 +0000 (07:26 +0000)]
Replace the fastforward path with tryforward which does not require a
sysctl and will always be on. The former split between default and
fast forwarding is removed by this commit while preserving the ability
to use all network stack features.
gonzo [Thu, 5 Nov 2015 03:46:54 +0000 (03:46 +0000)]
Refactor mailbox property API to make it usable for /dev/vcio driver:
- Add bcm2835_mbox_property for generic property request, it accepts
pointer to prepared property chan message and its size, forwards
it to MBOX and copies result back
- Make all bcm2835_mbox_XXX functions that use property channel go
through bcm2835_mbox_property path. Do not accept device_t as
an argument, it's not required: all DMA operatiosn should go
through mbox device, and all API consumers should report errors
on their side.
projects / people / liuw / freebsd.git / log