Warner Losh [Sun, 3 May 2015 22:51:29 +0000 (22:51 +0000)]
When merging the floating point type attribute, and reporting an error
when things don't match, report which file has them and which one
doesn't correctly.
Adrian Chadd [Sun, 3 May 2015 22:28:42 +0000 (22:28 +0000)]
Remove old iv_bss entry from the node table
This may happen on RUN -> SCAN -> RUN -> SCAN state transition:
1. RUN -> SCAN: in ieee80211_sta_join1(): iv_bss will be moved to obss,
refcnt will be reduced by 2 (default minimum).
Now, if old iv_bss have some extra references (for example, from
unacknowledged probe responses), it will not be freed and will stay
in the node table.
2. SCAN -> RUN.
3. If old iv_bss will not be deleted by the time when the next RUN -> SCAN
state transition occurs, then sta_leave() will reduce it's reference
counter once more. As a result, two last users will free it -> this will
lead to kernel panic.
In this patch old iv_bss entry is explicitly removed from the node table in
ieee80211_sta_join1() (as a result, it will not be processed by sta_leave()).
Use the BCM2835_MBOX_CHAN_PROP mbox channel to setup the framebuffer,
remove DMA code (its now done in bcm2835_mbox.c).
Also adjust the color palette when bcm2708_fb.fbswap is set. The
firmware used on RPi 2 uses this mode.
Tested on: RPi-B and RPi 2 with 16, 24 and 32bpp
Note: The 32bpp mode on RPi-B has the red and blue swapped, this
is a know problem (not a driver problem).
Add the routines to query and setup the framebuffer state using the
BCM2835_MBOX_CHAN_PROP channel. The old channel (BCM2835_MBOX_CHAN_FB)
seems deprecated on recent firmware versions and is causing a freeze on
RPi 2.
The actual changes in the framebuffer drivers will follow in subsequent
commits.
Gleb Smirnoff [Sat, 2 May 2015 20:37:40 +0000 (20:37 +0000)]
After r281643 an #ifdef IFT_FOO preprocessor directive returns false,
since types became a enum C type. Some software uses such ifdefs to
determine whether an operating systems supports certain interface type.
Of course, such check is bogus. E.g. FreeBSD defines about 250 interface
types, but supports only around 20.
However, we need not upset such software so provide a set of defines. The
current set was taken to suffice the dhcpd.
Gleb Smirnoff [Sat, 2 May 2015 20:27:37 +0000 (20:27 +0000)]
Instead of reading, validating and adjusting value of the vm.swap_async_max
in the main swapper work cycle, do it in the sysctl handler. This removes
extra mutex acquisition from the main cycle and makes the sysctl knob return
error on an invalid value, instead of accepting and fixing it.
Reviewed by: kib
Sponsored by: Netflix
Sponsored by: Nginx, Inc.
Approved, oprócz użycie RESTORE_ERRNO() do ustawiania errno.
Change the nvlist_recv() function to take additional argument that
specifies flags expected on the received nvlist. Receiving a nvlist with
different set of flags than the ones we expect might lead to undefined
behaviour, which might be potentially dangerous.
Update consumers of this and related functions and update the tests.
Approved by: pjd (mentor)
Update man page for nvlist_unpack, nvlist_recv, nvlist_xfer, cap_recv_nvlist
and cap_xfer_nvlist.
Bjoern A. Zeeb [Sat, 2 May 2015 08:31:16 +0000 (08:31 +0000)]
Fix an off-by-one bug in string/array handling which lead to memory overwrite
and follow-up assertion errors on at least ARM after r282257,
with nvp_magic being 0x6e7600:
Assertion failed: ((nvp)->nvp_magic == 0x6e7670), function nvpair_name, file .../subr_nvpair.c, line 713.
Mark Johnston [Sat, 2 May 2015 00:26:38 +0000 (00:26 +0000)]
Don't drop the idr lock before verifying that the newly-inserted element
is present in the tree. Otherwise there exists a window during which the
element could be removed by another thread, triggering an incorrect
assertion failure.
Andreas Tobler [Fri, 1 May 2015 20:35:09 +0000 (20:35 +0000)]
Fix the loader for sparc64.
The commit from r279799 missed the adaptation of md_load in
sparc64/loader/main.c. This resulted in Memory Address Align issues at the
early boot stage.
Add a state transition call to show that we have entered TIME_WAIT.
Although this is not important to the rest of the TCP processing
it is a conveneint way to make the DTrace state-transition probe
catch this important state change.
Ed Maste [Thu, 30 Apr 2015 23:48:05 +0000 (23:48 +0000)]
crunchide: Restore local EM_AARCH64 constant for bootstrapping
Most of the EM_* constants are available in all supported host branches,
but EM_AARCH64 was added relatively recently. Add it back to fix
building HEAD on 10.x.
Rename macros to use prefix ERRNO. Add macro ERRNO_SET. Now
ERRNO_{RESTORE/SAVE} must by used together, additional variable is not
needed. Always use ERRNO_{SAVE/RESTORE/SET} macros.
John Baldwin [Thu, 30 Apr 2015 18:23:38 +0000 (18:23 +0000)]
Various fixes to the stats in igb(4), ixgbe(4), and ixl(4).
- Use hardware counters for ifnet stats in igb(4) when possible. This
ensures these stats include packets that bypass the regular stack via
netmap.
- Don't derefence values off the end of the igb(4) VF stats structure.
Instead, add a dedicated if_get_counter method for igb(4) VF interfaces.
- Report missed packets on igb(4) as input queue drops rather than an
input error.
- Report bug_ring drop counts as output queue drops for igb(4) and ixgbe(4).
- Export the buf_ring drop stats for individual rings via sysctl on
ixgbe(4).
- Fix a typo that in ixl(4) that caused output queue drops to be reported
as input queue drops and input queue drops to be unreported.
John Baldwin [Thu, 30 Apr 2015 15:48:48 +0000 (15:48 +0000)]
Remove support for Xen PV domU kernels. Support for HVM domU kernels
remains. Xen is planning to phase out support for PV upstream since it
is harder to maintain and has more overhead. Modern x86 CPUs include
virtualization extensions that support HVM guests instead of PV guests.
In addition, the PV code was i386 only and not as well maintained recently
as the HVM code.
- Remove the i386-only NATIVE option that was used to disable certain
components for PV kernels. These components are now standard as they
are on amd64.
- Remove !XENHVM bits from PV drivers.
- Remove various shims required for XEN (e.g. PT_UPDATES_FLUSH, LOAD_CR3,
etc.)
- Remove duplicate copy of <xen/features.h>.
- Remove unused, i386-only xenstored.h.
Differential Revision: https://reviews.freebsd.org/D2362
Reviewed by: royger
Tested by: royger (i386/amd64 HVM domU and amd64 PVH dom0)
Relnotes: yes
Much of the code was common to begin with. There is one nit, which is likely
not an issue at all. With the old code, the AIM machdep would __syncicache()
the entire kernel core at setup. However, in the unified setup, that seems to
hang on the MPC7455, perhaps because it's running later than before. Removing
this allows it to boot just fine. Examining the code, the FreeBSD loader
already does syncicache of the full kernel, and each module loaded, so this
doesn't appear to be an actual problem.
Phabricator: enable "history.immutable":
With certain arc workflows leaving history.immutable as false
results in using the upstream template instead of our usual
commit template. Since the git workflow issues alluded to in my
prior commit message can be worked around, set history.immutable
once again.
Sean Bruno [Thu, 30 Apr 2015 00:17:32 +0000 (00:17 +0000)]
Fix native-xtools breakage when building a gcc enabled target on a clang
enabled host. Build a one-off gperf and put it in the PATH for the rest
of the target so the ONE call to gperf by the gcc build picks it up and
DTRT.
Neel Natu [Wed, 29 Apr 2015 23:44:28 +0000 (23:44 +0000)]
Re-implement RTC current time calculation to eliminate the possibility of
losing time.
The problem with the earlier implementation was that the uptime value
used by 'vrtc_curtime()' could be different than the uptime value when
'vrtc_time_update()' actually updated 'base_uptime'.
Fix this by calculating and updating the (rtctime, uptime) tuple together.
Those functions are problematic, because there is no way to report
memory allocation problems without complicating the API, so we can
either abort or potentially return invalid results. None of which is
acceptable.
In most cases the caller knows the size of the name, so he can allocate
buffer on the stack and use snprintf(3) to prepare the name.
After some discussion the conclusion is to removed those functions,
which also simplifies the API.