[AArch64] PR28877: Don't assume we're running after legalization when
creating vcvtfp2fxs
Summary:
The DAG combine transformation that was generating the
aarch64_neon_vcvtfp2fxs node was assuming that all inputs where legal
and wasn't accounting that the input could be a v4f64 if we're trying
to do the transformation before legalization. We now bail out in this
case.
All illegal types besides v4f64 were already rejected.
Fixes https://llvm.org/bugs/show_bug.cgi?id=28877
Reviewers: jmolloy
Subscribers: aemerson, rengolin, llvm-commits
Differential Revision: https://reviews.llvm.org/D23261
This fixes several ports on AArch64.
Requested by: andrew
MFC after: 3 days
- Correctly refresh Rx filter when AP (IBSS) vap is created after STA vap.
- Block any RCR updates during TSF correction (IBSS mode).
- Set CBSSID* bits during vap creation, not when it was started / stopped.
- Cache current state to prevent unnecessary register reads.
Tested with RTL8188CE, STA + AP mode.
* extend the keycache flag word to be 32 bits, not 16 bits
* add new key flags for transmit:
+ IEEE80211_KEY_NOIV: Don't insert IV in the payload when transmitting data frames;
+ IEEE80211_KEY_NOIVMGT: Don't insert IV in the payload when transmitting MIC frames;
+ IEEE80211_KEY_NOMIC: Don't insert MIC in the payload when transmitting data frames;
+ IEEE80211_KEY_NOMICMGT: don't insert MIC in the payload when transmitting management
frames.
* teach ieee80211_crypto_demic() about hardware decrypted frames:
+ if frames are hardware decrypted and the frame has failed MIC, treat it as a
michael failure.
+ if frames are hardware decrypted and the frame has stripped MIC, we can't check the
MIC in the payload - we don't have anything to compare it against.
This is only part of the work required to successfully transmit/receive
hardware crypto frames such as the qualcomm atheros 11ac offload chips.
There will be further work in the transmit and receive path before this
can be done by default.
Reviewed by: avos
Differential Revision: https://reviews.freebsd.org/D8364
adapter to work around bugs in TSO handling at this speed.
em_init_locked is called during first boot of the adapter and will
see that link_speed is unitialized, effectively turning off tso for
all cards at all speeds, which I believe was *not* the intent.
Move the handling of TSO deactivation to the link handler where we can
more effectively make the decision about what to do. In addition,
completely purge the TSO capabilities instead of disabling just CSUM_TSO.
Thanks to jhb for explanation of the hw capabilites api.
Thanks to royger and cognet for testing the 100Mbit failure case to
ensure that their adapters do indeed still work.
MFC after: 1 week
Sponsored by: Limelight Networks
from struct stat. We don't necessarily have permissions to see the
generation number and the host OS may not have st_gen in struct stat
anyway. Since the kernel assigns random numbers, there's nothing
meaningful about the generation that requires us to preserve it when
the file system image is created. With this change, all generation
numbers come from random() and that makes it easier to add support
for reproducible builds at some time in the future (i.e. by adding
an argument to makefs that changes the behaviour of random() so that
it always returns 0 or some predictable sequence).
Differential Revision: https://reviews.freebsd.org/D8418
The previous calculation used an approximation which was only valid in
cases where the means being compared were similar; this resulted in very
odd claims being made, e.g. that 0 +/- 0 is a difference of -100% +/- 1%
from 100 +/- 1.
The new calculation scales sample standard deviations by the means, and
yields approximately correct percentage difference bounds providing that
the reference population is bounded away from zero. (In the case where
the values being compared are not sufficiently bounded away from zero,
the distribution of ratios becomes much harder to calculate, and is not
likely to be useful anyway.)
Note that when ministat is used for its intended purpose of determining
whether two samples are statistically different, this change is unlikely
to have any noticeable effect; in such cases the means will be similar
enough that the correction applied here will be minimal.
In r297602, which included a __FreeBSD_version bump to 1100105, we changed
sed 'i' and 'a' from discarding whitespaces to conform with what GNU and
sysvish sed do.
There are arguments in favor of keeping the old behavior but the new
behavior is also useful for migration purposes. It seems important to at
least consider the case of developers depending on the previous behavior,
so add a CFLAG to enable the old behaviour.
PR: 213474
MFC after: 5 days
Compiler-rt and LLVM's libunwind provide a suitable replacement for
libgcc.a, libgcc_eh.a, and libgcc_s.so.
Remove the now-unused LLVM_LIBUNWIND block from gnu/lib/libgcc.
PR: 213480 [exp-run]
Reviewed by: brooks, ed
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D8189
is not going to recover until the system is reset. Treat it as a special
case and don't allow it to fall through to quasi-success.
Reviewed by: ken, imp
Obtained from: Netflix
MFC after: 3 days
the components are reset). Therefore retries are pointless. This is very
visible in SATL systems, for example an LSI SAS controller and a SATA HDD/SSD.
Reviewed by: ken
Obtained from: Netflix
MFC after: 3 days
Bay Trail has three banks of GPIOs exposed to userland as /dev/gpiocN,
where N is 1, 2, and 3. Pins in each bank are pre-named to match names
on boards schematics: GPIO_S0_SCnn, GPIO_S0_NCnn, and GPIO_S5_nn.
Controller supports edge-triggered and level-triggered interrupts but
current version of the driver does not have interrupts support
Requests which cannot be satisfied by allocators at boot time often
have unrealizable parameters. Waiting for the pagedaemon' start would
hang the boot if done in the thread0 context and just never succeed if
executed from another thread. In fact, for very early stages, sleep
attempt panics with obscure diagnostic about the scheduler state, and
explicit panic in vm_wait() makes the investigation much shorter by
cut off the examination of the thread and scheduler.
Theoretically, some subsystem might grab a resource to exhaustion, and
free it later in the boot process. If this unlikely scenario does
appear for real, the way to diagnose the trouble can be revisited.
Reported by: emaste
Reviewed by: markj
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Differential revision: https://reviews.freebsd.org/D8421
The audio controller in the H3 is more or less the same as A10/A20 except
some registers are shuffled around. The mixer interface, however, is
completely different between SoCs. Separate a10_mixer_class and
h3_mixer_class implementations are now made available. This will also make
adding support for other SoCs easier in the future.
Reviewed by: andrew, ganbold
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D8425
they shouldn't be.
I used this during driver bring-up to find that the Linux driver holds a
whole lot of locks whilst doing their equivalent of busdma operations.
If this works out well, it should be added to the other architecture busdma
implementations to aid in similar debugging.
Tested:
* bounce buffer and dmar busdma, Lenovo X230 laptop, all the internal
hardware
* ath(4) too
Discussed with: jhb
This is a long time coming. The general pieces have been floating around
in a local repo since circa 2012 when I dropped the net80211 support
into the tree.
This allows the per-chain RSSI and NF to show up in 'ifconfig wlanX list sta'.
I haven't yet implemented the EVM hookups so that'll show up; that'll come
later.
Thanks to Susie Hellings <susie@susie.id.au> who did the original work
on this a looong time ago for a company we both worked at.
Now that the changes to the dirname(3) function had some time to settle,
let's go ahead and use the same approach for replacing basename(3) by a
simple implementation that modifies the input string, thereby making it
thread-safe and guaranteed to succeed.
Unlike dirname(3), this function already had a thread-safe variant
basename_r(3). This function had its own set of problems, like having an
upper bound on the pathname length. Keep this function around for
compatibility, but remove most references from the man page. Make the
man page more similar to that of dirname(3).
As the basename_r(3) function is only provided by FreeBSD (and Bionic),
depending on its use is even more implementation defined than assuming
that basename(3) is thread-safe.
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D8382
be accessed by root. It uses unsigned ints instead of size_t
to preserve the ABI.
PR: 207627
Submitted by: ryan@ryanday.net (with slight tweaks)
MFC after: 1 month
