| Age | Commit message (Collapse) | Author |
|---|
|
From Mario Limonciello
ce258c74f8d95e81ce65f53775fcdcbc8ca090da in linux 5.15.y/5.15.16
63ad5371cd1e379519395c49a4b6a652c36c98e5 in mainline linux
|
|
OK mpi@
|
|
|
|
descriptors for explicit fencing
tested with libdrm's amdgpu_test syncobj timeline tests and vkcube on
intel broadwell with Mesa 21.3 (which hangs without sync file support
after the 'anv: Assume syncobj support' Mesa commit)
feedback and ok visa@
|
|
the leaf (<=cpuid_level) and always pass the subleaf. Delete the
CPUID_LEAF() calls made superfluous by always passing the subleaf.
ok mlarkin@
|
|
of two, but can be a multiple of two. Make use of this to achieve card
clock frequencies closer to the target frequency.
ok visa@
|
|
kern.wxabort=1 logs and kills programs after W^X violations.
At least sigexit() -> coredump() as well as the non-atomic increment of
ps_wxcounter require protection, so grab the big lock for the entire block.
This is part of the effort to unlock mmap(2)'s MAP_ANON case.
Feedback mvs claudio kettenis deraadt
OK kettenis
|
|
|
|
this may further help people with recent gen intel machines
|
|
-Wunneeded-internal-declaration when building drm with clang
|
|
|
|
The swap code path in uvm_aio_aiodone() is not holding the corresponding
page lock and shouldn't as long as anons are locked inside uvm_page_unbusy()
to handle the PG_RELEASED case.
Reported by Ralf Horstmann on bugs@
|
|
|
|
If the first mbuf of a chain in m_pullup is a cluster, check if the
cluster is read-only (shared or an external buffer). If so, don't
touch it and create a new mbuf for the pullup data.
This restores original 4.4BSD m_pullup, that not only returned
contiguous mbuf data of the specified length, but also converted
read-only clusters into writeable memory. The latter feature was
lost during some refactoring.
from ehrhardt@; tested by weerd@; OK stsp@ bluhm@ claudio@
|
|
we have passed a modified version of the contents in the capabilities
register if we wanted to override what sdhc(4) would otherwise read.
Unfortunately there's a second capabilities register that we did not
yet take into consideration, which is why to disable DDR50 support we
created a quirk flag in the softc. The sdhc(4) ACPI nodes have a way
to mask and set bits in both of the capabilities register, which is a
flexible approach to solving that issue and using that for our sdhc(4)
drivers even improves readability.
ok kettenis@
|
|
this is consistent with other drivers when they report their
underlying device being detached.
|
|
|
|
Tested on a PolarFire SoC.
|
|
Make `cpu' signed so that the possible return value -1 from
plic_get_cpuid() gets handled correctly in the (cpu < 0) condition.
This prevents plic_attach() from updating sc_contexts[] out of bounds.
When plic_get_cpuid() returns -1, ignore the entry and continue
processing. The error is not fatal. It is normal that secondary CPUs
are not found when running a non-MULTIPROCESSOR kernel on
a multiprocessor machine.
OK kettenis@
|
|
Adjust end condition so that sfcc_cache_wbinv_range() would not spin
when (pa + len) is not cache line aligned.
While here, fix parameter types of sfcc_cache_wbinv_range() as
suggested by kettenis@.
OK kettenis@ (earlier version), miod@
|
|
There is no functionnal change as the former is just a wrapper around the
latter. However upper layer of UVM do not need to mess with the internals
of the page allocator.
This will also help when a page cache will be introduced to reduce contention
on the global mutex serializing acess to pmemrange's data.
ok kettenis@, kn@, tb@
|
|
our 5.10 drm used (via a task) by changing from timeout_set() to
timeout_set_proc()
irq work is supposed to be in interrupt context but some path used on
inteldrm with gen 9 graphics sleeps.
Reported by Agnosto Dvonik on kaby lake and jcs@ and aja@ on comet lake.
For Agnosto Dvonik this occured when running firefox or mpv, for jcs@
when playing a video in firefox, for aja@ after 5 min of running gnome.
aja@ has run gnome for many hours with this change without problem.
|
|
|
|
The previous limit of VM_PHYSSEG_MAX ranges (16) was proving too small for
newer machines. This diff reorganizes the hibernate signature block to allow
for 22 ranges by removing the kernel version comparison and replacing it
with a SHA of several unique kernel features (the version string and several
addresses of functions not inside the same .o).
Reported by claudio@, who also helped fix some issues in the diff. Input
from deraadt@ as well.
Tested by myself and claudio on a variety of machines. Only compile tested on
i386 as I have no more S4-capable i386 hardware anymore.
ok claudio@
|
|
removed from other archs in 2020 before riscv64 was imported
|
|
From Nathan Chancellor
48d56b00c35266b00ec92aaf0db5b71ce1f27702 in linux 5.15.y/5.15.15
2e70570656adfe1c5d9a29940faa348d5f132199 in mainline linux
|
|
|
|
aside from getting rid of a use of the ifq_deq_begin/commit/rollback
stuff, it simplifies the code a lot. im also sick of having the
diff in my tree.
|
|
this builds on the mpsafe kq/kevent work visa has been doing.
normally kevents are notified by calling selwakeup, but selwakeup
needs the KERNEL_LOCK. because bpf runs from all sorts of contexts
that may or may not have the kernel lock, the call to selwakeup is
deferred to the systq which already has the kernel lock. while this
avoids spinning in bpf for the kernel lock, it still adds latency
between when the buffer is ready for a program and when that program
gets notified about it. now that bpf kevents are mpsafe and bpf_wakeup
is already holding the necessary locks, we can avoid that latency.
bpf_wakeup now checks if there are waiting kevents and notifies
them immediately. if there are no other things to wake up, bpf_wakeup
avoids the task_add (and associated reference counting) to defer
the selwakeup call.
selwakeup can still try to notify waiting kevents, so this uses the
hint passed to knote() to differentiate between the notification
from bpf_wakeup and selwakeup and returns early from the latter.
ok visa@
|
|
loop and adjust another polling loop. This makes accessing the cs42l83
audio codec work. Probably removing the reset is enough to fix this
but the other changes bring us in line with the Linux driver which had
some decent RE done to it recently.
ok patrick@
|
|
pointed out by Brad
|
|
So don't provide atomic64_cmpxchg() on powerpc and on other architectures
make its implementation similar to atomic64_xchg().
This makes the tree build again on macppc.
suggested by & ok jsg@
|
|
new hardware support includes
Intel
ehl/Elkhart Lake (embedded)
jsl/Jasper Lake (atom)
rkl/Rocket Lake (desktop)
AMD
van gogh APU (gfx1033)
yellow carp / rembrandt APU (gfx1035?)
Ryzen 6000 APU
navy flounder / navi 22 (gfx1031)
RX 6700, RX 6700 XT, RX 6700M, RX 6800M, RX 6850M XT
dimgrey cavefish / navi 23 (gfx1032)
Pro W6600, Pro W6600M, RX 6600, RX 6600 XT, RX 6600M,
RX 6600S, RX 6650M, RX 6650M XT, RX 6700S, RX 6800S
beige goby / navi 24 (gfx1034)
RX 6500 XT, RX 6400, RX 6500M, RX 6300M
Thanks to the OpenBSD Foundation for sponsoring this work
niklas@ for helping with ttm and amdgpu and patrick@ for adapting
rockchip drm.
|
|
Use bd_mtx to serialize bpf knote handling. This allows calling the
event filter without the kernel lock.
OK mpi@
|
|
The lookup should not fail because the kernel lock should prevent
simultaneous detaching on the vnode layer. However, most other device
kqfilter routines check the lookup's outcome anyway, which is maybe
a bit more forgiving.
OK mpi@
|
|
address space of the SPMI PMU to prevent the machine from immediately
starting up again.
The implementaton makes aplpmu(4) provide powerdownfn(), which sets the
magic bit and then chains into cpuresetfn(). It also makes aplsmc(4)
provide cpuresetfn() to reset the machine via the SMC. Resetting via
the watchdog works as well (and will powerdown the machine if the magic
bit is set) but letting the SMC handle things might do some other
required steps.
ok patrick@
|
|
OK stsp@ (without assuming any responsibility for NFS)
|
|
|
|
ok kettenis@
|
|
and remaining time to battery full and empty and feed these values to
apm(4)
ok kettenis@
|
|
|
|
wide and some _DSD properties depend on it.
ok kettenis@
|
|
The original purpose of ieee80211_find_node_for_beacon() was to avoid
storing duplicate nodes with the same source MAC address in a hash table.
Later on, our node table data structure was changed from a hash table
to an RB tree. The RB tree can only store a single node per MAC address.
However, find_node_for_beacon() was kept regardless, now documented to
serve a different purpose.
Its new purpose is to tell apart different nodes which happen to use
the same MAC address and hence cannot both be stored in the RB tree.
The idea is to filter such duplicate nodes out during a scan. But colliding
nodes are told apart by RSSI and channel, and either may change over time.
So this does not really prevent duplicate MAC addresses from causing issues.
The code which decides which node is "better" can erroneously match an
AP against itself, in case the AP uses a hidden SSID. This caused
workarounds for hidden SSID to pile up over time.
Just a bit further down, the code looks up the same node again and
performs all of the intended node state updates. Simply skipping the
ieee80211_find_node_for_beacon() check makes such state updates work.
ok tobhe@
|
|
|
|
ok kettenis@ deraadt@
|
|
|
|
|
|
fixes accessing eMMC on Acer Swift 1 SF114-34
problem reported and fix tested by Sven Wolf
|
|
ok dv@ mlarkin@
|
|
ok phessler@, deraadt@, miod@
|
