Changes in arm

Changes in arm http://nxr.netbsd.org/rss/src/sys/arch/arm/ en Copyright 2005 Java gicv3: don't wait for RWP to clear when unblocking irqs The RWP bit does not track writes to ISENABLER<n>, only ICENABLER<n>. So don't bother waiting for it to clear when unblocking IRQs. This is ok; pic_block_irqs is expected to be synchronous (no IRQs allowed after the call) but not pic_unblock_irqs (some time can pass before an IRQ is raised). Sun Sep 13 15:00:00 UTC 2026 jmcneill Enable sdhc on Google i.MX8MQ Phanbell Fri Sep 11 01:00:00 UTC 2026 rxg s/contorol/control/ in comments and log messages. Tue Sep 01 17:00:00 UTC 2026 andvar s/virutal/virtual/ in comments. Tue Sep 01 18:00:00 UTC 2026 andvar uvm: Fix missed wakeups and reduce lock contention a little bit. This addresses two problems under heavy load: (a) uvm_wait would sometimes miss wakeups, causing various processes system to hang but then recover from `call wakeup(uvmexp+0x10)' in ddb: PR kern/58964: uvm: missing wakeup on uvmexp.free (b) Contention on uvmpd_lock would lead to so much time spent at IPL_SOFTBIO softint context in uvm_pageout_done spinning for uvmpd_lock that it would trip heartbeat panics: PR kern/60029: panic: cpu0: softints stuck for 16 seconds There are three intertwined parts to this, which I tried to split into separate commits, but I eventually decided it wasn't worth the trouble: 1. New ticket idiom for uvm_wait is needed to avoid missing a wakeup in the face of multiprocessing or kernel preemption: top: ticket = uvm_wait_prepare(); pg = uvm_pagealloc(...); if (pg == NULL) { uvm_wait("foo", ticket); goto top; } Without this, a wakeup by the page daemon that arrives after uvm_pagealloc has failed and before we have called uvm_wait will be lost, so we might hang indefinitely. Internally, the ticket is a 64-bit generation number, so it can't possibly overflow, that is advanced on any wakeup(&uvmexp.free). For LP32 platforms, we manage it with a seqlock/Lamport-type algorithm, so that reading it is cheap and unlocked. This may also help reduce contention on uvmpd_lock under heavy paging load by having uvm_wait skip the lock if there has been concurrent paging since uvm_wait_prepare. 2. Split uvmpd_lock into three different locks for different purposes: - uvmpd_kick_lock to coordinate threads or interrupt handlers wanting the page daemon to work with the page daemon wanting to sleep (poor beast, woken by any thread or interrupt handler, hence IPL_VM), - uvmpd_waiter_lock to coordinate the page daemon thread or paging completion soft interrupt handler threads waiting for memory to be freed up (thread/softint, hence IPL_SOFTBIO), and - uvmpd_pool_drain_lock to coordinate the page daemon thread with the pool drain thread (thread-to-thread, hence IPL_NONE). This way, anyone trying to kick the page daemon, and the page daemon's internal coordination with the pool drainer, won't incur any contention on the same lock that uvm_pageout_done is trying to take on paging completion to notify threads that they can try allocation again. 3. Use Dekker-synchronized atomic_swap fast paths to reduce time spent spinning on these locks when they are under contention. This way, if multiple callers (such as uvm_pageout_done and a thread trying to allocate memory) want to wake the page daemon, only one of them needs to take the lock most of the time. And if there's a lot of pageout activity because many threads want to allocate memory, uvm_pageout_done need only take the lock once to trigger a wakeup until the threads that have been waiting have noticed it. Additionally, move one call to uvm_availmem out from under any locks, to match the other calls to it -- this should help reduce the time anyone spends holding the locks, and thus the time everyone else spends waiting for it. Empirically, this seems to enable a heavily loaded bulk build to complete when before the softbio completions would starve softclk leading to heartbeat panics. Even without the heartbeat panics, we measured over 8sec running time for softint_dispatch for softbio, and extremely heavy contention on uvmpd_lock in uvm_pageout_done and uvm_wait. This isn't really a solution to the contention, however -- it just mitigates the contention. To scale to more CPUs in parallel, we'll probably need to devise a per-CPU or per-NUMA-cluster page daemon system, with each parallel daemon splitting up the work in some reasonable way that doesn't add more contention. Note: Before, all access to uvm_pagedaemon_waiters was serialized by uvmpd_lock. Now, all _writes_ are serialized by uvmpd_lock, but _reads_ are allowed unlocked. So the writes have to be issued with atomic_store_*, but there is no need for atomic_r/m/w_*; hence we can increment it with atomic_store_relaxed(&uvm_pagedaemon_waiters, atomic_load_relaxed(&uvm_pagedaemon_waiters) + 1); rather than needing the often costlier atomic_inc_uint(&uvm_pagedaemon_waiters); Sun Aug 23 22:00:00 UTC 2026 riastradh evbarm/ti_edma: clear events before transfer The sdmmc controller on the am18xx fires dma events even in PIO mode. Clear these event before starting a DMA transfer. Wed Aug 19 09:00:00 UTC 2026 yurix evbarm/am18xx: sdmmc driver A first version of the SD card driver for the TI am1808. The driver works fine, but the performance is lower than expected. Wed Aug 19 09:00:00 UTC 2026 yurix vchiq: Back port of the change commit 49bec49fd7f273ec114e2e533c1bb8f21a654aaf Author: Michael Zoran <mzoran@crowfest.net> Date: Sun Oct 30 05:55:07 2016 -0700 staging: vc04_services: remove vchiq_copy_from_user The vchiq_copy_from_user function is not portable and is consider "bad practice." Replace this function with a callback based mechanism that is passed downward on the stack. When it is actually time to copy the data, the callback is called to copy the data into the message. This callback is provided internally for userland calls through ioctls on the device. NOTE: Internal clients will need to be modified to work with the new internal API. Test Run: vchiq_test -p 1 vchiq_test -f 10 Both tests pass. Internal API Changes: Change vchi_msg_queue to: int32_t vchi_msg_queue(VCHI_SERVICE_HANDLE_T handle, ssize_t (*copy_callback)(void *context, void *dest, size_t offset, size_t maxsize), void *context, uint32_t data_size ); Remove: vchi_msg_queuev_ex vchi_msg_queuev These functions were not implemented anyway so no need to fix them. It's easier to just remove them. Signed-off-by: Michael Zoran <mzoran@crowfest.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Sun Aug 16 19:00:00 UTC 2026 skrll s/forcably/forcibly/ and couple more typos in comments. Sat Aug 15 11:00:00 UTC 2026 andvar s/staring/starting/ in comments and log messages. Fri Aug 14 09:00:00 UTC 2026 andvar evbarm/am18xx: expose the PSC as powerdomain Some peripherals in the DTB expect to manage clocks as power domains. Wed Aug 12 09:00:00 UTC 2026 yurix evbarm/ti_edma: start using fdt_dma In preparation for the am18xx sdmmc driver. Wed Aug 12 09:00:00 UTC 2026 yurix evbarm/ti_edma: enable on am18xx SoCs The linux dt-bindings allow two different ways to specify power/clock domains for the edma3. This commit adds support for the way used on the am18xx (powerdomains). Wed Aug 12 09:00:00 UTC 2026 yurix KNF comments Fri Jul 31 05:00:00 UTC 2026 skrll RCSId police Fri Jul 31 05:00:00 UTC 2026 skrll fix cvs revision comment not expanding Thu Jul 30 15:00:00 UTC 2026 yurix evbarm/am335x: initialize usb phy This fixes the USB devices not being detected part of PR port-evbarm/60404. Wed Jul 29 08:00:00 UTC 2026 yurix Does not use i2c_bitbang, so should not depend on it. Fri Jul 17 17:00:00 UTC 2026 thorpej Remove unused variable. Wed Jul 15 06:00:00 UTC 2026 martin Now that bad144 handling is isolated to the places that care about it, garbage-collect it from all of the various machine-dependent files that have been cargo-culting it around needlessly for 20+ years. Version bump to 11.99.7 because this changes the size of struct cpu_disklabel on some platforms, and that structure is exposed in the module ABI. Tue Jul 14 13:00:00 UTC 2026 thorpej